Coseismic Deformation of Chi-Chi Earthquake as Detected by Differential Synthetic Aperture Radar Interferometry and GPS Data

517

Terr. Atmos. Ocean. Sci., Vol. 17, No. 3, 517-532, September 2006

Coseismic Deformation of Chi-Chi Earthquake as Detected by

Differential Synthetic Aperture Radar Interferometry and GPS Data

Chia-Sheng Hsieh1 and Tian-Yuan Shih1,

*

(Manuscript received 24 March 2006, in final form 30 June 2006)

ABSTRACT

1

Department of Civil Engineering, National Chiao-Tung University, Hsinchu, Taiwan, ROC

*

A rupture in the Chelungpu fault caused an M_w 7.6 earthquake on 21 September 1999 near Chi-Chi in central Taiwan. This earthquake was the most destructive experienced in Taiwan for the past century along this fault. In this study, we examined the earthquake-induced surface deformation pattern using differential synthetic aperture radar interferometry (D-InSAR) combined with global positioning system (GPS) data regarding the foot-wall of the Chelungpu fault. Six synthetic aperture radar (SAR) scenes, approximately 100 × 100 km each, recorded by the European Remote Sens-ing Satellite 2 (ERS-2), spannSens-ing the rupture area, were selected for study. The data were used to generate a high-resolution, wide-area map of dis-placements in flat or semi-flat areas. Interferograms show radar line con-tours indicating line-of-sight (LOS) changes corresponding to surface dis-placements caused by earthquake ruptures. These results were compared to synthetic interferograms generated from GPS data. Displacements shown by GPS data were interpolated onto wide-area maps and transformed to coincide with the radar LOS direction. The resulting coseismic displace-ment contour map showed a lobed pattern consistent with the precise GPS-based displacement field. Highly accurate vertical displacement was deter-mined using D-InSAR data using the coordinate transform method, while GPS data was effective in showing the horizontal component. Thus, this study confirmed the effectiveness of the D-InSAR method for determining the coseismic deformation caused by the Chi-Chi earthquake at the foot-wall of the Chelungpu fault.

Terr. Atmos. Ocean. Sci., Vol. 17, No. 3, September 2006

518

1. INTRODUCTION

Synthetic Aperture Radar Interferometry combines complex images recorded by anten-nas at different locations or at different times. The resulting interferograms permit the deter-mination of differences in the range direction. This technique has been used widely to produce highly accurate digital elevation model (DEM) and topographic maps (Zebker and Goldstein 1986; Madsen et al. 1993) and to measure displacement fields or terrain motions (Massonnet et al. 1993; Zebker et al. 1994). This technique also has been applied to ocean current measurements, hazard mapping and glacier motion detection (Massonnet and Feigl 1998; Rosen et al. 2000).

A radar sensor above the Earth detects tiny changes on the ground by accurately measur-ing changes in the time delay, or phase, of a radar echo. D-InSAR and its spatially dense, accurate deformation measurements have advanced studies of the Earth’s crust. Importantly, this technique provides a comprehensive view of detectable motion for the entire area affected. Results from such analyses can supplement ground-based measurements taken at a limited number of locations. Because radar satellites make global observations, we can study defor-mation processes anywhere at little additional expense once a satellite is in operation. The great advantage of the radar interferometer system is that it allows deformation measurements at very fine spatial spacings, creating a visual image of the deformation distribution (Zebker 2000).

A disastrous M_w7.6 earthquake struck central Taiwan at 01:47 (local time) on 21 Sep-tember 1999. The Seismology Center of the Central Weather Bureau located the epicenter near the town of Chi-Chi (23.85°N, 120.78°E) in Nantou County. To monitor fault activity, it is important to characterize the crustal deformation caused by this quake. Following the earthquake, measurements were taken at 99 GPS stations. These measurements suggested horizontal displacements of 0.1 to 8.5 m at the sites (Chang 2000). Yang et al. (2000) studied the three-dimensional displacements of 285 geodetic control stations using GPS observations collected before and after the earthquake. Yu et al. (2001) applied annually repeated GPS data from 1992 to 1999 to predict the crustal deformation rate; data within 3 months after the main shock were used to estimate the coseismic displacements.

In addition, Suga et al. (2001) used D-InSAR data to analyze deformation caused by the Chi-Chi earthquake. The land displacement pattern extracted from D-InSAR data was gener-ally consistent with results of a GPS survey conducted by researchers Yang et al. (2000). Pathier et al. (2003) corrected interferograms using GPS data to create a precise map of the D-InSAR component of coseismic displacement in the LOS direction. Chang et al. (2004) and Liu et al. (2004) applied the interferometric method to detect earthquake displacement. The interfero-metric results were precisely examined using GPS data in the LOS direction.

Major earthquake damage results from the vertical and horizontal components of fault motion. The advantage of the D-InSAR technique is that it can be used to characterize coseismic deformation over a larger area than that covered by GPS surveys. However, the inverse trans-formation from LOS displacement to vertical and horizontal displacement is not unique. In this study, we attempted to obtain the vertical displacement component using the coordinate transform method and the horizontal component using GPS horizontal data. The results

con-Chia-Sheng Hsieh & Tian-Yuan Shih

519

2. GEOLOGICAL BACKGROUND

The Chi-Chi earthquake was caused by a rupture of the Chelungpu fault. This earthquake, the most destructive in Taiwan in a century, struck the west-central island on 21 September 1999 (Kao and Chen 2000). For approximately 90 km along the fault, extensive surface rup-tures with vertical thrust and left lateral strikes-slip offsets occurred (Angelier et al. 2001; Chen et al. 2001; Lee et al. 2002). The vertical offsets averaged approximately 2 m along the southern half of the fault and about 4 m along the northern segment (Yang et al. 2000; Yu et al. 2001). Horizontal offsets of 5 to 8 m occurred along the northern part of the major fault (Yang et al. 2000; Yu et al. 2001). Tens of thousands of buildings collapsed, resulting in 10000 injuries and over 2300 fatalities.

Central Taiwan has several series of N-S trending and east-dipping thrust faults located in the Western Foothills; from east to west, they are Shuangtung fault, Chelungpu fault and Changhua fault, respectively (Fig. 1). The major rupture of the Chelungpu fault is near N-S trending that follows the boundary between the Western Foothills and the Taichung piggy-back basin (Chen et al. 2001). The earthquake fault of the Chi-Chi earthquake has been attrib-uted to the Chelungpu fault in the western part; however, fault branches east of the Fengyuan area trending nearly NE-SW were also suggested to form a Cholan segment (Chen et al. 2001). For simplification, we consider this segment as part of the Chelungpu fault.

The original boundary of the foreland basin was the Shuangtung fault, but due to continu-ing northwestward collision from the Philippine Sea plate, the foreland basin moved westward, causing the Chelungpu fault and the Changhua fault to develop in sequence. Consequently, the Deformation Front in the central Taiwan is now located in the Changhua fault, west of the Chelungpu fault.

The Chelungpu fault extends along the western front of the Taiwan Mountain belt in central Taiwan. The footwall of the Chelungpu fault is almost flat or semi-flat, including some urban areas. Thus, these areas are suitable for D-InSAR techniques. In contrast, the hanging wall of the Chelungpu fault lies in a steep mountainous area covered by dense vegetation.

3. METHODS

3.1 SAR Interferometry

When two radar scans have the same viewing angle but different viewing times, a small change in the target position can create a detectable change in the phase of the reflected signal. The phase is exactly proportional to the measured time delay and effective path length of the signal. The path differences of two signals can be determined to sub-wavelength accuracy by observing the phase differences of echoes. Phase differences may be caused by topographic