Long wavelength DInSAR and PSInSAR for the detection of landslides : an experience in the Romanian Subcarpathians

Provost, F. (1), Malet, J.-P. (1), Doubre, C. (1), Puissant, A. (2), Micu, M. (3)

(1) Institut de Physique du Globe de Strasbourg, CNRS UMR 7516, Université de Strasbourg, 5 rue Descartes, F-67084 Strasbourg Cedex
(2) Laboratoire Image, Ville, Environnement, CNRS UMR 7362, Université de Strasbourg, France
(3) Romanian Academy of Science, Bucarest, Romania

Landslide is one of the common natural hazards in Romania, especially in the Curvature area of the Romanian Subcarpathians, characterized by various conditionning factors. In this region, landslides cause considerable damages to critical infrastructures, build-up environment and cultivated areas. Most of the slopes are affected by translational and rotational landslide types. The objective of this work is to locate and inventory landslides in the Buzau County, and possibly to characterize their dynamics. As the vegetation in abundant in the study area, series of L-band ALOS/PALSAR images are processed using advanced multi-temporal differential SAR interferometry (DInSAR & PSInSAR). To analyze the DInSAR results, an object-oriented segmentation method is proposed to identify possible landslide candidates in the interferograms ; to analyze the PSInSAR results, a statistical method is used to identify PS characterized by the same evolution pattern in the time series. Both techniques have proved to be able to detect unexpected active landslides in the area, and allow to complete existing geomorphological inventories. But that does not mean that applying SAR interferometry is a sufficient tool to build exhausting inventories, and depending on the characteristics of the images (frame/track, baseline), the characteristics of the terrain (landcover, slope gradient, geomorphology) and the characteristics of the landslide (size, displacement rate), only a certain percentage (roughly less than 50%) of the already known landslides are detected with DInSAR results. Moreover, because of the large variety of landcover (urban, forest, cultivated areas, bare soils) and slope morphology in the study area, DInSAR and PSInSAR techniques provide very different results depending on local conditions.