Landslides in Eastern Honshu Induced by the 2011 Off the Pacific

Transcription

Landslides in Eastern Honshu Induced by the 2011 Off the Pacific
Bulletin of the Seismological Society of America, Vol. 103, No. 2B, pp. 1503–1521, May 2013, doi: 10.1785/0120120128
Ⓔ
Landslides in Eastern Honshu Induced by the 2011 Off the Pacific Coast
of Tohoku Earthquake
by Joseph Wartman, Lisa Dunham, Binod Tiwari, and Daniel Pradel
Abstract
In this article we discuss the character and spatial pattern of coseismic
landslides from the eastern Honshu region of Japan, which was strongly shaken in the
2011 Tohoku earthquake. We developed a detailed geospatial database of 3477 landslides based on postearthquake field surveys and examination of high-resolution satellite imagery across a 28;380 km2 landslide study area. Analysis of the database
shows that a substantial majority (80%) of landslides occurred in Quaternary soil
and Neogene rock units. Despite their abundance in the study area, relatively few
landslides occurred in pre-Neogene rocks (i.e., older than 23 Ma). Further examination of the data showed that the most common types of landslides were (1) disrupted
landslides in Neogene sedimentary rocks and (2) lateral spreading in Quaternary sediments. However, we found that coseismic landslide erosion (i.e., debris mobilization)
was almost fully dominated by lateral spreading within Quaternary sediments. When
comparing the landslide inventory with ground motions recorded by dense regional
seismic arrays, we found no statistically significant correlation between landslide
intensity and ground motion within the study area.
Online Material: Derivation of empirical parameters used in the area to volume
transformations for lateral spreads.
Introduction
The Mw 9.0 2011 Off the Pacific Coast of Tohoku Earthquake struck Japan on 11 March 2011 at 05:46 UTC (14:46
local time; U.S. Geological Survey [USGS], 2011;
see Data and Resources). The earthquake’s rupture zone
was situated northeast of Tokyo within a highly active subduction trench that has generated other large earthquakes in
historical times (Koketsu et al., 2011). The event triggered a
widely known tsunami that killed many thousands of people
and resulted in tens of billions of dollars in damage. The
earthquake also had other important coseismic effects including landslides, which additionally caused both human and
capital losses. Whereas the intensity of landsliding was low
compared with other recent seismic events in Japan (e.g.,
2008 Iwate–Miyagi earthquake; Miyagi et al., 2011), the
Tohoku earthquake triggered landslides that significantly
impacted the highly developed mesoseismal region. At least
14 fatalities have been directly attributed to coseismic landslides; additionally, landslides mobilized large amounts of
debris that damaged buildings, temporarily obstructed transportation networks, and impeded surface-water flow within
fluvial networks.
In this article, we discuss the character and number of
coseismic landslides, and examine spatial patterns of landsliding across the region most strongly shaken by the earth-
quake. Additionally, we provide a high-quality geospatial
database of 3477 individually mapped and classified landslides from a 28;380 km2 area of northeast Japan (see Ⓔ
available as an electronic supplement to this article). As one
of the largest magnitude, best recorded (10,000+ groundmotion stations; Midorikawa et al., 2012), and most carefully
documented seismic events in history, the Tohoku Earthquake is a benchmark event for the research community. The
geospatial database presented in this article is, to date, the
largest and most carefully documented record of landslides
initiated by a great subduction earthquake; as such, it directly
addresses a recognized research priority (e.g., Wasowski
et al., 2011). Our work focuses on landslides in both natural
and modified terrain; slope instability of major earth structures such as dams has been presented elsewhere (Pradel
et al., 2012a) and will not be discussed here.
Owing to the large magnitude of the earthquake, landslides were distributed across a large region for which the
documented limits encompass approximately 65;000 km2
(Miyagi et al., 2011). However, we observed that a substantial majority of landslides were contained within the region
of northeast Japan that coincided with the fault-rupture zone.
Not coincidently, this region corresponds to the part of Japan
that experienced the highest levels of ground shaking (Fig. 1).
1503
1504
J. Wartman, L. Dunham, B. Tiwari, and D. Pradel
per km2 ) observed during the field reconnaissance. We additionally considered the distribution of ground shaking and
also reviewed low-resolution satellite imagery made available immediately after the earthquake. Later, we made a
more precise delineation of the formal study area based on
the availability of high-resolution, clear-visibility satellite
imagery captured soon after the earthquake. The resulting
study area encompasses an approximately 70-km-wide zone
that stretches from 40.0° N to 35.8° N along the Pacific coast
of the Japanese island of Honshu (Fig. 1). The study area
crosses the Prefectures (i.e., subnational jurisdictions) of
Iwate, Miyagi, Fukushima, Tochigi, and Ibaraki. Within the
study area are two major population centers, Sendai and Mito,
as well many moderate-density coastal communities such as
Iwaki, Hitachi, and Soma.
Geology and Landforms
Figure 1. Map showing seismological features of the Tohoku
earthquake (after Koketsu et al., 2011) and the location of the
eastern Honshu landslide study area. The color version of this figure
is available only in the electronic edition.
This observation helped define our study area, which spans
a range of landforms and geomorphic settings (Prima et al.,
2006). We collected landslide field data during a nine-day,
multiteam postearthquake reconnaissance and later expanded
our database by visually examining high-resolution satellite
imagery of the region. Although time intensive, the visual
examination of imagery allowed us to manually identify,
confirm, measure, and classify individual landslides, thus
producing a detailed, high-quality database. Our analysis
of the database shows that a majority of the landslides
occurred in the youngest (Neogene) geologic units of the
region. Moreover, we find that coseismic landslide erosion
(i.e., debris mobilization) was controlled by lateral spreading
within Quaternary sediments. We also compared the landslide database with ground-motion recordings, but found no
correlation between landslide intensity and ground shaking
within the study area.
Eastern Honshu Landslide Study Area
Setting
We initially defined the general landslide study region
based on relative landslide concentrations (i.e., landslides
Honshu is located within a highly active subduction region characterized by active volcanism, seismicity, orogeny,
and mass wasting. East of the island lies the Japan trench,
where the Pacific plate subducts below the North American
plate at the relatively high rate of 10 cm=year (Hasegawa
et al., 2000). The crust of Honshu was initially generated by
Paleozoic to Mesozoic accretionary processes, and later
by voluminous intrusion of granitic rocks during the late
Cretaceous. The subsequent rifting of Honshu from the
Eurasian continent from 25 to 15 Ma was accompanied by
emplacement of volcanic materials over portions of the
island (Sato et al., 2002). Today, Honshu includes several
prominent north–south oriented geologic features that are
common in island-arc settings (Yoshii, 1979). These include:
(1) an active volcanic front located across the middle/eastern
portion of the island, (2) fore- and backarc zones along the
Pacific Ocean and the Sea of Japan, respectively, and (3) large
accretionary complexes and zones of regional metamorphic
rocks (Hashimoto, 1991). Additionally, recent (i.e., Quaternary) erosion has deposited unconsolidated materials
along Honshu’s coastal margin. A consequence of this complex geologic history is that a wide range of landforms is
found across the forearc (eastern) portion of Honshu (Prima
et al., 2006). As shown in Figure 2 (and summarized in
Table 1), major landforms include the Kitakami Mountains,
Sendai plain, the Ou Range, the Abukuma Mountains, and
Kanto plain. Additionally, a narrow (∼10 km) coastal margin
is located along the eastern edge of the island. For comparison, Figure 2 also shows a simplified geologic map and the
distribution of ground shaking in the study area.
The study area’s most rugged terrain lies in the Kitakami
Mountains, which are divided into distinct north and south
segments by the Hayachine eastern boundary fault (Fig. 2b).
The northern segment of the Kitakami Mountains largely
comprises Mesozoic sedimentary rocks that are part of a
large accretionary complex (Mikoshiba et al., 2004). The
southern segment consists of shallow-marine Silurian to Cretaceous sedimentary rocks, along with Paleozoic ultramafic
Landslides in Eastern Honshu Induced by the 2011 Off the Pacific Coast of Tohoku Earthquake
1505
Figure 2. (a) Major landforms of the eastern Honshu landslide study area (see Table 1); (b) simplified geologic map (based on reclassification of units defined by the Geological Survey of Japan AIST, 2009); (c) distribution of peak ground acceleration (PGA)-based Tohoku
earthquake ground-motion recordings (USGS, 2011; see Data and Resources). The color version of this figure is available only in the electronic edition.
rocks, and a smaller amount of high-pressure/low-temperaturetype metamorphic rocks (Ujiie-Mikoshiba et al., 2006). Additionally, the Kitakami Mountains have been intruded by
several large Cretaceous plutons, which occupy about a quarter of the area (Mikoshiba et al., 2004). Landforms in the
Kitakami Mountains consist of low mountains and smooth
to rough hills based on the terrain classification scheme of
Dragut and Eisank (2012). The Kitakami Mountains are considered to have a moderate degree of relative relief (∼300 to
∼500 m; Yonekura et al., 2001).
The Ou Range lies west of the Kitakami Mountains across
a narrow river valley. The Ou Range defines the volcanic front
of Honshu (Ujiie-Mikoshiba et al., 2006): a 500-km ridge with
many Quaternary stratovolcanoes. Since the late Miocene, voluminous magmatic intrusions have been uplifting the Ou
Range along its adjoining Uwandaira (east) and Senya (west)
reverse faults (Sato et al., 2002). In consequence, the base
(ridge) elevation of the volcanoes is quite high (∼750 to
∼1000 m) and many volcanic peaks exceed 2000 m. Within
the study area, landforms of the Ou Range classify as low
mountains near the volcanoes and as smooth to rough hills
along the central ridge (Dragut and Eisank, 2012).
In the southern portion of the study area lies the
Abukuma Mountains, an uplifted peneplain. The bedrock of
the Abukuma Mountains primarily consists of Cretaceous
plutonic rocks (granite) and also includes large regions of
Cretaceous high-pressure/low-pressure regional metamorphic rocks (gneiss; Miyashiro, 1958; Faure et al., 1986).
Landforms across the granitic portions of the Abukuma
Mountains classify as smooth low hills, but display more
relief as high hills in the metamorphic rock area (Dragut and
Eisank, 2012). The mountains have relatively low relief
(100–300 m; Yonekura et al., 2001).
The study area also includes the Sendai and Kanto
plains, which comprise unconsolidated Quaternary sediments. The Sendai plain is a low-lying (elevation < 10 m)
region that has experienced rapid progradation over the past
5000 years (Matsumoto, 1985). It features a series of sandy
beach ridges parallel to the coastline and back marshes
containing fine-grained organic sediments (Sugawara et al.,
2012). The Abukuma, Natori, and Nanakita rivers provide a majority of the sediment to the Sendai plain
(Tamura and Masuda, 2005). The southern portion of the
study area contains the Kanto plain, which consists of late
1506
J. Wartman, L. Dunham, B. Tiwari, and D. Pradel
Table 1
Geophysical Characteristics of Major Landforms in the Eastern Honshu Landslide Study Area
Landform
Region
Area (Percentage
of Total
Study Area)
Mean Elevation
(±1 Standard
Deviation)
Landform
Classification*
Relative
Relief
Scale†
Kitakami
mountains
km2
8425
(31%)
388 m
(92–682 m)
Low mountains,
and smooth to
rough hills
Moderate
(300–500 m)
Ou Range
2610 km2
(9%)
406 m
(79–733 m)
Low mountains,
and smooth to
rough hills
Slightly high
(> 500 m) to
medium
(300–500 m)
Abukuma
mountains
5426 km2
(19%)
222 m
(10–434 m)
High hills, and
smooth low hills
Low
(100–300 m)
11;690 km2
(41%)
25 m
(2–48 m)
Flat plains
Unclassified
(i.e., ∼flat)
Sendai and
Kanto plains
Geological Characteristics
Northern portion of region
includes pre-Neogene
sedimentary units
(accretionary complex),
whereas the south
contains pre-Neogene
sedimentary units
(shallow marine rocks),
with ultramafic rocks,
and small area of
metamorphic rocks.
The region also
contains plutons.
Typical rocks include
sandstone, mudstone,
conglomerate, schist,
and granite and
granodiorite.
Volcanic front, including
Quaternary
stratovocanoes; active
Mt. Zao volcano located
within study area.
Typical rocks include tuff,
volcanic breccia,
sandstone, mudstone,
and conglomerate.
Pre-Neogene plutonic
rocks, and large regions
of pre-Neogene
metamorphic rocks.
Typical rocks include
granite and gneiss.
Unconsolidated
Quaternary sediments.
Peak Ground
Acceleration
(PGA) in the Tohoku
Earthquake (Mean)
(±1 Standard Deviation)
0:50g (0:34–0:67g)
0:80g (0:31–1:29g)
0:72g (0:33–1:11g)
0:69g (0:28–1:10g)
*Based on objective terrain classification scheme of Dragut and Eisank (2012).
†
Based on mountain range relative classification scheme of Yonekura et al. (2001).
Pleistocene upland terraces (elevation ∼25 m) and Holocene
alluvial lowlands (elevation < 5 m; Ishihara et al., 2012).
The Kanto plain has been strongly influenced by the Tone
River, which has the largest catchment area in Japan (Okada
et al., 2011). The Tone river region of the Kanto plain has
undergone significant anthropogenic modification over the
past 400 years, including major alterations to its fluvial
network (Pradel et al., 2012b). Landforms in both Sendai
and Kanto plains classify as flat plains (Dragut and Eisank,
2012).
Between the Sendai and Kanto plains lies a narrow
coastal margin located immediately east of the Abukuma
Mountains. This small region is comprised of Neogene
sedimentary rocks that have been uplifted and eroded to
a low-relief landscape. Additionally, Quaternary alluvial and
fluvial deposits are found near the shoreline.
Climate and Antecedent Precipitation
Although climate varies in Japan as a function of both
latitude and elevation, conditions are relatively consistent
across eastern Honshu. Summers, which span July through
September, are typically marked by warm, wet weather
resulting from tropical airflows from the Pacific Ocean (EastAsian Monsoon). These moisture-laden airflows drop significant amounts of rain when they make landfall (Dolan
and Worden, 1992). Throughout winter, cold air fronts traveling from the west bring freezing temperatures and snowfall
to the higher elevation portions of the region. At the centrally
located Sendai weather observatory (elevation 40 m), the
30 year mean annual precipitation is 1254 mm and the mean
annual temperature is 12.4° C (mean monthly temperature
ranges from a low of 1.6° C to a high of 24.2° C; Japan
Meteorological Agency, 2012a; see Data and Resources).
Landslides in Eastern Honshu Induced by the 2011 Off the Pacific Coast of Tohoku Earthquake
1507
Figure 3. Comparison of monthly precipitation for the year prior to the Tohoku earthquake with 30 year mean monthly precipitation for
observatories located within the Kitakami Mountains (Iyako station), Sendai plain (Sendai station), and the Ou Range (Yamagata station;
Japan Meteorological Agency 2012b; see Data and Resources). Data are presented chronologically for the 12-month period concluding in
March 2011 (i.e., the month of the earthquake).
Antecedent precipitation has influenced the style,
number, and distribution of landslides in previous earthquakes in Japan (e.g., Uzuoka et al., 2005). We examined
local meteorological observatory records to see the degree to
which antecedent precipitation affected coseismic landsliding. Figure 3 compares monthly precipitation for the year
prior to the Tohoku Earthquake with 30 year mean monthly
precipitation for observatories located within the Kitakami
Mountains, Sendai plain, and the Ou Range (Japan Meteorological Agency, 2012b; see Data and Resources). The data is
presented chronologically for the 12 month period ending in
March 2011 (i.e., the month of the earthquake). The 30 year
mean monthly data exhibits seasonal variation and shows
that March falls within the historically dry part of the year.
In the year preceding the earthquake, precipitation was
highly variable and December 2010, in particular, is noted as
having been an atypically wet month. Nevertheless, by
March 2011, three-month cumulative precipitation, which
commonly influences landsliding, was close to or below
30 year mean values. Together, these data indicate the earthquake occurred during a dry season that was typical for the
region. This concurs with our field observations that, aside
from lateral spreads, landslides typically occurred in unsaturated materials, and additionally, that flow-type failures in
saturated masses were virtually nonexistent.
Earthquake Ground Motion
The earthquake ruptured a low-angle subduction interface within the Japan trench. Horizontal surface displacements captured by the GEONET network indicate slip
occurred along two regions located roughly concentric to and
south of the epicenter (Fig. 1). A unified source model based
on joint inversion of teleseismic, strong motion, and geodetic
data indicates that fault rupture took place in three distinct
episodes: (1) an initial phase (0–50 s) with slow northeast–eastward propagation, followed 20 s later by (2) slow
westward propagation coincident with the maximum slip,
and finally, (3) rapid southward propagation from 90 to
100 s (Koketsu et al., 2011). This complex rupture pattern
resulted in variation in the recorded waveforms along the
length of the slip zone. In the northern portions of the study
area (i.e., ∼38° N and further north), two distinct phases of
strong shaking were recorded, whereas in the south a single
long-duration phase of motion was measured (Furumura
1508
J. Wartman, L. Dunham, B. Tiwari, and D. Pradel
et al., 2011; Midorikawa et al., 2012). Despite the variation
in the waveforms, the duration of strong shaking (Trifunac
and Brady, 1975) was generally similar across the study area,
ranging from 80 to 110 s across all but the Kanto plain,
which experienced shorter durations of shaking (∼50 to
∼80 s; Midorikawa et al., 2012).
As shown in Figure 2c, strong ground shaking was felt
throughout the study area. We considered shaking intensity
in terms of peak horizontal ground acceleration (PGA) as this
ground-motion parameter is directly related to coseismic
landslide initiation (Terzaghi, 1950). It is noted that the distribution of PGA within the study region also resembles that
of other common ground-shaking-intensity measures such
as peak ground velocity (PGV) and Japan Meteorological
Agency (JMA) seismic intensity (Goto and Morikawa, 2012).
Although the westward attenuation of ground motions is
apparent at the scale of northeast Japan (Fig. 1), the decay
of ground shaking was not significant across the ∼70-kmwide study area (Fig. 2c). Moreover, as the study area closely
parallels the full length of the rupture zone, there was little
north–south attenuation of ground motion. Nevertheless,
there exists significant variation in ground motion due to site
effects and other factors (Furumura et al., 2011). Remarkably
high PGA values (> 1g ) were recorded at several locations
within the Sendai and Kanto plains, where soft Quaternary
sediments amplified ground motion. By contrast, PGA values
were generally lower (∼0:3g to ∼0:7g) in the mountainous
portions of the study area. An exception is noted for a localized area of higher intensity shaking in the central Abukuma
Mountains (Fig. 2c).
Landslide Database
The landslide database was developed based on field
mapping conducted during the postearthquake reconnaissance, and additionally from examination of satellite imagery.
A multiteam landslide field reconnaissance of the mesosiesmal region took place from 22 to 30 April 2011 and included
ground surveys through portions of the Prefectures of Miyagi,
Fukushima, Tochigi, Gumma, Ibaraki, and Chiba. Some of
these prefecture areas were not included within our delineated
study area due to their low landslide concentrations. A boatbased survey of islands in Matsushima Bay was also undertaken to study landslides in this coastal region. During the
field reconnaissance we observed and/or mapped approximately 250 landslides that were subsequently used to verify
the analyses of satellite imagery.
High-resolution (typically 0.5–2.5 m) postearthquake
satellite imagery used in our research was provided by
Google Earth as part of their crisis response to the earthquake
and tsunami (Google.co.jp., 2011; see Data and Resources).
A majority of the imagery was captured between 11 March
and 5 April 2011 by satellites operated by GeoEye and DigitalGlobe, among other providers (Google Maps and Earth
Team, 2011; see Data and Resources). The study area was
generally well vegetated, which often allowed landslides to
be first identified based on changes in the appearance of
vegetation or other land surface features. Both large (i.e.,
long and/or wide) ground cracks and en echelon shear zones
with smaller cracks were visible in the imagery and aided
the identification of landslides, especially lateral spreads.
In some regions, identification and characterization of landslides was further supported by the availability of multiple
postevent imagery scenes and, additionally, by postevent terrestrial imagery captured as part of Google Street View’s
event documentation project (Google, 2011; see Data and
Resources). Landslide features in the low-lying coastal areas
severely impacted by the tsunami were not considered and
are not represented in the database.
We verified the seismogenic origin of landslides by
comparing the postearthquake imagery with scenes from
Google Earth’s historical archives. Our colleagues in Japan
reported to us that, with one notable exception, aftershocks
did not significantly contribute to landsliding and thus we
believe that nearly all of the landslides in the database were
triggered by the mainshock (the exception was a shallowdepth aftershock on 11 April 2011 that triggered a localized
cluster of additional landslides at a known location; see Pradel et al., 2011). Once identified, landslides were assigned a
reference number, delineated, measured, and classified based
on the landslide categorizations defined by Keefer (1999), as
summarized in Table 2. Delineation of the landslides was
aided by the terrain feature of Google Earth, which allowed
visualization of the landscape to be enhanced by draping
imagery over a Shuttle Radar Topography Mission (SRTM)derived 30 m digital elevation model (DEM).
Because of their subtle morphological features and often
small displacements, it is difficult to identify lateral spreads
in lower resolution satellite and/or aerial imagery from past
earthquakes. Additionally, recent inventories sometimes
censor data from flat terrain regions where lateral spreads
most commonly occur (e.g., Parker et al., 2011). As a result,
few landslide inventories include data related to lateral
spreading (Keefer, 2002). However, following with Keefer
(1999), we consider lateral spreads to be an important type
of coseismic landslide that results in mass wasting (i.e.,
downslope movement of soil or rock, resulting in permanent
ground displacement). For this reason we opted to include
this commonly neglected type of coseismic landslide in our
database.
The high-resolution imagery allowed us to resolve landslide features such as scarps and bulges at the toes of slopes
that were larger than 1–4 m. The accuracy of our landslide
measurements (both location and size) is a function of several factors, which vary across the study area including nadir
(angle of satellite view), cloud cover, image resolution, and
size of feature, among others. As such, it is not possible to
provide a single accuracy or resolution value for the entire
28;380 km2 study area; nevertheless, our comparison of satellite and field-mapping-derived data suggest that a majority
of the landslides have a location accuracy of 8 m, and a
measurement accuracy of 5 m. Imagery in the study area
Landslides in Eastern Honshu Induced by the 2011 Off the Pacific Coast of Tohoku Earthquake
1509
Table 2
Summary of Landslide Classifications (Adapted from Keefer, 1999; as Modified by Keefer, 1984)
Landslide Category
Included Landslide Subcategories
Characteristics
Disrupted
Rock falls, rock slides, rock avalanches, as
well as debris (i.e., soil) falls and slides
Coherent
Rock slumps, rock block slides, earth and
debris slumps, and slow earth slides
Lateral Spreads
Earth and debris lateral spreads and flows
Landslide movements typically rapid. Movement by free fall, sliding, and/or
rolling of debris. Debris masses typically dry and brittle. Deposition
occurs along or at the base of slope. Often involves long run-out
displacement. High to very high levels of internal disruption.
Landslide movements typically slow to moderate. Movement by slumping,
translation, and/or rotational sliding of mass. Debris masses often are
partially to fully saturated. Displacements typically moderate (< 2 m), but
can be large in many instances. Slight to moderate levels of internal
disruption.
Landslide movements typically rapid. Movement by translational movement
of a viscous mass, or of a stiffened crust of dry soil moving over a viscous
surface. Movement can also involve flow. Debris masses often are
partially (in case of crust) to fully saturated. Displacements typically
moderate (< 2 m), but can be large, especially in the case of flow failures.
Moderate to very high levels of internal disruption.
was not notably affected by hillslope shadows and as a result
the measured landslide locations are largely free of systematic directional bias. Overall, we are confident that our inventory represents a comprehensive cataloging of disrupted and
coherent landslides having features larger than 1–4 m. The
morphological features of lateral spreads were subtler, and
additionally, permanent deformation of these landslides was
often less than the 1–4 m identification threshold. Therefore,
it is likely that our mapping does not contain a complete
inventory of lateral spreads that are small or that have no
obvious surface manifestation. It is impossible to accurately
quantify the degree of under sampling of lateral spreads;
however, based on comparisons of field and imagery-derived
data, we believe it to be small. It is important to note that
lateral spread displacements tend to diminish in the direction
of movement (i.e., undergo ductile displacement response,
with deformations progressively approaching zero near the
toe). When delineating lateral spreads, we encircled only
those areas where deformation was visible (rather than inferred). Therefore, our delineations represent a conservative
(i.e., lower bound) size interpretation of lateral spreads.
Distribution of Landslides and Associated Erosion
Figure 4 presents the point locations of the 3477 landslides in the study area. Landslides appear to be generally
well distributed across the study area, although high concentrations are found at two locations: (1) an approximately
2200 km2 rectangular area located within the Sendai plain,
extending west to the Ou Range, and (2) an approximately
120-km-long segment of the narrow coastal margin immediately east of the Abukuma mountains, extending westward
near the southern portion of this area. Both the Sendai plain
and the area southeast of the Abukuma Mountains roughly
coincide with zones of moderately high shaking intensity
(Fig. 2c); however, this is not the case for the Ou Range area,
for which shaking intensity was about average for the overall
study area. Figure 4 also shows that the Kitakami and
Abukuma Mountains generally have few landslides. This is
surprising as both of these mountain regions are characterized by rugged terrain and also experienced relatively high
levels of ground shaking.
Figure 5 presents the data in a different context by additionally considering size (i.e., estimated volume) of the landslides, which can be related to equivalent local levels of
erosion (i.e., mobilization of debris and sediment), or landscape lowering (LL). Because it captures both the number
and size of landslides, the LL metric more meaningfully represents the intensity of coseismic landslide-erosion processes
within a localized area. Moreover, LL has important implications for landscape evolution (e.g., Hovius et al., 1997),
land productivity (e.g., Blaschke et al., 2000), and ecological
systems (e.g., Montgomery et al., 2000). Additionally, because the size of a landslide is typically directly related to
its effects on the built environment, LL is relevant to infrastructure system damage, emergency management, and postdisaster recovery.
We estimated LL by first relating the measured surface
area of each landslide to equivalent landslide volume using
the empirical relationship (Guzzetti et al., 2009; and references therein):
V αAγ ;
(1)
where V = landslide volume (m3 ); A = landslide surface area
(m2 ), and α and γ are empirical constants.
Based on a comprehensive review of over 150 published
landslide area–volume relationships, Larsen et al. (2010)
found that empirical constants α and γ vary regionally for
landslides in soil and in weathered and/or closely jointed
rock. As these types of geologic materials dominate the
inventory of disrupted and coherent landslides, we adopted
region-specific empirical constant values (α 0:25, γ 1:36) based on surveys of 236 soil and bedrock landslides
1510
J. Wartman, L. Dunham, B. Tiwari, and D. Pradel
Figure 5.
Estimated landslide erosion (i.e., debris mobilization)
for all 3477 landslides presented as equivalent local levels of landscape lowering (LL). The color version of this figure is available
only in the electronic edition.
Figure 4. Point locations of the 3477 landslides in the eastern
Honshu landslide study area. The color version of this figure is
available only in the electronic edition.
in Honshu (Yamagishi and Iwahashi, 2007). This γ value is
less than 1.5, the threshold corresponding to the self-similar
behavior (i.e., the situation in which landslides have the same
3D scaling of geometry, regardless of size). Instead, it indicates self-affine behavior whereby larger landslides are shallow in comparison with smaller landslides (Klar et al., 2011),
which is consistent with surveys from our field reconnaissance. Mechanistically, lateral spreads are largely controlled
by cyclically induced pore-pressure generation. This contrasts
with disrupted and coherent landslides, which are more directly governed by shaking-induced inertial forces. For this
reason, area–volume relationships between these fundamentally different types of landslides may not be identical.
Accordingly, we developed empirical parameters specific
to lateral spreads based on detailed examination of 13 representative sites within the study area where the depth of lateral
spreading could be reliably inferred based on subsurface investigation data (Cox et al., 2012; Kawabe et al., 2012). Regression of this data yielded empirical parameter values of
α 0:74 and γ 1:19 (R2 0:95) for landslide areas spanning over three orders of magnitude (100–50;000 m2 ; for a
complete summary of the regression data, see Ⓔ Table S1
and Figure S1 in the electronic supplement). We are aware
of no other published area–volume relationship for lateral
spreads, but note that our value of the controlling parameter
γ is reasonably close to γ 1:38 estimated by Innes (1983)
for debris flows, the initiation and subsequent deformation
response for which are likewise governed by raised pore
pressures. The difference between γ values for lateral spreads
(γ 1:19) and disrupted/coherent landslides (γ 1:36)
indicate that lateral spreads are shallower than disrupted
and coherent landslides, which otherwise have an identical
surface area.
We applied equation (1) to landslides in the database to
estimate volume, and then summed these volumes across a
Landslides in Eastern Honshu Induced by the 2011 Off the Pacific Coast of Tohoku Earthquake
grid of 3-km-square quadrats. Volumes within individual
quadrats were subsequently normalized by area (9 km2 ) to
compute localized equivalent values of LL. As shown in
Figure 5, the relative magnitude of LL values generally
parallels the landslide concentrations visually inferred from
Figure 4. However, exceptions are noted for some locations
such as the near-coast areas of Sendai and Kanto plains,
which appear to produce disproportionately high levels of
localized landslide erosion.
Disaggregation by Landslide Classification
and Geologic Unit
To better understand the spatial trends shown in
Figures 4 and 5, we subdivided the database based on landslide category (i.e., disrupted, coherent, or lateral spread).
Additionally, we broadly classified rock and soil units in
a manner that reflects their geologic history and associated
engineering characteristics. The resulting simplified geologic
map (Fig. 2b) consists of six units divided by age and
geologic origin. Our simplified geologic map was based on
reclassification of units defined on a detailed geologic map
developed by the Geological Survey of Japan AIST (2009),
additionally, Wakita et al. (2008). We make a fundamental
distinction between pre-Neogene (age > 23 Ma) and
Neogene and Quaternary units as this separates materials
of Eurasian continental origin from those formed after rifting
had commenced (e.g., Sato et al., 2002). This division also
reflects the Japanese Civil Engineering Code that considers
all pre-Neogene rocks to be well lithified, whereas Neogene
rocks are defined as being weak materials (JSCE, 2001).
Additionally, this follows the work of Okamoto et al. (1981),
whose compilation of data from across Japan reveals an
order-of-magnitude difference in the unconfined compressive strength between rocks of the Oligocene (i.e., preNeogene epoch) and Miocene (Neogene) periods. We further
subdivided the pre-Neogene rocks based on geologic group
(igneous, metamorphic, or sedimentary). We subdivided
more recent materials (age < 23 Ma) based both on period
(Quaternary, which includes unconsolidated, soil-like materials, versus consolidated Neogene rocks) and geologic
group (i.e., igneous and sedimentary rocks; younger metamorphic rocks are nonexistent).
Figure 6 shows disaggregations of landslide count and
erosion data based on landslide classification and simplified
geologic unit. Figure 6a summarizes overall count data,
which reflect both the measured number of landslides and the
extent of geologic units in the study area. The leftmost row
of Figure 6a shows that most landslides occurred within
Quaternary sediments and Neogene sedimentary rocks.
Lesser but nevertheless significant numbers additionally occurred in Neogene and pre-Neogene igneous units. Overall,
the inventory is almost completely dominated by Neogene
rock and Quaternary units, which account for 80% of
the total number of landslides. Further disaggregation of
the data (rows 2–4 of Fig. 6a) indicates that disrupted land-
1511
slides in Neogene sedimentary rocks and lateral spreading
in Quaternary sediments were in almost equal measure
the two principal types of landslides. Secondary modes of
landsliding include disrupted landslides in Quaternary
sediments, disrupted landslides in both pre-Neogene and
Neogene igneous units, and lateral spreads in Neogene sedimentary units.
Figure 6b shows the landslide counts normalized by the
area of the respective geologic units. These data indicated
that when considered relative to their distribution, Neogene
sedimentary units (< 10% of study area) largely dominate the
landslide counts. In this same context, the more widely distributed Quaternary sediments unit (32% of area) has a lower
relative count. Further disaggregation of the data (rows 2–4)
shows that disrupted failures in Neogene sedimentary units
dominate the normalized landslide count data.
Figure 6c considers the data in the context of landslide
erosion and again controls for proportion by normalizing
estimated landslide-erosion volume by the total area of each
geologic unit within the study area. The leftmost row of
Figure 6c shows that an overwhelming majority of landslides
erosion occurred within Quaternary sediments and Neogene
sedimentary rocks. The further disaggregated data (rows
2–4) shows landslide erosion to be dominated by lateral
spreading within Quaternary sediments. Disrupted landsliding in Neogene sedimentary units also plays an important,
but secondary, role in landslide erosion. (Although not
shown here, it is noted that the same trends are also found
the non-normalized landslide-erosion data.) Differences in
dominant modes of landsliding between Figure 6b,c reflect
the number and size characteristics of the landslides. As discussed below, disrupted landslides in Neogene sedimentary
units were common but usually small, whereas lateral
spreads in Quaternary sediments were fewer but larger.
Landslide Characteristics
The data presented in Figure 6 reflect average trends
across the study area. To better understand the spatial pattern
of landsliding, we developed maps depicting local variations
in the data. Figure 7 shows localized measures of landside
concentration (LC = landslides/km2 ) and landslide erosion
(LL) for each landslide category. We computed local measures of LL across a grid of 3-km-square quadrats by dividing
erosion volume by area (9 km2 ) and present the results as
equivalent landscape lowering in standard units (mm). As
with the data of Figure 6, localized dissimilarities in relative
LC and LL intensities reflect differences in number and size
characteristics within a respective landslide category.
Disrupted Landslides
High concentrations of disrupted landslides were found
in three areas (Fig. 7): (1) across a 45-km-long east–west
trending zone of Neogene sedimentary rock located east of
the Ou Range, continuing into Matsushima Bay; (2) within
1512
J. Wartman, L. Dunham, B. Tiwari, and D. Pradel
Figure 6. Disaggregations of landslide frequency and erosion data based on classification and simplified geologic unit. (a) Overall
landslide-count data, (b) landslide-count data normalized by the area of the respective geologic units, (c) landslide erosion normalized
by total area of each geologic unit. Geologic units correspond to those shown in Figure 2c. The color version of this figure is available
only in the electronic edition.
small (5 km), highly concentrated east–west area of Neogene
sedimentary rock located in the Abukuma Mountains, and
(3) within a more dispersed ∼10 km circular region of preNeogene granitic rocks of the Abukuma Mountains. Relative
LC and LL intensities are generally similar across the study
area suggesting that disrupted landslides contributed to erosion in proportion to their total number. A notable exception
is the small high-erosion area located immediately south
of the Kitakami Mountains, where comparatively large
disrupted landslides occurred in pre-Neogene sedimentary
rocks.
The distribution of measured sizes for disrupted landslides was approximately log-normal with a median area of
314 m2 , and 25% and 75% percentile (quartile) values of 135
and 703 m2 , respectively. These areas translated (via equation 1) to estimated volumes for which distributions were
also log-normal, but heavier tailed, as suggested by the
spread in quartile range values (median volume = 625 m3 ,
25% and 75% quartiles = 198, and 1870 m3 , respectively).
We observed the geometry of disrupted landslides to vary as
a function of local topography and geologic unit, but in rock
units these were generally oblong and parallel to the slope.
A majority of these disrupted landslides appear to have
originated at or near the crests of steep slopes, suggesting
that the topographic modification of ground motion played
a role in their initiation (e.g., Meunier et al., 2008). Disrupted
Landslides in Eastern Honshu Induced by the 2011 Off the Pacific Coast of Tohoku Earthquake
Figure 7.
1513
Localized measures of landside concentration (LC = landslides/km2 ) and landslide erosion (LL) for each landslide category.
Values shown are local measures of LC and LL computed across a grid of 3-km-square quadrats, and normalized by area. Localized dissimilarities in relative LC and LL intensities reflect differences in landslide frequency and size characteristics within a respective category. The
color version of this figure is available only in the electronic edition.
1514
Figure 8. Disrupted landslide located immediately south of
the Abukuma Mountains (Latitude 36.520228° N, Longitude
140.468244° E). The color version of this figure is available only
in the electronic edition.
landslides occurred most commonly in weakly cemented,
highly weathered, and/or well-jointed (∼10- to ∼50-cm spacing) rocks. Their debris reflected the character of the parent
source material, but generally became increasingly disaggregated as it traveled along steep slopes. When they occurred
in unconsolidated (dry, gravelly, and sandy materials) or otherwise weak materials (e.g., saprolite), disrupted landslides
tended to have surface areas that were more rounded or elliptical in shape and typically extended below the root depth
of vegetation. We observed the consequences of disrupted
landslides in undeveloped terrain to include production, mobilization, and transport of debris, denudation of slopes, and
over steepening of slopes near landslide source areas. In metropolitan areas, the debris from disrupted landslides blocked
or impeded transportation corridors and in several instances
impacted structures.
Figure 8 depicts a large (∼2800 m3 ) disrupted landslide
located immediately south of the Abukuma Mountains. The
70-m-wide landslide occurred within a 9-m-high, 40° slope.
The lateral limits of the landslide appeared to be controlled
by local relief, which decreased to about 6 m near its edges.
The landslide occurred in weakly cemented, moderately
jointed (spacing ∼20–80 cm) Neogene sedimentary rock
(siltstone). The rock mass was highly weathered across the
crest of slope, but become blocky with depth. Water was
not observed within the landslide source area. The landslide
appeared to initiate within the weathered zone at the slope
crest. Landslide debris, which consisted of many 30- to
70-cm-rock blocks mixed with large gravel fragments, traveled downslope before becoming entwined in large vegetation (trees and shrubs). A minor amount of debris reached the
backs of several houses located at the bottom of the slope, but
caused only minor, nonstructural damage. The area immediately behind the landslide source area was denuded and over
steepened; unless mitigated, this will likely become a location having increased landsliding and active erosion in the
future.
J. Wartman, L. Dunham, B. Tiwari, and D. Pradel
Figure 9.
Typical example of a disrupted landslide located Matushima Bay (Latitude 38.336762° N, Longitude 141.115812° E).
The color version of this figure is available only in the electronic
edition.
Figure 9 shows another example of a disrupted landslide
located just offshore of the Kanto plain in Matushima Bay.
The landslide was approximately 50-m wide and appeared
to originate near the crest of a 65° slope. The slope consisted
of a 50-cm layer of highly jointed (spacing < 10 cm) and
poorly indurated tuff that became stiffer with depth. Landslide debris consisted of 10- to 20-cm blocky fragments
mixed with a large fraction of gravel and sand. Having occurred along the periphery of an island, the landslide produced
debris that was transported directly to Matsushima Bay.
Lateral Spreads
Lateral spreads occurred within Quaternary sediments at
multiple locations in the Sendai and Kanto plains. Lateral
spreads in the Sendai plain had generally similar LC and
LL values, whereas those in the Kanto plain show LL intensities that are high in comparison with LC intensities. This
indicates that lateral spreads were fewer but larger in the
Kanto plain, an observation supported by size statistics for
the two regions: the median area of lateral spreads in the
Kanto plain was 1465 m2 , whereas in the Sendai plain the
median area was 444 m2 . We attribute these size contrasts
to differences in morphology between the plain areas. The
Kanto plain is characterized by uniform, flat expanses of
low-lying land, whereas Sendai plain, although still lowlying and flat, exhibits a higher degree of small-scale relief
and slightly more rugged terrain. We adopted the standard
deviation of SRTM-derived elevation (σ) as a simple proxy
for terrain roughness and computed significantly higher values for areas of landsliding in the Sendai plain (σ 19:0 m)
versus the Kanto plain (σ 3:2 m). We speculate that in flat,
broad expanses of ground, such as that found near the Tone
River in Kanto plain, lateral spreads can coalesce and grow
into extraordinarily large mass movements. In contrast, local
Landslides in Eastern Honshu Induced by the 2011 Off the Pacific Coast of Tohoku Earthquake
variations in relief inhibit propagation and coalescence,
which limits the size of lateral spreads. It is also plausible
that surface roughness may also be associated with a higher
degree of subsurface heterogeneity, and thus the spatial extent of liquefaction; however, we have no data to support this
hypothesis.
Although the majority of lateral spreads occurred in
Quaternary sediments, a small number were also located in
units indicated as Neogene sedimentary rocks on geologic
maps. Lateral spreads at these locations were associated
with anthropogenic activity (e.g., reclamation using hydraulic fills, conversion of old river beds to agricultural
production, construction of artificial lakes, etc.), and thus
occurred in soil masses rather than consolidated “rock”
units. Overall, aggregate data for lateral spreads across
the entire study area was log-normally distributed with a
median area of 458 m2 and 25% and 75% quartile values
of 139 and 1437 m2 . These areas translated to estimated volumes with more heavily tailed distributions (median, and
25% and 75% quartile volumes = 1110, 267, and 4344 m3 ,
respectively).
Ground inclinations near lateral spreads were typically
low (i.e., < 5°) and the resulting movement was largely translational towards a steeper free face (typically the bank of
a waterway). Lateral spreading is related to seismically
induced pore-water pressure generation, and thus it may be
inferred that groundwater was located at or above the landslide basal surface. This agrees with our reconnaissance
observations of open water in close proximity to most laterally spread sites. We witnessed many locations where soil
liquefaction (confirmed by sand boils) occurred in ground
where free faces were absent and therefore were unaccompanied by lateral spreading (Pradel et al., 2012b). Because
these sites did not undergo permanent lateral ground deformation, we do not regard them as landslides and they have
not been included in the database. Many lateral spreads had
horizontal displacements that reached only 0.5–4.0 m.
However, in areas of greater local relief, lateral spreads occasionally appeared to have transitioned to large deformational
(> 10 m) flow-type movements.
Lateral spreads significantly affected both the natural
and built environments. In less developed areas, lateral
spreads fissured the ground, locally altered morphology,
narrowed channels, and mobilized and transferred sediment
directly to rivers, canals, lakes, and ponds. In developed
areas, lateral spreads damaged flood-protection systems
(e.g., artificial levees), interrupted road networks, severed
subsurface utilities, and tilted and deformed buildings and
structures (Pradel et al., 2012b). During a return visit to
the region 12 months after the earthquake, we observed that
earthworks construction and repairs had been undertaken to
restore function to major infrastructure systems damaged
by lateral spreads. Nevertheless, some areas remained abandoned due to lateral-spreading-induced damage to subsurface utilities.
1515
Figure 10. Typical example of lateral spreading within
Quaternary sediments of Kanto plain. The site was situated along
the Jukken-Gawa River near its confluence with the Tone River
(Latitude 35.898830° N, Longitude 140.499374° E). The color
version of this figure is available only in the electronic edition.
Figure 10 shows a typical example of lateral spreading
within Quaternary sediments of the Kanto plain. The site was
situated along the Jukken-Gawa River near its confluence
with the Tone River. Lateral spreads occurred along both
banks of the river, with the lateral spread on the right bank
being larger (800 m3 ) than that on the left (225 m3 ). The surrounding development along with the straight channelized
form of the river indicates a high degree of anthropogenic
modification in the immediate area. This is consistent
with findings of Pradel et al. (2012b), who reviewed historic
maps of the Kanto plain and concluded that liquefactionsusceptible fill materials had been placed in this area.
Although the morphological features of the lateral spread
are subtle and partially obscured by vegetative cover, the
overall ground-deformation patterns are obvious. During the
reconnaissance we measured horizontal ground displacements of 2+ m at the site and noted the consequent constriction of the river channel. Additionally, we observed tilting of
adjacent structures and disruption of nearby utility lines,
suggesting that the lateral spread extended upslope of the
immediate river channel. Kawabe et al. (2012) conducted
a subsurface investigation of the site and found 5 m of low
density silts (thought to be fill), over 3 m of slightly higher
density alluvial sands. Stiffer fine sands were encountered at
a depth of 8 m.
Figure 11 outlines a large lateral spread (area =
12;910 m2 , estimated volume = 59;600 m3 ) located near the
Port of Hitachinaka in the Kanto plain. The site is comprised
of unconsolidated materials and situated immediately west
of a Neogene sedimentary unit that prevented the landward
encroachment of the lateral spread. The ground deformation
patterns appeared to be influenced by the presence of a large,
deep-founded structure at the port. The lateral spread had
a slumping morphology, with ground deformations being
predominantly vertical along the 230-m-long back-scarp area
and gradually becoming more horizontal (and smaller)
1516
J. Wartman, L. Dunham, B. Tiwari, and D. Pradel
Figure 11.
Outline of a large lateral spread (Latitude 36.344817° N, Longitude 140.603443° E). The color version of this figure is
available only in the electronic edition.
across the length of the lateral spread. Terrestrial Light Detection And Ranging (LiDAR) surveys of the site by Kayen
et al. (2011) indicate that vertical deformations were as great
as 2 m, but averaged 1 m and varied in a normally distributed
manner across the scarp of the lateral spread.
Coherent landslides
We encountered few coherent landslides in the study
area, mostly within or near the Ou Range. Figure 6 shows
relative LC and LL intensities to be generally similar. The
size distribution for coherent landslides was approximately
log-normally distributed with a median area of 1017 m2
and 25% and 75% percentile (quartile) values of 400 and
2286 m2 , which correspond to estimated median, and quartile volumes of 3092, 869 and 9298 m3 , respectively. We
observed the geometry of coherent landslides to vary with
local topography but most were elliptically shaped and
appeared to have near-circular or circular-translational basal
surfaces. Coherent landslides usually extended from a headscarp region to the toe of a slope and typically reached depths
of 4 m or more. The displacements varied widely depending
on the apparent degree of stability of the landslide mass (as
suggested by the steepness of the terrain), but were typically
in the range of 3–6 m. In a limited number of cases shakinginduced displacements appeared to be followed by the onset
of higher velocity, dry-flow-type movement leading to deformation exceeding many tens of meters. We speculate that the
large displacement of these landslides was related to the high
porosity (and low density) of the Neogene source materials
(typically of volcanic origin). In the Sendai metropolitan
area, we observed several coherent landsides located within
hillslope fills placed for residential development. The modifications clearly served to destabilize nearby slopes and
thus increase susceptibility of coseismic landsliding, in some
cases with devastating after effects. For example, Figure 12
shows a coherent landslide that severely damaged five homes
in metropolitan Sendai. The landslide occurred at a location
where graded fill had been placed over a Neogene pyroclastic
unit, thus surcharging the adjacent 30°, 28-m-high hillslope.
The landslide had a deep circular-translational basal surface
and an estimated volume of 16;000 m3 .
The consequences of coherent landslides included
mobilization and transport of debris, denudation of slopes,
and oversteepening of slopes in landslide source areas. In
developed areas coherent landslides impacted, damaged, and
destroyed structures, blocked transportation corridors, and
caused bodily injuries and fatalities. In the absence of any
remediation efforts, these landslides are likely to enlarge and
affect ground proximal to these slopes, especially where
oversteepened head scarps are present.
Landslides in Eastern Honshu Induced by the 2011 Off the Pacific Coast of Tohoku Earthquake
1517
Figure 12. Coherent landslide at a hilltop residential development in metropolitan Sendai (Latitude 38.267447° N, Longitude
140.799143° E). The color version of this figure is available only in the electronic edition.
Relationship between Ground Motion and Landslide
Concentration
We investigated the potential relationship between
ground motion and LC by comparing disaggregated data
(Fig. 7) with local PGA values (Fig. 2c). We adopted simplified geologic unit as a simple first-order control on landslide
susceptibly and made separate comparisons on this basis.
Figure 13 presents the results as LC and PGA values for
grid points located within each respective geologic unit.
Differences in the PGA ranges between the six graphs reflect
variation in ground shaking across the geologic units within
the study area. The figure shows a high degree of scatter and
additionally, least-square regressions of the data for individual geologic units suggest little, if any, relationship between
PGA and LC. Further statistical analyses reinforced this
observation and indicated that, at least within the study area,
there is no statistically significant correlation between LC
and PGA (Pearson’s R range = −0:016 to 0.126; ρ > 0:05).
We subsequently repeated these analyses using LL in lieu of
LC and found again no correlation between ground motion
and landslide intensity.
Discussion
The pattern of landslides across the study area is
complex and fairly unexpected. For example, despite the
high ground motions and rugged terrain, the Kitakami and
Abukuma Mountains experienced very few landslides of any
type. Similarly, the very high level of landslide erosion in the
low-relief plain regions was unforeseen. Nevertheless, clear
trends emerge when the geospatial data is disaggregated
according to landslide type and geologic unit. Perhaps most
significant is the marked distinction in landslide intensity
between Neogene/Quaternary and pre-Neogene geologic
units. Neogene/Quaternary units occupy 53% of the study
area, yet account for 80% of the total number of landslides
and 90% of the overall landslide erosion. This simple geologic age distinction largely defines the terrain most susceptible to coseismic landsliding. It also explains clustering
of landslides in younger materials (Fig. 4), and conversely,
the scarcity of landslides in the mountain areas, which are
largely comprised of pre-Neogene units. Interestingly, the
data show that pre-Neogene units were almost equally resistant to coseismic landsliding regardless of geologic subgroup
(igneous, sedimentary, or metamorphic).
Further investigation of the data based on landslide
category provides additional insight to coseismic landslide
processes. Disrupted landslides were the most common
and widely distributed type of landslide in the study area,
yet these accounted for only 10% of the total sediment
production. By contrast, lateral spreads were fewer in number but were responsible for most of the landslide erosion.
Lateral spreads have an empirical γ area–volume value
that is low compared with that of other types of landslides.
1518
J. Wartman, L. Dunham, B. Tiwari, and D. Pradel
Figure 13. Comparison of landslide concentration (LC) and peak ground acceleration (PGA) values (based on USGS, 2011; see Data and
Resources) for 3-km grid points located within respective geologic units. Further analyses indicated that there is no statistically significant
correlation in the data.
Therefore, lateral spreads exhibit geometric characteristics
more self-affine (and less self-similar) than disrupted and coherent landslides. Thus, whereas lateral spreads were comparatively shallow, the broad, flat regions where these
occurred facilitated their propagation and coalescence into
large mass movements. This is reflected by their heavy
tailed, log-normal size distribution, which indicates that a
limited but nevertheless significant number of lateral spreads
were very large. As land development in the region has
favored the flat plain areas most susceptible to lateral
spreads, their effects had significant effects on the built environment. Lateral spreads have traditionally been neglected in
many landslide investigations; however, our analyses indicate
that they were the prevailing mode of coseismic landsliding in
eastern Honshu.
The lack of a statistically significant correlation between
landslide intensity and ground motion contrasts with established empirical relationships (e.g., Meunier et al., 2007) and
Landslides in Eastern Honshu Induced by the 2011 Off the Pacific Coast of Tohoku Earthquake
suggests, at least within Honshu, that geologic unit may
overshadow earthquake ground motion as the principal control on coseismic landsliding. In the Tohoku earthquake, this
may have been exacerbated by the high intensity of ground
motion, which so greatly exceeded the threshold-yield (triggering) acceleration of landslides that its influence was in
effect minimized. It is also possible that including data from
outside the study area would yield a more robust groundmotion–landslide intensity relationship. It should be recognized that our work considers ground motion only in the
context of PGA, which does not represent other potentially
important characteristics such as frequency content, duration, or the multiple phases of shaking recorded at some locations in Honshu. Clearly, the role of subduction ground
motion in coseismic landslide processes remains an important issue that warrants further investigation.
Conclusions
Coseismic landslides from the Tohoku earthquake produced, mobilized, and transported large amounts of debris,
which damaged buildings, obstructed transportation networks, modified fluvial networks, denuded slopes, and
caused human losses. We developed a detailed geospatial
database of 3477 landslides based on surveys conducted during a postevent field reconnaissance and examination of
high-resolution satellite imagery across a 28;380 km2 region
of Honshu. The Ⓔ database (see supplement) is to date the
largest and most detailed record of landslides initiated by a
great subduction earthquake.
A substantial majority (80%) of landslides occurred in
younger (i.e., Neogene) rock units and Quaternary sediments. Further examination of the data showed that the most
common types of landslides were (in equal measure): (1) disrupted landslides in Neogene sedimentary rocks and (2) lateral spreading in Quaternary sediments. Other common but
secondary modes of landsliding included disrupted landsliding in Quaternary sediments, pre-Neogene igneous rocks,
and Neogene igneous rocks. When normalized by the area
of the respective geologic units, Neogene sedimentary units
were shown to dominate the landslide counts.
When considered in the context of erosion, lateral
spreading within Quaternary sediments governs the landslide
inventory. We believe this reflects both the weak nature of
unconsolidated materials and the topographic settings where
these occurred, which allowed lateral spreads to propagate
and coalesce into very large mass movements. This was reflected by their heavy tailed, log-normal size distribution,
which indicated that a small but nevertheless significant
number of lateral spreads were very large. Lateral spreads
have been overlooked in many coseismic landslide investigations; however, our research has shown these to be the
governing mode of landsliding, and one that significantly
affected both the natural and built environment in eastern
Honshu.
1519
The dense seismograph network in Japan provides an
extraordinary opportunity to examine the relationship between coseismic landsliding and strong ground motion.
Nevertheless, our comparisons found no statistically significant correlation between landslide intensity and ground
shaking (i.e., PGA) across the study area. For the Tohoku
earthquake, this suggests that geologic unit exerted a strongly
dominant control on coseismic landsliding that largely overshadowed ground motion. This remains an important issue
that warrants further investigation.
Data and Resources
Google (2011). Experience the tsunami-affected area
of Japan through Street View. From Google Lat Long Blog:
http://google‑latlong.blogspot.com/2011/12/experience‑
tsunami‑affected‑areas‑of.html (last accessed September
2012).
Google Maps and Earth Team (2011). Google Lat Long
Blog. From http://google‑latlong.blogspot.com: http://google
-latlong.blogspot.com/2011/03/post-earthquake-images-ofjapan.html (last accessed September 2012).
Google.co.jp. (2011). Google Crisis Response. From
Resources related to the 2011 Japan Crisis: http://www.
google.co.jp/intl/en/crisisresponse/japanquake2011.html
(last accessed September 2012).
Japan Meteorological Agency (JMA; 2012a). Monthly
mean and monthly total tables retrievd from Japan Meteorological Agency at http://www.data.jma.go.jp/obd/stats/data/
en/smp/index.html (last accessed September 2012).
Japan Meterological Agency (JMA; 2012b). Climate statistics retrieved from Japan Metorological Agency at http://
www.data.jma.go.jp/obd/stats/data/en/normal/normal.html
(last accessed September 2012).
USGS (2011). Shakemaps and stastical information retrived from the USGS site: M 9.0—Near the East Coast of
Honshu, Japan. From USGS Earthquake Hazards Program:
http://earthquake.usgs.gov/earthquakes/dyfi/events/us/
c0001xgp/us/index.html (last accessed September 2012).
Acknowledgments
We acknowledge support of the American Society of Civil Engineers
(ASCE) for funding our reconnaissance mission in April 2011. We thank the
Japan Landslide Society for hosting our postearthquake visit and for kindly
providing logistical support during our field reconnaissance. Finally, we
thank Google Earth for processing and openly providing high-resolution
imagery and Street View surveys of the damaged area.
References
Blaschke, P., N. Trustrum, and D. Hicks (2000). Impacts of mass movement
erosion on land productivity: A review, Progr. Phys. Geogr. 24,
21–52.
Cox, B. R., C. M. Wood., R. Deschenes, and M. Pearson (2012). University
of Arkansas Preliminary Data Report for: Surface Wave Testing in
Urayasu and other Key Locations Affected by the M 9.0 Tohoku,
Japan Earthquake, University of Arkansas, 1–86.
1520
Dolan, R., and R. Worden (1992). Japan: A Country Study, Washington:
GPO for the Library of Congress, 610 pp.
Dragut, L., and C. Eisank (2012). Automated object-based classification
of topography from SRTM data, Geomorphology 141–142,
21–33.
Faure, M., F. Lalevee, Y. Gusokujima, J. T. Iiyama, and J. P. Cadet (1986).
The pre-Cretaceous deep-seated tectonics of the Abukuma massif and
its place in the structural framework of Japan, Earth Planet. Sci. Lett.
77, 384–398.
Furumura, T., S. Takemura, S. Noguchi, T. Takemoto, T. Maeda, K. Iwai,
and S. Padhy (2011). Strong ground motions from the 2011 offthe Pacific-Coast-of-Tohoku, Japan (Mw 9:0) earthquake obtained from a dense nationwide seismic network, Landslides 8,
333–338.
Geological Survey of Japan AIST (2009). Seamless digital geological map
of Japan 1:200,000, Dec 15, 2009 version, Research Information
Database DB084, Geological Survey of Japan, National Institute of
Advanced Industrial Science and Technology.
Goto, H., and H. Morikawa (2012). Ground motion characteristics during
the 2011 off the Pacific coast of Tohoku earthquake, Soils Found.
52, no. 5, 769–779.
Guzzetti, F., F. Ardizzone, M. Cardinali, M. Rossi, and D. Valigi (2009).
Landslide volumes and landslide mobilization rates in Umbria, central
Italy, Earth Planet. Sci. Lett. 279, 222–229.
Hasegawa, A., A. Yamamoto, N. Umino, S. Miura, S. Horiuchi, D. Zhao,
and H. Sata (2000). Seismic activity and deformation process of the
overriding plate in the northeastern Japan subduction zone, Tectonophysics 319, 225–239.
Hashimoto, M. (1991). Geology of Japan, Japan: Terra Scientific Publishing
Company, Tokyo, 253 pp.
Hovius, N., C. Stark, and P. Allen (1997). Sediment flux from a mountain
belt derived by landslide mapping, Geology 25, 231–234.
Innes, J. (1983). Lichenometric dating of debris-flow deposits in the Scottish
Highlands, Earth Surf. Process. Landf. 8, 579–588.
Ishihara, T., T. Sugai, and S. Hachinohe (2012). Fluvial response to
sea-level changes since the latest Pleistocene in the near-coastal
lowland, central Kanto Plain, Japan, Geomorphology 147–148,
49–60.
JSCE (2001). Geological Factors to be Considered in the Selection of
Preliminary Investigation Areas for HLW Disposal, Sub-Committee
on the Underground Environment, Committee of the Nuclear Power
Facilities, Japan Society of Civil Engineers.
Kawabe, S., Y. Tsukamoto, T. Kokusho, and R. Takahashi (2012).
Soil Liquefaction Observed at Katori City located along the
lower stream of Tonegawa River during 2011 Great East Japan Earthquake, in Proc. of the International Symposium on Engineering
Lessons Learned from the 2011 Great East Japan Earthquake, Tokyo,
739–746.
Kayen, R., Y. Tanaka, H. Tanaka, T. Sugano, I. Estevez, S. Cullenward,
W. Yeh, and D. Thomas (2011). LiDAR and Field Investigation of
the March 11, 2011 M 9.0 Great Tohoku Offshore Earthquake, and
April 7, 2011 M 7.4 Aftershock, Geotechnical Extreme Events Reconnaissance (GEER).
Keefer, D. K. (1984). Landslides caused by earthquakes, Bull. Geol. Soc.
Am. 95, 406–421.
Keefer, D. K. (1999). Earthquake-induced landslides and their effect on
alluvial fans, J. Sediment. Res. 69, 84–104.
Keefer, D. K. (2002). Investigating landslides caused by earthquakes—A
historical review, Surv. Geophys. 23, 473–510.
Klar, A., E. Ahoronov, B. Kalderon-Asael, and O. Katz (2011).
Analytical and observational relations between landslide volume and
surface area, J. Geophys. Res. 116, F02001, 10 pp., doi: 10.1029/
2009JF001604.
Koketsu, K., Y. Yokota, N. Nishimura, Y. Yagi, S. Miyazaki, K. Satake,
K. Y. Fujii, H. Miyake, S. Sakai, Y. Yamanaka, and T. Okada
(2011). A unified source model for the 2011 Tohoku earthquake, Earth
Planet. Sci. Lett. 310, 480–487.
J. Wartman, L. Dunham, B. Tiwari, and D. Pradel
Larsen, I. J., D. R. Montogomery, and O. Korup (2010). Landslide erosion
controlled by hillslope material, Nature Geosci. 3, 247–251.
Matsumoto, H. (1985). Beach ridge ranges and the Holocene sea-level
fluctuations on alluvial coastal plains, northeast Japan, Sci. Rep.
Tohoku Univ. 7th series (Geography) 35, 15–46.
Meunier, P., N. Hovius, and A. Haines (2007). Regional patterns of
earthquake-triggered landslides and their relation to ground motion,
Geophys. Res. Lett. 34, L20408, doi: 10.1029/2007GL031337.
Meunier, P., N. Hovius, and A. Haines (2008). Topographic site effects and
the location of earthquake induced landslides, Earth Planet. Sci. Lett.
275, 221–232.
Midorikawa, S., H. Miura, and T. Atsumi (2012). Strong motion records
from the 2011 off the Pacific coast of Tohoku earthquake,
in International Symposium on Engineering Lessons Learned from
the 2011 Great East Japan Earthquake, Tokyo, 297–304.
Mikoshiba, M. U., S. Kanisawa, Y. Matsuhisa, and S. Togashi
(2004). Geochemical and isotopic characteristics of the Cretaceous
Orikabe Plutonic Complex, Kitakami Mountains, Japan: Magmatic
evolution in a zoned pluton and significance of a subductionrelated mafic parental magma, Contrib. Mineral. Petrol. 146,
433–449.
Miyagi, T., D. Higaki, H. Yagi, S. Doshida, N. Chiba, J. Umemura, and G.
Satoh (2011). Reconnaissance report on landslide disasters in northeast Japan following the M 9 Tohoku earthquake, Landslides 8,
339–342.
Miyashiro, A. (1958). Regional metamorphism of the Gosaisho Takanuki
district in the Central Abukuma plateau, J. Facul. Sci. Univ. Tokyo,
Sec. 2 11, 219–272.
Montgomery, R., K. Schmidt, H. Greenberg, and W. Dietrich (2000). Forest
clearing and regional landsliding, Geology 28, 311–314.
Okada, T., K. McAneney, and K. Chen (2011). Estimating insured residential losses from large flood scenarios on the Tone River, Japan
—A data integration approach, Nat. Hazards Earth Syst. Sci. 11,
3373–3382.
Okamoto, R., K. Kojima, and R. Yoshinaka (1981). Distribution and engineering properties of weak rocks in Japan, in Proceedings of the
International Symposium on Weak Rock, 21–24.
Parker, R., A. Densmore, N. Rosser, M. de Michele, Y. Li, R. Huang,
S. Whadcoat, and D. N. Petley (2011). Mass wasting triggered by the
2008 Wenchuan earthquake is greater than orogenic growth, Nature
Geosci. 4, 449–452.
Pradel, D., B. Tiwari, and J. Wartman (2011). Landslides triggered by
the Tohoku earthquake: Preliminary observations, GeoStrata 10,
28–32.
Pradel, D., J. Wartman, and B. Tiwari (2012a). Failure of Fujinuma Dam
during the 2011 Tohuku Earthquake, in Joint Proc., 9th International
Conference on Urban Earthquake Engineering/4th Asia Conference
on Earthquake Engineering, Tokyo.
Pradel, D., J. Wartman, and B. Tiwari (2012b). Impact of anthropogenic
changes on liquefaction along the Tone River during the 2011 Tohoku
Earthquake, Nat. Hazards Rev., doi: 10.1061/(ASCE)NH.15276996.0000097.
Prima, O., A. Echigo, R. Yokoyama, and T. Yoshida (2006). Supervised
landform classification of Northeast Honshu from DEM-derived
thematic maps, Geomorphology 78, 373–386.
Sato, H., N. Hirata, T. Iwasaki, M. Matsubara, and T. Ikawa (2002).
Deep seismic reflection profiling across the Ou Backbone
range, northern Honshu Island, Japan, Tectonophysics 355,
41–52.
Sugawara, D., F. Imamura, K. Goto, H. Matsumoto, and K. Minoura (2012).
The 2011 Tohoku-oki Earthquake Tsunami: Similarities and Differences to the 869 Jogan Tsunami on the Sendai Plain, Pure Appl.
Geophys., doi: 10.1007/s00024-012-0460-1.
Tamura, T., and F. Masuda (2005). Bed thickness characteristics of innershelf storm deposits associated with a transgressive to regressive
Holocene wave-dominated shelf, Sendai coastal plain, Japan, Sedimentology 52, 1375–1395.
Landslides in Eastern Honshu Induced by the 2011 Off the Pacific Coast of Tohoku Earthquake
Terzaghi, K. (1950). Mechanism of landslides, in Applications of Geology to
Engineering Practice, Berkley Volume, S. Paige (Editor), Geological
Society of America, 83–123.
Trifunac, M., and A. Brady (1975). A study of the duration of strong
earthquake ground motion, Bull. Seismol. Soc. Am. 65, 581–626.
Ujiie-Mikoshiba, M., N. Imai, S. Terashima, Y. Tachibana, and T. Okai (2006).
Geochemical mapping in northern Honshu, Japan, Appl. Geochem. 21,
492–514.
Uzuoka, R., N. Sento, M. Kazama, and T. Unno (2005). Landslides during
the earthquakes on May 26 and July 26, 2003 in Miyagi, Japan, Soil
Found. 45, 149–163.
Wakita, K., T. Igawa, S. Takarada, and Y. Fusejima (2008). Creation of
seamless geological map of Japan at the scale of 1:200,000 and its
distribution through the web, Synthesiology 1, 82–93.
Wasowski, J., D. Keefer, and C. Lee (2011). Toward the next generation of
research on earthquake-induced landslides: current issues and future
challenges, Eng. Geol. 122, 1–8.
Yamagishi, H., and J. Iwahashi (2007). Comparison between the two
triggered landslides in Mid-Niigata, Japan by July 13 heavy rainfall and October 23 intensive earthquakes in 2004, Landslides 4,
389–397.
Yonekura, N., S. Kaizuka, M. Nogami, and K. Chinzei (Editors) (2001).
Introduction to Japanese Geomorphology, University of Tokyo Press,
Tokyo (in Japanese).
Yoshii, T. (1979). A detailed cross section of the deep seismic zone beneath
northeast Honshu Japan, Tectonophysics 55, 349–360.
Department of Civil and Environmental Engineering
University of Washington
201 More Hall, Box 352700
Seattle, Washington 98195-2700
[email protected]
[email protected]
(J.W., L.D.)
Department of Civil and Environmental Engineering
California State University, Fullerton
800 N State College Blvd., E-419
Fullerton, California 92834
[email protected]
(B.T.)
University of California, Los Angeles
Civil and Environmental Engineering Department
5732 Boelter Hall, Box 951593
Los Angeles, California 90095-1593
[email protected]
(D.P.)
Manuscript received 6 April 2012
1521