1. EARTHQUAKE-INDUCED LANDSLIDES 25 YEARS LATER
David Keefer

2. CHRONOLOGICAL OUTLINE
BEFORE 25 YEARS AGO
Before 1783
1783 through 1947
1948 through 1983
THE LAST 25 YEARS

3. TOPICAL OUTLINE Post-earthquake investigations and inventories
Worldwide and regional-scale synthesis of data
Hazard mapping and evaluation
Geomorphology and landscape evolution
Paleoseismology
NOTE: NOT discussing site-specific seismic slope-stability analyses, laboratory testing of materials, etc.

4. ANCIENT TIMES (BEFORE 1783)
Landslides mentioned as earthquake effects as early as 1789 BCE in China (Hansen and Franks, 1991), BUT
Data for any given earthquake are very incomplete
Descriptions of locations and characteristics are vague
Earthquakes are pre-instrumental, so epicenters, etc. poorly constrained
At worst, “data” are misleading

5. 1783-1947: GROUND-BASED SCIENTIFIC POST-EARTHQUAKE INVESTIGATIONS
Characteristics
Ground-based surveys of large areas
Landslides documented along with other earthquake effects
Large (but not necessarily complete) sampling of landslide localities
More precise and detailed descriptions of characteristics than before this time
Notable Examples
1783 Calabria, Italy (M 7) (Sarconi, 1784; Lyell, 1874)
New Madrid, Missouri USA M (Fuller, 1912)
1886 Charleston, South Carolina, USA (M 6.8) (Dutton, 1989)
1897 Assam, India (M 8.3) (Oldham, 1899)
1906 San Francisco, California, USA (M 7.8) (Lawson, 1908)
1934 Bihar, India-Nepal (M 8.1) (Geological Survey of India, 1939)

6. 1948 THROUGH 1983 Use of aerial photography to map landslides
Extensive to “complete” landslide inventories with detailed data on locations and characteristics

7. COMPLETE LANDSLIDE INVENTORIES THROUGH 1983
EARTHQUAKE
DATE M
NUMBER OF LANDSLIDES
REFERENCES
Daly City, Calif. USA
22-May-1957
5.3 23
Bonilla, 1960
Guatemala
4-Feb-1976
7.5
~50,000
Harp et al., 1981
Mt. Diablo, Calif. USA
24-Jan-1980
5.8
103
Wilson et al., 1985
Mammoth Lakes, Calif. USA
25-May-1980
6.2
5253
Harp et al.,1984
Coalinga, Calif. USA
2-May-1983
6.5
9389
Harp and Keefer, 1990

8. 1964 ALASKA M 9.2 (> 60 reports)
Area too large (269,000 km2) and much of it too remote for complete inventory
“Thousands of landslides,” at least
Nevertheless, much detailed information on landslides in numerous reports
Landslides caused 56% of economic loss and at least 45 of the 130 deaths
Detailed site-specific analyses of large landslides in Anchorage, Alaska’s largest city
Turnagain Heights Landslides, Anchorage

9. 1970 PERU M 7.9 (Cluff, 1971; Plafker et al., 1971)
First extensive airphoto inventory for large earthquake
Large landslides mapped throughout central area of 8300 km2
Nevados Huascarán rock avalanche killed >25,000 people
(From Plafker et al., 1971)

10. OTHER NOTABLE POST-EARTHQUAKE INVESTIGATIONS: 1948-1983
1959 Hebgen Lake, Mont. USA M 7.1 (Hadley, 1964)
1971 San Fernando, Calif. USA M 6.7 (Morton, 1971, 1975)
1976 Friuli, Italy M (Govi, 1977, Govi and Sorzana, 1977)
1978 Izu-Oshima Kinkai, Japan (Geographical Survey Institute of Japan, 1978; Okusa and Anma, 1980)
1980 Irpinia, Italy M 6.9 (Agnesi et al. 1983; Carrera et al., 1986; Del Gaudio et al. 2000; Wasowski et al., 2002; Del Gaudio and Wasowski, 2004)

11. NEWMARK SEISMIC SLOPE STABILITY ANALYSIS
HAZARD EVALUATION
NEWMARK SEISMIC SLOPE STABILITY ANALYSIS
Developed by Newmark (1965)
Field verified by Wilson and Keefer (1983)
Applied on regional scale by Wieczorek et al. (1985)
(From Wilson and Keefer, 1983; Jibson et al. 2000)

12. 1948 THROUGH 1983 Post-earthquake investigations and inventories
Worldwide and regional-scale synthesis of data
First worldwide compilation (for liquefaction-induced landslides; Seed, 1967 Terzaghi Lecture)
First regional compilation (for northern California, including 1906 San Francisco earthquake; Youd and Hoose, 1978)
Hazard mapping and evaluation
Geomorphology and landscape evolution
First landscape evolution study (for Southern Alps, New Zealand; Adams, 1980)
Paleoseismology
Simonett, pioneered statistical method, using 1935 New Guinea earthquake (M 7.9)

13. 25 YEARS AGO (1984): First general synthesis of worldwide data

14. 25 YEARS AGO: ANALYSIS OF WORLDWIDE DATASET (Keefer, 1984) Sample of 40 worldwide earthquakes M > 5.2 (Supplemented several hundred small US events to determine minimum M for triggering)
Classification
Abundance
DISRUPTED SLIDES AND FALLS
Rock falls Soil falls
Rock slides Disrupted soil slides
Rock avalanches Soil avalanches
COHERENT SLIDES
Rock slumps Soil slumps
Rock block slides Soil block slides
Slow earth flows
LATERAL SPREADS AND FLOWS
Soil lateral spreads
Rapid soil flows
Subaqueous landslides

15. DISRUPTED SLIDES AND FALLS
Highly disrupted masses (small blocks, fragments, or particles)
Steep slopes (may run out onto gentle slopes)
High velocity
Move long distances
Shallow (< 3 m deep), except for rock avalanches
Moderate to high casualty potential for rock (low for soil)
Low to moderate historical damage costs
Soil falls Rock falls
Disrupted soil slides Rock slides
Soil avalanches Rock avalanches

16. COHERENT SLIDES Relatively coherent masses (one or a few large blocks)
Moderate slopes
Low velocity
Move short distances
Deep (> 3 m)
Low to moderate casualty potential (higher for soil)
Moderate to high historical damage costs
Soil slumps Rock slumps
Soil block slides Rock block slides
Slow earth flows

17. LATERAL SPREADS AND FLOWS
Mostly liquefied mass (soil flow)
Blocks moving on subsurface liquefied zone (lateral spread)
Gentle slopes to level ground
High velocity
Move long distances (soil flows)
High casualty potential for soil flows
High historical damage costs for lateral spreads
Soil lateral spreads
Rapid soil flows
Subaqueous landslides

18. EARTHQUAKE MAGNITUDE VS. AREA AFFECTED BY LANDSLIDES

19. Disrupted Coherent Disrupted Coherent
MAGNITUDE VS. MAXIMUM EPICENTRAL DISTANCE
MINIMUM MODIFIED MERCALLI INTENSITY
Disrupted
Coherent
Disrupted
Coherent
Lateral Spreads and Flows
Upper Bounds
Lateral Spreads and Flows

20. THE LAST 25 YEARS: 1984-Present
Improvements in remote sensing
Use of Geographic Information Systems (GIS)

21. COMPLETE LANDSLIDE INVENTORIES, 1984-PRESENT

22. 1999 CHI-CHI, TAIWAN M 7.5
Most extensive landslide studies of any earthquake ever. Many, many studies and publications covering virtually all aspects of earthquake-induced landslides.
26,000 landslides mapped from aerial photographs (Wang et al., 2002); 9272 large landslides mapped from SPOT satellite images (first use) (Liao and Lee, 2000)
> 400 free-field strong motion recordings for correlation

23. LOMA PRIETA, CALIF. EARTHQUAKE M 6. 9 Landslide Concentration vs
LOMA PRIETA, CALIF. EARTHQUAKE M 6.9 Landslide Concentration vs. Slope and Material
(From Keefer, 2000)

24. LANDSLIDE CONCENTRATION VS. DISTANCE LOMA PRIETA, CALIF. EARTHQUAKE M 6.9
(From Keefer, 2000)

25. WORLDWIDE AND REGIONAL ANALYSES SINCE 1984
Worldwide (Rodríguez et al., 1999)
Landslide Damage (Bird and Bommer, 2004)
New Zealand (Hancox et al., 1997, 2002)
Italy (Prestininzi and Romeo, 2000)
Greece (Papadopoulos and Plessa, 2000)
Central America (Bommer and Rodríguez, 2002)

26. MAGNITUDE VS. AREA AFFECTED BY LANDSLIDES (Rodríguez et al., 1999)
Keefer, 1984

27. MINIMUM INTENSITIES FOR LANDSLIDES (Rodríguez et al., 1999)
Number of earthquakes

28. Northridge, California area (from Jibson et al., 2000).
REGIONAL-SCALE EVALUATION OF SEISMIC SLOPE STABILITY USING NEWMARK METHOD
Northridge, California area (from Jibson et al., 2000).
(Area is ~2.5x4 km)

29. TIME-PROBABILISTIC EVALUATION OF EARTHQUAKE-INDUCED LANDSLIDE HAZARD (Del Gaudio et al., 2000, 2003; Romeo, 2000)
(From Del Gaudio et al., 2003)
(From Romeo, 2000)

30. HAZARD ANALYSES USING STATISTICAL METHODS
Probability of earthquake-induced slope failure, upper Serchio River basin, Italy (Luzi et al., 2000)
Probability of earthquake-induced slope failure, Kouhsing Quadrangle, Taiwan (Lee et al., 2008)

31. HAZARD ANALYSIS USING COMPUTING WITH WORDS, OR “FUZZY LOGIC”
“Comprehensive Areal Model of Earthquake-Induced Landslides (CAMEL)
( Miles, 2004; Miles and Keefer 2007, 2008, 2009)
“Possibility Module”
“Intensifier Module”, Probability Determination

32. SAMPLE RESULTS FROM CAMEL APPLIED TO 1906 SAN FRANCISCO EARTHQUAKE, SANTA ROSA AREA, CALIFORNIA (Swank, 2007)
Rock falls
Soil slumps

33. Divide by Area to Calculate Erosion Rate
Seismic Moment vs. Total Volume of Landslide Material
Combine
with
Earthquake
History
to get
Rate of
Seismic
Moment
Release
GEOMORPHOLOGY: CALCULATING EROSION RATES FROM EARTHQUAKE-INDUCED LANDSLIDES (Keefer, 1994)
Calculate Total Volume of
Landslide Material Produced Over Time
Divide by Area to Calculate Erosion Rate

34. REGIONAL EROSION RATES FROM EARTHQUAKE-INDUCED LANDSLIDES

35. MODELING EROSION RATE FROM EARTHQUAKE-INDUCED LANDSLIDES (Malamud et al., 2004)

36. GEOMORPHOLOGY AND LANDSCAPE EVOLUTION: TWO TYPES OF STUDIES
Post-earthquake sediment
transport, Chi-Chi, Taiwan (Dadson et al., 2004)
Distribution of landslides related to topography and earthquake source:
Chi-Chi, Finisterre, and Northridge earthquakes (Meurnier et al., 2007)

37. PALEOSEISMOLOGY-ROTATIONAL SLUMP IN PYRAMID AT CARAL, PERU
From Sandweiss et al., 2009
For comprehensive discussion of using landslides as paleoseismic indicators, see Jibson, 1996
Evidence of contemporaneous repairs, so dates to ~2000 BCE
Dry (hyperarid) conditions, and no likely cause other than earthquake
Distribution and severity of this and other earthquake effects suggests M>7.8

38. SO... IN THE LAST 25 YEARS MUCH HAS BEEN ACCOMPLISHED TO INCREASE OUR UNDERSTANDING OF EARTHQUAKE-INDUCED LANDSLIDES
This includes advances in the areas I’ve talked about, as well as advances in areas(such as site-specific seismic slope-stability analysis) that I haven’t talked about.
Much of the current state of knowledge will be presented in the papers at this Conference during the next 4 days
HOWEVER...

39. WE CURRENTLY HAVE ONLY 18 (+/-) “COMPLETE” INVENTORIES WORLDWIDE
13 of those inventories were produced in the last 25 years
During this time, according to the US Geological Survey, there were:
450 earthquakes of M > 7,
3800 earthquakes of M > 6, and
36,775 earthquakes of M > 5
Comprehensive inventories are important because they provide the basic data for all other studies,
Only with complete inventories can we differentiate between types of slopes that failed and types of slopes that didn’t
Large number of variables involved in earthquake-induced slope failure means that a large dataset is necessary for statistically-significant analysis

40. THE GREATEST RESEARCH NEED I SEE IS FOR COMPLETE INVENTORYING AND COMPREHENSIVE STUDY OF LANDSLIDES GENERATED BY MANY MORE EARTHQUAKES IN THE FUTURE.