Stepwise approach to INSAR processing of multitemporal datasets

Multi-temporal DInSAR processing techniques aim at monitoring millimetric displacements through periodic detection of vertical movements of the earth surface, on either point or distributed scatterers. Many such techniques benefit from the availability of a set of SAR acquisitions which is completely connected through suitable interferometric pairs. Moreover, many of the above methods require as a pre-processing step the co-registration of long series of SAR images to sub-pixel accuracy. Automation of this processing step is highly desirable for effective applications to surface displacement monitoring. In this paper an approach to determine the best procedure to connect multi-temporal InSAR datasets is investigated. The method consists in adopting an a priori measure of image interferogram quality, and then build a minimum spanning tree (MST) connecting all the image points in the space determined by the spatial and temporal baselines. As a priori measure, a natural choice is that of coherence. We model spatial decorrelation as mainly due to the wavenumber shift effect, and temporal decorrelation as a combination of a seasonal effect plus a decrease of coherence with time. Various examples of MSTs computed using different parameters for these models are shown. Results on a real multi-temporal DInSAR dataset are reported and evaluations of both processing efficiency as well as final data quality are presented. Experimental tests and results are provided with reference to an ERS-1/2 dataset over the Italian landslide site of Caramanico Terme, involved in the ESA AO3-313 project.