You are on page 1of 38

INVESTIGATING LANDSLIDES CAUSED BY EARTHQUAKES A

HISTORICAL REVIEW
DAVID K. KEEFER
U.S. Geological Survey, 345 Middlefield Road MS 977, Menlo Park, CA 94025, USA
E-mail: dkeefer@usgs.gov

(Received 17 January 2002; Accepted 11 February 2002)

Abstract. Post-earthquake field investigations of landslide occurrence have provided a basis for
understanding, evaluating, and mapping the hazard and risk associated with earthquake-induced
landslides. This paper traces the historical development of knowledge derived from these investigations. Before 1783, historical accounts of the occurrence of landslides in earthquakes are typically
so incomplete and vague that conclusions based on these accounts are of limited usefulness. For
example, the number of landslides triggered by a given event is almost always greatly underestimated.
The first formal, scientific post-earthquake investigation that included systematic documentation of
the landslides was undertaken in the Calabria region of Italy after the 1783 earthquake swarm. From
then until the mid-twentieth century, the best information on earthquake-induced landslides came
from a succession of post-earthquake investigations largely carried out by formal commissions that
undertook extensive ground-based field studies. Beginning in the mid-twentieth century, when the use
of aerial photography became widespread, comprehensive inventories of landslide occurrence have
been made for several earthquakes in the United States, Peru, Guatemala, Italy, El Salvador, Japan,
and Taiwan. Techniques have also been developed for performing retrospective analyses years or
decades after an earthquake that attempt to reconstruct the distribution of landslides triggered by the
event. The additional use of Geographic Information System (GIS) processing and digital mapping
since about 1989 has greatly facilitated the level of analysis that can applied to mapped distributions
of landslides. Beginning in 1984, syntheses of worldwide and national data on earthquake-induced
landslides have defined their general characteristics and relations between their occurrence and
various geologic and seismic parameters. However, the number of comprehensive post-earthquake
studies of landslides is still relatively small, and one of the most pressing needs in this area of research
is for the complete documentation of landslides triggered by many more earthquakes in a wider
variety of environments.
Keywords: Debris flows, earthquakes, ground failure, historical landslides, landslides, landslide
inventories, lateral spreads, liquefaction, review, rock falls, seismic slope stability, slope failure

1. Introduction
Much has been learned about earthquake-induced landslides since the first formal,
scientific post-earthquake investigation was undertaken following the earthquake
swarm in Calabria, Italy in 1783. Knowledge has come from field studies following
many earthquakes throughout the world, and these studies have generally become
more detailed and comprehensive with time as increased resources and new tools,
Surveys in Geophysics 23: 473510, 2002.
2002 Kluwer Academic Publishers. Printed in the Netherlands.

474

DAVID K. KEEFER

such as aerial photography and Geographic Information Systems (GIS) processing,


have become available.
The primary purpose of this paper is to review the history of the most comprehensive investigations of earthquake-induced landslides. The review concentrates
on (1) those earthquakes for which the landslides have been documented most
comprehensively, and (2) those studies that have synthesized the primary data
from post-earthquake investigations to determine general characteristics of landslide occurrence. Studies of both these types have provided the basis for increasing
numbers of seismic landslide-hazard and slope-stability analyses on a variety of
scales, but no attempt is made to review those analyses here. This paper, however,
does additionally review less extensive but significant recent findings on three
selected topics: the occurrence of landslides at great distances from earthquake
sources, field verification of Arias intensity thresholds, and delayed activation of
landslide movement by earthquake shaking. This selection of topics and studies is
subjective and not exhaustive; much has also been learned from other studies that
are omitted from this review because of limitations of time and space.
The first extensive synthesis of data on the occurrence of landslides in earthquakes was completed in 1984, for earthquakes that had occurred through May of
1980 (Keefer, 1984). The present paper thus initially discusses investigations of
landslides caused by earthquakes through May of 1980. Results from that initial
synthesis are then summarized. Following that, comprehensive landslide inventories from events that occurred after 1980 are described, and then more recent studies
synthesizing worldwide and national data are discussed. The final main section
reviews the findings on landslides at great distances, Arias intensity thresholds,
and delayed landslide activation.

2. From ancient times through May 1980


Landslides caused by earthquakes have been documented from at least as early as
1789 BCE in China (Hansen and Franks, 1991) and 373 or 372 BCE in Greece
(Seed, 1968). However, before about 200 years ago, accounts of earthquakeinduced landslides in historical documents are generally incomplete for any given
event, imprecise concerning locations of landslides, and vague about the characteristics of material and movement. While sometimes valuable from a historical
perspective or for local hazard analysis, such accounts give at best an incomplete
and at worst a misleading picture of the extent, nature, and abundance of landslides
caused by any given seismic event.

INVESTIGATING LANDSLIDES CAUSED BY EARTHQUAKES

475

2.1. E ARLY FORMAL INVESTIGATIONS


The earthquake swarm in Calabria, Italy in 1783 (estimated maximum magnitude,
MI 1 = 7) led to the first formal scientific study of a large earthquake when the
Neapolitan Academy of Sciences and Fine Letters appointed a commission to conduct field studies in the epicentral area (Sarconi, 1784, quoted by Cotecchia et al.,
1986; Davison, 1936; Cotecchia et al., 1986). These reports, a subsequent synthesis
by Sir Charles Lyell, in his Principles of Geology (1874), and retrospective studies
by Cotecchia and Melidoro (1974) and Cotecchia et al. (1986) indicate that landslides, including several that were exceptionally large, were widespread. Among
other effects, at least 215 landslide-dammed lakes were documented (Vivenzio,
1788, quoted by Cotecchia et al., 1986).
Throughout the nineteenth and early twentieth centuries, various other scientific
commissions carried out field investigations after several other major earthquakes,
and landslides were commonly documented along with other earthquake effects.
Although undertaken without the benefit of aerial observation or aerial photography, several of these investigations provided extensive information on the
locations, though not always on the characteristics, of landslides caused by the
earthquakes. Examples of these early reports include those on the MI 7.3 Charleston, South Carolina, earthquake of 1886 (Dutton, 1889), the M 8.3 Assam, India,
earthquake of 1897 (Oldham, 1899), the M 7.8 San Francisco, California, earthquake of 1906 (Lawson, 1908), and the M 8.1 Bihar, India-Nepal, earthquake of
1934 (Geological Survey of India, 1939). All of these reports contain descriptions
of landslides occurring throughout the relatively large areas that were surveyed
on the ground, although not all localities that probably produced landslides were
described. Notable for its detailed descriptions of landslides is the report of Lawson
(1908) on the San Francisco earthquake. From this and other contemporary reports
on the earthquake (Youd and Hoose, 1978), it is possible to infer that thousands
of landslides occurred throughout an area of 32,000 km2 . Many localities are also
described in enough detail to determine the types and source characteristics of the
landslides.
2.2. R ETROSPECTIVE STUDIES
Some other early earthquakes have been the subjects of retrospective studies
conducted decades after the events that have sought to reconstruct the landslide
occurrence. Even field investigations conducted soon after an earthquake may
encounter problems separating out landslides triggered by the earthquake from,
for example, landslides triggered by recent rainstorms. Retrospective studies face
1 Earthquake magnitude designations are as follows: M signifies a Richter surface-wave mags

nitude, ML a Richter local magnitude, M a moment magnitude, MI a magnitude estimated from


intensity information, and M a general magnitude determination of type not specified in the original
source. Moment magnitude, M, is used preferentially throughout this paper.

476

DAVID K. KEEFER

the additional challenges of separating earthquake-induced landslides from landslides possibly triggered by other, unrelated events. The confidence in identifying
landslides caused by an earthquake several decades in the past can be increased
if even incomplete reports or eyewitness accounts recorded soon after the event
are available; these can be used to calibrate landslide identification by establishing
apparent geomorphic ages and degrees of surface alteration for landslides triggered
by that earthquake.
Early retrospective studies include the investigation of a 1935 earthquake in
New Guinea (M = 7.9), for which Simonett (1967) pioneered a statistical method to
differentiate presumed earthquake-induced landslides from landslides presumably
triggered by other events, such as rainstorms. He concluded that a distribution in
which the concentration of landslides decreases away from a central point or zone
indicates earthquake triggering, and this conclusion is confirmed by typical landslide distributions associated with recent earthquakes. Other retrospective studies
were used to identify landslides caused by several earthquakes in northern California (Youd and Hoose, 1978), landslides in British Columbia caused by an Ms
7.2 earthquake in 1946 (Matthews, 1979; Rogers, 1980), and landslides caused
by earthquakes in 1929 and 1968 in New Zealand (Adams, 1980; Pearce and
OLoughlin, 1985; Hancox et al., 2002).
The retrospective approach has been applied most extensively to the 18111812
New Madrid, Missouri, earthquake sequence (MI 8.1, 7.8, and 8 for the three
main shocks; Johnston, 1996). Contemporary accounts, such as those in Mitchill
(1815) and Penick (1981), indicate that many landslides occurred on riverbanks
and bluffs of the Mississippi River during the earthquake. Fuller (1912) synthesized
information from the available contemporary accounts and carried out field studies
in the affected region in 1904 and 1905. He found abundant evidence of relatively
recent landslides along a 56-km-long stretch of the bluffs. Noting that upright
trees on landslide surfaces were also of a fairly uniform age of a little less than
100 years, he concluded that the landslides had occurred during the 18111812
earthquake sequence.
To test the hypothesis that these landslides were a result of the earthquake
sequence, Jibson (1985) and Jibson and Keefer (1988, 1989, 1993) prepared a
landslide inventory and performed statistical and geotechnical analyses. The inventory identified about 200 large, complex landslides that appeared to be about
the same age, and statistical analysis showed significant correlation between the
concentrations of these landslides and distances from the inferred hypocenters of
the largest three 18111812 earthquakes (Jibson, 1985; Jibson and Keefer, 1988,
1989). Additionally, detailed slope-stability analyses of two representative landslides showed that neither would have moved under any reasonable combination
of non-seismic conditions whereas inferred ground shaking from the 18111812
events would have induced large displacements in both (Jibson, 1985; Jibson and
Keefer, 1993). The convergence of evidence from inventory mapping, statistical

INVESTIGATING LANDSLIDES CAUSED BY EARTHQUAKES

477

analysis, and geotechnical investigations led to the conclusion that these landslides
almost certainly moved during the 18111812 earthquakes.
2.3. C OMPREHENSIVE LANDSLIDE INVENTORIES THROUGH 1980
Aerial photographs, which started becoming generally available in the 1930s and
1940s, have become the single most important tool for documenting the occurrence of landslides in earthquakes. They were probably used systematically for the
first time to document earthquake effects immediately following the M 7.3 Fukui,
Japan, earthquake in 1948 (Collins and Foster, 1949). Where landslides are visible
from the air, systematic and complete post-earthquake aerial photography potentially allows landslides from a seismic event to be completely documented and
precisely located. This in turn permits landslide occurrence to be correlated with
topographic, geologic, and seismic parameters, thus providing the basic data for
general assessments of landslide susceptibility and hazard. Aerial photographs do,
of course, have limitations for documenting landslide occurrence. These include a
lower limit on the size of landslides that can be detected, lack of visibility in heavily
vegetated areas, difficulty in differentiating landslides triggered by an earthquake
from landslides triggered by other events, difficulty in discerning landslide types,
and lack of complete airphoto coverage of affected areas after most earthquakes.
For these reasons, ground-based field studies have also been crucial in preparing
most comprehensive landslide inventories.
Probably the first essentially complete post-earthquake inventory of landslides
was produced for the relatively small 1957 Daly City, California, earthquake (ML =
5.3). For this event, Bonilla (1960) plotted the locations of all 23 known coseismic
landslides, which occurred in an area of about 10 km2 . He reported the dimensions, type of material, and type of movement for each landslide, thus providing an
excellent model for later post-earthquake investigations.
The much larger 1964 Alaska earthquake (M 9.2) provided much data on landslides even though the area affected was so large (269,000 km2 ; Plafker et al.,
1969) and much of it so remote that a complete inventory of the landslides was not
attempted. Nevertheless, thousands of landslides were documented by many investigators, and results of the investigations of landslides along with other earthquake
effects were published in an eight-volume report by the National Academy of
Sciences (National Research Council Committee on the Alaska Earthquake, 1968
1973) and many other reports and articles (for a list, see Keefer and Tannaci, 1981).
Detailed descriptions and analyses were carried out on a wide variety of landslides,
including large rock avalanches, subaqueous landslides, and several landslides that
caused considerable damage in Anchorage, the largest city in Alaska.
The M 7.9 earthquake in the Rio Santa region of Peru on May 31, 1970, produced the most destructive single earthquake-induced landslide in historic times
the rock avalanche from Nevados Huascarn in the Andean Cordillera Blanca,
which killed more than 18,000 people (Plafker et al., 1971). The earthquake also

478

DAVID K. KEEFER

generated thousands, or possibly tens of thousands of other landslides throughout


an onshore area of about 30,000 km2 (Plafker et al., 1971). An extensive airphoto
inventory, the first of its kind produced for a large earthquake, was prepared for an
area of 8,300 km2 in the most heavily affected region; it showed the locations of
about a thousand of the sources of the largest landslides as well as one extensive
area of numerous rock falls (Plafker et al., 1971). Landslides of different types were
differentiated by use of different map symbols, and the overwhelming majority of
the landslides were identified as rock falls, rock slides, or disrupted soil slides
(Plafker et al., 1971). The report accompanying the map (Plafker et al., 1971) contained much additional data on the types of landslides, the distribution of landslides
either outside the limits of airphoto coverage or too small to be discerned, and the
geologic environments of the landslides sources.
Several more post-earthquake inventories prepared between 1970 and 1980,
inclusive, substantially expanded knowledge of the occurrence of landslides in
earthquakes. The first of these was prepared after the M 6.7 San Fernando, California, earthquake of February 9, 1971 (Morton, 1971, 1975) and showed the
landslides that occurred throughout about 250 km2 of the most heavily affected
region. Morton (1975) noted that several thousand landslides occurred in this area,
but only the larger ones and those whose displacement was sufficient to produce
landslide morphology or to give rise to bare areas stripped of vegetation could be
mapped from the aerial photographs. Additionally, Morton (1975) indicated that
innumerable minor rockfalls in roadcuts occurred outside the mapped area, and
other sources indicated that landslides from this earthquake occurred throughout
an area of 3400 km2 (Keefer, 1984). Nevertheless, this study provided detailed
data on earthquake-induced landslides, including types of movement and material involved. The study also concluded that the types of landslides produced by
the earthquake were different from the large rotational or complex landslides
previously found to be characteristic of this area (Morton, 1971, 1975).
Probably the first attempt to map essentially all of the landslides produced by
a large earthquake was the inventory prepared by Harp et al. (1981) after the February 4, 1976, Guatemala earthquake (M 7.5). Mapping from aerial photographs
covering an area of 16,000 km2 and ground-based field studies revealed more than
10,000 landslides (Harp et al., 1981), a number later estimated by E.L. Harp (unpublished data) to be about 50,000. Nearly all of these landslides were rock falls or
disrupted soil slides, and 90 percent of these occurred in weakly-cemented Pleistocene pumice deposits or the residual soils developed on them. Analyses by Harp et
al. (1981) related the landslide occurrence to bedrock lithology, slope inclination,
topographic amplification of ground motion, seismic intensity, and regional-scale
fracture systems in the bedrock. Harp et al. (1981) also found that only one of
the many pre-existing, large and deep-seated landslides mapped in the region was
reactivated by the earthquake shaking.
Landslides from the 1976 Friuli, Italy, earthquake sequence (MS of the largest
shock = 6.36.5) were also documented in detail (Ambraseys, 1976; Govi, 1977a,

INVESTIGATING LANDSLIDES CAUSED BY EARTHQUAKES

479

1977b; Govi and Sorzana, 1977) from airphoto interpretation and ground-based
field work. Landslides, which were almost entirely rock falls, were mapped
throughout an area of about 500 km2 in the epicentral region, with sources, travel
paths, dimensions, and locations of deposits shown for masses as small as single
boulders. Such detailed and complete mapping permitted analyses relating landslide occurrence and hazard to topographic and geologic parameters with a high
degree of confidence (Govi, 1977b). In all, landslides were produced by the earthquake sequence throughout an area of 2100 km2 (Keefer, 1984), and landslides
throughout this larger area were documented in a detailed ground-based field study
by Ambraseys (1976).
A complete inventory of landslides (Wilson et al., 1985) was produced after
the January 1980 Mount Diablo earthquake sequence in the San Francisco Bay
region of California (maximum M = 5.8). This inventory, produced from extensive
ground-based field work and limited observation from light aircraft, showed the
locations and types of 103 mostly small landslides that occurred throughout an
area of 500 km2 .
Another virtually complete inventory of landslides was produced after the May
1980 earthquake sequence at Mammoth Lakes (maximum M = 6.2) in the eastern
Sierra Nevada of California. The steep and high, glaciated slopes in this region
produced abundant rock falls and rock slides and a few landslides of other types
throughout an area of 1220 km2 , virtually all of which were documented from a
combination of airphoto interpretation, aerial observation, and ground-based field
work. A map produced by Harp et al. (1984) shows 5244 landslides in the epicentral area, and an additional 9 rock falls and rock slides were reported from Yosemite
Valley, immediately to the West (Wieczorek and Jger, 1996).

3. 1984 synthesis of data from earthquakes through 1980


In an attempt to provide a general framework for understanding the occurrence of
landslides in earthquakes, Keefer (1984) analyzed data from a sample of historical
earthquakes ranging in time from the 18111812 New Madrid earthquake sequence
through the May 1980 Mammoth Lakes earthquake sequence. The primary data
were taken from a sample of 40 worldwide earthquakes having magnitudes in
the range between 5.2 and 9.5, chosen to represent a wide variety of geographic,
geologic, and seismic environments. The earthquakes chosen included all those
for which comprehensive landslide inventories were available as well as others for
which data on landslides were relatively extensive or for which some aspect of
the landslide occurrence (such as landslide damage or landslide mechanisms) was
especially significant. To determine the smallest earthquakes and lowest shaking
intensity levels associated with landslides, intensity reports from several hundred
mostly small earthquakes in the United States were also studied.

480

DAVID K. KEEFER

Earthquake-induced landslides were classified into fourteen individual types


and three main categories on the basis of type of material, type of landslide movement, degree of internal disruption of the landslide mass, and geologic environment
(Table I). The first major category Disrupted Slides and Falls, or Disrupted
Landslides includes highly to very highly disrupted landslides, consisting of
boulders and masses of rock fragments, small blocks of soil, and (or) individual soil
grains that move downslope by falling, bouncing, and (or) rolling (rock falls and
soil falls), translational sliding (rock slides and disrupted soil slides), or complex
mechanisms involving both sliding and fluid-like flow (rock avalanches and soil
avalanches). Landslides in this category typically originate on steep slopes, travel
relatively fast, and are capable of transporting material far beyond the bases of the
steep slopes on which they originate. Except for rock avalanches, all of which have
volumes greater than 0.5 106 m3 , landslides in this category are also typically
thin, with initial failure depths of less than 3 m (Table I).
The second major category Coherent Slides, or Coherent Landslides includes translational slides (rock block slides and soil block slides), rotational slides
(rock slumps and soil slumps), and slow earth flows, which move by a combination
of translational sliding and flow. These landslides exhibit a slight to moderate
amount of internal disruption, typically consisting of a few moving blocks, each
of which may be little deformed except for localized internal fissuring. These types
of landslides occur most commonly on moderately steep slopes, typically move
relatively slowly, and displace material less than 100 m. These landslides are also
relatively thick, with typical initial failure depths greater than 3 m (Table I).
The third major category Lateral Spreads and Flows includes those
landslides for which fluid-like flow is the predominant movement mechanism.
Landslides in this category initiate only in soil materials and involve either blocks
of relatively intact material moving on a subsurface liquefied zone (soil lateral
spreads) or more completely liquefied masses that move by fluid-like flow throughout (rapid soil flows). In many cases, these landslides are the results of soil
liquefaction in saturated sands, gravels, or silts; occasionally they result from
seismically-induced disturbance in sensitive (i.e., thixotropic) clays. Also included
in this category are all underwater (subaqueous) landslides, most of which are
complex and can involve elements of rotational or translational sliding as well as
lateral spreading or flow. Landslides in this category commonly initiate and move
on gentle to nearly level slopes, move rapidly, and can transport material large
distances (Table I).
Keefer (1984) gave order-of-magnitude estimates for the numbers of landslides
of each type triggered by each earthquake studied. Total numbers of landslides generally increased with earthquake magnitude, ranging from a few tens of landslides
at most for earthquakes with M < 5.5 to several thousands at least for earthquakes
with M > 8.0, although numbers were highly variable among individual events
within each magnitude range. This variability almost certainly resulted from real
variations in the numbers of landslides triggered as well as from the incomplete

Bouncing, rolling, free fall

Soil falls

Rotational sliding

Translational sliding

Rotational sliding

Translational sliding

Translational sliding and


internal flow

Coherent landslides
Rock slumps

Rock block slides

Soil slumps

Soil block slides

Slow earth flows

Slight

Slight or moderate

Slight or moderate

Slight or moderate

Slight or moderate

Deep

Deep

Generally shallow;
occasionally deep

Deep

Deep

Shallow

Very high

Complex, involving sliding,


flow, and occasionally free
fall

Soil avalanches

Shallow

Shallow

Deep

Shallow

High

Shallow

Typical
depths

Water content
D U PS S

Disrupted soil slides Translational sliding

High or very high

Very high

Complex, involving sliding,


flow, and occasionally free
fall

Rock avalanches

High or very high

Internal
disruption

High

Bouncing, rolling, free fall

Type of
movement

Disrupted rock slides Translational sliding

Rock falls

Disrupted landslides

Name

10

15

15

25

15

40

25

35

40

Most less than 1 104 m3 ;


maximum reported 2
109 m3

Most less than 1 104 m3 ;


maximum reported 2
107 m3

Typical
volumes

maximum reported 4.8


107 m3

Most less than 1,000 m3 ;


maximum volumes not well
documented
Most less than 1 104 m3 ;

Very slow to moderate;


occasionally, with very rapid
surges

Slow to very rapid

Slow to rapid

Slow to rapid

Slow to rapid

Most between 100 and 1


106 m3 ; maximum reported
between 3 107 and 6
107 m3

Most between 100 and a few


million m3 ; maximum at
least tens of millions of m3
Most between 100 and a few
million m3 ; maximum at
least tens of millions of m3
Most between 100 and 1
105 m3 ; occasionally 1
105 to several million m3
Most between 100 and 1
105 m3 ; maximum reported
1.12 108 m3

Very rapid to extremely rapid Volumes not well


documented; maximum
reported 1.5 108 m3

Moderate to rapid

Extremely rapid

Very rapid to extremely rapid 5 105 2 108 m3 or


more

Rapid to very rapid

Extremely rapid

Minimum Typical
slope ( ) velocities

TABLE I
Characteristics of earthquake-induced landslides (Modified from Keefer 1984, 1999)

Typically less than 100 m;


maximum displacements not
well documented

Typically less than 100 m;


maximum displacements not
well documented

Typically less than 10 m;


occasionally 100 m or more

Typically less than 100 m;


maximum displacements not
well documented

Typically less than 10 m;


occasionally 100 m or more

May slide to base of steep


source slope and several tens
or hundreds of meters farther,
on relatively gentle slopes
Several tens of meters to
several kilometers beyond
steep source slopes

Most come to rest at or near


bases of steep source slopes

Several kilometers

May slide to base of steep


source slope and several tens
or hundreds of meters farther,
on relatively gentle slopes

May fall to base of steep


source slope and move as far
as several tens or hundreds of
meters farther, on relatively
gentle slopes

Typical
displacements

INVESTIGATING LANDSLIDES CAUSED BY EARTHQUAKES

481

Generally lateral spreading or Generally high or very high;


flow; occasionally sliding
occasionally moderate or
slight

Subaqueous landslides

Shallow

Variable

? ? ?

Variable

Typical
depths

Water content
D U PS S

0.5

2.3

0.3

Typical
volumes

Most between 100 and 1


105 m3 ; largest reported 9.6
106 m3 .
Very rapid to extremely rapid Volumes not well
documented; largest are at
least several million m3 .
Generally rapid to extremely Volumes not well
rapid; occasionally slow to
documented; largest are at
moderate
least tens of millions of m3 .
Very rapid

Minimum Typical
slope ( ) velocities

Not well documented, but


some move more than 1 km

A few m to several km

Typically less than 10 m;


maximum reported 600 m

Typical
displacements

cemented aggregates of particles that may or may not contain organic materials. Internal disruption: slight signifies landslide consists of one or a few
coherent blocks; moderate signifies several coherent blocks; high signifies numerous small blocks and individual soil grains and rock fragments; very
high signifies nearly complete disaggregation into individual soil grains or small rock fragments. Depth: shallow signifies generally < 3 m deep; deep
signifies generally > 3 m deep. Water content: D = dry; U = moist but unsaturated; PS = partly saturated; S = saturated. Velocity: very slow = 1 106 3
106 m/min; slow = 3 106 3 105 m/min; moderate = 3 105 0.001 m/min; rapid = 0.0010.3 m/min; very rapid = 0.3180 m/min; extremely
rapid = >180 m/min. (Terminology after Varnes, 1978).

Notes: Names: rock signifies bedrock that is relatively firm and intact prior to landslide initiation, and soil signifies loose, unconsolidated or poorly

Very high

Flow

Rapid soil flows

Internal
disruption

Generally moderate;
occasionally slight or high

Type of
movement

Lateral spreads and flows


Soil lateral spreads
Translation on fluid basal
zone

Name

TABLE I
Continued

482
DAVID K. KEEFER

INVESTIGATING LANDSLIDES CAUSED BY EARTHQUAKES

483

TABLE II
Relative abundance of earthquake-induced landslides (from Keefer, 1984)

Landslide type, listed in order of decreasing total numbers


Very abundant: > 100,000 in the 40 historical earthquakes
Rock falls
Disrupted soil slides
Rock slides
Abundant: 10,000 to 100,000 in the 40 historical earthquakes
Soil lateral spreads
Soil slumps
Soil block slides
Soil avalanches
Moderately common: 1,000 to 10,000 in the 40 historical earthquakes
Soil falls
Rapid soil flows
Rock slumps
Uncommon: 100 to 1,000 in the 40 historical earthquakes
Subaqueous landslides
Slow earth flows
Rock block slides
Rock avalanches

reporting from some of the earthquakes in the sample. Three types of disrupted
landslides rock falls, disrupted soil slides, and rock slides were the most abundant types of earthquake-induced landslides (Table II). These landslides comprised
about 80 percent of all the reported landslides (35, 26, and 20 percent, respectively),
and the six types of disrupted landslides together comprised about 86 percent of
all the landslides reported. Coherent slides comprised about 8 percent, and lateral
spreads and flows comprised about 6 percent of the total number of landslides
reported from the 40 earthquakes.
Types of slopes and geologic environments that produced earthquake-induced
landslides varied widely, ranging from overhanging slopes in well-indurated bedrock to unconsolidated sediments with nearly level surfaces. Minimum slope
inclinations for the various types of landslides ranged from 0.3 to 40 (Table I).
In general, materials most susceptible to earthquake-induced landslides were (1)
weakly cemented, weathered, sheared, intensely fractured, or closely jointed rocks,
(2) better-indurated rocks having prominent discontinuities, (3) sandy residual or

484

DAVID K. KEEFER

colluvial soils, (4) saturated volcanic soils containing sensitive clay, (5) loess, (6)
cemented soils, (7) granular deltaic sediments, (8) granular flood-plain alluvium,
and (9) uncompacted, or poorly compacted, granular artificial fill.
Approximate magnitudes of the smallest earthquakes causing landslides of various types were 4.0 for rock falls, rock slides, soil falls, and disrupted soil slides;
4.5 for soil slumps and soil block slides; 5.0 for soil lateral spreads, rapid
soil flows, subaqueous landslides, rock slumps, rock block slides, and slow earth
flows; 6.0 for rock avalanches; and 6.5 for soil avalanches. Because all types of
earthquake-induced landslides can also be triggered by nonseismic agents, Keefer
(1984) noted the possibility that landslides of all types could occasionally occur in
earthquakes smaller than those indicated.
The areas affected by landslides in the earthquakes (Figure 1a) correlated with
earthquake magnitude, although such other factors as focal depth, specific groundmotion characteristics of individual earthquakes, and geologic conditions were
also important (Keefer, 1984). However, as far as could be determined from the
available data, regional variation in seismic attenuation was not a significant factor.
A non-linear upper bound enclosing the data showed the maximum size of the
area likely to be affected by landslides increased from nil at M 4 to 500,000
km2 at M = 9.2 (Figure 1a). A subsequent study (Keefer and Wilson, 1989) using
data from this sample and seven additional earthquakes, related area affected by
landslides, A, to magnitude, M, with the least-squares linear regression mean,
log10 A = M 3.46(0.47)
for 5.5 < M 9.2, where A is in square kilometers. Maximum distances from
epicenters and fault ruptures were also determined for landslides in each major
category for each earthquake. These distances also correlated with magnitude, and
non-linear upper bounds enclosing all the data were determined for each landslide
category (Figures 2 and 3).
Minimum shaking intensities on the Modified Mercalli (MMI) scale for landslides in each category were determined for each earthquake by comparing maps
showing landslide distribution with isoseismal maps (Figure 4). Substantial scatter in these data presumably reflected differing application of the intensity scale
by various investigators, among other factors. Nevertheless, disrupted landslides
occurred on average at one intensity grade lower than landslides in the other categories; for them, the predominant minimum intensity was MMI VI, and the lowest
reported intensity in any earthquake was MMI IV; whereas, for landslides in the
other two categories the predominant minimum intensity was MMI VII, and the
lowest reported intensity was MMI V (Keefer, 1984; Figure 4).
Landslides were responsible for highly variable, but in many cases significant, numbers and proportions of casualties and levels of economic damage in the
various earthquakes. In at least eight of the earthquakes, landslides caused half
or more of the deaths (Keefer, 1984). The greatest number of deaths caused by
landslides occurred during the M 7.8 Kansu, China, earthquake in 1920, in which

INVESTIGATING LANDSLIDES CAUSED BY EARTHQUAKES

485

Figure 1a. Relations between area affected by landslides and earthquake magnitude. Circles are
data from earthquakes discussed by Keefer (1984) and Keefer and Wilson (1989). Dashed line is
approximate upper bound from Keefer (1984), curved to approach A = 0 at M = 4. Solid line is
least-squares linear regression mean from Keefer and Wilson (1989). Magnitude determinations for
individual earthquakes are given in Keefer (1984) and Keefer and Wilson (1989); most magnitudes
smaller than 7.5 are Richter surface-wave magnitudes (MS ), and most magnitudes of 7.5 or larger
are moment magnitudes (M).

rapid flows in loess, some of which traveled several kilometers, killed most of the
240,000 people who died in that seismic event (Wang and Xu, 1984). The 1970
rock avalanche from Nevados Huascarn, in the Andes of Peru, which transported
at least 50 106 m3 of material 16 km at an average velocity of 280 km/hr, killed
about 18,000 people (Plafker et al., 1971; Plafker and Ericksen, 1978). Rapid soil
flows and a rock avalanche also caused substantial loss of life in the MS 7.6 Khait
earthquake in Tajikistan in 1949. There, flows in loess probably triggered by a
combination of the main shock, aftershocks, and heavy rainfall killed an estimated
15,000 people, and a rock avalanche triggered by the main shock buried the town of
Khait, killing at least 3,000 people (Seed, 1968; A.M. Sarna-Wojcicki, unpublished
data).
Significant numbers and proportions of deaths were also attributable to landslides in many other historical earthquakes (Keefer, 1984). Although all types of
landslides posed some degree of hazard to human life, at least 90 percent of the
deaths in the 40 earthquakes studied were caused by rock falls, rock avalanches,
and rapid soil flows (Keefer, 1984), even though the latter two types of landslides
occur in relatively small numbers (Table II). Rock avalanches and rapid soil flows

486

DAVID K. KEEFER

Figure 1b. Relations between area affected by landslides and earthquake moment magnitude. Circles
are data from earthquakes discussed by Rodrguez et al. (1999), plotted using moment magnitude
(M); open circle is 1988 Saguenay, Quebec earthquake. Solid line is upper bound of Keefer (1984;
see Figure 1a). Dashed line is upper bound of Rodrguez et al. (1999). Triangle is datum from 1963
Peria, New Zealand, earthquake, for which area exceeds upper bounds, plotted using Richter local
magnitude (ML ), from Hancox et al. (1997, 2002).

are particularly dangerous because they can transport material several kilometers
at high velocities over gently sloping ground (Table I); also, rock avalanches are
all relatively large. The high risk from rock falls derives both from their transport
of rock materials at high velocities (Table I) and their abundance (Table II).
Landslides also caused significant economic losses in many of the earthquakes.
Examples include the 1964 earthquake in Niigata, Japan (M 7.5), where landslides
and related soil-liquefaction phenomena caused an estimated $US 800 million in
damage2 (Lee et al., 1977), the 1964 Alaskan earthquake (M 9.2), in which landslides caused an estimated $US 279 million in damage (W.R. Hansen, unpublished
data), and the 1970 Peru earthquake (M 7.9) in which landslides buried thousands
of residences and other buildings and destroyed substantial parts of the regional
transportation and utility systems (Plafker et al., 1971).

2 Damage estimates are in $US of the time of the earthquakes.

INVESTIGATING LANDSLIDES CAUSED BY EARTHQUAKES

487

Figure 2. Approximate upper bounds for maximum epicentral distances of landslides in three major
categories related to earthquake magnitude, from Keefer (1984), and data from earthquakes for which
maximum epicentral distances of landslides exceed the relevant upper bound. Dashed line is upper
bound for disrupted landslides, dot-dash line is upper bound for coherent landslides, and dotted line
is upper bound for lateral spreads and flows. Magnitudes used in determining upper bounds were
typically moment magnitude (M) for M 7.5 and Richter surface-wave magnitude (MS ) for M <
7.5. Diamonds and squares are data for disrupted landslides. Diamonds are data for earthquakes in the
Colorado Plateau region of the western United States; sources of data and magnitude determinations
given in text. Filled square is datum from 1988 Saguenay, Quebec, earthquake, plotted using moment magnitude, from Rodrguez et al. (1999). Open square is datum for 1963 Peria, New Zealand,
earthquake, plotted using Richter local magnitude (ML ), from Hancox et al., (1997). Circles are data
for coherent landslides. Open circle is datum from1988 Saguenay, Quebec, earthquake, plotted using
moment magnitude, from Rodrguez et al. (1999). Filled circle is datum from 1992 St. George, Utah,
earthquake in the Colorado Plateau region, plotted using moment magnitude, from Jibson and Harp
(1996).

4. Comprehensive landslide inventories for earthquakes since 1980


A few additional comprehensive landslide inventories for earthquakes that have
occurred since 1980 have been prepared using aerial photography, exhaustive
ground-based field work, and, in one case, satellite imagery. The first event for
which such an inventory was prepared (Harp and Keefer, 1990) was the 1983 Coalinga, California, earthquake (M 6.5). Mapping from aerial photographs identified
9389 landslides (as subsequently counted by Keefer and Wilson, 1989; Table III)
throughout an area of 650 km2 , and additional landslides too small to distinguish on
the aerial photographs were identified during ground-based field studies (Harp and

488

DAVID K. KEEFER

Figure 3. Approximate upper bounds for maximum distances of landslides in three major categories
from fault-rupture zones related to earthquake magnitude, from Keefer (1984), and distance to disrupted landslide, triggered by 1988 Saguenay, Quebec earthquake (square), that exceeds upper bound
for that landslide category, plotted using moment magnitude, from Rodrguez et al., (1999). Dashed
line is upper bound for disrupted landslides, dot-dash line is upper bound for coherent landslides,
and dotted line is upper bound for lateral spreads and flows. Magnitudes used in determining upper
bounds were typically moment magnitude (M) for M 7.5 and Richter surface-wave magnitude
(MS ) for M < 7.5.

Keefer, 1990). More than 97 percent of the landslides were rock falls, rock slides,
or soil falls; most of those involving bedrock initiated from slopes steeper than 60
formed along steeply dipping joint surfaces in weakly cemented sedimentary rocks
(Harp and Keefer, 1990). In contrast, the earthquake caused little renewed activity
of any of the numerous large, pre-existing slumps or earth flows in the affected
region (Harp and Keefer, 1990).
Rymer (1987) and Rymer and White (1989) used aerial photographs and
ground-based observations to map landslides produced by the M 5.7 San Salvador,
El Salvador, earthquake in 1986. The inventory was estimated to be about 95 percent complete, as it excluded any landslides having volumes less than 50 m3 , those
that traveled less than 2 to 3 m, and those that may have occurred in heavily forested
areas. Additionally, heavy rainfall before the earthquake could have caused at least
some of the landslides in part of the affected area. Excluding those landslides, 200
sources of disrupted landslides (some of which may have produced more than one
landslide) were mapped, along with 12 rock or soil slumps, 4 rapid soil flows, and
stretches of slope totaling about 3 km long where abundant landslides occurred.

INVESTIGATING LANDSLIDES CAUSED BY EARTHQUAKES

489

Figure 4a. Minimum, or threshold shaking intensity on the Modified Mercalli scale at which
disrupted landslides were reported in historical earthquakes. Black bars are for data from Keefer
(1984), and gray bars are for data from Rodrguez et al. (1999).

Figure 4b. Minimum, or threshold shaking intensity on the Modified Mercalli scale at which coherent landslides were reported in historical earthquakes. Black bars are for data from Keefer (1984),
and gray bars are for data from Rodrguez et al. (1999).

22 May 1957
4 Feb 1976
24 Jan 1980
25 May 1980
2 May 1983
10 Oct 1986
17 Oct 1989
17 Jan 1994
17 Jan 1995
26 Sep 1997
21 Sep 1999

Daly City, Calif. USA


Guatemala
Mt. Diablo, Calif. USA
Mammoth Lakes, Calif. USA

Coalinga, Calif. USA


San Salvador, El Salvador
Loma Prieta, Calif. USA
Northridge, Calif. USA
Hygoken-Nanbu, Japan
Umbria-Marche, Italy
Chi-Chi, Taiwan

6.5
5.7
6.9
6.7
6.9
6.0
7.7

5.3
7.5
5.8
6.2

Earthquake
magnitude

9,389
>216
1,500
>11,000
674747
100124
22,000

23
50,000
103
5,253

Number of
landslides

>97
>93
74#
>90
81 to 83
61
>85

48
Most
83
>98
<2
<5
26#
<9
13 to 15
34
11 to 15

30
?
17
<1
<1
<2
0#
<1
3 to 4
5
<4

22
?
0
<1

Approximate percentage by category


Disrupted (%)
Coherent (%)
Lateral spreads
and flows (%)

Bonilla, 1960
Harp et al., 1981; E. L. Harp, unpublished data
Wilson et al., 1985; Keefer and Wilson, 1989
Harp et al., 1984; Keefer and Wilson, 1989;
Wieczorek and Jger, 1996
Keefer and Wilson (1989); Harp and Keefer (1990)
Rymer, 1987; Rymer and White, 1989
Keefer and Manson, 1998
Harp and Jibson, 1995, 1996
Sassa et al., 1995; Fukuoka et al., 1997; Okimura and Torii, 1999
Bozzano et al., 1998; Esposito et al., 2000; I. LaRosa, unpublished data
Lin and others, 2001; Sitar and Bardet, 2001

Reference

# Percentages determined in central area only.

Magnitudes in bold are moment magnitudes; others are Richter surface-wave (Mt. Diablo) or Richter local (Daly City) magnitude.

Earthquake
date

Earthquake location

TABLE III
Numbers of documented landslides produced by earthquakes

490
DAVID K. KEEFER

INVESTIGATING LANDSLIDES CAUSED BY EARTHQUAKES

491

Figure 4c. Minimum, or threshold shaking intensity on the Modified Mercalli scale at which lateral
spreads and (or) flows were reported in historical earthquakes. Black bars are for data from Keefer
(1984), and gray bars are for data from Rodrguez et al. (1999).

Within the area of possible rainfall triggering were another 68 disrupted landslides
and 10 km of slopes with abundant landslides. The most abundant landslides were
soil falls and disrupted soil slides that initiated in poorly consolidated Pleistocene
tuffs similar to those that produced abundant landslides in the 1976 Guatemala
earthquake.
In 1989, 1994, and 1995 three earthquakes that caused large economic losses in
heavily developed urban areas led to intensive studies of earthquake effects, including landslides. Studies of these earthquakes are notable for the first application of
digital, GIS-based mapping techniques, which have greatly facilitated the analysis
of landslide occurrence in these and subsequent earthquakes.
The first of these three earthquakes was the M 6.9 Loma Prieta, California,
earthquake of October 17, 1989. Because much of the mountainous area around
the epicenter was covered by heavy, undisturbed vegetation that obscured visibility
from the air, landslide inventories were prepared almost entirely from groundbased fieldwork, which involved more than 60 investigators (Weber and Nolan,
1989; Plant and Griggs, 1990; Spittler and Harp, 1990; Manson et al., 1992; Keefer,
1998, 2000). In an area of about 2,000 km2 around the epicenter, 1280 landslides
were mapped; an additional 80 landslides occurred along adjacent coastal bluffs,
and several dozen to possibly a few hundred were scattered throughout an additional 13,000 km2 (Table III). Soil lateral spreads and other soil-liquefaction effects

492

DAVID K. KEEFER

Figure 5a. Relation between landslide concentration (number of landslides per square kilometer) and
epicentral distance for landslides in the southern Santa Cruz Mountains triggered by the 1989 Loma
Prieta, California, earthquake, from Keefer (2000).

throughout this larger area were also documented by ORourke (1992) and Holzer
(1998). In the mountainous epicentral region, 74 percent of the landslides were
disrupted landslides mostly rock falls, rock slides, and disrupted soil slides and
the remaining 26 percent were slumps and block slides, many of which involved
highway fills (Keefer and Manson, 1998). Statistical analysis of the landslide distribution showed a strong correlation between landslide concentration, on the one
hand, and distance from the epicenter (Figure 5a), distance from the fault rupture (Figure 5b), and slope inclination, on the other (Keefer, 2000). Correlation
between landslide concentration and bedrock lithology was more complex, but
well-indurated igneous and metamorphic rocks had significantly lower landslide
concentrations than weakly or moderately indurated sedimentary rocks (Keefer,
2000).
Landslides from the January 17, 1994 Northridge, California, earthquake (M
6.7), the second of these damaging urban earthquakes, were mapped by Harp and

INVESTIGATING LANDSLIDES CAUSED BY EARTHQUAKES

493

Figure 5b. Relation between landslide concentration (number of landslides per square kilometer)
and distance from the surface projection of the updip edge of the fault plane for landslides in the
southern Santa Cruz Mountains triggered by the 1989 Loma Prieta, California, earthquake, from
Keefer (2000).

Figure 5c. Relation between percentage of landslides and distance from fault plane for landslides
triggered by the 1995 Hyogoken-Nanbu, Japan, earthquake, from Fukuoka et al. (1997).

494

DAVID K. KEEFER

Figure 5d. Relation between number of landslides and epicentral distance for 99 classified landslides
triggered by the 1997 Umbria-Marche, Italy earthquake, from Esposito et al. (2000).

Figure 5e. Relation between percentage of landslides and distance from assumed fault plane for
landslides triggered by the 1997 Umbria-Marche, Italy earthquake, from Bozzano et al. (1998).

INVESTIGATING LANDSLIDES CAUSED BY EARTHQUAKES

495

Jibson (1995, 1996) from aerial photographs and ground-based fieldwork. The
earthquake triggered more than 11,000 landslides throughout an area of 10,000
km2 ; most were in a central area of about 1,000 km2 of steep slopes underlain
by clastic sedimentary rocks that have little cementation. Most of the landslides
were rock falls, rock slides, soil falls, and disrupted soil slides (Table III); a few
tens to hundreds were slumps and block slides, and at least one was a soil lateral
spread (Harp and Jibson, 1995, 1996). Several other soil lateral spreads and related
ground failures, not visible on the aerial photographs were also reported (Hall,
1994; Hecker et al., 1995; Stewart et al., 1996). The inventory of Harp and Jibson
(1995, 1996) served as the basis for extensive analytical (Jibson et al., 1998; Jibson et al., 2000) and statistical studies (Parise and Jibson, 2000) relating landslide
occurrence to slope inclination and aspect, bedrock lithology, seismic ground motion, and slope-failure potential as predicted by the Newmark (1965) slope-stability
analysis.
Two landslide inventories prepared using aerial photographs and ground-based
field work following the third of these earthquakes the1995 Hyogoken-Nanbu,
Japan, earthquake (M = 6.9) showed that about 700 landslides were triggered
throughout an area of about 700 km2 (Sassa et al., 1995; Fukuoka et al., 1997;
Okimura and Torii, 1999; Table III). The predominant types of landslides were
rock falls, disrupted soil slides, rock slides, and avalanches of mixed rock and soil
in weathered granite (Ochiai et al., 1996; Fukuoka et al., 1997), but about 100
were coherent slides (mostly slumps) in human-built fill, 24 were liquefactioninduced lateral spreads, and 2 were rapid soil flows (Kamai, 1995; Sassa et al.,
1995). Additional soil lateral spreads occurred at several localities outside of this
mapped area (Japan Geotechnical Society, 1996). Landslide abundance showed an
exponential decrease with increasing distance from the fault-rupture zone (Figure
5c) but not with increasing epicentral distance.
Landslide inventories prepared from ground-based fieldwork indicated that the
September-October, 1997 Umbria-Marche earthquake sequence in central Italy
(maximum M = 6.0) produced about 100 landslides throughout an area of 1400
km2 (Bozzano et al., 1998; Esposito et al., 2000; I. Larosa, unpublished data).
Disrupted landslides comprised 61 percent of the total number; 34 percent were
coherent slides, and 5 percent were rapid soil flows (Esposito et al., 2000; Table
III). Landslide abundance was highest adjacent to the epicenter and fault rupture,
and decreased progressively with distance (Figures 5d and 5e).
The September 21, 1999 Chi-Chi, Taiwan, earthquake (M = 7.7) was the largest
earthquake since 1970 for which extensive landslide inventories were prepared.
The earthquake triggered abundant landslides from the steep, heavily vegetated
slopes in the Central Mountain Range. Inventories prepared from aerial photographs showed as many as about 22,000 landslides (Cheng et al., 2000; Kamai
et al., 2000; Lin et al., 2001; Table III), and an inventory prepared from SPOT
satellite imagery recorded 9272 landslides having surface areas larger than 625 m2
(Liao and Lee, 2001). These numbers do not include soil lateral spreads and other

496

DAVID K. KEEFER

effects due to soil liquefaction that occurred in several low-lying localities on the
coastal plain to the West of the mountains (Stewart, 2001). Statistical analysis of
a sample of 436 landslides (Lin et al., 2000, quoted in Sitar and Bardet, 2001)
indicated that 63 percent of the landslides were disrupted soil slides and 22 percent
were rock falls (Table III). Ninety-eight percent of the landslides originated on
slopes steeper than 30 , and 90 percent originated on slopes steeper than 45 (Lin
et al., 2000, quoted in Sitar and Bardet, 2001). The earthquake also triggered two
particularly large rock avalanches, having volumes of 120 and 70 million m3 , that
together killed at least 50 (Lin et al., 2001) and possibly as many as 119 (Kamai et
al., 2000) people. Strong ground-motions from this earthquake were recorded by
more than 400 free-field stations, the most ever for a major earthquake (Lee, 2001),
making possible a detailed correlation of seismic ground-motion parameters with
landslide initiation.
Landslide numbers recorded in the most complete inventories from 1957
through 1999 are summarized in Table III and Figure 6. These data show that even
moderate-sized earthquakes can produce large numbers of landslides, that the number of landslides generally increases as the earthquake magnitude increases, and
that disrupted landslides predominate in the earthquakes inventoried since 1980,
much as they did in the earlier earthquakes (Keefer, 1984).

5. Worldwide and national data synthesis since 1984


Since 1980, one additional synthesis of worldwide data has been completed
(Rodrguez et al., 1999), and national syntheses have been prepared for New Zealand (Hancox et al., 1997, 2002), Greece (Papadopoulos and Plessa, 2000), and Italy
(Prestininzi and Romeo, 2000). The worldwide synthesis of Rodrguez et al. (1999)
included data from 36 earthquakes with magnitudes ranging from 5.3 to 8.1 that
occurred between 1980 and 1997, inclusive. The information available for many of
these earthquakes did not permit reliable determination of the numbers of various
types of landslides, so order-of magnitude estimates for total numbers of reported landslides were made for 24 of these earthquakes. The numbers were highly
variable (Rodrguez et al., 1999); this variability probably reflects differences in
completeness of reporting as well as differences in susceptibility of affected slopes
and in ground-motion characteristics of individual earthquakes.
Rodrguez et al. (1999) also determined areas affected by landslides and maximum distances of landslides from epicenters and fault ruptures. One earthquake
the 1988 Saguenay, Quebec, earthquake (M 5.9) caused landslides over an area
(Figure 1b) and at distances (Figures 2 and 3) that were exceptionally large. Except
for this earthquake, data on maximum landslide distances were similar to the data in
Keefer (1984), and so Rodrguez et al. (1999) did not propose making any revisions
in the upper bound curves of Keefer (1984; Figures 2 and 3). However, seven of
the other 19801997 events, which had moment magnitudes between 5.3 and 6.9,

Figure 6. Relation between total number of reported landslides and earthquake magnitude for earthquakes with comprehensive inventories of landslides.
Data are for earthquakes listed in Table III. The solid line is the least-squares linear regression mean, with R 2 being the square of the correlation coefficient.

INVESTIGATING LANDSLIDES CAUSED BY EARTHQUAKES

497

498

DAVID K. KEEFER

caused landslides throughout areas larger than indicated by the upper bound curve
of Keefer (1984; Figure 1a); Rodrguez et al. (1999) therefore presented a revised
upper bound for this magnitude range (Figure 1b). They also noted a report of a
landslide from an M 2.9 earthquake in China in 1984 and determined the minimum Modified Mercalli shaking intensities at which landslides were reported in
the various earthquakes (Figure 4). The ranges in these minimum intensities were
identical to the ranges determined by Keefer (1984) except for the lack of reports of
any landslides in areas of MMI IV in the 19801997 database (Figure 4). However,
Rodrguez et al. (1999) found different modal values of threshold intensity (Figure
4): MMI VII for disrupted landslides, MMI VI for coherent landslides, and MMI
VIVII for lateral spreads and flows.
Hancox et al. (1997, 2002) synthesized data on landslides from 22 earthquakes in New Zealand with magnitudes ranging from 4.65.1 to 8.2 that occurred
between 1848 and 1995, inclusive. They found that areas affected by landslides
in most New Zealand events were small relative to the worldwide data (possibly
because of the limited range of geologic environments present in New Zealand),
although one small (ML 4.9) earthquake caused landslides throughout an area that
exceeded the worldwide upper bounds (Figure 1b). A linear-regression line plotted
through the New Zealand data by Hancox et al. (1997, 2002) related the total area
affected by landslides, A, to earthquake magnitude, M, by
Log10 A = 0.96(0.16)M 3.7(1.1)
where A is in square kilometers. Likewise, maximum epicentral distances for landslides generated by New Zealand earthquakes were also relatively small except for
the ML 4.9 event (Figure 2). Most landslides caused by the New Zealand earthquakes were disrupted landslides; the most numerous were rock falls and soil falls.
The minimum Modified Mercalli shaking intensity for those landslides was MMI
VI, and the most common intensities for significant landslide occurrence were
MMI VII to MMI VIII. The typical threshold intensity for soil lateral spreads was
VIII, but they occurred occasionally at MMI VII in highly susceptible materials
(Hancox et al., 1997). Landslide size was also related to earthquake magnitude,
shaking intensity, and epicentral distance. For example, only landslides having
volumes < 104 m3 occurred in New Zealand earthquakes having M < 6, whereas
landslides having volumes > 108 m3 occurred only in earthquakes having M > 7.5
and intensities of MMI IX or higher (Hancox et al., 2002).
Papadopoulos and Plessa (2000) compiled data on maximum epicentral distances of landslides in 47 earthquakes in Greece that occurred between 1650 and
1995. The authors noted that, for earthquakes before 1911, errors in epicentral
location were as great as about 30 km, and errors in magnitude were as great as 0.5
units. Thus, for many of the earthquakes, errors in epicentral location were of the
same order as the measured epicentral distances of landslides, and the error range
in magnitude determination was also significant. Nevertheless, with the exception

INVESTIGATING LANDSLIDES CAUSED BY EARTHQUAKES

499

of an MS 3.8 aftershock (which could be considered as part of an earthquake sequence with a main shock MS of 5.5), Papadopoulos and Plessa (2000) showed
that the maximum epicentral distances for landslides are in the same range as the
worldwide data. They proposed a straight-line upper bound for the Greek data,
log(Re ) = 2.98 + 0.75MS
for a range of magnitudes between approximately 4.5 and 7.5, where Re is the
maximum epicentral distance to a landslide, in kilometers.
Prestininzi and Romeo (2000) analyzed data on landslides in earthquakes that
have occurred in Italy since 461 BCE. Most of these earthquakes were from preinstrumental times and thus were associated with three sources of uncertainty.
First, pre-instrumental magnitudes were determined by converting from epicentral
intensities. Second, uncertainties in epicentral location were as large as 50 km for
the older events. Third, the reliability of reporting of landslides was questionable
for most events, and so the reported numbers of landslides were probably only a
small fraction of those that actually occurred. These effects caused large uncertainties in the determination of such measures as the maximum epicentral distances to
landslides. Nevertheless, most of these data were also in the same range as the other
worldwide data. A straight-line upper bound was developed relating epicentral intensity to distance to landslides, and minimum shaking intensities for landslides
were determined. These shaking intensities had a range of IV to XI and a modal
value of IX on the Mercalli-Cancani-Sieberg (MCS) scale.

6. Selected other recent findings


6.1. L ANDSLIDES AT GREAT DISTANCES
Five earthquakes in intraplate regions of North America have produced landslides
at distances significantly greater than indicated by the upper bounds shown in
Figures 2 and 3. One of these earthquakes was in the Saguenay, Quebec, region
of eastern North America, as was discussed above. Four of the earthquakes were in
the Colorado Plateau region of western North America. On the Colorado Plateau,
the ML 5.3 San Rafael Swell, Utah, earthquake of August 14, 1988, produced
possibly hundreds of rock falls (Case, 1988). Most of these were within about
40 km of the epicenter, but isolated rock falls were reported from at least as far
as 113 km and possibly as far as 129 km from the epicenter (Case, 1988; Figure
2). The M 5.7 St. George, Utah, earthquake of September 2, 1992 generated about
10 m of displacement on a 14 106 m3 block slide 44 km from the epicenter
(Jibson and Harp, 1996; Figure 2). An ML 4.6 earthquake in western Colorado
on September 13, 1994 triggered rock falls 87 km from the epicenter (Jibson and
Harp, 1996; Figure 2). The Western Arizona earthquake of April 29, 1993 (ML
= 5.3) generated one rock fall 169 km from the epicenter (Harp et al., 1993; D.

500

DAVID K. KEEFER

K Keefer and E. L. Harp, unpublished data; Figure 2). This earthquake produced
few other landslides and none in the area immediately around the epicenter (D.
K. Keefer and E. L. Harp, unpublished data.). The most distant rock fall detached
from a vertical cliff face along a prominent vertical joint that had been open before
the earthquake as indicated by the presence of live tree roots in place along the
preserved joint surface (D.K. Keefer, unpublished data).
The reasons for the triggering of these landslide at great distances are not well
known, but may involve anomalous seismic attenuation in the intraplate region
of North America, extraordinary susceptibility of some sites in these regions of
low seismicity, or a combination of both. For example, the most distant landslides
triggered by the Saguenay, Quebec, earthquake involved very sensitive clays and
fill embankments founded on loose, saturated materials and inclined at near their
angles of repose (Rodrguez et al., 1999). Additionally, shaking in the failure localities may have been amplified by the arrival of shear waves reflected from interfaces
in the lower crust simultaneously with direct radiation from the source (Rodrguez
et al., 1999). The block slide triggered by the St. George, Utah, earthquake was
one of only two coherent slides triggered by the earthquake and involved a slope
in clay, saturated by above-normal pre-earthquake precipitation, that was being
undercut by active fluvial erosion (Jibson and Harp, 1996). The condition of the
slope producing the rock fall in the Western Arizona earthquake indicated that the
rock mass that failed was probably partly detached prior to the earthquake and thus
highly susceptible to falling during even relatively weak shaking.
6.2. A RIAS INTENSITY THRESHOLDS
An instrumentally-based measure of seismic intensity developed by Arias (1970)
was first used for analyzing the occurrence of landslides by Wilson and Keefer
(1985), and its use has become relatively widespread for that purpose since that
time. This Arias intensity for any given strong-motion recording is expressed as:


Td

Ia = /2g

[a(t)2 ]dt
0

where Ia is the Arias intensity, t is the time, a(t) is the ground acceleration as
a function of time, Td is the total duration of the strong-motion record, and g is
the acceleration of gravity. Arias intensity is thus measured in units of velocity.
Based on theoretical considerations, statistical analysis of strong-motion attenuation, and empirical data on landslide limits in historical earthquakes, Wilson and
Keefer (1985) suggested threshold Arias intensity values of 0.15 m/s for disrupted landslides and 0.5 m/s for coherent slides, lateral spreads, and flows (Table
IV). Additional analysis by Keefer and Wilson (1989) suggested threshold Arias
intensity values of 0.11 m/s for disrupted landslides, 0.32 m/s for coherent slides,
and 0.54 m/s for lateral spreads and flows, whereas Wilson (1993) found the best-

INVESTIGATING LANDSLIDES CAUSED BY EARTHQUAKES

501

fit threshold Arias intensity value for disrupted landslides to be 0.10 m/s (Table
IV).
Harp and Wilson (1995) used data from two seismic events in southern California the October 1, 1987 Whittier Narrows earthquake (M 6.1) and the
double-shock October 24, 1987 Superstition Hills event (M 6.2 and 6.6) to correlate the mapped limits of landslide occurrence with Arias intensity values calculated
from strong-motion records. In the Superstition Hills event, most of the landslides
were soil falls and disrupted soil slides, and most were generated by the M 6.6
shock. Interpolation of data from 19 strong-motion records showed that the mean
threshold Arias intensity for the occurrence of these disrupted landslides was 0.3
m/s, with a range in thresholds of 0.1 to 0.5 m/s (Harp and Wilson, 1995; Table IV).
All of the landslides generated by the Whittier Narrows earthquake were rock falls,
rock slides, soil falls, or disrupted soil slides (Harp and Wilson, 1995). Interpolation of data from 83 strong-motion records showed a wider range of threshold Arias
intensities for these disrupted landslides, which varied with rock type (Harp and
Wilson, 1995). For weakly cemented sandstones and conglomerates of Miocene
and Pliocene age, threshold Arias intensity values were in the range between 0.06
and 0.7 m/s (Table IV). In an apparent paradox, threshold Arias intensities for
landslides in well cemented Mesozoic and Precambrian rocks were substantially
lower, with a range of 0.01 to 0.07 m/s and a mean of 0.04 m/s (Harp and Wilson,
1995; Table IV). The discrepancy in threshold values between the two classes of
rocks was explained by the pervasive occurrence of prominent, open fissures and
the abundance of loose rocks perched on steep slopes in areas underlain by the
older, better cemented rocks (Harp and Wilson, 1995).
6.3. D ELAYED INITIATION OR REACTIVATION OF LANDSLIDE MOVEMENT
Although many earthquake-induced landslides initiate in dry materials (Keefer,
1984; Keefer and Wilson 1989), seismic shaking also commonly combines with
existing or transiently elevated pore-water pressures to destabilize slopes. This
combination of agents can lead to the initiation or reactivation of landslide
movement, especially in coherent slides, several hours or days after strong groundshaking has stopped. However, only a few such cases have been documented.
Probably the earliest well-documented example was that of a large earth flow, 120
to 240 m wide and 800 m long, that was reactivated 5 days after the main shock of
the M 7.3 Hebgen Lake, Montana, earthquake in 1959 (Hadley, 1964). This earth
flow moved about 30 m during the month after the earthquake, and Hadley (1964)
inferred that the reactivated movement was due to an increase in ground-water
flow and a coincident increase in the local slope of the ground surface caused by
coseismic tectonic deformation.
The most numerous reports of delayed initiation or reactivation of landslides
were associated with the November 23, 1980 Irpinia, Italy, earthquake (M = 6.9). In
the region affected by that earthquake, several large earth flows and other coherent

0.15
0.11
0.10
0.5
0.32
0.5
0.54

Modeling studies
Disrupted
Disrupted
Disrupted
Coherent
Coherent
Lateral spreads and flows
Lateral spreads and flows
24 Oct. 1987 Superstition Hills, California earthquake
Disrupted
0.3 (mean)
Disrupted
0.10.5 (range)
1 Oct. 1987 Whittier Narrows, California earthquake
Disrupted
0.060.7 (weakly cemented sandstones and conglomerates)
Disrupted
0.04 (mean, well-cemented rocks)
Disrupted
0.010.07 (range, well-cemented rocks)

Arias intensity threshold (m/s)

Landslide category

TABLE IV
Arias intensity thresholds for landslides

Harp and Wilson, 1995


Harp and Wilson, 1995
Harp and Wilson, 1995

Harp and Wilson, 1995


Harp and Wilson, 1995

Wilson and Keefer, 1985


Keefer and Wilson, 1989
Wilson, 1993
Wilson and Keefer, 1985
Keefer and Wilson, 1989
Wilson and Keefer, 1985
Keefer and Wilson, 1989

Reference

502
DAVID K. KEEFER

INVESTIGATING LANDSLIDES CAUSED BY EARTHQUAKES

503

slides, with volumes as great as 28 106 m3 , began moving a few hours to a few
days after the main shock (Agnesi et al., 1983; Cotecchia and Del Prete, 1984;
DElia et al., 1985; Carrera et al.,1986; Cotecchia, 1986; Del Gaudio et al., 2000).
The post-earthquake movement on these landslides was inferred to have been
caused by increased spring flow and pore-water pressures associated with tectonic
deformation of the area in which the landslides occurred (Cotecchia and Del Prete,
1984; Cotecchia, 1986; Wasowski et al., this volume). These hydrologic changes
were also probably partly responsible for an anomalous distribution of landslides
in part of the affected region, where landslide concentrations were relatively higher
at greater distances from the fault rupture (Wasowski et al., this volume).
Similar observations of delayed reactivation owing to changes in ground-water
conditions were reported by Jibson et al. (1994) from the epicentral area of the
M 7.1 April 29, 1991 earthquake in the Racha region of the Republic of Georgia.
There, two large earth flows showed only small displacements during the main
shock but began moving at velocities as great as several m/day 2 to 3 days after the
earthquake (Jibson et al., 1994).
Following the October 28, 1983 Borah Peak, Idaho, earthquake (M = 7.0),
Keefer et al. (1985) reported that a rapid soil flow with an estimated volume of
0.2 106 m3 had initiated 36 to 48 hours after the main shock. They attributed
this initiation to saturation of the source colluvium owing to increased surface and
subsurface water flow resulting from the earthquake, which Wood et al. (1985),
in turn, concluded was probably due to sudden release of elastic strain in aquifers
during the earthquake.

7. Summary
Although earthquake-induced landslides have been described in documents for
more than 3700 years, accounts from earthquakes before the late eighteenth century are incomplete concerning landslide numbers and vague concerning landslide
characteristics. They are thus typically misleading concerning the true abundance
of landslides and range of characteristics. Beginning with studies of the 1783
Calabria, Italy, earthquake sequence, more complete and precise data concerning
the occurrence of landslides in earthquakes have become available. The historical
development of knowledge concerning landslides triggered by earthquakes can be
divided into several periods. The first period, from 1783 until the first application
of aerial photography, was characterized by ground-based studies of earthquake
effects, typically carried out by formal scientific commissions. These formal studies typically identified a large, but not necessarily comprehensive, sampling of
localities where landslides had occurred. In some, but not all cases, landslide characteristics were also described in enough detail that the general range of landslide
characteristics could begin to be determined. More recently, some nineteenth to
mid-twentieth century earthquakes have been studied using retrospective analyses,

504

DAVID K. KEEFER

in which the landslide occurrences associated with the event are inferred years
to decades later, using contemporary accounts, mapping from aerial photographs,
statistical studies, and (or) geotechnical analyses.
The first use of aerial photographs to map earthquake effects immediately after
the event probably occurred in 1948. Since that time, the use of aerial photography
has greatly facilitated the compilation of post-earthquake landslide inventories, although because of the limitations of aerial photography, ground-based field studies
continue to be crucial in preparing accurate and comprehensive landslide maps.
Beginning with a small California earthquake in 1957, extensive to relatively complete inventories of landslides have been prepared for a relatively small number of
earthquakes. Through the 1960s and 1970s the best landslide inventories typically
were complete only for a central affected area, although the first virtually complete
inventory of a large earthquake was prepared for the M 7.5 Guatemala earthquake
in 1976.
Beginning in 1980, virtually complete landslide inventories have been prepared
for several additional earthquakes in California, El Salvador, Japan, Italy, and
Taiwan. Most of these used aerial photography in combination with ground-based
field studies, although studies of the most recent of these events, in Taiwan, have
also used satellite imagery, and three of the others (including the two smallest)
were compiled largely from ground-based field studies without aerial photography.
Since 1989, digital-mapping and GIS techniques have come into common use for
mapping earthquake-induced landslides, and the use of these techniques has greatly
enhanced the level of analysis that can be applied to earthquake-induced landslide
occurrence.
The first general synthesis of data on earthquake-induced landslides (Keefer,
1984) defined the general characteristics of these landslides, derived relations
between landslide occurrence on the one hand and geologic and seismic parameters
on the other hand, and identified the types of hazards associated with the landslides.
Since then, additional synthesis of worldwide data (Rodrguez et al., 1999) and
national data from New Zealand (Hancox et al., 1997, 2002), Greece (Papadopoulos and Plessa, 2000), and Italy (Prestininzi and Romeo, 2000) have provided
additional data on landslide characteristics and hazards and have extended, revised,
and refined these relations. Recently completed studies have also identified areas
with anomalous landslide distributions, have provided data for correlating the occurrence of landslides with a measure of local ground motion, and have verified the
occasional delayed triggering of landslides as a consequence of seismic shaking.
The documentation and synthesis of data on landslide occurrence in earthquakes has led to greatly increased understanding of the hazards associated with
earthquake-induced landslides and to the development of models and methods for
hazard mapping and evaluation. However, the number of earthquakes with relatively complete data on landslide occurrence is still small, and one of the most
pressing research needs is for complete landslide inventories for many more events
in a wider variety of environments. Whereas many additional inventories in all

INVESTIGATING LANDSLIDES CAUSED BY EARTHQUAKES

505

regions are needed, data from seismically active regions of continental Asia and the
Middle East, the Andean region of South America, the tropical western Pacific, and
intraplate regions worldwide would be especially desirable. Such additional data,
coupled with the increasing use of GIS and other current analytical tools should
lead to substantial additional refinements in models relating seismic shaking and
geologic conditions to slope failure, and thus to our ability to minimize damage
and loss of life from seismically generated landslides.

Acknowledgements
This article was improved by the thoughtful reviews of Vincenzo Del Gaudio, Lynn
Highland, Randall Jibson, and Janusz Wasowski.

References
Adams, J.: 1980, Contemporary Uplift and Erosion of the Southern Alps, New Zealand, Geological
Society of America Bulletin 91(1), 24 and (2), 1114.
Agnesi, V., Carrara, A., Macaluso, T., Monteleone, S., Pipitone, G., and Sorriso-Valvo, M.: 1983,
Elementi Tipologici e Morfologici dei Fenomeni di Instabilit dei Versanti Indotti dal Sisma del
1980 (Alta Valle del Sele), Geologia Applicata e Idrogeologia 18(1), 309341.
Ambraseys, N.N.: 1976, The Gemona di Friuli Earthquake of 6 May 1976, UNESCO Restricted
Technical Report RP/1975-76/2.222.3, Part II.
Arias, A.: 1970, A Measure of Earthquake Intensity, in R.J. Hansen (ed.), Seismic Design for Nuclear
Power Plants, Massachusetts Institute of Technology Press, Cambridge, Massachusetts, pp. 438
483.
Bonilla, M.G.: 1960, Landslides in the San Francisco South Quadrangle, California, U.S. Geological
Survey Open-File Report.
Bozzano, F., Gambino, P., Prestininzi, A., Scarascia Mugnozza, G., and Valentini, G.: 1998, Ground
Effects Induced by the Umbria-Marche Earthquakes of SeptemberOctober 1997, Central Italy,
Proceedings of the 8th International Congress of the International Association for Engineering
Geology and the Environment, Vancouver, Canada, 2125 September, 1998, A.A. Balkema,
Rotterdam, pp. 825830.
Carrara, A., Agnesi, V., Macaluso, T., Monteleone, S., and Pipitone, G.: 1986, Slope Movements
Induced by the Southern Italy Earthquake of November 1980, Geologia Applicata e Idrogeologia
21(2), 237250.
Case, W.F.: 1988, Geologic Effects of the 14 and 18 August, 1988 Earthquakes in Emery County,
Utah, Utah Geological and Mineral Survey, Survey Notes 22, pp. 814.
Cheng, J.D., Chen, B.K., Su, R.R., and Yeh, C.L.: 2000, Landslides Induced by the Disastrous
Earthquake on September 21, 1999 in Central Taiwan (Abs.), European Geophysical Society
25th General Assembly, Nice, France, 2000, Geophysical Research Abstracts.
Collins, J.L. and Foster, H.L.: 1949, The Fukui Earthquake Hokuriku Region, Japan, 28 June 1948,
Volume I, Geology, U.S. Army Office of the Engineer, General Headquarters, Far East Command.
Cotecchia, V.: 1986, Ground Deformations and Slope Instability Produced by the Earthquake of 23
November 1980 in Campania and Basilicata, Proceedings of the International Symposium on
Engineering Geology Problems in Seismic Areas, Bari, Italy vol. 5, pp. 31100.

506

DAVID K. KEEFER

Cotecchia, V. and Del Prete, M.: 1984, The Reactivation of Large Flows in the Parts of Southern Italy
Affected by the Earthquake of November 1980, with Reference to the Evolutive Mechanism,
Proceedings of the Fourth International Symposium on Landslides, Toronto, vol. 2, pp. 3337.
Cotecchia, V., Guerricchio, A., and Melidoro, G.: 1986, The Geomorphogenetic Crisis Triggered by
the 1783 Earthquake in Calabria (Southern Italy), Proceedings of the International Symposium
on Engineering Geology Problems in Seismic Areas, Bari, Italy, vol. 6, pp. 245304.
Cotecchia, C. and Melidoro, G.: 1974, Some Principal Aspects of the Landslides of Southern Italy,
Bulletin of the International Association of Engineering Geology 9, 2332.
Davison, C.: 1936, Great Earthquakes, Thomas Murby & Co., London.
Del Gaudio, V., Trizzino, R., Calcagnile, G., Calvaruso, D., and Pierri, P.: 2000, Landsliding in
Seismic Areas: The Case of the Acquara-Vadoncello Landslide (Southern Italy), Bull. Eng. Geol.
Env. 59, 2337.
DElia, B., Esu, F., Pellegrino, A., and Pescatore, T.: 1985, Some Effects on Natural Slope Stability
Induced by the 1980 Italian Earthquake, Proceedings of the Eleventh International Society of
Soil Mechanics and Foundation Engineering Conference, San Francisco.
Dutton, C.E.: 1889, The Charleston Earthquake of August 31, 1886, U.S. Geological Survey Ninth
Annual Report 18871888, pp. 203528.
Esposito, E., Porfido, S., Simonelli, A.L., Mastrolorenzo, G., and Iaccarino, G.: 2000, Landslides
and Other Surface Effects Induced by the 1997 Umbria-Marche Seismic Sequence, Engineering
Geology 58, 353376.
Fukuoka, H., Sassa, K., and Scarascia-Mugnozza, G.: 1997, Distribution of Landslides Triggered by
the 1995 Hyogo-ken Nanbu Earthquake and Long Runout Mechanism of the Takarazuka Golf
Course Landslide, Journal of Physics of the Earth 45, 8390.
Fuller, M.L.: 1912, The New Madrid Earthquake, U.S. Geological Survey Bulletin 394, reprint
edition.
Geological Survey of India: 1939, The Bihar-Nepal Earthquake of 1934, Memoirs of the Geological
Survey of India 73.
Govi, M.: 1977a, Carta delle Frane Prodotte dal Terremoto (Map Showing the Landslides Triggered
by the Earthquake), Rivista Italiana di Paleontologia e Stratigrafia 83, Plate 1.
Govi, M.: 1977b, Photo-interpretation and Mapping of the Landslides Triggered by the Friuli
Earthquake (1976), International Association of Engineering Geology Bulletin 15, 6772.
Govi, M. and Sorzana, P.F.: 1977, Effetti Geologici del Terremoto, Frane, Rivista Italiana di
Paleontologia e Stratigrafia 83, 329368.
Hadley, J.B.: 1964, Landslides and Related Phenomena Accompanying the Hebgen Lake Earthquake
of August 17, 1959, U.S. Geological Survey Professional Paper 435, pp. 107138.
Hall, J.F. (ed.): 1994, The Northridge Earthquake January 17, 1994 Preliminary Reconnaissance
Report, Earthquake Engineering Research Institute Publication 94-01.
Hancox, G.T., Perrin, N.D., and Dellow, G.D.: 1997, Earthquake-induced Landsliding in New Zealand and Implications for MM Intensity and Seismic Hazard Assessment, Institute of Geological
and Nuclear Sciences, Ltd., GNS Client Report 43601B.
Hancox, G.T., Perrin, N.D., and Dellow, G.D.: 2002, Recent Studies of Historical EarthquakeInduced Landsliding, Ground Damage, and MM Intensity in New Zealand, Bulletin of the New
Zealand Society for Earthquake Engineering, vol. 35, no. 2, pp. 5994.
Hansen, A. and Franks, C.A.M.: 1991, Characterisation and Mapping of Earthquake Triggered
Landslides for Seismic Zonation, Proceedings of the Fourth International Conference on Seismic
Zonation, Stanford, California, vol. 1, pp. 149195.
Harp, E.L. and Jibson, R.W.: 1995, Inventory of Landslides Triggered by the 1994 Northridge,
California Earthquake, U.S. Geological Survey Open-File Report 95-213.
Harp, E.L. and Jibson, R.W.: 1996, Landslides Triggered by the 1994 Northridge, California,
Earthquake, Bulletin of the Seismological Society of America 86(1B), S319S332.

INVESTIGATING LANDSLIDES CAUSED BY EARTHQUAKES

507

Harp, E.L., Jibson, R.W., and Keefer, D.K.: 1993, Seismically-induced Landslides Triggered at Extraordinary Distances: Evidence from the Springdale, Utah, Landslide, Geological Society of
America Abstracts with Programs, vol. 25, no. 6, p. A32.
Harp. E.L. and Keefer, D.K.: 1990, Landslides Triggered by the Earthquake, in M.J. Rymer and W. L.
Ellsworth (eds.), The Coalinga, California, Earthquake of May 2, 1983, U.S. Geological Survey
Professional Paper 1487, pp. 335347.
Harp, E. L., Tanaka, K., Sarmiento, J., and Keefer, D. K.: 1984, Landslides from the May 2527,
1980, Mammoth Lakes, California, Earthquake Sequence, U.S. Geological Survey Miscellaneous
Investigations Series Map I-1612.
Harp, E.L. and Wilson, R.C.: 1995, Shaking Intensity Thresholds for Rock Falls and Slides: Evidence
from 1987 Whittier Narrows and Superstition Hills Earthquake Strong-motion Records, Bulletin
of the Seismological Society of America 85, 17391757.
Harp, E.L., Wilson, R.C., and Wieczorek, G.F., 1981, Landslides from the February 4, 1976,
Guatemala Earthquake, U.S. Geological Survey Professional Paper 1204-A.
Hecker, S., Ponti, D.J., Garvin, C.D., and Hamilton, J.C.: 1995, Characteristics and Origin of Ground
Deformation Produced in Granada Hills and Mission Hills During the January 17, 1994 Northridge, California, Earthquake, California Department of Conservation, Division of Mines and
Geology Special Publication 116, pp. 111131.
Holzer, T.L. (ed.): 1998, The Loma Prieta, California, Earthquake of October 17, 1989 Liquefaction, U.S. Geological Survey Professional Paper 1551-B.
Japan Geotechnical Society, 1996, Special Issue on Geotechnical Aspects of the January 17, 1995
Hyogoken-Nambu Earthquake, Soils and Foundations, The Japanese Geotechnical Society,
Tokyo.
Jibson, R.W.: 1985, Landslides Caused by the 1811-12 New Madrid Earthquakes, Ph.D. Dissertation,
Stanford University.
Jibson, R.W. and Harp, E.L.: 1996, The Springdale, Utah, Landslide: An Extraordinary Event,
Environmental & Engineering Geoscience 2, 137150.
Jibson, R.W., Harp, E.L., and Michael, J.A.: 1998, A Method for Producing Digital Probabilistic
Seismic Landslide Hazard Maps: An Example from the Los Angeles, California, Area, U.S.
Geological Survey Open-File Report 98-113.
Jibson, R.W., Harp, E.L., and Michael, J.A.: 2000, A Method for Producing Digital Probabilistic
Seismic Landslide Hazard Maps, Engineering Geology 58, 271289.
Jibson, R.W. and Keefer, D.K.: 1988, Landslides Triggered by Earthquakes in the Central Mississippi
Valley, Tennessee and Kentucky, U.S. Geological Survey Professional Paper 1336-C.
Jibson, R.W. and Keefer, D.K.: 1989, Statistical Analysis of Factors Affecting Landslide Distribution
in the New Madrid Seismic Zone, Tennessee and Kentucky, Engineering Geology 27, 509542.
Jibson, R.W. and Keefer, D.K.: 1993, Analysis of the Seismic Origin of Landslides: Examples from
the New Madid Seismic Zone, Geological Society of America Bulletin 105, 521536.
Jibson, R.W., Prentice, C.S., Borissoff, B.A., Rogozhin, E.A., and Langer, C.J.: 1994, Some Observations of Landslides Triggered by the 29 April 1991 Racha Earthquake, Republic of Georgia,
Bulletin of the Seismological Society of America 84, 963973.
Johnston, A.: 1996, Seismic Moment Assessment of Earthquakes in Stable Contiental Regions III.
New Madrid 18111812, Charleston 1886, and Lisbon 1755, Geophysical Journal International
126, 314344.
Kamai, T.: 1995, Landslides in the Hanshin Urban Region Caused by the 1995 Hygoken-Nanbu
Earthquake, Japan, Landslide News 9, 1213.
Kamai, T., Wang, W.N., and Shuzui, H.: 2000, The Landslide Disaster Induced by the Taiwan ChiChi Earthquake of 21 September 1999, Landslide News 13, 812.
Keefer, D.K.: 1984, Landslides Caused by Earthquakes, Geological Society of America Bulletin 95,
406421.

508

DAVID K. KEEFER

Keefer, D. K. (ed.): 1998, The Loma Prieta, California, Earthquake of October 17, 1989 Landslides:
U.S. Geological Survey Professional Paper 1551-C.
Keefer, D.K.: 1999, Earthquake-induced Landslides and Their Effects on Alluvial Fans, Journal of
Sedimentary Research 69, 84104.
Keefer, D.K.: 2000, Statistical Analysis of an Earthquake-induced Landslide Distribution the 1989
Loma Prieta, California Event, Engineering Geology 58, 213249.
Keefer, D.K. and Manson, M.W.: 1998, Regional Distribution and Characteristics of Landslides
Generated by the Earthquake, in D.K. Keefer (ed.), The Loma Prieta, California, Earthquake
of October 17, 1989 - Landslides, U. S. Geological Survey Professional Paper 1551-C, pp. 732.
Keefer, D.K. and Tannaci, N.E.: 1981, Bibliography on Landslides, Soil Liquefaction, and Related
Ground Failures in Selected Historic Earthquakes, U.S. Geological Survey Open-File Report
81-572.
Keefer, D.K. and Wilson, R.C.: 1989, Predicting Earthquake-induced Landslides, with Emphasis on
Arid and Semi-arid Enviroments, in P.M. Sadler, and D.M. Morton (eds.), Landslides in a Semiarid Environment, Inland Geological Society of Southern California Publications, Riverside,
California, vol. 2, part 1, pp. 118149.
Keefer, D.K., Wilson, R.C., Harp, E.L., and Lips, E.W.: 1985, The Borah Peak, Idaho Earthquake of
October 28, 1983 Landslides, Earthquake Spectra 2, 91125.
Lawson, A.C. (ed.): 1908, The California Earthquake of April 18, 1906: Report of the California
State Earthquake Investigation Commission, Carnegie Institution, Washington, D.C., Publication
87, reprint edition.
Lee, K.L., Marcuson, W.F. III, Stokoe, K.H. II, and Yokel, F.Y. (eds.): 1977, Research Needs and Priorities for Geotechnical Earthquake Engineering Applications, Report of Workshop at University
of Texas, Austin, National Science Foundation Grant No. AEN77-09861.
Lee, W.H.K.: 2001, Strong-motion Instrumentation and Data, in J. Uzarski and C. Arnold, (eds.),
Chi-Chi, Taiwan, Earthquake of September 21, 1999 Reconnaissance Report, Earthquake Engineering Research Institute Publication 2001-02, Earthquake Spectra, Supplement A to Vol. 17,
pp. 518.
Liao, H.W. and Lee, C.T.: 2000, Landslides Triggered by the Chi-Chi Earthquake, Asian
Association on Remote Sensing, Asian Conference on Remote Sensing ACRS 2000,
http://www.gisdevelopment.net/aars/acrs/2000/ts8/hami0007.htm.
Lin, J.C., Jen, C.H., and Jou, T.C.: 2001, Earthquake Landslide Hazard: Case Study of Chi-Chi
Earthquake, Taiwan, Paper Presented at International Conference on Geomorphology, Tokyo,
(Abs.), p. 471.
Lin, M.L., Wang, K.L., and Chen, C.T.: 2000, Characteristics of the Slope Failure Caused by the ChiChi Earthquake, International Workshop on Annual Commemoration of the Chi-Chi Earthquake,
September 1820, 2000, vol. 3, pp. 199209.
Lyell, C.: 1874, Principles of Geology, D. Appleton, New York, 11th edition, vol. 2, pp. 113144.
Manson, M.W., Keefer, D.K., and McKittrick, M.A., (eds.): 1992, Landslides and Other Geologic
Features in the Santa Cruz Mountains, California, Resulting from the Loma Prieta Earthquake of
October 17, 1989, State of California Division of Mines and Geology Open-File Report 91-05.
Matthews, W.H.: 1979, Landslides of Central Vancouver Island and the 1946 Earthquake, Seismological Society of America Bulletin 69, 445450.
Mitchill, S.L.: 1815, A Detailed Narrative of the Earthquake which Occurred on the 16th Day of
December, 1811, Literary and Philosophical Society of New York, Transactions 1, 281307.
Morton, D.M.: 1971, Seismically Triggered Landslides in the Area Above the San Fernando Valley,
U.S. Geological Survey Professional Paper 733, pp. 99104.
Morton, D.M.: 1975, Seismically Triggered Landslides in the Area Above the San Fernando Valley,
State of California Division of Mines and Geology Bulletin 196, pp. 145154.

INVESTIGATING LANDSLIDES CAUSED BY EARTHQUAKES

509

National Research Council Committee on the Alaska Earthquake: 19681973, The Great Alaska
Earthquake of 1964, National Academy of Sciences, Washington, D.C., National Research
Council Publication 1603.
Newmark, N.M.: 1965, Effects of Earthquakes on Dams and Embankments, Geotechnique 15, 139
160.
Oldham, R.D.: 1899, Report on the Great Earthquake of 12th June 1897, Memoirs of the Geological
Survey of India, vol. 29.
Ochiai, H., Kitahara, H., Sammori, T., and Abe, K.: 1996, Landslides Triggered by the 1995 HyogenKen Nanbu Earthquake in the Rokko Mountains, in K. Senneset (ed.), Landslides, Proceedings
of the Seventh International Symposium on Landslides, Balkema, Rotterdam, pp. 10071012.
Okimura, T. and Torii, N.: 1999, A Study on Slope Failures due to Hyogoken-Nanbu Earthquake
and Post-earthquake Rainfalls, Sino-Japan Second Workshop on Seismic Hazard and Mitigation,
Taiwan National University, Taipei, Taiwan, pp. 6265.
ORourke, T.D. (ed.): 1992, The Loma Prieta, California, Earthquake of October 17, 1989 Marina
District: U.S. Geological Survey Professional Paper 1551-F.
Papadopoulos, G.A. and Plessa, A.: 2000, Magnitude-Distance Relations for Earthquake-induced
Landslides in Greece, Engineering Geology 58, 377386.
Parise, M. and Jibson, R.W.: 2000, A Seismic Landslide Susceptibility Rating of Geologic Units
Based on Analysis of Characteristics of Landslides Triggered by the 17 January, 1994 Northridge,
California Earthquake, Engineering Geology 58, 251270.
Pearce, A.J., and OLoughlin, C.L.: 1985, Landsliding During a M 7.7 Earthquake: Influence of
Geology and Topography, Geology 13, 855858.
Penick, J. L., Jr.: 1981, The New Madrid Earthquakes of 18111812, University of Missouri Press,
Columbia and London.
Plafker, G. and Ericksen, G.E.: 1978, Nevados Huascarn Avalanches, Peru, in B. Voight (ed.),
Rockslides and Avalanches, 1 Natural Phenomena, Elsevier, Amsterdam, pp. 277314.
Plafker, G., Ericksen, G.E., and Fernndez Concha, J.: 1971, Geological Aspects of the May 31,
1970, Per Earthquake, Bulletin of the Seismological Society of America 61, 543578.
Plafker, G., Kachadoorian, R., Eckel, E.B., and Mayo, L.R.: 1969, Effects of the Earthquake of March
27, 1964 on Various Communities, U.S. Geological Survey Professional Paper 542-G.
Plant, N. and Griggs, G.B.: 1990, Coastal Landslides Caused by the October 17, 1989 Earthquake
Santa Cruz County, California, California Geology 43, 7584.
Prestininzi, A. and Romeo, R.: 2000, Earthquake-induced Ground Failures in Italy, Engineering
Geology 58, 387397.
Rodrguez, C.E., Bommer, J.J., and Chandler, R.J.: 1999, Earthquake-induced Landslides: 1980
1997, Soil Dynamics and Earthquake Engineering 18, 325346.
Rogers, G.C.: 1980, A Documentation of Soil Failure During the British Columbia Earthquake of 23
June, 1946, Canadian Geotechnical Journal 17, 122127.
Rymer, M.J.: 1987, The San Salvador Earthquake of October 10, 1986 Geologic Aspects,
Earthquake Spectra 3, 435463.
Rymer, M.J. and White, R.A.: 1989, Hazards in El Salvador from Earthquake-induced Landslides,
in E.E. Brabb and B.L. Harrod (eds.), Landslides: Extent and Economic Significance, Balkema,
Rotterdam, pp. 105109.
Sassa, K., Fukuoka, H., Scarascia-Mugnozza, G., Irikura, K., and Okimura, T.: 1995, Landslides
Triggered by the Hyogoken-Nanbu Earthquake, Landslide News 9, 25.
Sarconi, M.: 1784, Osservazioni Fatte Nelle Calabrie e Nella Frontiera di Valdemone sui Fenomeni
del Tramoto del 1783 e Sulla Geografia Fisica di Quelle Regioni, Reali Accademia delle Scienze
e Belle Lettere. Napoli.
Seed, H.B.: 1968, Landslides During Earthquakes Due to Soil Liquefaction, American Society of
Civil Engineers, Journal of the Soil Mechanics and Foundations Division 94, 10531122.

510

DAVID K. KEEFER

Simonett, D.S.: 1967, Landslide Distribution and Earthquakes in the Bewani and Torricelli Mountains, New Guinea a Statistical Analysis, in J.N. Jennings and J.A. Mabbutt (eds.), Landform
Studies from Australia and New Guinea, Cambridge University Press, Cambridge, United
Kingdom, pp. 6484.
Sitar, N. and Bardet, J.P.: 2001, Landslides, in J. Uzarski and C. Arnold (eds.), Chi-Chi, Taiwan,
Earthquake of September 21, 1999 Reconnaissance Report, Earthquake Engineering Research
Institute Publication 20012002, Earthquake Spectra, Supplement A to Vol. 17, pp. 6176.
Spittler, T. E. and Harp, E.L.: 1990, Preliminary Map of Landslide Features and Coseismic Fissures,
in the Summit Road Area of the Santa Cruz Mountains, Triggered by the Loma Prieta Earthquake
of October 17, 1989, U.S. Geological Survey Open-File Report 90-688.
Stewart, J.P.: 2001, Soil Liquefaction, in J. Uzarski and C. Arnold, (eds.), Chi-Chi, Taiwan, Earthquake of September 21, 1999 Reconnaissance Report, Earthquake Engineering Research Institute
Publication 20012002, Earthquake Spectra, Supplement A to Vol. 17, pp. 3760.
Stewart, J.P., Seed, R.B., and Bray, J.D.: 1996, Incidents of Ground Failure from the 1994 Northridge
Earthquake, Bulletin of the Seismological Society of America 86(1B), S300S318.
Varnes, D.J.: 1978, Slope Movement Types and Processes, in R.L. Schuster and R.J. Krizek (eds.),
Landslides Analysis and Control, National Academy of Sciences Transportation Research
Board Special Report 176, pp. 1133.
Vivenzio, G.: 1788, Istoria de Tremouti Avvenuti Nella Provincia della Calabria Ulteriore, e Nella
Citt di Messina Nellanno 1783. E di Quanto Nella Calabria Fu Fatto per lo Suo Risorgimento
Fino al 1787. Preceduta da una Teoria ed Istoria Generale dei Tremouti. II Edizione, Stamperia
Reale, Napoli.
Wang, G. and Xu, B: 1984, Brief Introduction of Landslides in Loess in China, Proceedings of the
Fourth International Symposium on Landslides, Toronto, vol. 1, pp. 197207.
Wasowski, J., Del Gaudio, V., Pierri, P., and Capolongo, D.: 2002, Factors Controlling Seismic
Susceptibility of the Sele Valley Slopes: The Case of the 1980 Irpinia Earthquake Re-examined,
Surveys in Geophysics (this volume).
Weber, G.E. and Nolan, J.M., 1989, Landslides and Associated Ground Failure in the Epicentral Region of the October 17, 1989 Loma Prieta Earthquake, Final Technical Report to U.S. Geological
Survey under Contract 14-08-0001-G1861.
Wieczorek, G.F. and Jger, S.: 1996, Triggering Mechanisms and Depositional Rates of Postglacial
Slope-movement Processes in the Yosemite Valley, California, Geomorphology 15, 1731.
Wilson, R.C.: 1993, Relation of Arias Intensity to Magnitude and Distance in California, U.S.
Geological Survey Open-File Report 93-556.
Wilson, R.C. and Keefer, D.K.: 1985, Predicting Areal Limits of Earthquake-induced Landsliding, in
J.I. Ziony, (ed.), Earthquake Hazards in the Los Angeles Region An Earth-science Perspective,
U.S. Geological Survey Professional Paper 1360, pp. 317345.
Wilson, R. C., Wieczorek, G. F., Keefer, D. K., Harp, E. L., and Tannaci, N.E.: 1985, Map Showing Ground Failures from the Greenville/Mount Diablo Earthquake Sequence of January 1980,
Northern California, U.S. Geological Survey Miscellaneous Field Studies Map MF 1711.
Wood, S.H., Wurts, C., Lane, T., Ballenger, N., Shaleen, M., and Totorica, D.: 1985, The Borah Peak,
Idaho Earthquake of October 28, 1983 Hydrologic Effects, Earthquake Spectra 2, 127150.
Youd, T.L. and Hoose, S.N.: 1978, Historic Ground Failures in Northern California Associated with
Earthquakes, U.S. Geological Survey Professional Paper 993.

You might also like