Monitoring of Tectonic Deformation by Mining Satellite Image Time Series

In this paper, an original approach for analyzing Interferometric Synthetic Aperture Radar (InSAR) time series is presented and applied to the monitoring of creep along a major active fault. InSAR data are computed from satellite acquisitions and allow ground deformations occurring between different dates to be measured with high precision over large areas. However, measurement precision remains limited by the atmospheric noise, due to variations in space and time of atmospheric conditions. The proposed approach is designed to process InSAR data in an unsupervised way. It handles varying atmospheric conditions and provides end-users with spatiotemporal maps describing ground deformation evolutions. Experimental results on an ENVISAT InSAR time series covering the Haiyuan fault in the north-eastern boundary of the Tibetan plateau are presented. The maps obtained reveal continuous, aseismic shallow slip along a section of the fault, which is consistent with geophysical models. These results illustrate the potential of the proposed approach which combines advanced InSAR processings and data mining techniques.