Exploiting COSMO-Skymed Data and Multi-Temporal Interferometry for Early Detection of Landslide Hazard: A Case of Slope Failure and Train Derailment Near Marina Di Andora, Italy.

Abstract:

The improving temporal and spatial resolutions of new generation space-borne X-Band SAR sensors such as COSMO-SkyMed (CSK) constellation, and therefore their better monitoring capabilities, will guarantee increasing and more efficient use of multi-temporal interferometry (MTI) in landslide investigations. Thanks to their finer spatial resolution with respect to C-band data, X-band InSAR applications are very promising also for monitoring smaller landslides and single engineering structures sited on potentially unstable slopes. This work is focused on the detection of precursory signals of an impending slope failure from MTI time series of ground deformations obtained by exploiting 3 m resolution CSK data. We show the case of retrospectively captured pre-failure strains related to the landslide which occurred on January 2014 close to the town of Marina di Andora. The landslide caused the derailment of a train and the interruption of the railway line connecting north-western Italy to France. A dataset of 56 images acquired in STRIPMAP HIMAGE mode by CSK constellation from October 2008 to May 2014 was processed through SPINUA algorithm to derive the ground surface deformation map and the time series of displacement rates for each coherent radar target. We show that a cluster of moving targets coincides with the structures (buildings and terraces) affected by the 2014 landslide. The analysis of the MTI time series further shows that the targets had been moving since 2009, and thus could have provided a forewarning signal about ongoing slope or engineering structure instability. Although temporal landslide prediction remains difficult even via in situ monitoring, the presented case study indicates that MTI relying on high resolution radars such as CSK can provide very useful information for slope hazard mapping and possibly for early warning.

Acknowledgments

DIF provided contribution to data analysis within the framework of CAR-SLIDE project funded by MIUR (PON01_00536).