Deformation monitoring using scanning synthetic aperture radar interferometry

This dissertation provides the first demonstration of scanning synthetic aperture radar (ScanSAR) advanced interferometry processing for measuring surface deformation. ScanSAR data are synthesized from ERS-1/2 stripmap SAR images over known deformation in Phoenix, Arizona. The strategy is to construct a burst pattern similar to Envisat ScanSAR data and to create a realistic variable-burst synchronization scenario in which any image pair has at least 50% burst overlap. The Small Baseline Subsets technique is applied to the synthesized data to demonstrate ScanSAR time series analysis for a scenario generally conducive for interferometry. The same processing approach is employed with the stripmap data to validate the results. The differences in ScanSAR and stripmap velocities have a mean and standard deviation of 0.02±0.02 cm/year. 96.3% and 99.1% of the velocity differences are within ±0.1 cm/year and ±0.2 cm/year, respectively. The RMS deviations between the ScanSAR and stripmap displacement estimates are 0.40±0.30 cm. 68.5% and 94.6% of the differences are within ±0.5 cm and ±1.0 cm, respectively. The Permanent Scatterer (PS) technique also is adapted and applied to the synthesized data to demonstrate the presence of PS in ScanSAR data. The atmospheric and nonlinear motion phase derived from a PS analysis of stripmap data are removed from the ScanSAR interferograms. Even for this idealized scenario, the final PS identification yields fewer ScanSAR PS (10 PS/km²) than the stripmap PS results (312 PS/km² or 15.6 PS/km² at the ScanSAR pixel resolution). Based on the calculated likelihood of finding multiple stripmap PS within a ScanSAR pixel, it is concluded that the ScanSAR single scatterer PS model is flawed. A model is introduced that considers multiple PS within a ScanSAR pixel. The search for two PS per pixel yields 120 PS/km². The ScanSAR and stripmap PS velocity differences mean is zero and standard deviation is 0.02 cm/year. However, while the differences between the ScanSAR and stripmap PS DEM error estimates are zero-mean, they have a 7-meter standard deviation. One possible explanation for this relatively large deviation is the differencing of the wrong ScanSAR and stripmap PS as the result of a misalignment between the ScanSAR and stripmap images.