Document Actions

ijege-13_bs-uchida-et-alii.pdf

background image
Italian Journal of Engineering Geology and Environment - Book Series (6) www.ijege.uniroma1.it © 2013 Sapienza Università
Editrice
211
DOI: 10.4408/IJEGE.2013-06.B-18
RAINFALL THRESHOLDS FOR DEEP-SEATED RAPID LANDSLIDES
T
aro
UCHIDA
(*)
, a
Tsushi
OKAMOTO
(*)
, J
un
'
ichi
KANBARA
(*)
& K
azumasa
KURAMOTO
(**)
(*)
National Institute for Land and Infrastructure Management - Ibaraki, Japan
Postal address: Asahi 1, Tsukuba, Ibaraki 3050804, Japan - E-mail: uchida-t92rv@nilim.go.jp
(**)
Chuden Engineering Consultants Co., Ltd. - Hiroshima, Japan
K
ey
words
: deep-seated catastrophic landslide, rainfall thre-
shold, early warning
INTRODUCTION
In steep mountainous regions, soils and weathered
bedrock have been known to slide simultaneously
(e.g., J
iTosono
et alii, 2008; u
chida
et alii, 2011).
These landslides sometimes move rapidly and trigger
debris flows (e.g., n
ishiguchi
et alii, 2012), and such
landslides can have serious impacts on human lives
and infrastructure (e.g., E
vans
et alii, 2007; J
iTosono
et alii, 2008; T
aniguchi
, 2008; s
hiEh
et alii, 2009). In
this study, we refer to these landslides as deep-seated
rapid (catastrophic) landslides (hereafter, DCLs). This
study excludes slow, mall displacement, failures, such
as slower deep-seated landslides, including earth-
flow, slump (e.g., s
idlE
& o
chiai
, 2006), deep-seated
gravitational slope deformation (e.g., a
gliardi
et alii,
2009), or rock flow, which are distinct from DCLs.
It has been proposed that early warning systems
and the development of countermeasures for DCLs
are important tools for disaster risk reduction. Ac-
cordingly, u
chida
et alii (2011) recently proposed a
method for mapping the susceptibility of a landscape
to DCLs. In fact, the Japanese government initiated a
nationwide survey to map susceptibility to DCL using
this method (u
chida
et alii, 2012).
In 2005, the Japanese government also initiated
a new nationwide early warning system for disas-
ters associated with landslides and debris flows. The
ABSTRACT
Soils and weathered bedrock have been known
to slide simultaneously, with the resulting landslides
sometimes moving rapidly and triggering debris flows.
In this study, we refer to these landslides as deep-seat-
ed rapid (catastrophic) landslides (hereafter, DCLs).
DCLs can result in serious damage, although the fre-
quency of such disasters is generally low. Therefore,
early warning systems and the construction of coun-
termeasures for DCLs are important tools for disas-
ter risk reduction. We analyzed the characteristics of
recent storms that triggered DCLs in Japan. We found
that several of these storms triggered multiple DCLs
(multi-DCL events), although most triggered only a
single DCL (single-DCL events). For short-term (<24
h) rainfall intensity, there was no significant differ-
ence in maximum rainfall intensities between storms
that triggered single and multiple DCLs. Conversely,
for long-term (48 or 72 h) rainfall amounts, storms
that triggered multiple DCLs exhibited considerably
higher rainfall amounts than storms that triggered
no DCLs or a single DCL. In particular, more than
90% of storms that triggered multiple DCLs recorded
rainfall of more than 600 mm per 48 h. Our results
suggest that the 48-h or 72-h rainfall amounts were
effective for assessing temporal changes in DCL sus-
ceptibility, but not the 1-h to 24-h rainfall amounts.
This indicates that the occurrence of DCLs is strongly
controlled by long-term rainfall amounts but less
strongly by short-term rainfall intensity.
background image
T. UCHIDA, A. OKAMOTO, J. KANBARA & K. KURAMOTO
212
International Conference Vajont 1963-2013. Thoughts and analyses after 50 years since the catastrophic landslide Padua, Italy - 8-10 October 2013
ture, including journals, conference proceedings, and
event and technical reports. Most of the literature
used was written in Japanese. The PWRI included
only landslides triggered by rainfall or snowmelt and
those that occurred after 1868 (the beginning of the
Meiji era). Thus, the database excluded earthquake-
triggered landslides and volcanic landslides. The
PWRI confirmed areas of landsliding using current
topographic maps, current and historical aerial pho-
tographs, and field surveys, and did not include land-
slides that could not be confirmed.
Landslides with volumes greater than 10
6
m
3
or
areas greater than 10
5
m
2
were compiled in the da-
tabase. The PWRI excluded slow failures of a more
chronic nature (such as slower deep-seated land-
slides, deep-seated gravitational slope deformation,
and rock flow) from the database, similarly to K
o
-
rup
et alii (2007). Landslides that turned into debris
flows (Fig. 1a) and slid in many fragments (Fig. 1b)
were also included in the database. Moreover, if more
than half the volume of landslide runout from land-
slide scar, this landslide was included in the database
(Fig. 1c). Around 150 deep catastrophic landslides
were included in the database. For each landslide,
system primarily involves setting a criterion for the
occurrence of debris flows and shallow landslides
based on several rainfall indices (e.g., o
sanai
et alii,
2010). Moreover, many studies have been conducted
to clarify rainfall thresholds for use in the prediction
of landslide occurrence (e.g., c
ainE
, 1980; g
uzzETTi
et alii, 2008; s
aiTo
et alii, 2010). However, since
the frequency of DCLs is low, it is very difficult to
collect a large quantity of data describing the char-
acteristics of rainfall-triggered DCLs. Consequent-
ly, the rainfall threshold for DCLs has not yet been
clarified. In this study, we compiled rainfall data and
landslide inventories to clarify the rainfall thresh-
old for DCL occurrence. Additionally, we examined
several rainfall indices to select the most effective
rainfall indices for assessing temporal changes in
DCL susceptibility.
METHODS
LANDSLIDE DATA
We used a database of DCLs in Japan, published
by the volcanic and debris flow team of the Public
Works Research Institute (PWRI). The PWRI com-
piled landslide information from the existing litera-
Fig. 1 - Deep-seated rapid cata-
strophic landslides in Japan,
triggered by Typhoon Tales
in September 2011. (a) Ki-
tamata landslide, Nara Pre-
fecture, turned into a debris
flow. (b) Akatani landslide,
Nara Prefecture, slid in
many parts and caused a
landslide dam. (c) Iya land-
slide, Wakayama Prefecture.
More than half of the volume
of landslide runout from
landslide scar and caused
landslide dams
background image
RAINFALL THRESHOLDS FOR DEEP-SEATED RAPID LANDSLIDES
Italian Journal of Engineering Geology and Environment - Book Series (6) www.ijege.uniroma1.it © 2013 Sapienza Università
Editrice
213
DATA ANALYSIS
If a single DCL was known to be triggered by a
given storm, we collected rainfall data for the rain
gauge closest to the landslide (Fig. 2a). Conversely,
if several DCLs were triggered by a single storm,
we collected data at a rain gauge located around the
center of the area of landslide distribution (Fig. 2b). If
a DCL occurred more than 20 km away from the rain
gauge closest to the center of area landslide distribu-
tion, we collected additional rain gauge data for the
DCL,apart from more than 20 km (Fig. 2c). So, we
collected two sets of rainfall data 20 km apart In the
latter case, we assumed the DCL to be triggered by
two (or more) distinct events.
Consequently, we analyzed data from 43 rain-
fall events. Thirteen of these rainfall events triggered
more than two DCLs; hereafter, we refer to these
rainfall events as multi-DCL events. Conversely, 30
rainfall events triggered only one DCL; hereafter, we
refer to these as single-DCL events. Moreover, eleven
of these single-DCL events, but no multi-DCL events,
occurred in snowmelt season.
We analyzed rainfall data for four days: the day
of landslide occurrence, and the three days prior
(Fig. 3). We refed these four days as “analyzed pe-
riod”, Several rainfall indices have been proposed
for assessing temporal change in landslide suscep-
tibility (e.g., o
sanai
et alii, 2010). Here, we simply
varied the calculated total rainfall amount for a given
duration to search for effective rainfall indices to as-
sess temporal changes in DCL susceptibility. In this
study, we used seven durations (1, 3, 6, 12, 24, 48,
and 72 h) for calculation. Using hourly rainfall data,
we calculated total rainfall amount for a given dura-
tion, such as 1, 3, 6, 12, 24, 48, and 72 h, for every
hour. Then, we searched for the maximum total rain-
the information stored includes the following: (i)
geographical location; (ii) landslide geometrical
properties, including length, width, depth, area, and
volume; (iii) date of occurrence; and (iv) triggering
phenomena (e.g., snowmelt, rainfall). Since DCLs
that occurred in 2011 were not included in the data-
base, we added these DCLs for our study.
RAINFALL DATA
Because it is difficult to obtain rainfall data for
old landslides, we used information for landslides that
occurred after 1976. We used rainfall data from the
Automated Meteorological Data Acquisition System
(AMeDAS), operated by the Japan Meteorological
Agency (JMA). AMeDAS was developed in 1976,
and has around 1,700 rain gauges; therefore, the mean
density of rain gauges is approximately one per 220
km
2
. We used hourly rainfall data.
Fig. 2 - Schematic illustration describing location of ana-
lyzed rainfall gauges
Fig. 3 - Schematic illustration describing the periods used
for calculation of rainfall indices
background image
T. UCHIDA, A. OKAMOTO, J. KANBARA & K. KURAMOTO
214
International Conference Vajont 1963-2013. Thoughts and analyses after 50 years since the catastrophic landslide Padua, Italy - 8-10 October 2013
est 1-h rainfall value for multi-DCL events (20 mm/h).
Generally, the difference in rainfall amount between
single-DCL and multi-DCLs events became more pro-
nounced with increasing duration. Thus, if the calcu-
lated duration was shorter than 24 h, the difference be-
tween rainfall amounts for single-DCL and multi-DCL
events was minor (Fig. 5b-5d). Conversely, the 48-h
rainfall amounts for multi-DCL events were clearly
larger than those for single-DCL events (Fig. 5f). Al-
though 48-h rainfall was less than 400 mm for one of
the multi-DCL events, the remaining twelve multi-DCL
events involved rainfall of more than 600 mm. Moreo-
ver, all 48-h rainfall amounts of single-DCL events in-
volved rainfall of less than 600 mm. There were also
clear differences between single-DCL and multi-DCLs
events for the maximum 72-h rainfall (Fig. 5g).
Figure 6 shows the relationship between the number
of storms in 1976-2010 and the number of DCLs for
each criterion (see Tab. 1). For example, around 230
storms had 48-h rainfall greater than 600 mm, while
around 90 DCLs occurred during such events (Fig. 6).
The plot in the upper left of Figure 6 indicates the ratio
of number of DCLs to number of rain storms , which
satisfied the criterion, likes 600 mm/48-h, was high, in-
dicating that that if rain storms satisfied such criteria,
there was high probability of DCL occurrence. Con-
versely, the plot in the lower right of Figure 6 indicates
that even if rain storms satisfied these criteria, likes 50
mm/1-h, there was low probability of DCL occurrence.
This figure also suggests that, for a set number of rain-
fall events, the numbers of DCLs occurring increases
with increasing duration. Moreover, if the number of
rainfall was the same, multi-DCLs event increased with
the increase of calculated duration (from brawn plots to
light blue plots in Fig. 7). We found no clear differences
fall amount for a given duration in the analyzed pe-
riod for each rainfall event (Fig. 3).
Moreover, we used all of the rainfall data in AMe-
DAS from 1976 to 2011 to calculate the numbers of
rainfall events for each criterion. Criteria are summa-
rized in Tab. 1.
RESULTS AND DISCUSSIONS
Figure 4 shows the relationships between dura-
tion and maximum rainfall amount for each rainfall
event. For single- DCL events, the maximum rainfall
amounts increased with increasing duration if the
duration was shorter than 24 h. Conversely, if the
duration was longer than 24 h, the maximum rainfall
amounts of most single-DCL events remained almost
constant (Fig. 4a). In contrast, the maximum rain-
fall amounts for multi-DCLs events increased with
increasing duration, at least when the duration was
shorter than 72 h (Fig. 4b). The maximum rainfall
amounts were generally very small for single-DCL
events affected by snowmelt (Fig. 4c).
There was no clear difference in maximum 1-h
rainfall amount between single-DCL and multi-DCL
events (Fig. 5a). Approximately half of the 1-h rainfall
values for single-DCL events were larger than the low-
Fig. 4 - Relationship between calculated duration and maximum rainfall amounts for analyzed periods: (a) single-DCL
events, (b) multi-DCL events, (c) single-DCL events affected by snowmelt
Tab. 1 - Criteria for counting numbers of rainfall events
and DCLs
background image
RAINFALL THRESHOLDS FOR DEEP-SEATED RAPID LANDSLIDES
Italian Journal of Engineering Geology and Environment - Book Series (6) www.ijege.uniroma1.it © 2013 Sapienza Università
Editrice
215
multi-DCL events occurring. These results indicate that
the occurrence of DCLs is strongly controlled by long-
term rainfall amounts, but not short-term rainfall inten-
sity. This agrees with existing evidence, which suggests
that DCLs tend to occur near the ends of storms.
in these relationships between the 48-h and 72-h rain-
fall amounts; this suggests that using 48-h or 72-h rain-
fall amounts, but not 1-h to 12-h amounts, can provide
an opportunity to assess the probability of single- and
Fig. 5 - Numbers of rain storm events triggerd DCL according to maximum rainfall amounts for (a) 1 h, (b) 3 h, (c) 6 h, (d)
12 h, (e) 24 h, (f) 48 h, and (g) 72 h
Fig. 6 - Relationship between number of rainfall events
(from 1976 to 2010 in Japan) and number of
DCLs for each criterion
Fig. 7 - Relationship between number of rainfall events
(from 1976 to 2010) and number of multi-DCL
events for each criterion
background image
T. UCHIDA, A. OKAMOTO, J. KANBARA & K. KURAMOTO
216
International Conference Vajont 1963-2013. Thoughts and analyses after 50 years since the catastrophic landslide Padua, Italy - 8-10 October 2013
As shown in Figure 8, the probability of occur-
rence of multi-DCL events is not particularly high,
even when the 48-h rainfall is used to assess temporal
changes in DCL susceptibility. For example, if we set
600 mm in 48 h as a threshold value for DCL occur-
rence, the probability of DCL occurrence is less than
3%. This indicates that most rainfall events will not
trigger any DCLs, even if total rainfall exceeds the
threshold. Therefore, to establish an effective early-
warning system for DCLs, it is necessary to combine
rainfall information with other tools, such as suscepti-
bility mapping (e.g., u
chida
et alii, 2011), hydrologi-
cal and sediment transport monitoring (e.g., F
uJiTa
et
alii, 2010; o
KamoTo
et alii, this issue), and seismic
wave observation (e.g., Y
amada
et alii, 2012).
CONCLUSIONS
We investigated the rainfall threshold for the
occurrence of deep-seated rapid (catastrophic) land-
slides (DCLs) by studying around 50 rainfall events
that triggered DCLs in Japan. We found that 48-h or
72-h rainfall are more effective in assessing tempo-
ral changes in DCL susceptibility 1-h to 24-h rainfall
amounts. Our results suggest that the occurrence of
DCLs is strongly controlled by long-term rainfall
amounts, but not by short-term rainfall intensity.
Fig. 8 - Relationship between 48 h rainfall amounts, ra-
tio of multi-DCL events to all rainfall events from
1976 to 2010, and number of all rainfall events
from 1976 to 2010
REFERENCES
a
gliardi
F., c
rosTa
g.B., z
anchi
a. & r
avazzi
c. (2009) - Onset and timing of deep-seated gravitational slope deformations in
the eastern Alps, Italy. Geomorphology, 103: 113-129.
c
ainE
N. (1980) - The rainfall intensity-duration control of shallow landslides and debris flows. Geografiska Annaler., Series
A, 62: 23-27.
E
vans
s.g., g
uThriE
r.h., r
oBErTs
n.J., & B
ishop
n.F. (2007) - The disastrous 17 February 2006 rockslide-debris avalanche
on Leyte Island, Philippines: a catastrophic landslide in tropical mountain terrain. Nat. Hazards Earth Syst. Sci., 7: 89-101.
F
uJiTa
m., o
hshio
s. & T
suTsumi
d. (2010) - A prediction method for slope failure by means of monitoring of water content in
slope-soil layer. J. Disaster Res., 5: 296-306.
g
uzzETTi
F., p
Eruccacci
s., r
ossi
m. & s
TarK
c. (2008) - The rainfall intensity-duration control of shallow landslides and debris
flows: an update. Landslides, 5: 3-17.
J
iTosono
T., s
himoKawa
E. & T
EramoTo
Y. (2008) - Debris flow induced by deep-seated landslides at Minamata City, Kumamoto
Prefecture, Japan in 2003. Int. J. Erosion Cont. Eng., 1: 5-10.
K
orup
o., c
laguE
J.J., h
Ermanns
r.l., h
EwiTT
K., s
Trom
a.l. & w
EidingEr
J.T. (2007) - Giant landslides, topography, and
erosion. Earth Planet. Sci. Lett., 261: 578-589.
n
ishiguchi
Y., u
chida
T., T
aKEzawa
n., i
shizuKa
T. & m
izuYama
T. (2012) - Runout characteristics and grain size distribution of
large-scale debris flows triggered by deep catastrophic landslides. Int. J. Erosion Control Engng., 5: 16-26.
o
KamoTo
a., u
chida
T., h
aYashi
s., s
uzuKi
T., Y
amashiTa
s., T
agaTa
s. & F
uKumoTo
a. (this issue) - Possibility of early-
warning for large-scale landslide using hydrolical and sedment transport observation in mountain rivers.
o
sanai
n., s
himizu
T., K
uramoTo
K., K
oJima
s. & n
oro
T. (2010) - Japanese early-warning for debris flows and slope failures
using rainfall indices with Radial Basis Function Network. Landslides, 7: 1-14.
s
aiTo
h., n
aKaYama
d. & m
aTsuYama
h. (2010) - Relationship between the initiation of a shallow landslide and rainfall
intensity-duration threshold in Japan. Geomorphology, 118: 167-175.
s
hiEh
c-l., w
ang
c-m., l
ai
w-c., T
sang
Y-c., & l
EE
s-p. (2009) - The composite hazard resulted from Typhoon Morakot in
Taiwan. J. Jpn. Soc. Erosion Control Engng., 62 (4): 61-65.
s
idlE
r.c. & o
chiai
h. (2006) - Landslides: processes, prediction and land use. 312 pp. American Geophisical Union.
background image
RAINFALL THRESHOLDS FOR DEEP-SEATED RAPID LANDSLIDES
Italian Journal of Engineering Geology and Environment - Book Series (6) www.ijege.uniroma1.it © 2013 Sapienza Università
Editrice
217
T
aniguchi
Y. (2008) - Sediment disasters caused by typhoon No. 14, 2005, in Miyazaki Prefecture. Int. J. Erosion Cont. Eng., 1:
11-19.
u
chida
T., Y
oKoYama
o., s
uzuKi
r., T
amura
K. & i
shizuKa
T. (2011) - A new method for assessing deep catastrophic landslide
susceptibility. Int. J. Erosion Control Engng., 4: 32-42.
u
chida
T., Y
oKoYama
o., T
aKEzawa
n. & i
shizuKa
T. (2012) - Assessment for susceptibility of deep catastrophic landslide
induced disasters in Japan. Proceedings of 12
th
INTERPRAEVENT, 1: 609-618.
Y
amada
m., m
aTsushi
Y., c
higira
m., & m
ori
J. (2012) - Seismic recordings of the Landslides caused by Typhoon Talas.
Geophy. Res. Lett., 39: L13301, doi:10.1029/2012GL052174.
background image
Statistics