10:45 AM - 11:00 AM
[SGD02-06] Mapping coseismic crustal deformation in eastern Taiwan by InSAR and GNSS to construct a fault model
Keywords:Taiwan, InSAR, GNSS, fault, earthquake, inversion
Taiwan is formed by the collision of the Philippine Sea plate and the Eurasian plate. The inter-plate convergence rate is ~7 cm/yr (Seno et al., 1993). The Longitudinal Valley Fault (LVF) in eastern Taiwan is part of the collision boundary where earthquakes occur frequently. On March 22, 2022 (UTC), an earthquake (Mw 6.4) occurred in eastern Taiwan and two more earthquakes (Mw 6.4 and 6.7) occurred in September 2022. The aftershock distribution and the geological structure indicate that the crustal deformation associated with the collision is extremely complex. To unravel this complexity, in this study, we detected crustal deformation caused by the March earthquake (Mw 6.4) using InSAR and GNSS data and then estimated a fault model.
The GNSS stations are very sparse in the mountainous area of eastern Taiwan due to high altitude and steep topography. In addition, InSAR analysis using short-wavelength satellite data is difficult to detect crustal deformation because the study area is covered with dense vegetation. In this study, we created five InSAR pairs using L-band SAR data acquired by ALOS-2. The GNSS data are processed at National Central University using GIPSY-X.
The L-band SAR data are strongly affected by ionospheric disturbances. To remove it, two correction methods for ionospheric disturbances were used: (1) approximate the long wavelength phase changes by a polynomial function and subtract it, (2) the Split Spectrum Method (SSM) (Bricic et al., 2010; Rosen et al., 2010; Gomba et al., 2016), which extracts the effects of ionospheric disturbances by utilizing the dispersion nature of the ionosphere. We examined the validity of the above-mentioned methods by comparing the corrected InSAR images with GNSS data. The results show that the polynomial method is effective when the phase change associated with ionospheric disturbances is spatially smooth. Although the accuracy of the SSM depends on the radar bandwidth, on the other hand, it is effective even when the phase change due to the ionospheric disturbance shows a complex spatial pattern. The InSAR images successfully corrected indicate a short wave-length displacement pattern associated with fault movement. Using the above approaches, even in steep mountainous regions, the phase changes associated with fault movement and noise were distinguished.
To explain the displacement field given by thus corrected interferograms and GNSS data, we estimated the slip distribution on the fault plane using inversion analysis. We used a Green's function for elastic half-space (Okada, 1985). First, the fault geometry is estimated with a uniform-slip model by a grid-search method, which shows the best model dips 70 degrees down to the southeast. Next, we estimate the slip distributions on the fault plane which best explains the InSAR and GNSS data. The estimated slip vectors mostly consist of reverse slip components. Our fault model brings the following additional information: (1) The earthquake occurred at a depth range of 10–30 km below the LVF, and the maximum slip on the fault plane was about 40 cm, (2) The seismic moment of fault motion estimated from InSAR and GNSS is 1.42 × 1019 Nm, which is larger than the seismic moment estimated from seismic waves (5.73 × 1018 Nm) by Central Weather Bureau, indicating that aseismic slip likely occurred.
Our result is consistent with the previous report that the LVF is a southeast-dipping reverse fault (Shyu et al., 2005), suggesting that the March earthquake contributed to the uplift of the Coastal Range located east of the fault. To clarify the details of the deformation pattern and subsurface structure around the collision boundary, we are applying the above analysis flow to two earthquakes (Mw 6.4 and 6.7) that occurred in September 2022 around the LVF.
The GNSS stations are very sparse in the mountainous area of eastern Taiwan due to high altitude and steep topography. In addition, InSAR analysis using short-wavelength satellite data is difficult to detect crustal deformation because the study area is covered with dense vegetation. In this study, we created five InSAR pairs using L-band SAR data acquired by ALOS-2. The GNSS data are processed at National Central University using GIPSY-X.
The L-band SAR data are strongly affected by ionospheric disturbances. To remove it, two correction methods for ionospheric disturbances were used: (1) approximate the long wavelength phase changes by a polynomial function and subtract it, (2) the Split Spectrum Method (SSM) (Bricic et al., 2010; Rosen et al., 2010; Gomba et al., 2016), which extracts the effects of ionospheric disturbances by utilizing the dispersion nature of the ionosphere. We examined the validity of the above-mentioned methods by comparing the corrected InSAR images with GNSS data. The results show that the polynomial method is effective when the phase change associated with ionospheric disturbances is spatially smooth. Although the accuracy of the SSM depends on the radar bandwidth, on the other hand, it is effective even when the phase change due to the ionospheric disturbance shows a complex spatial pattern. The InSAR images successfully corrected indicate a short wave-length displacement pattern associated with fault movement. Using the above approaches, even in steep mountainous regions, the phase changes associated with fault movement and noise were distinguished.
To explain the displacement field given by thus corrected interferograms and GNSS data, we estimated the slip distribution on the fault plane using inversion analysis. We used a Green's function for elastic half-space (Okada, 1985). First, the fault geometry is estimated with a uniform-slip model by a grid-search method, which shows the best model dips 70 degrees down to the southeast. Next, we estimate the slip distributions on the fault plane which best explains the InSAR and GNSS data. The estimated slip vectors mostly consist of reverse slip components. Our fault model brings the following additional information: (1) The earthquake occurred at a depth range of 10–30 km below the LVF, and the maximum slip on the fault plane was about 40 cm, (2) The seismic moment of fault motion estimated from InSAR and GNSS is 1.42 × 1019 Nm, which is larger than the seismic moment estimated from seismic waves (5.73 × 1018 Nm) by Central Weather Bureau, indicating that aseismic slip likely occurred.
Our result is consistent with the previous report that the LVF is a southeast-dipping reverse fault (Shyu et al., 2005), suggesting that the March earthquake contributed to the uplift of the Coastal Range located east of the fault. To clarify the details of the deformation pattern and subsurface structure around the collision boundary, we are applying the above analysis flow to two earthquakes (Mw 6.4 and 6.7) that occurred in September 2022 around the LVF.