This article was downloaded by: [Florida Atlantic University]

On: 16 November 2014, At: 11:14

Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer
House, 37-41 Mortimer Street, London W1T 3JH, UK

Geographic Information Sciences: A Journal of the

Association of Chinese Professionals in Geographic
Information Systems
Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/tagi19

Preliminary Results of Satellite Radar Differential

Interferometry for the Co-seismic Deformation of the
12 May 2008 Ms8.0 Wenchuan Earthquake
a a a a a a
Linlin Ge , Kui Zhang , Alex Ng , Yusen Dong , Hsing-chung Chang & Chris Rizos
a
Cooperative Research Centre for Spatial Information School of Surveying & Spatial
Information Systems, The University of New South Wales , Sydney , NSW , 2052 ,
AUSTRALIA
Published online: 18 Aug 2009.

To cite this article: Linlin Ge , Kui Zhang , Alex Ng , Yusen Dong , Hsing-chung Chang & Chris Rizos (2008) Preliminary
Results of Satellite Radar Differential Interferometry for the Co-seismic Deformation of the 12 May 2008 Ms8.0 Wenchuan
Earthquake, Geographic Information Sciences: A Journal of the Association of Chinese Professionals in Geographic
Information Systems, 14:1, 12-19, DOI: 10.1080/10824000809480634

To link to this article: http://dx.doi.org/10.1080/10824000809480634

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of
the Content. Any opinions and views expressed in this publication are the opinions and views of the authors,
and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied
upon and should be independently verified with primary sources of information. Taylor and Francis shall
not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other
liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or
arising out of the use of the Content.

This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions
 Linlin Ge el al.: Preliminary Results of Satellite Radar Differential Interferometry for the Co-seismic Deformation
-2 of the 12 May 2008 Ms8.O Wenchuan Earthquake

Preliminary Results of Satellite Radar Differential Interferometry for the

Co-seismic Deformation of the 12 May 2008 Ms8.0 Wenchuan Earthquake
Linlin Ge, Kui Zhang, Alex Ng, Yusen Dong, Hsing-chung Chang, Chris Rizos

Cooperative Research Cenlre for Spatial Information School of Survoying & Spatial Information Systems
Tho University of New Sourh Wales Sydney, NSW 2052 AUSTRALIA
E-mail: I.ge@unsw.edu.au

Abstract
Satellite differentialS A R interferometryhas been widely accepted as a powerful twl to map co-, post- and inter-seismic defonnation
since its successful application to the 1992 Landers Ewhquake. As soon as the Ms8 0 Wenchuan Earthquake occurred on 12 May
2008 in the Sichuan Province of southwestem China, the Japan Aerospace Exploration Agency tasked its Advanced Land Observing
Satellite (ALOS) to respond to the disaster by collecting images. This paper presents the preliminary DInSAR results of co-seisnuc
deformation of the quake observed from two satellite paths of the onboard ALOSPALSAR sensor with post-seismic images acquired
on 19 and 24 May. Results from pixel offsetanalysis and difference of coherence will also be discussed. The radar mapping is still on-
going because the ruptured seismic fault is more than 300km in length. Each swath of the PALSAR fme beam covers only about a
75km segment of the fault, and it takes 46 days for ALOS to revisit the sene site.
Downloaded by [Florida Atlantic University] at 11:14 16 November 2014

Keywords
Siclinan Earthquake, SAR, ALOSPALSAR, co-seismic deformatton

I. IN"RODUCI'1ON

The devastating Ms8.0 Wenchuan Earthquake that struck Data relay satellite
China's southwestern Sichuan province occurred at 14:28:01. communication antenna
42 UTC on 12 May 2008 in an area that is defoiming as a result
of the collision between two tectonic plates, the Indo-
Australian plate and the Eurasian plate. As of 7 June 2008, it is
Solar may paddle
estimated that 69,134people weie killed as a direct result of
the quake(CEA, 2008). Since records were kept, there have
been eight major earthquakes with magnitude larger than 7.0
within 200km of the cuiTent epicentre, with the Ms 7.5 in 1933 nvm1R-2
being the largest (CSIN,2008).

ALOS,the Advanced Land Observing Satellite, is a followon

J1
Velocity
Nadir

Japanese satellite mission to the JEW-1 and the Advanced Earth Figure 1. The ALOS satellite (EORC JAXA, 2 W )
Observing Satellite (ADEOS) missions. ALOS was launched on
2d January 2006 from the Tanegashima Space Centre by the Japan with a high accuiacy.
Aerospace Exploration Agency. The satellite is designed to
provide high quality, low-cost Earth observation data for terrain PALSAR provides better performance than the JERS-1's
mapping, disaster monitoring and climate change studies. ALOS synthetic aperture radar (SAR). PALSAR operates in five
bas t h m remote sensing instruments, namely, the Panchromatic observation modes: Fine Beam Single mS),Fine Beam Dual
Remote-sensing Instrument for Stereo Mapping (PRISM) for (FBD), Direct Transmission (DT), ScanSAR, and polarimetry
digitalelevation mapping, the Advanced Visible andNwIufrared (POL). It has finer spatialresolution (down to 10 metres in FBS
Radiometer type2 (AVNIR-2) for precise land coverage mode) than JERS-I, with variable look angles.
observation, and the Phased Array type L-hand Synthetic
Aperture Radar (PALSAR) for day-and-night and all-weather In ScanSAR mode it is possible to acqum up to a 250 to 350
land observation (Figure 1). kilometres wide swath of SAR, but at the cost of spatial
resolution. This swath is three to five times wider thaii
The ALOS was designed with two advanced technologies: conventional SAR images. PALSAR can provide global
the first is the high speed and large data capture capacity, and coverage t h e e times per year in FBS/FBD mode and once a
the second is the accurate spacecraft position and attitude year in ScanSAR mode (EORC JAXA, 2006).
determination capability - using an onhoard GPS receiver and
high precision star trackers to determine position and attitude For InSAR applications, the coherence is an important
1082-4006/08/14(01)-12$5.00
02008 The International Association of Chinese Professionals
in Geographic Information Science (CPGIS)
 Geographic Information Sciences Wt, 14, No. 1, June 2008 13
parameter quantifying the accuracy of the interferometric The aim of this study was to investigate the co-seismic
measurements. The coherence can be affected hy many factors, deformation due to the most recent Wenchuan quake by
including geometric decorrelation, Doppler Centroid difference satellite radar interferometry using ALOSPALSAR data.
between the SAR images, impacts of volume scattering,
temperature of sensors during data acquisition, temporal
decorrelation, and so on (Zhang, Prinet, 2004). Unlike the C- 11. RESULTS FROM DIFFERENTIAL SAR
band SAR sensor, the L-band PALSAR sensor is suitable for INTERFEROMETRY
observations in vegetated areas. Temporal decorrelation can
occur if there is a difference in the terrain (e.g. the trees lose In order to map co-seismic deformation, PALSAR images for
their leaves, or snow covers the ground) between the two each path acquired before the quake must exist. As shown in
acquisitions, which can he caused by the change in weather Figure 2, the full Wenchuan fault is so long that it has to be
or season. Low frequency radars with a longer wavelength covered with ALOSlPALSARacquisitionsfrom seven adjacent
(e.g. L-band S A R ) are usually more sensitive to larger objects pathsltracks. In the figure the SRTM digital elevation model
than high frequency radars (e.g. C-hand SAR).Since it is easier (Fan& Kobrick, 2000) was plotted as background. The location
for environmental effects to changdmove smaller objects, the of the main quake is indicated by the star.
temporal decorrelation is expected to be a stronger affect for
radars with shorter wavelength than radars with long Fortunately, this indeed is the case so that as soon as a post-
wavelength (Ge, et al., 2007a; Ge,et al., 2004). Because of these seismic PALSAR image was acquired, co-seismic deformation
factors, ALOS is expected to produce much more coherent can be derived in near real-time (i.e. within 24 hours of image
Downloaded by [Florida Atlantic University] at 11:14 16 November 2014

S A R images than the JERS-1 and other previous C-band SAR capture) from the pair formed with the pre- and post-seismic
satellites, which therefore makes it more suitable for DInSAR SAR images. The first post-seismic PALSAR acquisition
applications. occurred on 19 May 2008 from Path 473. At the time this paper
was prepared, a second acquisition happened on 24 May 2008
S A R has the remarkable ability of imaging the Earth's surface from Path 476. Path 476 covers the epicentre, while Path 473
day or night, and even when the surface is cloud-covered. covers the worst-hit Beichuan City. This paper will focus on
This is particirlarly intporlarztfor monitorirrg the Weiiclirimt results generated from the co-seismic pairs from both paths.
quake because rains followed the mairr shock otrd rnost
quake-stricken regions were covered by clouds. The Figure 3 shows the DInSAR results from ALOSlPALSAR
advantage of monitoring crustal deformation using a S A R - co-seismic pairs from Paths 473 (17 February - 19 May) and
based remote sensing system is its large spatial coverage and 476 (8 April - 24 May). It can be seen clearly that the quake was
cost effectiveness. In particular it is possible to investigate so powerful, and hence the devastation was so widespread,
areas affected by large-scale seismic events. that the SAR coverage appeared to he limited and several
images from the same path are needed in order to include
Satellitedifferential SAR interfernmetryhas been widely accepted areas of least impact, which can he used as reference for phase
as a powerful tool to map co-, post- and inter-seismic deformation unwrapping. For example, from the result it can he seen that
since its successful applications to the 1992Landers Eaahquake the ground surface at the bottom right comer of Path473 was
(e.g. MassOMet, et al., 1993: Massonnet, et al., 1994;MassoMet, displaced the least but a few more frames of images further
et al., 1996; Jonsson, et al., 2002, Jonsson, et al., 2003; Fialko, et down are needed in order to find a stable reference point.
al., 2005; Fialko, 2006,Wang, et al., 2007).
Also obvious is the much better coherence n the Sichnan
A few studies using PALSAR data for co-seismic deformation Basin at the bottom right half than in the mountainous region
monitoring have been reported since the launch of ALOS. at the top left half. By zooming in on the top left portion of the
Matsuoka (2006) demonstratedALOS's ability to monitor areas interferograms, it is evident that topographic residuals exist in
damaged by earthquake activity by comparing the correlation the DInSAR results because of the poor quality of the SRTM
coefficient of a pre-event images pair and a pair of images 3 arcsecond DEM in the rough terrain (Chang, et al., 2004).
spanning the seismic event (Matsuoka, 2006; Ge and Yonemwa, These results can be improved significantly by using a more
2004). The results correlate well with the damage assessment accurate and higher resolution DEM equivalent to that of the
reports from field survey and high-resolution optical satellite PALSAR image, i.e. 1Om.
data. Results of co-seismic deformation from InSAR
measurements using PALSAR data have also been reported. Note also that the low spatial frequency fringes in the lower
PALSAR data acquired before and after an earthquake can be part of Path 473 interferogram. They are so similar to orhit
used to detect the co-seismic deformation using the DInSAR fringes that only when DInSAR results from neighhouring
technique. Ground surveys have been carried out to verify paths, i.e. 472 and 474, become available can they be ascertahed
the results. The measured deformations are compared to those as true deformation.
induced from a fault model and they are highly correlated
(Miyagi, et al., 2W7). On the other hand, the terrain was deformed so much along
the fault, where the aftershocks are located, that fringes were
 1
Downloaded by [Florida Atlantic University] at 11:14 16 November 2014

Figure 3. DInSAR results for Paths 473 and 476 overlaid on Landsat image
 Geographic Information Sciences Vol. 14, No. I, June 2008 15
totally lost. This signal saturation (Ge, et al., 2007h) would sampling of the correlation function between the images is
lead to co-seismic deformation beiig underestimated, as shown done in order achieve sub-pixel level accuracy (better than
in Figure 4. 0.2 pixels). Aquadratic polynomial is then used to remove trends
in the estimated offsets in order to obtain the range/azimuth
Figure 4 shows RELATIVE ground movement iu the radar- residuals. These residuals are then converted to ground surface
looking direction for each pathfiomunwrappinginterferograms displacement in both azimuth and range directions.
of Paths 473 and 476. From this figure it is estimated that the
ground surface displaced a maximum of 75cm and 130cm away Figure 5 shows the results of plxel offset analysis for azimuth
from and towards the satellite respectively. But preliminary (left) and range (right). Companng Figure 5(b) with Figure 4(b)
modelling results from seismic observations indicate around it can be seen that the hvo agree well with each other in general,
5m displacement (Qiao, 2008). Therefore, other techniques such hut pixel offset analysis reveals much larger ground
as pixel offset analysis have to he applied in order to resolve displacement around the epicentre of the Ms 8.0 main shock,
large displacement of the older of a few metres along the fault, where the DInSAR signal is saturated. Hence pixel offset
while DInSAR is used as a highly accurate technique (-cm analysis can be used as a technique in cases of large dynamic
level) with limited dynamic range (-1-2m) (Ge et al., 2007a). range to complement DInSAR results.

On the other hand, there rue difficulties to fully explain the

III. RESULTS FROM PIXEL OFFSET ANALYSIS azimuth result. One possible explanation is that differential
atmospheric delay has caused the pattern (Janssen, et al., 2w4;
Downloaded by [Florida Atlantic University] at 11:14 16 November 2014

In order to measure large ground displacement along the fault, Xu,et al., 2006). Therefore, in Figute 6 only range offset results
pixel offset analysis (Wang, et al., 2007; Tohita, et al., 2001) for both paths were collocated.
was carried out in order to calculate ground surface movement
in both the azimuth and range directions. Satellite radar scans
an area of interest in both the satellite flight direction ( i s . IV. RESULTS FROM COHERENCE DIFFERENCE
azimuth direction) and the radar looking dnection (i.e. slant
range direction). The range and azimuth offsets are first Coherence map is a by-product of the radar interferometry
estimated using the intensity cross-correlation method for and measures the similarity between two radar images. In
each grid of approximately 16 and 32 pixels respectively. Over- general, the coherence between the two images taken hefore
 Linlm Oe e l al.: Preliminary Results of Solellite Radar Differential Inteiierometry for the Co-seismic Defonuation
Downloaded by [Florida Atlantic University] at 11:14 16 November 2014

Figure 5. : direction
 Geoeraohie Information Sciences Vol. 14. No. 1. June 2008 17
the quake, namely 2 January and 17 February acquisitions for
Path 473, is high. On the other hand, the coherence between
one image taken before the quake, 17 February acquisition, V. CONCLUDING REMAKES AND FUTURE WORK
and one taken after the quake, i.e. 19 May acquisition, is
reduced in damaged areas and remains high in undamaged Preliminary results of satellite radar differential interferometry
regions. Therefore, the difference between the two coherence for the co-seismic deformation of the 12 May 2008 Ms8.0
maps, namely (2 January and 17February+(l7 February and Wenchuan Fmihquake were denved from ALOSPALSAR data
19 May), represents the level of quake-induced damage for Paths 473 and 476, with two images acquired before the
(Matsuoka, 2006;Ge, Yonezawa, 2004). quake and one after the event for each path. Three products
were described in this paper. Among them, the DInSAR result
Figure 7 shows the quake damage assessment based on the is highly accurate but with limited dynamic iange. Pixel offset
difference of DInSAR coherence for Path 473. In the figure, analysis can be used as a technique in cases such as this
0.7 represents severe damage (hence rescue is most needed); where there is large dynamic range, to complement DInSAR.
0.0 means no damage (location can be used as safe island); The difference ofcoherence between pre- and co-seismic pairs,
and 4 . 4 means rubble accumulated in the location (indicating a by-product of DInSAR, can be used for quantitative quake
possible blockage of roads and river channels). Figure 8 damage assessment.
presents a similar result for Path 476.
Downloaded by [Florida Atlantic University] at 11:14 16 November 2014

InsAR Cohm

4.4
I

1M?WE tO3WD.E

Figure 7. Coherence difference between pre- and co-seismic pairs for Path 413
 Linlin Ge et al.: Preliminary Results of Satellite Radar Differential Interferometry far the Ca-seismic Deformation
.8 of the 12 May 2008 Ms8.0 Wenohum Earthquake
iwlm IWWW 101WT 1 M W E
Downloaded by [Florida Atlantic University] at 11:14 16 November 2014

Figure 8. Coherence difference between pre- and co-seismic pairs for Path 476

In addition to refining the preliminary results reported here,

future work will include: We wish to thank the Earth Remote Sensing Data Analysis
Mapping the whole seismic fault; Center (ERSDAC) for providing ALOS PALSAR data. We also
Inverse modelling of the cc-seismic deformation h m bath acknowledge the strong support from the International
DInSAR and pixel offset analysis, and comparing these Association of Geodesy Sub-Commission 4.4 "'Applications
with results from seismic observation; of Satellite & Airborne Imaging Systems" and the IAG
Comparing SAR observed co-seismic deformation with Consortium for Mine Subsidence Monitoring.
ground survey results; and
Correlating coherence difference with damage assessment MET1 and JAXA have the ownership of the ALOS PALSAR
from survey on the ground. original data. ThePALSARLevel-1.1 pioducts were produced
and provided to the CRC-SI/UNSW by ERSDAC, Japan.

The authors wish to thank Prof Hui Lin of the Chinese

University of Hong Kong for providing the opportunity to
This research work has been supported by the Cooperative publish the preliminary results in such a timely manner.
Research Centre fur Spatial Information through Project 4.09,
whose activities are funded hy the Australian Commonwealth's
Cooperative Research Centres Programme. The Australian R
EEE
Rmm
Research Council and the Australian Coal Association
Research Program have been funding radar related studies of 111 Chang H C., L. Ge, C. Rizos, 2004. Assessment of digitaI
the team in the last few years. elevation models using RTK GPS. Joirmol of CeospotiolEiig, 6
 Geographic Informalion Sciences Val. 14. No. 1, June 2008 19
~~

(1): 13-20. [14] Jonsson S., P. Segall, R. Pedersen, G. Bjomssoo, 2003, Post-
[2] China Earthquake Administration (CEA), 2008. http:/lww.cea. earthquake ground movements correlated to pore-pressure
go~.cn:99/manage/htd8a8520balaWb8fW 1laWcdf697000.V transients. Notiwe. 424(694S): 179-183.
index.html, last accessed on 10/06/2008. [I51 Mmonnet D., el al., 1993, The displaceoient field of the Landers
[3] China Seismic Infomiation Network (CSEV)),2008. http://www. earthquake mapped by radar interferometry. Nntrire, 364(6433):
csi.ac.cnlsichuanlsichuanO805lZ_his.htm, last accessed on 10/ 138-142.
06/2008. 1161 Massonnet D., K. Feigl. M. Rossi, F. Adragna, 1994, Radar
[4] EORC JAXA, 2006, About ALOS, EORC JAXA, http://www. interferonietric mapping of deformation in the year after the
eorc,jaxa.jp/ALOS/about/about~index.litm, last accessed on I01 Landers eanhquake. Nafror,369(6477): 227-230.
1212007. 1171 Massonnet D., Thatcher W., Vadon, H., 1996, Detection of
IS] Fan, T., M. Kobrick, 2MM. Shuttle Rada Topography Mission postseismic fault-zone collapse following the Landers
produces a wealth of data. AGUEos 81: 583-585. earthquake. Nature, 382(6592): 6 1 2 4 1 6 .
161 Fialko Y., 2006, Interseismic strain accumulation and the [ I S ] Matsuoka, M., 2006, Use of ALOSRALSAR imagery for
exlhquake potential on the southern San Andrea fault system. monitoring m a s damaged due torecent nahlml disasters, disaster
Nfltirr-e,JJl(7096): 968-97 I . forew.aming diagnostic iiiethods and management. Proc, of SPIE,
171 Fialko Y., D. Sandwell. M. Sinions, P. Rosen, 2005, Three- VoI.6412,ID641204, 2006.11.
dimensional deformation caused by the Bam. Iran, earthquake [I91 Miyagi Y., Y. Nishimura, H. Takahashi, M. Shiniada, 2007,
and the origin of shallow slip deficit. Nflhrre,435(7040): 295- Cnistal deformation caused by earthquake detected by InSAR
299. technique using ALOSIPALSAR data. Eos Trmis. AGU. 88(52):
181 Ge L., C. Yonezawa, 2004, Radar interferometry for disater Fall Meet. Suppl.
prevention and mapping, pres. liileninrio~~nl Workshop 011 Earth [20] Qiao X.. 2008, Personal communications.
Downloaded by [Florida Atlantic University] at 11:14 16 November 2014

Obsewntioii Tecliriologg arid Applicntio~r(CEOS 2004). (211 Tobita M., M. Murakami H. Nakagawa H., et al., 2001, 3-D
November 16-17, Beijing, China (invited presentation). surface deformation of the 2000 Usu eruption measured by
[Y] Ge L.. H C. Chang, C. Rizos, 2W7a. Mine subsidence monitoring matching of SAR images. Groplrgsicnl Research Letter; 28(22):
using multi-source satellite SAR images. Phntograriitiretr-ic 42914294.
Engineering and Reirrore Serrsing, 73(3): 259-266. [22] Wang H., L. Ge, C. Xu. 2W7. 3-D coseisiilic displacement field
[lo] Ge L.. H. Wang, H C. Chmg, 2007b. Linear combination for of the 2005 Kashniir earthquake inferred from satellite radar
differential radar interferometry. The X X N General Asseiribly imagery. Endi Plaiiers & Space, S9(5): 343-349.
of tire I,iteriinriorial Unioii of Geodesy nrtd Geopli.ssicr (IUGG [23] Wang H., C. Xu, L. Ge, 2007, Coseismic deformation and slip
2007). Perugia. Italy, 2-13 July. distribution of the 1997 mw7.S Manyi, Tibet, earthquake from
[I I] Ge L., X. Li, C. Rizos. M. Omum, 2004, GPS and GIS assisted InSAR measurements. Joiw~rolof Geodyiaiiiics, 44(3-5):
m d a interferometry. PliofograerrnetricE?rgirieeringaiidReiirore 200-212.
Sensing, 70(10): 1173-1 178. [24] Xu C., H. Wang, L. Ge, C. Yonezawa, P. Cheng, 2006, InSAR
[I21 Janssen V.,L. Ge, C. Rizos, 2004, Tropospheric corrections to tropospheric delay mitigation by GPS observations: A c a e study
SAR interferometry from GPS obsemations. GPS Sohrlioris, 8 in Tokyo area. Jnrrrnal of Ahrrosplrrric a i d Solnr-Ter-rrstrirrl
(3): 140-151. Physics, 68(6): 6 2 9 4 3 8 .
[I31 J6nsson S., H. Zebker, P. Segall, 2002. Fault slip distribution of [25] Zhang Y., V . Prinet, 2004, InSAR coherence estimation.
the 1999 Mw7.1 Hector Mine, California, earthquake, estimated Geosrierrce arid Reviole Sensing Sy,ripnsirriii (IGARSS 2004
from satellite radar and GPS measurements. Balletin of the Prnceedipigs), September 2 k 2 4 , 5: 3353-3355.
Seisrriological Sociely nfhrerim, 92(4): 1377-1 389.