Siddique, Muhammad Adnan, Hajnsek, Irena, Frey, Othmar, Mallorqui, Jordi J., Eineder, Michael, Crosetto, Michele, and Fornaro, Gianfranco
Persistent scatterer interferometry (PSI) is a synthetic aperture radar (SAR) signal processing technique for the measurement of land surface deformation. As a conceptual derivative of differential interferometry, it strives to extract the interferometric phase variation induced by the line-of-sight component of the deformation. PSI limits the interferometric analysis to the so-called {persistent} (or synonymously {coherent}) scatterers (PS). These are single dominant scatterers exhibiting point-like, quasi-deterministic behavior. On the one hand, it implies reduced susceptibility to temporal and geometric decorrelation, but on the other hand, it incurs the limitation that range-azimuth resolution cells containing multiple scatterers are rejected in the PSI processing, even if they are individually coherent. Side-looking geometry of SAR sensors results in frequent layovers, whereby multiple scatterers situated at different elevations fall in the same resolution cell. Consequently, deformation coverage with PSI processing may remain limited in layover-affected areas. SAR tomography is a means to alleviate the aforementioned limitation. It allows 3-D reconstruction of the scene reflectivity -- a feature that offers the potential to resolve the layover problem. The coherent scatterers that are interfering in the same cell can be separated along the elevation. Additionally, differential SAR tomographic methods allow a joint spatio-temporal inversion of the coherent scatterers in layover, i.e., the position along the elevation axis as well as the deformation velocity of the interfering scatterers are simultaneously estimated. Therefore, differential SAR tomography can be used as an add-on to PSI techniques to improve deformation coverage in layover-affected areas. This dissertation provides a comprehensive assessment of the utility offered by SAR tomography as an add-on to PSI. Several aspects of a tomographic processing framework, such as different phase models for tomography, phase calibration of the interferometric stack, statistical detection of coherent scatterers, etc., need to be investigated. To this end, three core investigations have been performed. In each case, a prior PSI solution has been used as a starting point. It serves not only as a reference to compare with, but is also shown to be a natural precursor to tomographic processing. An interferometric data stack comprising 50 TerraSAR-X stripmap-mode acquisitions over an urban zone in the city of Barcelona, Spain, has been used in the first investigation. The phase models for classical SAR tomography (3-D SAR), differential tomography with the assumption of linear deformation over time, and the one further extended to simultaneously model thermal expansion, are compared against each other with respect to their suitability in resolving layovers. The results confirm that modeling thermal expansion of the scatterers, in addition to linear deformation and elevation, is indeed critical for effective layover separations, especially in the case of high-rise buildings. The quality of the scatterers obtained with tomography has been evaluated in terms of the dispersion of the residual phase and compared against the quality of the PS identified in the prior PSI processing. The results show a trade-off between the quantity and the quality of the scatterers. The second investigation focuses on the problem of phase calibration for a potential application of SAR tomography in mountainous regions. It is a case study that assesses a regression-kriging approach to functionally model height-dependent atmospheric phase variations and lateral phase trends, and consider the turbulent mixing effects in a stochastic sense. The study has been performed on a stack comprising 32 Cosmo-SkyMed acquisitions over Matter Valley in the Swiss Alps. Phase corrections with the kriging approach extend the deformation coverage to parts of a mountainside (in layover) where no PS were identified in the prior PSI processing. However, a very few double scatterers are detected on the whole. The third investigation explores how to perform scatterer detection for tomography extending the same quality considerations as used in the prior PSI processing. The outcome of this work is a detection strategy whereby quality parameters (in terms of the statistics of the phase residue or ensemble coherence) are used to determine the thresholds for hypothesis testing. The detection strategy is tested on the same data stack as for the first investigation to detect single and double scatterers in an urban area. An empirical analysis of the probability of false alarm is also provided. As a whole, this dissertation covers several aspects that collectively highlight how the synergistic use of PSI and tomography can lead to improved deformation coverage.