Document Actions


background image
Italian Journal of Engineering Geology and Environment - Book Series (6) © 2013 Sapienza Università
DOI: 10.4408/IJEGE.2013-06.B-49
& G
Shanghai Jiao Tong University - Department of Civil Engineering - Shanghai 200240, P.R. China
Kyoto University - Disaster Prevention Research Institute - Uji, 611-0011 - Kyoto, Japan
In the construction of dams, the landslides on the
banks of the reservoir are always a challenging problem
to both geologists and engineers. This kind of landslide
can not only affect the function of the dams, but also re-
sult in great disaster to the downstream of the dam. For
example, the catastrophic failure of the left bank slope
of the Vajont Reservioir resulted in more than 2600 cas-
ualties (M
, 1964), sweeping away several villages
completely. More recently, it has been reported that a
huge number of landslides were triggerd or reactivated
in the resovoir area of the Three Gorges Dam due to the
impudement, causing great loss on both lives and econ-
omy, threntening the safe function of the dam, and forc-
ing the government to relocate another some 100,000
people (Y
, 2004). Although the instability problem of
bank slopes had been widely studied since the Vajont
event, understanding on this kind of bank slope instabil-
ity is still limited. This can be examplified by a big land-
slide on the upper stream of the dam of Laxiwa Hydro-
power station. This landslide seems to be very similar to
the case of Vajoint dam and is directly threatening the
safety of the dam and people living downstream. The
local authoriteis had paid great concern to this problem,
and further survey had been conducted to better under-
stand the possible deformatin/sliding mechanism. In this
study, we used satellite remote sensing-based method to
detect the deformation of the slope before and after the
impudement. Some preliminary investigation results
will be introduced in this paper.
Laxiwa Hydropower Station is situated on the
main reach of the Yellow River, China. The construc-
tion was started in October 2001, and impoundment
started in March 2009. However, from May 2009, the
right-bank slope of the reservoir, about 700 m high
and 1000 m wide about 500m far from the dam, was
found to be deforming greatly and continuously. Al-
though this slope had been identified as an old land-
slide, the survey before the construction of the dam
concluded that this slope is stable and will be stable
even after the impoundment. Thus, the slope was not
monitored before the visible deformation occurred
after the impoundment. To identify the relation-
ship between the slope deformation and impound-
ment, we utilized D-InSAR and ALOS Prism data
to analyze the slope deformation before and after
impoundment. We found that there was no identifi-
able deformation before the impoundment, and the
maximum horizontal displacement reached approxi-
mately 7.5m during the period of 3 April 2009 to 22
May 2010 after impoundment. Although the poten-
tial sliding surface has not been identified irrespec-
tive of the performance of a lot of survey works, we
concluded that the instability of this high and steep
slope will be great threaten to the dam as well as to
the downstream residents.
: remote sensing, D-InSAR, ALOS prism, hydropo-
wer station, slope deformation, impoundment, risk assessment
background image
International Conference Vajont 1963-2013. Thoughts and analyses after 50 years since the catastrophic landslide Padua, Italy - 8-10 October 2013
and five exploration tunnels were newly dug into the
slope to identify the potential sliding surface. All the
monitored results revealed that the slope is deform-
ing greatly and continuously. However, borings failed
to reach the target depth, because the deforming land-
slide mass made the drilling impossible. The deform-
ing slope is about 700 m high (from the original river
bed to the crest of the sliding slope) and 1000 m wide
in average, and also very close to the dam site (about
500 m far from the dam). Therefore, if collapse failure
occurred to this slope, the dam and all other facilities
will be greatly damaged or destroyed, and over-topping
reservoir water may flood the wide downstream area,
greatly threatening the people living on downstream
and the safe function of another big hydropower sta-
tion (Lijiaxia) 73 km downstream. Namely, this dam is
facing the potential occurence of catastrophic disaster
similar to that occured in Vajont dam (M
, 1964).
After the setup of monitoring system, the deforma-
tion/displacement of the slope had been monitored con-
tinuosly, and the deformation/sliding patten and tenden-
cy with increase of reservior water level had been made
clear. Nevertheless, the defromation/sliding mechanism
has not been made clear, because (1) the deformation/
sliding feature of the downslope area was not clear due
to submersion under water, (2) those borings or explo-
ration tunnels were not deep or long enough, such that
no evidence for the identification of sliding surface was
found, and (3) no monitoring data was availabe for the
slope before the impoundment. To understand the initia-
tion mechanism of instability and then provide proper
countermeasures for lowering or releasing the risk of
collapse failure, it is necessary to undestand the possible
developing history of this landslide and also the spatial
information of the slope deformation. As part of a project
for investigating the movement mechanism, this study
aimed at using interferometric analysis of synthetic ap-
erture radar (InSAR) and also satellite optical images
with very high-resolution to grasp the possible deform-
ing areas and also to analyze the ground displacement
level before and shortly after the impundment.
The landslide is located on the right bank upstream
from the Laxiwa Hydropower station, 500-1700 m far
from the dam (Fig. 2). The valley here is narrow and
steep with the right slope’s angle ranging from 38 to 46
degrees. The original water surface was 2,254 ~ 2,257
Laxiwa Hydropower Station is a large on-con-
struction hydropower station on the main reach of the
Yellow River. It locates between Guide County and
Guinan County, Qinghai Province, about 134 km far
from Xining City, Qinghai Province (Fig. 1). This sta-
tion is the second cascade hydropower project to the
Longyangxia Hydropower project on the upper stream
of Yellow River. The dam is a 250 m high concrete
double-curvature arch dam with its crest length being
about 460 m. The primary purpose of the dam is hy-
droelectric power generation and it is designed to sup-
port a 4200 MW power station. The normal reservoir
water level is 2452m, dead water level is 2440m, the
total reservoir capacity is about 1.08 billion m
, and
the active reservoir capacity is 0.15 billion m
The construction of Laxiwa Hydropower Project
was strated in October 2001. In April 2006, the first
concrete was cast and on March 1, 2009, impound-
ment began, and on May 18, 2009, the first two elec-
tric generators were put into function. However, soon
after the impoundment (in late May 2009), the right-
bank slope (Guobu slope) on the upper stream of the
dam showed remarkable deformation (Fig. 2). Many
cracks appeared and was keeping widening on the crest
of the slope. Also rocks dropped off the slope into the
river frequently. Recongnizing the risk of occurence of
catastrophic landslide, the deformation of the slope had
been monitored by installing a deformation-monitoring
network immeadialty since July 2009. This monitoring
network included many monitoring techniques, such as
airborne topographic laser scanner for obtaining topog-
raphy of the crest and the slope in high precision, total
station (39 observation points were newly setup on the
slope for direct measuring). Six borings were drilled
Fig. 1 - Location of Laxiwa dam site
background image
Italian Journal of Engineering Geology and Environment - Book Series (6) © 2013 Sapienza Università
had been conducted during the period from 1991 to
1997. The results showed that there was no obvious
displacement. On the other hand, no visible cracks or
deformation appeared on the crest before 2003 (i.e.,
before the impoundment).
Tension cracks appeared on the crest of Guobu
m a.s.l. Fig. 3a shows a view of the whole area from
Google Earth that was based on the satellite images of
May 18, 2004, while Fig. 3b shows the handmade photo
of the crest that were taken in 1989. The crest of Guobu
Slope is a triangular in plan view and relatively flat ‘ter-
race’, 750 m long and 50~290 m wide at 2930~2950 m
a.s.l., ~700 m above the river bed. It is bounded by a dis-
tinct scarp, and talus can be seen in two areas outlined
by the circle and the square in Fig. 3a.
According to the geological map of this area, the
bedrocks mainly consist of Mesozoic granites, Terti-
ary sandstone, Triassic slate and Quaternary sediments,
while Laxiwa dam site and the landslide area are Meso-
zoic granite (Fig. 4).
There is a stepped topography on the crest of the
slope indicating that Guobu slope might be an old
deep-seated landslide. Detailed geological survey had
been performed on this slope since 1989, which gave a
conclusion that this landslide is an ancient one. How-
ever, only four simple observation piles were installed
to monitor the slope deformation, and the monitoring
Fig. 4 - Geology map of Laxiwa dam site area
Fig . 3 - Oblique view of Guobu slope(a) from Google Earth, and the views of crest of Guobu slope at different time (b)
Fig. 2 - View of Guobu slope on the upper stream of Laxiwa dam after impoundment (Photo on 2010/1/14). Arrows show the
boundary of the landslide
background image
International Conference Vajont 1963-2013. Thoughts and analyses after 50 years since the catastrophic landslide Padua, Italy - 8-10 October 2013
the displacement rates showed good correlation with
the variation of water level. It is noted that although a
lot of monitoring data had been obtained, they are not
available at present, and will be open only after counter-
measures being performed. Here we present two photos
(Fig. 5) from which the scale of the displacement can be
briefly understood through the subsidence appearing on
the crest of the slope. The crest had a settlement of about
20 m by 12 March 2010 (Fig. 5a). A further settlement
of 6 m was identified seven months late (Fig. 5b).
The topography map with a contour interval of 10
m was surveyed by airborne laser scanner after May
2009 (Fig. 6a). A longitudinal section along the line
connecting point ‘P1’ and ‘P2’ in this figure is given in
Fig. 6b, where the sliding surface is just an inferred one.
As mentioned above, there is yet not clear evidence
showing the location of sliding surface. If this sliding
surface be correct, the total volume of the potentially
unstable massif will be approximately 120-150 Mm
This volume will be about one half of that of the Vajont
landslide (M
, 1964; G
& G
, 2005).
Differential Synthetic Aperture Radar Interferometry
(D-InSAR), is a microwave remote sensing technology
based on the Interferometric Synthetic Aperture Radar. It
has the advantage of high-accuracy, high-resolution, all-
weather, low-cost and wide-range, which enables us to
analyze very small ground movement and to cover in con-
tinuity large areas. So INSAR becomes very useful in de-
tecting ground movement and studying landslide hazard.
We use interferometric synthetic aperture radar
(InSAR) to obtain spatially detailed maps of ground-
surface deformation. This technique has been applied
slope in May, 2009, about two months after the im-
poundment. These cracks extended along the direction
of North-South with linear features, being parallel to the
main cliff. The terrace showed a settlement of 1.0~1.5
m. Also, there were some small-scale rockfalls on dif-
ferent parts of the slope, with rocks dropping into the
reservoir. Acknowledging the potential of catastrophic
landsliding, urgent reconnaissance had been conducted
to the slope, including the setup of slope deformation
monitoring system, geophysical survey, drilling explo-
ration, tunnel investigation at differing altitudes and site
reconnaissance. The deformation monitoring system
started to work in August 2009. The monitored data
showed that the whole slope was deforming significant-
ly and continuously with the increase of the reservoir
water level. Some monitoring points indicated a daily
displacement of several centimeters, and the total dis-
placement in five months (from middle August 2009
to middle January 2010) reached several meters, and
Fig. 6 - (a) Topography of Guobu slope area (a), and (b) longitudinal section along line P-P' in Fig.6a
Fig. 5 - Deformation on the crest of Guobu slope after the
background image
Italian Journal of Engineering Geology and Environment - Book Series (6) © 2013 Sapienza Università
sor have a resolution of 2.5 m. Considering the avail-
able period and the resolutions of these images, we
collected ENVISAT ASAR images from the later half
of 2007, and PASAR images from 2008 to present.
In the present study, we use ALOS PALSAR as the
main image source to interpret the radar images (Ta-
ble 1). We analyzed the slope deformation during two
periods (December 2007~ June 2008; December 2009
~ February 2010). Table 2 shows the combination and
baselines of 4 pairs of images.
Concerning the coverage ranges, image quality and
phase distribution, ALOS PRISM panchromatic images
are finally chosen to be the image source for interpret-
ing images (Tab. 3). For the period of December 2007
~ June 2008, four images that were taken on December
9, 2007, January, 24 2008, March 10, 2008, June 10,
2008, were used. The first 3 images have a time interval
of 46 days (one revisit cycle) between each of them,
and the last two have a time difference of 92 days (2
revisit cycles). For the period of December 2009~Feb-
ruary 2010, four images for two pairs were selected and
analyzed. One pairs were taken on November 14, 2009
and December 31, 2009 respectively; the other pairs
were taken on January 29, 2010 and February 15, 2010,
respectively, at a differing orbit. Table 4 shows the basic
information of these images.
We processed three pairs of interference images
taken from 2007 to 2008. The processing procedure in-
volves mainly seven stages: image focusing, geometry
calibration, generation of differential interferograms,
phase unwrapping, atmospheric effect assessment and
reduction, deformation map generation, and result ge-
ocoding from radar coordinates to Universal Traverse
Mercator. The final results shown in Fig. 7 indicate that it
is feasible to apply D-InSAR to monitor the deformation
of Guobu slope, irrespective of the fact that the slope is
steep. The selected four images have good coherence.
As shown in Figs.7a and b, the pairs of 20071209-
20080124 and 20080124-20080310 show a high coher-
ence, while the coherence in the final pair of 20080310-
20080610 (Fig. 7c) is relatively lower. The deformation
value at the platform (within the polygon) corresponds
to light green-light blue on the color bar. Therefore, we
estimated that the deformation at the slope platform was
quite small within the range of allowable error during
the period from 10 March 2008 to June 10 2008. Name-
ly, no obvious deformation occurred during this period
before the impoundment. It is noted that we also tried
previously to investigate earthquakes (M
et alii,
1993), volcanoes (M
et alii, 1995), land subsid-
ence (M
et alii, 1997; F
et alii, 1998;
et alii, 1998), and also landslide acitivities
et alii, 2004; s
et alii, 2005; C
& w
2006; r
& w
2008; C
et alii, 2010; Y
et alii, 2010a, 2010b).
Remote sensing images are the basis of monitor-
ing deformation. The investigation and selection of
image sources play a crucial role in interpreting the
images. It is understood that appropriate monitoring
time and spatial scales should be carefully combined
with existing SAR images to fit different types of sur-
face deformation monitoring.
We checked all the synthetic aperture radar (SAR)
images acquired by the European Earth Remote-Sens-
ing (ERS) satellites from 2003 to present, and found
that most of the images are ascending-orbit ones, and
no image covered Laxiwa dam site area until late
2007. However, the satellite images taken by the Ad-
vanced Land Observing Satellite (ALOS) of the Japan
Aerospace Exploration Agency (JAXA) from 2008
to present are available for the dam site area. These
ENVISAT images of ERS have a resolution of 30 m,
whereas the ALOS images have higher resolution.
The unipolar PALSAR images have a resolution of 10
m, the images taken by the AVNIR-2 sensor have a
resolution of 10 m, and those taken by the PRISM sen-
Tab. 1 - ALOS PALSAR Image list
Tab. 2 - 2007-2008 ALOS PALSAR Image Pairs
Tab. 3 - ALOS PRISM Image list
background image
International Conference Vajont 1963-2013. Thoughts and analyses after 50 years since the catastrophic landslide Padua, Italy - 8-10 October 2013
to use two pairs of images (20091214 and 20100129;
20091231 and 20100215) to examine the slope defor-
mation after the impoundment, but we failed to obtain
useful results, because (a) there are many shadows and
overlapping in the slope area along the riverside, and (b)
these two pairs of images showed very low coherence,
probably due to the fact that the deformation was simply
too large for the D-InSAR.
We analyzed high-resolution optical images to ver-
ify the results from D-InSAR analyses before the im-
poundment, and also to understand the slope deforma-
tion after the impoundment. It is well known that from
optical imagery slope deformation can also be estimat-
ed through image processing applied to multi-temporal
data. We used ALOS-PRISM images for the interpreta-
tion of slope deformation. Therefore, the phases of two
images were made to match each other at first. And then
by means of multi-bands synthesizing method, the pan-
chromatic-bands images in the later phase were put to
the blue channel, and the panchromatic-bands images
in the former phase were put to red and green channel.
Finally, false-color images were synthesized by using
these three-color channels on the basis of RGB prin-
ciple. The gray value of the area without deformation
would be close to it before synthesis. In contrast, the
area where deformation occurred will show blue or yel-
low color in different phases. By comparing and ana-
lyzing the changes in color, the area with deformation
can be identified and positioned accurately.
Using the ALOS PRISM images listed in Tab. 3,
we obtained three interpreting results as shown in Fig.
8, where Fig. 8a presents the result for the period of 14
November 2007 ~1 July 2008, Fig. 8b for the period of
1 July 2008 to 3 April 2009, and Fig. 8c for the period of
3 April 2009 - 22 May 2010. From Fig. 8a, it can be no-
ticed that on the crest part (the platform) of Guobu slope,
the gray value after synthesis is close to that before the
synthesis, indicating that no obvious deformation oc-
curred on this part during this period. The blue area on
the toe part of the slope (within the red cycle) may result
from the local failure accompanying the river incision.
From Fig. 8b, it can be noticed that on the platform
of Guobu slope, the gray value after synthesis is also
close to the one before synthesis. This indicates that the
slope deformation, if any, should be smaller than the
value that could be detected by one pixel. Because the
resolution of the images is 2.5 m, we conclude that the
deformation was less than 2.5 m during the period be-
fore and soon after the impoundment. The topographic
change indicated by the yellow color resulted from the
water level change due to impoundment. However, from
Fig. 8c it can be noticed that remarkable change occurred
during the period from 3 April 2009 to 22 May 2010.
Large blue areas appeared on the slope and platform
(marked by yellow circle), indicating that the whole
slope showed remarkable deformation. Comparing two
images of different phases in the same window enabled
us to capture a maximum deformation of about 3 pixels
(about 7.5 m) on the leading edge of the platform, di-
recting toward NW280º~300º. This result showed good
consistency with the ground-based monitoring results.
It is noted that the vertical deformation of the
landslide can also be identified by comparing the
Google Earth images shown in Fig. 2b (taken on Oc-
tober 4, 2010) and Fig. 3a (Taken on May 18, 2004).
Many fractures spreading along the slope outside the
initial limits of the settled block are also visible on the
recent Google Earth. The photos taken from the upper
cliff at different dates (Fig. 9) showed that the crest of
the landslide deformed greatly. The terrace that was
relatively flat before the impoundment became lumpy
due to the occurrence of wide cracks and great settle-
ment (Fig. 9a). The crest had been shaped later, but
new cracks can be easily identified from the photo
shown in Fig. 9b. Through these deformation occurred
on the crest, we may infer that the whole slope is not
deforming or sliding as an en-mass.
Fig. 7 - Interpreted deformation for different periods. (a)
For the period of 2007/12/9 to 2008/1/24; (b) for
the period of 2008/1/24 to 2008/3/10; (c) for the
period of 2008/3/10 to 2008/6/10
background image
Italian Journal of Engineering Geology and Environment - Book Series (6) © 2013 Sapienza Università
submerging part, and then lower the stability of the
whole slope. If the sliding surface is not in such a lower
location, the increased water level will have little, if any,
effect on the whole slope. Nevertheless, concerning our
reference, further investigation, such as monitoring the
displacement of submerging slope or other geophysical
survey, will be needed. It is also noted that although the
present study presents only some preliminary informa-
tion of the slope deformation, it may provide valuable
information for the plan of countermeasures to lower/
mitigate the risk of catastrophic landslide.
In this study, we used ALOS PALSAR and ALOS-
PRISM images to examine the deformation of Guobu
slope in differing time with the purpose of examining
the relationship of Guobu landsliding and impound-
ment. The results can be summarized as follows.
(1) The results of differential interferometry indicate
that the deformation of Guobu slope during this
period from November 2007 to June 2008 was
very small. The image interpreting results from
ALOS PRISM also shows that no obvious defor-
mation occurred on Guobu slope during the period
from November 2007 (before impoundment) till
By now, although the displacement of Guobu land-
slide had been monitored and the movement features
had been made clear, the sliding surface had been keep-
ing unclear. The possible failure mechanisms of this
slope are still in discussion. This impedes the conduc-
tion of proper countermeasures. There are differing
opinions on the location of sliding surface. One opinion
is that the sliding surface may be in the high position of
around 2700 m in Fig. 6b, another one is around 2550
m. The sliding surface shown in Fig. 6b is the third one,
namely, the worst case. However, from the features of
the terraced crest of Guobu landslide and also the moni-
tored displacement data, we inferred that this landslide
is a gravitational deep-seated landslide, and the sliding
surface should have reached the lowest part of the slope,
namely near the river bed, because (1) in granite area,
weathering in deeper location of the slope is normally
serious, which will results the frequent occurrence of
shallow landslides, and also large deep-seated land-
slides within the weakened rock massif, (2) reactivation
of an old landslide during the impoundment normally
results from the increase of buoyancy. This increasing
buoyancy will lower the effective normal stress and then
lower the shear resistance that could be provided by the
Fig. 8 - ALOS-PRISM image interpreting results for the period of 2007/12/14 to 2008/7/1 (a), 32008/7/1 to 2009/4/3 (b),
and2009/4/3 to 2010/5/22 (c), respectively
Fig. 9 - Deformed crest of the landslide at different dates
background image
International Conference Vajont 1963-2013. Thoughts and analyses after 50 years since the catastrophic landslide Padua, Italy - 8-10 October 2013
on the terrace crest of the slope and the continu-
ing deformation reveals that the slope may not be
sliding as a whole.
The research work described herein was funded
by the National Nature Science Foundation of China
under Grant No. 40772187. The financial supports are
gratefully acknowledged.
April 2009 (one month after the impoundment).
(2) Through comparing the ALOS PRISM images,
it is found that Guobu slope had obviously de-
formed (with the horizontal displacement up to
7.5 m) during the period from April 2009 (after
the impoundment) to May 2010.
(3) The response of sliding to the increasing water
lever suggests that the sliding surface may be
locating near the river bed. The cracks occurring
G, C
n, F
P, l
D, n
G, s
, t
D (2004) Ground-based SAR interferometry for
monitoring mass movements. Landslides, 1: 21-28.
r, r
, F
(2000) Synthetic aperture radar interferometry to measure Earth's surface topography and
its deformation. Annual Review of Earth and Planetary Sciences, 28: 169-209
n, C
F, v
, l
G (2010) Monitoring, prediction, and early warning using ground-based radar
interferometry. Landslides, 7: 291-301.
F., C
n., e
l., r
G. & M
G. (2005) - Landslide hazard and riskmapping at catchment scale in the
Arno River basin. Landslides, 2: 329-342.
C. & w
j. (2006) - Investigating landslides with space-borne Synthetic Aperture Radar (SAR) interferometry.
Engineering Geology, 88 (3-4): 173-199.
r. & G
M. (2005) - The 1963 Vaiont Landslide. Giornale di Geologia Applicata 1: 41-52. doi: 10.1474/
e.j., b
r.G. & G
r.M. (1998) - Rapid subsidence over oil fields measured by SAR interferometry.
Geophysical Research Letters, 25: 3215.
D.l., h
k.w., i
s.e., P
s.P., P
G., r
F. & r
P.a. (1998) - InSAR detection of
system compaction and land subsidence, Antelope Valley, Mojave Desert, California. Water Resources Res., 34: 2573-2585.
G.a. (1964) - Vaiont reservoir disaster. Civil Engineering, 34: 32-39.
D., r
M., C
C., a
F., P
G., F
k. & r
t. (1993) - The displacement field of the
Landers earthquake mapped by radar interferometry. Nature, 364: 138-142.
D., b
P. & a
a. (1995) - Deflation of Mount Etna monitored by spaceborne radar interferometry. Nature,
375: 567-570.
D., h
t. & v
h. (1997) - Land field, California, observed using SAR interferometry. Geophysical Research
Letters, 24: 901-904.
l. (1964) - The rock slide in the Vaiont Valley. Rock Mech Engng Geol, 2: 148-212.
b. & w
a. (2008) - InSAR processing for the recognition of landslides. Adv Geosci, 14: 189-194.
e. & G
M. (2000) - History of the 1963 Vaiont slide: the importance of geological factors. Bull Eng Geol Env,
59: 87-97.
l.P., v
C.j., C
r.P.k. & P
P. (2005) - Accuracy assessment of InSAR derived inputmaps for
landslide susceptibility analysis: a case study from the Swiss Alps. Landslides, 2: 221-228.
a.w. (1966) - Bedding-plane slip, residual strength and the Vaiont landslide. Geotechnique, 16: 82-84.
b. & F
C. (1982) - Frictional heat and strength loss in some rapid landslides. Geotechnique, 32: 43-54.
Y. (2004) - Major geological hazards and the prevention on relocation sites of the Three Gorges Reservoir, the Yangtze River.
Geological Publishing House, Beijing (in Chinese).
Y., z
w., l
Y., z
j. & l
x (2010) - Integration of GPS with InSAR to monitoring of the Jiaju landslide in Sichuan,
China. Landslides, 7: 359-365.
Y., w
h., G
Y. & l
x. (2010) - Real-time monitoring and early warning of landslides at relocated Wushan Town, the
Three Gorges Reservoir, China. Landslides, 7: 339-349.