Document Actions

IJEGE-11_BS-Stock-&-Bellugi

background image
Italian Journal of Engineering Geology and Environment - Book www.ijege.uniroma1.it © 2011 Casa Editrice Università La Sapienza
1013
DOI: 10.4408/IJEGE.2011-03.B-110
AN EMPIRICAL METHOD TO FORECAST THE EFFECT OF STORM IN-
TENSITY ON SHALLOW LANDSLIDE ABUNDANCE
J
onatHan
d. STOCK
(*)
& d
ino
BELLUGI
(**),(***)
(*)
U.S. Geological Survey, 345 Middlefield Road, Menlo Park, CA, 94025, USA; jstock@usgs.gov
(**)
University of California, Department of Earth & Planetary Science, Berkeley, CA, USA
(***)
University of California, Department of Computational Science and Engineering, Berkeley, CA, USA
alii, 2009), and involve failure of soil and weathered
bedrock layers. They are commonly fractions of a me-
ter to several meters deep, rarely exceeding ~10 m in
width or length. A fraction of these failures mobilize
as debris flows, mixtures of soil, water and rock that
sweep down steep valleys at high velocities and destroy
homes and infrastructure with little warning.
Varved sediments in the Santa Barbara basin,
southern California, may contain the geologic record
of the largest storms over the past millennium (s
CHim
-
melman
et alii, 2003). Layers interpreted as storm
deposits have a periodicity of several hundred years,
not unlike earthquake recurrence intervals for the San
Andreas Fault. These deposits are substantially thick-
er than those associated with storms of January 1969,
the most recent historic events to generate widespread
landslides in southern California. If layer thickness
scales with storm intensity, these layers imply that
southern California has experienced storms that are
massive compared to our recent historical experience.
Unlike seismic hazard maps available for some large
faults, we cannot begin to quantify the magnitude of
landslide hazards that will accompany such a storm.
Put simply, we cannot answer the question of how
many shallow landslides California's largest storms
would trigger
We are beginning to understand more about what
such massive storms might look like. Recent work
(R
alPH
et alii, 2006; n
eiman
et alii, 2008) illustrates
that some of California’s largest storms are atmos-
ABSTRACT
We hypothesize that the number of shallow land-
slides a storm triggers in a landscape increases with
rainfall intensity, duration and the number of unstable
model cells for a given shallow landslide susceptibil-
ity model of that landscape. For selected areas in Cali-
fornia, USA, we use digital maps of historic shallow
landslides with adjacent rainfall records to construct
a relation between rainfall intensity and the fraction
of unstable model cells that actually failed in historic
storms. We find that this fraction increases as a power
law with the 6-hour rainfall intensity for sites in south-
ern California. We use this relation to forecast shallow
landslide abundance for a dynamic numerical simula-
tion storm for California, representing the most extreme
historic storms known to have impacted the state.
K
ey
words
: landslide susceptibility, storm intensity, landslide
frequency
INTRODUCTION
Intense rainfalls from historic storms have trig-
gered widespread shallow landslides in the coastal
mountains of California, USA (e.g., R
iCe
& f
oGGin
,
1971; t
ayloR
& b
Rabb
, 1972; C
amPbell
, 1975; n
ilsen
& t
uRneR
, 1975; t
ayloR
et alii, 1975; n
ilsen
et alii,
1976a,b; w
entwoRtH
, 1986; e
llen
& w
ieCzoRek
,
1988; G
odt
et alii, 1999; C
oe
& G
odt
, 2001). These
landslides may occur during or soon after the most in-
tense rainfall (e.g., C
amPbell
, 1975, m
ontGomeRy
et
background image
J.D. STOCk & D. BELLUGI
1014
5th International Conference on Debris-Flow Hazards Mitigation: Mechanics, Prediction and Assessment Padua, Italy - 14-17 June 2011
A number of recent workers have used intensive site
characterizations to calibrate physically-based models
of slope stability (e.g., G
odt
et alii, 2008). Where they
have been successfully calibrated at field sites, such
models represent the optimal method to forecast land-
slide abundance from simulation storms over small
areas. Extending such forecasts across broad regions
(states or nations) may currently be beyond our ca-
pacity, because of the profound difficulty in capturing
spatial variations in influential model parameters (e.g.,
soil depth, transmisivity, friction angle and cohesion)
An alternate approach would use the historical
record of landslide occurrences to create threshold
values of rainfall intensity for given durations (e.g.,
Caine, 1980; C
annon
& e
llen
, 1985, 1988; w
ilson
&
w
ieCzoRek
, 1995; C
asadei
et alii, 2003; G
odt
et alii,
2006; G
uzzetti
et alii, 2007; 2008). The exact loca-
tion of these thresholds remains problematic, in part
because rainfall analyses do not account for important
state and material property differences including ini-
tial moisture conditions, the rate at which rain enters
the soil, the rate at which it leaks into the underlying
bedrock, the soil strength properties, and the cohesion
due to vegetation roots. They also do not account for
variations in topographic convergence and slope that
influence soil stability through the routing of sub-
surface water and the magnitude of the downslope
force. All of these variables influence the stability of
hillslopes, and we expect them to vary spatially. As a
consequence, there is great uncertainty as to the ex-
act location of the threshold between rains that cause
pheric rivers of moisture that originate in the trop-
ics and convey vast amounts of water vapor towards
California in a narrow jet (e.g., 100-200 km wide) of
moisture. Where these jets strike California, they can
deliver record rainfalls and river discharges over the
course of 1-3 days. Consequently, depending on their
intensity and occupation time, they can cause wide-
spread landsliding (e.g., Southern California storm of
January 1969). The storms of 1861/62, largest in Cali-
fornia’s recorded history, may represent the most re-
cent great atmospheric river to hit California, and may
be an example of the kind of storms that are recorded
in the Santa Barbara basin deposits.
As part of an exercise to simulate the hazards
from a great storm, d
ettinGeR
et alii (2011) simulated
an atmospheric river storm (Arkstorm) using 1969 and
1985 atmospheric data as initial and boundary condi-
tions in a dynamic meteorological model. They used
a global climate model (GCM) to capture the large-
scale dynamics with a 150-km grid cell domain. With-
in the GCM, they nested a more detailed model (WRF,
or weather research and forecast) within the GCM to
simulate the storm over California. WRF model out-
put includes 2-km cells of hourly rainfall over a large
portion of southern California. Part of the motivation
for constructing this storm simulation was to under-
stand the hazards associated with such storms, includ-
ing shallow landslides.
Most shallow landslide susceptibility models tend
to overpredict the number of failures actually triggered
by historic storms (e.g., Fig. 1), even where calibrated.
Fig. 1 - January 2005 failures in
Ventura on 1-m contour
map from LiDAR. Areas
of instability represented
by green (least unstable),
yellow and red (most un-
stable) colors. Like most
models, this shallow slope
stability model (Shalstab)
vastly over-predicts the
areas of instability (~ 1%
of unstable areas actually
failed)
background image
AN EMPIRICAL METHOD TO FORECAST THE EFFECT OF STORM INTENSITY ON SHALLOW LANDSLIDE ABUNDANCE
Italian Journal of Engineering Geology and Environment - Book www.ijege.uniroma1.it © 2011 Casa Editrice Università La Sapienza
1015
2001), spring snowmelt (e.g., m
oRton
& C
amPbell
,
1978), and large winter storms. Winter storms include
atmospheric river events and are arguably responsible
for the most widespread historic landslide events (e.g.,
January 1969). Morton mapped shallow landslides
and debris flow runouts from winter storms using air
photos. Alvarez digitized landslide initiation points
and debris flow runouts by transferring air photo in-
terpretations to digital raster graphs of USGS 7.5’
quadrangles by hand. In the Santa Paula area, m
oRton
& a
lvaRez
mapped landslides that occurred during
storms in 1969, 1998, 2001 and 2005 (Fig 2). These
shallow landslides occur in weak Plio-Pleistocene
rocks with high clay and silt contents (e.g., Tan et al.,
2004). Landslides in these weak rocks tend to be wide
(> 3m), shallow (< 2m) failures of colluvium that ex-
tend to whole sections of hillslopes. In the Sunland
area of Los Angeles, they mapped landslides and de-
bris flows from 1998, and 2005 storms. Rocks in this
area are also predominately weak Plio-Pleistocene
sediments with high silt and clay fractions.
In the San Francisco Bay area of northern Cali-
fornia, USA, w
entwoRtH
(1986) used air photographs
to map shallow landslides and debris flows triggered
by storms in December 1955 and January 1982 in
the Montara area. In this detailed area, Wentworth’s
mapped failures are more accurately located than the
failures, and those that do not. Recent work (G
odt
et
alii, 2008; C
HleboRad
et alii, 2008) illustrates that
progress on these thresholds will require better data
on transient water content and soil properties..
We use elements of both approaches to attempt to
forecast landslide abundance for the Arkstorm simu-
lation. Despite uncertainty about the exact rainfall
thresholds for landslide generation, most believe that
storms with more intense, longer lasting rainfalls will
trigger more shallow landslides. We use historic rain-
fall records and digital maps of consequent shallow
failures to estimate a storm rainfall metric that scales
with shallow landslide abundance. We refine this scal-
ing by using topographically-based landslide suscepti-
bility models to account for regional variations in sta-
bility associated with varying topography (e.g., slope or
convergence). We present these findings as a graph of
the fraction of model cells predicted to be unstable as a
function of rainfall intensity. A regression through this
relation allows us to extrapolate it across the storm rain-
fall domain to forecast shallow landslide abundance.
METHODS
MAPPING
In southern California, USA, rainfall-induced
landslides have been triggered by short-duration, high-
intensity monsoonal storms (e.g., Morton and Hauser,
Fig. 2 - Map of Santa Paula analysis area, southern California showing historic landslides mapped by Morton and Alvarez.
Failures date from 1969 (red), 1998 (green), 2001 (blue), and 2005 (yellow). Rain gage locations shown by white
dots. Relative landslide susceptibility shown by yellow (low) to red( high) colors from Shalstab model with zero
cohesion and 45 degrees friction angle
background image
J.D. STOCk & D. BELLUGI
1016
5th International Conference on Debris-Flow Hazards Mitigation: Mechanics, Prediction and Assessment Padua, Italy - 14-17 June 2011
The nearest other rain gage with an hourly record is at
Pacoima dam, 5 km to the north. This hourly record
commences in 1997, and we use it to characterize
rainfall intensities during the 1998 and 2005 land-
slides in Sunland
Prior to the 2002 installation of an hourly sta-
tion (Spring Valley) near Pilarcitos, northen Califor-
nia, the closest hourly rainfall data for the Montara
analysis area was from San Francisco International
Airport (SFO). San Francisco Public Utilities Com-
mission (SFPUC) has operated four daily rain gages
in the Montara area, of which the Pilarcitos dam gage
(1916-present) is the closest. The other three gages
are in upper and lower Crystal Springs reservoir, and
San Andreas reservoir, along the San Andreas Fault.
We use the Pilarcitos record to rank the 24- and 48-
hour storm intensity since 1916, and the Spring Valley
RAWS station to calculate hourly data for the 2005
storm. We apply the pattern of rainfall at SFO hourly
gage to the Pillarcitos daily data to estimate 12-hour
rainfall intensity at Pillarcitos during the December
1955 and January 1982 storms. We use the daily totals
regional maps of landslide initiation points for the en-
tire Bay Area region in e
llen
& w
ieCzoRek
(1988).
Stock used digital imagery from before and after the
December 31, 2005 storm to add recent landslide
headscarps, lateral scarps and displaced mass (or ru-
nout). He used air photos from 1943 and 1973 to esti-
mate the number of shallow landslides within this area
(see Table 1). Stock also used 1-m LiDAR and time
sequence digital imagery to map shallow and deep-
seated landslides in the Ventura region (e.g., Fig. 1).
HISTORIC AND SIMULATED RAINFALL
Just to the south of the Santa Paula analysis area
in southern California, the Wheeler rain gage records
hourly and daily rain beginning in 1966, although it
lacks hourly data for a 2001 storm. The Santa Paula
gages record hourly data since 1961, and include all
of the storms of interest. We use the Wheeler record to
rank the 24- and 48-hour storm history, and the Santa
Paula gages for the hourly analysis. In the Sunland
area, the closest hourly rain gage at Hanson dam is in
a rain shadow from the Verdugo hills to its southeast.
Tab. 1 - Top 20 daily rainfall events in calibration areas
background image
AN EMPIRICAL METHOD TO FORECAST THE EFFECT OF STORM INTENSITY ON SHALLOW LANDSLIDE ABUNDANCE
Italian Journal of Engineering Geology and Environment - Book www.ijege.uniroma1.it © 2011 Casa Editrice Università La Sapienza
1017
allows estimates of the effect of soil strength (friction
and cohesion) on stability, but the spatial variability of
these parameters due to soil properties and vegetation
has resisted quantification. As a consequence, its use
on a regional area usually neglects variations in soil
strength and hydraulic properties. It also assumes a
steady-state balance between rainfall and subsurface
flow, a condition that is exceedingly unlikely during
the fluctuating rainfall intensities that accompany most
storms. A further limitation of the model is its neglect
of lateral forces that likely play a role in convergent
areas of the topography..
Despite these very real limitations, Shalstab offers
an ability to estimate relative differences in landslide
susceptibility due to differences in both slope and top-
ographic convergence. Where landslides are concen-
trated in hollows, it is particularly successful at pre-
dicting the locations of shallow failures. To the extent
that landslides occupy planar or divergent topography
as resolved by the DEM, Shalstab is less likely to cap-
ture instability. 10-m DEM’s tend to under-represent
both convergence and slope within 1-2 grid cells of
the ridgeline. As a consequence, landslides that initi-
ate close to ridge-tops are rarely predicted by 10-m
Shalstab runs. In some landscapes, this fraction may
approach 25% or so of the total landslide inventory.
We compiled 10-m DEM’s from the National
Elevation Dataset (NED), and removed sinks using
ArcGIS’s sink removal algorithm. We ran the default
version of Shalstab on 10-m USGS grids of Califor-
nia, assuming zero cohesion and a friction angle of 45
degrees (b
elluGi
et alii, this volume). These default
parameters have a tendency to maximize the areas of
likely instability, a useful result for comparing suscep-
tibility between landscapes using slope and conver-
gence as topographic metrics. They are not meant to
represent site-specific values of cohesion, whose use
would reduce the total area of instability.
RESULTS
RAINFALL
Table 1 ranks the top twenty daily rainfall events
in Santa Paula and Montara by their 48-hour totals
(peak rainfall day plus previous day). These days ap-
pear to encompass the major known landslide produc-
ing storms at both sites, although the Montara daily
totals omit a high intensity storm that caused a few
landslides observed in 1973 airphotos. Within these
from the surrounding SFPUC gages to evaluate the
presence or absence of orographic forcing in the 1955
and 1982 storms..
We used the daily data to calculate the anteced-
ent rainfall from 1
st
September up to the day of storm
(summer rainfall totals are commonly negligible in
these areas). Numerous previous studies (e.g., C
amP
-
bell
, 1975; n
ilsen
et alii, 1977; w
ilson
et alii, 1993)
reported that 10-15” (254-381 mm) of rainfall were
required before storms triggered shallow landslides.
Larger antecedent totals are presumed to correspond
to higher matric water content (i.e., lower suction val-
ues in the soil), reducing the amount of rain and time
required to generate pore water pressure in the soil.
We used daily totals to calculate the sum of rainfall to
the fieldsites from the previous 7 days, or continuous
rainfall period (days with non-zero rain). We hypothe-
size that high 7-day precursor rainfalls are likely to re-
sult in very low soil suction, so that additional rainfall
infills pores, generating positive pore water pressure.
We also calculated the rainfall total preceding the day
of peak rainfall intensity. We hypothesized that high
prior-day rainfalls in combination with high anteced-
ent rain might lead to rapid generation of positive pore
water pressures in many areas
For the Arkstorm simulation storm, d
ettinGeR
et
alii (2011) provided hourly rainfall totals on a 2-km
grid in southern California, and a 6-km grid over the
whole state. We imported their ASCII data into Arc-
GIS and created hourly 2.5 km grids of rainfall inten-
sity for storm simulation days 6-8, corresponding to
the January 26-28, 1969 period of most intense rain-
fall. We increased the grid-cell size to 2.5 km to cover
gaps in data introduced by gridding their lattice which
is oriented at ~ 45 degrees from N-S
We used Shalstab to estimate relative susceptibility
to shallow landslides initiated by storm rainfall. Shal-
stab (d
ietRiCH
et alii, 1995; 2001; m
ontGomeRy
& d
i
-
etRiCH
, 1994; b
elluGi
et alii, this volume) uses an infi-
nite slope approximation, and assumes steady input of
rainfall that infiltrates the soil and flows down-slope in
the subsurface, parallel to the gradient (Dupuit-Forche-
imer assumption). On a cell-by-cell basis, it calculates
the ratio of drainage area to the product of cell width
and sine of the slope, an approximation of the ratio of
rainfall rate (q) to the depth integrated hydraulic con-
ductivity (T). The q/T ratio controls saturation in the
soil column, a measure of relative instability. Shalstab
background image
J.D. STOCk & D. BELLUGI
1018
5th International Conference on Debris-Flow Hazards Mitigation: Mechanics, Prediction and Assessment Padua, Italy - 14-17 June 2011
top twenty rainfall events (all above 4 mm/hour 24-
hour intensity), landslide occurrences in both Santa
Paula and Montara have antecedent rainfalls ex-
ceeding 400 mm. This value is consistent with those
from previous studies (C
amPbell
, 1975; n
ilsen
et
alii, 1977) on large winter storms. Above this ante-
cedent rainfall, neither area has a consistent relation
between 24-hour intensity and landslide abundance.
Landslide abundance at Montara does scale with 48-
hour intensity, although this relation does not hold
for Santa Paula because of the short duration, high-
intensity 1998 storm. Landslide abundance at Santa
Paula scales with the 7-day prior rainfall totals, but
again this relation does not hold at Montara. Daily or
multi-day rainfall cannot be used alone as a proxy for
landslide abundance at these sites.
Table 2 lists the hourly rainfall intensities for the
peak rainfall intensity day associated with each of our
landslide inventory storms. Values for some of the an-
tecedent rainfalls are somewhat different from Table 1
where hourly data are available at Santa Paula. Peak
hourly intensities on these days range from 10-34 mm/
hour. There is no clear relation between these short
duration intensities and landslide abundance. For in-
stance, a peak hourly intensity of 25 mm/hr in 1998
produces fewer landslides at Santa Paula than the
lower maximum hourly intensity of 20 mm/hr from
the 1969 storm. Figure 3 summarizes intensity and du-
ration for landslide-producing storms at Santa Paula.
The figure illustrates that although the 1998 storm was
intense at short time scales, the 1969 storm has higher
intensities at all periods greater than 1-2 hours. At 3-8
hour rainfall durations, relative landslide abundance in
Table 1 scales with storm intensity at Santa Paula. The
scaling most closely mimics landslide abundance at
6-hour durations, where similar landslide abundances
mimic similar values for 1998 and 2005 storm
intensities. For this reason, we choose the 6-hour in-
tensities from Santa Paula as a storm metric that scales
with landslide abundance in this region.
Analysis of SFO and Spring Valley gages near
Montara indicated that intensities at SFO are system-
atically lower than at the higher elevation Spring Val-
ley gage, closest to the landslide analysis area. These
comparisons, in addition to those showing the strong
orographic effect on daily rainfall between Pilarcitos
and lower elevation gages for the 1955 event, indicate
that the 6-hour intensity pattern detected in the Santa
Paula data cannot be reconstructed for the Montara
area. For this reason, we restrict our simulations to the
southern California sites.
We used the hourly rainfall data from the Ark-
Tab. 2 - Hourly rainfall and SHALSTAB results in calibration areas
Fig. 4 - Map of maximum 6-hour rainfall intensi-
ty for the Jan 24-26 Arkstorm simulation.
Santa Paula and Sunland analyses areas
shown by white polygons. 2.5 km grid
cells are generated by averaging hourly
rainfalls over a 6-hour moving window,
and then extracting the maximum value
at any point on the grid
background image
AN EMPIRICAL METHOD TO FORECAST THE EFFECT OF STORM INTENSITY ON SHALLOW LANDSLIDE ABUNDANCE
Italian Journal of Engineering Geology and Environment - Book www.ijege.uniroma1.it © 2011 Casa Editrice Università La Sapienza
1019
maximum 6-hour rainfall intensity for the Arkstorm.
We derive a matching grid whose values represent the
number of unstable 10-m cells within each 2.5 km
2
cell. We multiply these grids together to obtain a grid
showing the number of shallow landslides for the
simulation storm.
The pattern of landslide numbers in Figure 7 mimics
the pattern of 6-hour maximum rainfall intensity of
Figure 3. Santa Ynez, Santa Monica and San Gabri-
el ranges are characterized by landslide densities of
1-3,000 per cell. The analysis areas of Santa Paula and
Sunland have densities of 200-1000 landslide per cell,
reflecting their role in generating (1). Lowland areas
have densities of 10 or less, largely concentrated at cut
banks on rivers or terraces.
DISCUSSION AND CONCLUDING RE-
MARKS
How generalizable is Figure 7? The weak rocks char-
acterizing the Santa Paula calibration area, and Sun-
land, have soils that fail at densities that tend to exceed
those observed in soils developed on older, stronger
lithologies. Widespread areas of very thin soils and
bedrock in the steepest parts of the Santa Ynez and
San Gabriel Mountains are also not likely to fail as
storm simulation (d
ettinGeR
et alii, 2011) to calculate
the grid of maximum 6-hour rainfall intensities across
the Southern California landscape (Fig. 4). This fig-
ure mirrors the peak hourly intensity grid in that the
highest rainfall intensities are distributed at mountain
fronts (e.g., Santa Ynez, Santa Monica, San Gabriels)
where orographic effects intensify rainfall rates. Note
the rainfall intensity low just to the south of the Sun-
land analysis area, which is where the Hanson dam
rain gage is located.
SHALSTAB LANDSLIDE MODEL
Figure 5 illustrates a subset of the the pattern of
shallow landslide susceptibility for southern Califor-
nia. At this small scale the visible pattern is dominated
by red and purple areas that correspond to steep, often
cliff-forming topography in the Santa Ynez and San
Gabriel ranges. To the extent that these areas are dom-
inated by rock outcrop and patchy colluvium, they
may actually be less likely to fail as shallow landslides
A MODEL FOR FORECASTING CRUDE LAND-
SLIDE FREQUENCY GIVEN RAINFALL SIMU-
LATIONS
Figure 6 plots the fraction of unstable Shalstab
cells that actually failed versus the maximum 6-hour
intensity. At the Santa Paula site, as 6-hour storm in-
tensity increased from ~ 5 mm/hr (2001) to 16 mm/hr
(1969), the number of failures increased and the frac-
tion of the unstable areas predicted by Shalstab that
actually failed increased from 0.0015 to 0.025 as the
power law:
f = .00001 I
6-hr
2.7
The 2005 data from Montara illustrate that, at the very
least, the intercept from (1) may not apply to northern
California sites. Consequently, we limit our analyses
to the southern California rainfall simulation area.
We apply (1) to each of the 2.5 km
2
cells representing
Fig. 5 - Map of Ventura and Santa Paula area illus-
trating Shalstab output for 45 degrees fric-
tion angle and zero cohesion on the 10-m
DEM. Relative landslide susceptibility
shown by yellow (low) to red (high) colors.
Note broad areas of red that correspond to
cliff-forming units in Tertiary sedimentary
units. Some of these areas are likely char-
acterized by rockfall processes, not shallow
landslides
Fig. 6 - Plot of 6-hour rainfall intensity vs. fraction of
unstable model cells that failed. Regression
from Santa Paula data
background image
J.D. STOCk & D. BELLUGI
1020
5th International Conference on Debris-Flow Hazards Mitigation: Mechanics, Prediction and Assessment Padua, Italy - 14-17 June 2011
provide an expectation for the rainfall intensities and
durations that generate large numbers of landslides.
Two such sites are currently active in the Bay Area as
part of the U.S. Geological Survey Bay Area Landslide
Task (BALT). They are providing data on soil mois-
ture, pore water pressure and local hourly rainfall that
will help us calibrate the next generation of forecasting
models for landslide thresholds and abundance.
ACKNOWLEDGEMENTS
We thank David Strong for compiling and merging the
10-m data. Maiana Hanshaw and Nichole Kneupprath
compiled much of the rain gage data. Bruce McGurk
of SFPUC provided the wonderfully long daily rain-
fall records of Pillarcitos and other SFPUC gages in
the Montara area. Mike Dettinger provided ASCII
files and Fortran code to extract the storm simula-
tion data. Geoff Phelps and Bruce Chuchel modified
this code to extract hourly values. This research used
resources of the National Energy Research Scientific
Computing Center, which is supported by the Office
of Science of the U.S. Department of Energy under
Contract No. DE-AC02-05CH11231. Further techni-
cal support was offered by the Berkeley UPC group.
We thank Brian Collins, Douglas Morton, Mark Reid,
and Carl Wentworth who provided insightful reviews.
shallow landslides during large rainfall events. Fig-
ure 7 shows the regions of weak Quaternary, Pliocene
and Miocene rocks in southern California that can be
expected to have similar, if not exactly the same at-
tributes as the calibration area. These hachured areas
encompass many of the residential areas of southern
California. High landslide numbers forecast within
these areas are a reflection of the abundance of failures
observed during the winter storms of 1969.
Figure 7 and equation 1 are crude, preliminary efforts at
forecasting the regional effects of large storms on shal-
low landslide abundance by combing rainfall metrics
with a shallow landslide susceptibility model. They do
not account for deep-seated landfslides, or steep areas
of bedrock or patchy soil where rockfall predominates.
As is evident from the lack of hourly data at Montara,
testing the generality of the form of (1) will require ac-
curately mapping the distribution and style of failures
in the vicinity of hourly rain gages. It is clear from re-
cent work (e.g., Baum et al., 2005; G
odt
et alii, 2008;
C
HleboRad
et alii, 2008) that understanding which
rainfall intensities generate positive pore water pres-
sures increasing the probability of failure will require
monitoring efforts in some of the historic mapping ar-
eas. In combination with geotechnical and hydraulic
properties of soils, these kinds of monitoring data will
Fig. 7 - Quaternary, Pliocene and Miocene sedimentary rocks (hachured area) superimposed on the forecast for land-
slide abundance resulting from the Arkstorm. Cells within the hachured area have rock types similar to the
calibration area at Santa Paula. Cells outside these zones likely overestimate the number of landslides be-
cause 1) rock units produce stronger soils and 2) different processes (e.g. rockfall) dominate erosion. Gray
areas have no unstable cells in 10-m data. Calibration (Santa Paula) and test (Sunland) areas shown by white
polygons
background image
AN EMPIRICAL METHOD TO FORECAST THE EFFECT OF STORM INTENSITY ON SHALLOW LANDSLIDE ABUNDANCE
Italian Journal of Engineering Geology and Environment - Book www.ijege.uniroma1.it © 2011 Casa Editrice Università La Sapienza
1021
REFERENCES
b
aum
R.l., G
odt
J.w., H
aRP
e.l., m
C
k
enna
J.P. & m
C
m
ullen
,s.R. (2005) - Early warning of Landslides for Rail Traffic
Between Seattle and Everett, washington, USA. In H
unGR
o., f
ell
R., C
outuRe
R. & b
eRnHaRd
e.,
eds
., Landslide Risk
Management, Proceedings of the 2005 International Conference on Landslide Risk Management: New York, A.A. Balkema,
p. 731-740.
b
elluGi
d., d
ietRiCH
we., s
toCk
J., m
C
k
ean
J., k
azian
b. & H
aRGRove
P. (this volume) - Spatially explicit shallow landslide
susceptibility mapping over large areas. Proceedings of the 5
th
International Conference on Debris-Flow Hazards Mitiga-
tion: Mechanics, Prediction and Assessment, Italian Journal of Engineering Geology and Environment.
C
aine
n. (1980) - The rainfall intensity-duration control of shallow landslides and debris flows. Geogr. Ann. 62A: 23–27.
C
amPbell
R.H. (1975) - Soil slips, debris flows, and rainstorms in the Santa Monica Mountains and vicinity, southern California.
U.S. Geological Survey Prof. Paper 851, 51 pp.
C
annon
s.H. & e
llen
s.d. (1985) - Rainfall conditions for abundant debris avalanches. San Francisco Bay Region, California.
Calif. Geol., 38: 12267–12272.
C
annon
s.H. & e
llen
s.d. (1988) - Rainfall that resulted in abundant debris-flow activity during the storm. In: e
llen
s.d. &
w
ieCzoRek
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. Geological Survey Prof. Paper 1434:27–34.
C
annon
s.H. (2005) - A NOAA–USGS demonstration flash-flood and debris-flow early-warning system. U.S. Geological Survey
Fact Sheet 2005-3104. Prepared in cooperation with National Oceanic and Atmospheric Administration (NOAA).
C
asadei
m., d
ietRiCH
w.e. & m
illeR
n.l. (2003) - Testing a model for predicting the timing and location of shallow landslide
initiation in soil mantled landscapes. Earth Surface Processes and Landforms, 28: 925-950.
C
HleboRad
a.f., b
aum
R.l., & G
odt
J.w. (2008) - A prototype system for forecasting landslides in the Seattle, washington, Area,
In: b
aum
R.l., G
odt
J.w. & H
iGHland
l.m.,
eds
., Engineering geology and landslides of the Seattle, washington area.
Geological Society of America Reviews in Engineering Geology, XX; 103-120, doi: 10.1130/2008.4020(06).
C
oe
J.a. & G
odt
J.w. (2001) - Debris flows triggered by the El Nino rainstorm of February 23, 1998, walpert Ridge and vicinity,
Alameda County, California. Reston, Virginia. U.S. Geological Survey Miscellaneous Field Studies Map MF-2384.
d
ettinGeR
m., R
alPH
f.m., m. H
uGHes
m., d
as
t., n
eiman
P., C
ox
d., e
stes
G., R
eynolds
d., H
aRtman
R., C
ayan
d. & J
ones
l. (2011) - Requirements and designs for a winter storm scenario for emergency preparedness and planning exercises in
California.
Weather, Climate and Society.
d
ietRiCH
w.e., R
eiss
R., H
su
m. & m
ontGomeRy
d.R. (1995) - A process-based model for colluvial soil depth and shallow
landsliding using digital elevation data. Hydrological Processes, 9: 383-400.
d
ietRiCH
w.e., b
elluGi
d., & R
eal
de
a
sua
R. (2001) - Validation of the shallow landslide model, Shalstab, for forest manage-
ment. In w
iGmosta
m.s. & b
uRGes
s.J. (
eds
), Land Use and watersheds: Human influence on hydrology and geomorpholo-
gy in urban and forest areas. Amer. Geoph. Union, Water Science and Application, 2: 195-227.
e
llen
s.d. & w
ieCzoRek
G.f. (1988) - Landslides, floods, and marine effects of the storm of January 3–5, 1982, in the San
Francisco Bay region, California. U.S. Geological Surve. Prof. Paper 1434.
G
odt
, J.w., b
aum
., R.l., & C
HleboRad
, a.f. (2006) - Rainfall characteristics for shallow landsliding in Seattle, washington,
USA, Earth Surface Processes and Landforms, 31:97–110.
G
odt
J.w., s
CHulz
w.H., b
aum
R.l. & s
avaGe
w.z. (2008) - Modeling rainfall conditions for shallow landsliding in Seattle, wa-
shington. In: b
aum
R.l., G
odt
J.w. & H
iGHland
l.m.,
eds
., Landslides and Engineering Geology of the Seattle, washington
area. Geological Society of America Reviews in Engineering Geology, XX: 137-152, DOI: 10.1130/2008.4020(08).
G
odt
J.w. (
editoR
), a
Rnal
C.H., b
aum
R.l., b
Rien
d.l., C
oe
J.a., d
e
m
outHe
J., e
llis
w.l., G
RaymeR
R.w., H
aRP
e.l.,
H
illHouse
J.w., H
oudRe
n., H
owell
d.G., J
ayko
a.s., l
aJoie
k.R., m
oRRissey
m.m., R
amsey
d.w., s
avaGe
w.z., s
CHu
-
steR
R.l., w
ieCzoRek
G.f. & w
ilson
R.C. (1999) - Maps showing locations of damaging landslides caused by El Nino
rainstorms, winter season 1997_98, San Francisco Bay region, California. Reston, Virginia: U.S. Geological Survey Mi-
scellaneous Field Studies Maps MF-2325- A-J. Available from: http://pubs.usgs.gov/mf/1999/mf-2325/, accessed on 25
November, 2008.
G
uzzetti
f., P
eRuCCaCCi
s., R
ossi
m. & s
taRk
C.P. (2007) - Rainfall thresholds for the initiation of landslides in central and
southern Europe. Meteorology and Atmospheric Physics, 98: 239-267.
background image
J.D. STOCk & D. BELLUGI
1022
5th International Conference on Debris-Flow Hazards Mitigation: Mechanics, Prediction and Assessment Padua, Italy - 14-17 June 2011
G
uzzetti
f., P
eRuCCaCCi
s., R
ossi
m. & s
taRk
C.P. (2008) - The rainfall intensity-duration control of shallow landslides and
debris flows: an update. Landslides, 5: 3-17.
m
ontGomeRy
d. R. & d
ietRiCH
w.e. (1994) - A physically-based model for topographic control on shallow landsliding. Water
Resources Research, 30(4): 1153-1171.
m
ontGomeRy
d.R., s
CHmidt
k.m., d
ietRiCH
w.e. & m
C
k
ean
J. (2009) - Instrumental record of debris flow initiation du-
ring natural rainfall: Implications for modeling slope stability. J. Geophy. Research- Earth Surface, 114: F01031,
doi:10.1029/2008JF001078, 15 pp.
m
oRton
d.m. & H
auseR
R.m. (2001) - A debris avalanche at Forest Falls, San Bernardino County, California, July 11, 1999:
U.S. Geological Survey Open-File Report 01-146. CD ROM and web; http://geopubs.wr.usgs.gov/open-file/of01-146.
m
oRton
D.M. & C
amPbell
R.H. (1978) - Cyclic landsliding at wrightwood, southern California - A preliminary report. U.S.
Geological survey Open File Report 78-1079.
n
ilsen
t.H., t
ayloR
f.a. & d
ean
R.m. (1976a) - Natural conditions that control landsliding in the San Francisco Bay region -
An analysis based on data from the 1968-69 and 1972-73 rainy seasons. U.S. Geological Survey Bull. 1424, 35 pp.
n
ilsen
t.H &t
uRneR
b.l. (1975) - Influence of rainfall and ancient landslide deposits on recent landslides (1950–71) in urban
areas of Contra Costa Country, California. U.S. Geological Survey Bull. 1388.
n
ilsen
t.H., t
ayloR
f.a.& b
Rabb
e.e. (1976) - Recent landslides in Alameda County, California (1940–71) an estimate of
economic losses and correlations with slope, rainfall, and ancient landslide deposits. U.S. Geological Survey Bull. 1398.
R
iCe
R.m. & f
oGGin
G.T. (1971) - Effects of high-intensity storms on soil slippage on mountainous watersheds in Southern
California. Water Resour Res, 7: 1485-1496.
s
CHimmelmann
a., l
anGe
C.b.& m
eGGeRs
b.l. (2003) - Palaeoclimatic and archaeological evidence for a ~200-yr recurrence
of floods and droughts linking California, Mesoamerica and South America over the past 2000 years. The Holocene, 13:
763-778.
t
an
s.s., C
laHan
k.b. & i
Rvine
P.J. (2004) - Geologic map of the Santa Paula 7.5-minute quadrangle, Ventura County, Califor-
nia. A digital database: California Geological Survey, Preliminary Geologic Maps, scale 1:24,000.
t
ayloR
f.a. & b
Rabb
e.e. (1972) - Maps showing distribution and cost by counties of structurally damaging landslides in the
San Francisco Bay region, California, winter of 1968-69. Reston, Virginia: U.S. Geological Survey Miscellaneous Field
Studies Map MF-327.
t
ayloR
f.a., n
ilsen
t.H. & d
ean
R.m. (1975) - Distribution and cost of landslides that have damaged manmade structures
during the rainy season of 1972-1973 in the San Francisco Bay region, California. Reston, Virginia: U.S. Geological Survey
Miscellaneous Field Studies Map MF-679.
w
ilson
R.C. & w
ieCzoRek
G.f. (1995) - Rainfall thresholds for initiation of debris flows at La Honda, California. Environ Eng
Geosci., 1: 11–27.
w
ilson
R.C., m
aRk
R.k. & b
aRbato
G. (1993) - Operation of a realtime warning system for debris flows in the San Francisco
Bay area, California. In: s
Hen
H.w., s
u
s.t. & w
en
f. (
eds
) Hydraulic Engineering ’93. Proc 1993 Environmental Geology
35 (2–3) August 1998 7 Q Springer-Verlag.
w
entwoRtH
C.m. (1986) - Maps of debris flow features evident after the storms of December 1955 and January 1982, Montara
Mountain area, California. U.S. Geological Survey Open-File Map 86-363, 2 sheets at 1:24,000 scale
Statistics