Influence of the external DEM on PS-InSAR processing and results on Northern Appennine slopes

Monday, 15 December 2014
Benedikt Bayer1, David A Schmidt2 and Alessandro Simoni1, (1)University of Bologna, Bologna, Italy, (2)University of Washington, Seattle, WA, United States
We present an InSAR analysis of slow moving landslide in the Northern Appennines, Italy, and assess the dependencies on the choice of DEM. In recent years, advanced processing techniques for synthetic aperture radar interferometry (InSAR) have been applied to measure slope movements. The persistent scatterers (PS-InSAR) approach is probably the most widely used and some codes are now available in the public domain. The Stanford method of Persistent Scatterers (StamPS) has been successfully used to analyze landslide areas. One problematic step in the processing chain is the choice of an external DEM that is used to model and remove the topographic phase in a series of interferograms in order to obtain the phase contribution caused by surface deformation. The choice is not trivial, because the PS InSAR results differ significantly in terms of PS identification, positioning, and the resulting deformation signal. We use four different DEMs to process a set of 18 ASAR (Envisat) scenes over a mountain area (~350 km2) of the Northern Appennines of Italy, using StamPS. Slow-moving landslides control the evolution of the landscape and cover approximately 30% of the territory. Our focus in this presentation is to evaluate the influence of DEM resolution and accuracy by comparing PS-InSAR results. On an areal basis, we perform a statistical analysis of displacement time-series to make the comparison. We also consider two case studies to illustrate the differences in terms of PS identification, number and estimated displacements. It is clearly shown that DEM accuracy positively influences the number of PS, while line-of-sight rates differ from case to case and can result in deformation signals that are difficult to interpret. We also take advantage of statistical tools to analyze the obtained time-series datasets for the whole study area. Results indicate differences in the style and amount of displacement that can be related to the accuracy of the employed DEM.