Cross-sensor SAR image offsets for deriving coseismic displacements: Application to the 2001 Bhuj (India) earthquake using ERS and Envisat data

Abstract:

Synthetic Aperture Radar (SAR) is a powerful imaging technique for measuring ground deformation, either through Interferometric SAR (InSAR) or image offset tracking. However, these methods are only applied to SAR images acquired by the same satellite, which limits the measurement capability for many earthquakes. Here we propose a novel approach that allows for calculating offsets between images acquired from the European ERS and Envisat satellites. To achieve this cross-sensor offset calculation, we first coregister pre-event (ERS) and post-event (Envisat) SAR images separately to generate averaged pre- and post-event SAR amplitude maps. We then compute the orbital offsets between these two maps in order to resample the ERS average map onto the grid of the averaged Envisat image. We finally calculate the cross-sensor image offsets based on cross-correlating selected sub-images distributed throughout the coregistered averaged SAR maps. Application to the 2001 Bhuj earthquake reveals, for the first time, its near-field coseismic displacement field right above the epicenter. We compare our measurements with the surface displacement field predicted from the published source model of Copley et al. [2011]. This model was derived from tele-seismic waveforms and limited far-field geodetic data. The comparison between the two displacement maps shows consistent displacement patterns, yet a systematic shift, which likely is due to the limited near-fault resolution of the data used in the previous model. We then perform a joint inversion using the newly derived SAR image offsets and tele-seismic waveforms. The preferred source model suggests a compact slip pattern at depths of 20-30 km with a peak slip of ~10 meters and a fairly short rise time (<3s). The large slip rate and low attenuation in the crust are likely responsible for the widely felt ground shaking despite of its compact source area. The result demonstrates that it is possible to correlate non-coherent SAR images acquired by different sensors to measure surface displacements. This approach extends further the possibility of mining the archive of SAR images for various types of earth-science studies.