Small Baseline Subset (SBAS) InSARAnalysis Using Sentinel-1 Data for Monitoring Landslide Deformation in the Alps
MetadataShow full item record
At: FRINGE 2017 ; Helsinki ; 5.6.2017 - 9.6.2017 ; New generations of radar satellites (e.g. TerraSAR-X, Cosmo-SkyMed, Sentinel-1A/B) with short repeat-pass cycles and high spatial resolutions have enhanced the capabilities for acquiring data over large areas shortly after major landslide events and for monitoring landslide activity at regular intervals. Over the past two decades, several studies have demonstrated the potential of differential synthetic aperture radar interferometry (DInSAR) for detecting and quantifying land surface deformation. The main challenges of the DInSAR technique include spatial and temporal decorrelation, an accurate estimation of the phase ambiguity in the phase unwrapping step and the presence of atmospheric artifacts. Permanent Scatterer Interferometry (PSI) and Small BAseline Subset (SBAS) are widely used to extract the phase information of the displacement component and mitigate the negative effects of the errors sources in the interferogram stack. The PSI technique is based on the pixels with a dominate radar backscattering in comparison to the background (i.e. high coherence over time). The PSs dependency of the PSI technique and lack of sufficient PSs in the natural terrains have limited the abilities of PSI in vegetated areas. For the vegetated areas, which is considered nearly as Gaussian scatterers, high temporal decorrelation caused by vegetation and reliable phase unwrapping are the main challenges of the surface deformation estimation within a given period of time. In this study, we present preliminary results of SBAS processing with Sentinel-1 IW data for monitoring the activity of two landslides located in the Alps: Corvara in Badia in the Autonomous Province of Bolzano-South Tyrol (Italy) and Schmirntal in Tyrol (Austria). Both study sites have displacement rates in the order of centimetres to meters per year. Two different approaches are tested using datasets acquired between 2015 and 2016. First, multi-looking data is processed to increase the Signal to Noise Ratio (SNR) and the reliability of the coherence estimation. Secondly, using the single look data to identify the isolated SBAS pixels (if surrounded by completely decorrelated pixels), which is limited when applying the multi-looking process due to the resulting coarser spatial resolution. For a more reliable phase unwrapping process particularly in areas with low coherent, in addition to the standard 2D unwrapping, a 3D phase unwrapping approach is tested. Remaining discontinuities caused by phase jumps in the unwrapped interferogram, are corrected by discarding the interferogram pairs showing low coherence while optimizing the adaptive filter strength. After the first estimation of the residual topography and displacement rate, the best fitting models corresponding to the landslide displacement behavior is selected for reprocessing the interferogram stack for improving and refining the final velocity rate. Finally, after applying low pass spatial and high pass temporal filters, the final displacement map is geocoded and is compared to the ground dGPS measurements. The results shows that the negative effects of the temporal decorrelation caused by the vegetation is still tangible and visible on the displacement map. We expect that in the further analysis, the higher frequency of the data due to the availability of additional Sentinel-1B data will reduce errors due to temporal decorrelation and allow describing the displacement rate.