Analysis of landslide surface deformation from the correlation of multiple Pleiades very-high resolution satellite images

Stumpf, A. (1,2), Malet, J.-P. (1), Allemand, P. (3), Jory, M. (2)

(1) Institut de Physique du Globe de Strasbourg, CNRS UMR 7516, Université de Strasbourg, 5 rue Descartes, F-67084 Strasbourg Cedex, France
(2) Laboratoire Image, Ville, Environnement, CNRS UMR 7362, Université de Strasbourg, France
(3) Laboratoire des Sciences de la Terre, CNRS UMR 5570, Université de Lyon and Ecole Normale Supérieure, Lyon, France

The recently launched Pleiades satellite constellation enables image acquisition of VHR resolution images with unprecedented short intervals of up to one image per day at submetric spatial resolution. The resulting enhanced availability of VHR optical images offers new opportunities for the frequent monitoring of small scale surface deformation induced by slow-moving landslides, whereas the agile acquisition geometry also poses new challenges for the application of digital image correlation (DIC) techniques. Since the satellite tasking does not allow for selecting specific view angles, it remains challenging to separate apparent parallax shift from real surface displacement with a single image pair. Additionally, at very-high spatial resolutions errors in the DEM used for orthorectification typically induce a stronger bias relative to the absolute pixel size. In this study, we applied DIC to characterize the surface deformation of three large landslides (La Valette, Poche, Super-Sauze ; South French Alps) from a triplet of Pleiades satellite images acquired in 2012. Image correlation was performed using a hierarchical approach operating on the first order derivate of the image. To evaluate the impact of view angle differences and orthorectification errors DIC measurements were applied on two pairs recorded at different geometries using different available DEMs for orthorectification. The respective obtained displacement vectors were compared with dGPS measurements. Current limitations resulting from variable view angles and the quality of the DEM used for orthorectification, as well as possible strategies for a joint inversion of 2D displacement vectors from measurements at two different view angles are being discussed