SAR Image Coregistration Based on Topography and Deformation

L-band repeat-pass InSAR has been getting scientists' attention for its penetration depth through dense vegetation, revealing the ground deformation under the canopy of forest. This makes it possible to achieve a good coverage of volcanic activities in tropical areas such as Hawaii, Galapagos, and Indonesia. Another advantage of L-band InSAR is its tolerance to a large baseline. The critical baseline scales with the wavelength, and practically a baseline of over 1 km often produces interferograms. However, the large baseline causes parallax in the presence of topography, which appears as pixel shift between master and slave images that sometimes leads to a severe decorrelation. Using ALOS PALSAR data of Java Island, Indonesia, and intermediate files produced by ROI_PAC, we correct this parallax before forming an interferogram to improve the interferometric coherence. We also test an automatic implementation of the 'rubber sheeting" coregistration (e.g. Yun et al., GRL, 2007) for localized large deformation that cannot be explained by polynomial fitting.