Inversion of deformation fields time-series from optical images, and application to the long term kinematics of slow-moving landslides in Peru

Slow-moving landslides are numerous in mountainous areas and pose a large threat to populations. Many observations show that their kinematics is driven by climatic forcings and earthquakes. In this study, we document the complex interaction between those two forcings on the slow-moving landslide kinematics, based on the retrieval of landslide displacements over 28-years using optical satellite images. To overcome the decorrelation effect over this large time-span, and possible misalignment between images, we develop a method that uses the redundancy of displacement fields from image pairs to derive a robust time-series of displacement. The method is tested on the 28-year long SPOT1/5-Pléiades archive, over an area in Peru affected by both earthquakes and rainfall. Errors are estimated on stable areas and by comparison with one 13-year long and eleven 3-year long GPS time-series on the Maca landslide. The methodology diminishes by up to 30% the uncertainty and reduces significantly the gaps due to decorrelation. The data set allows detecting 3 major landslides, moving at a rate of 35 to 50 m over 28 years, and smaller landslides with lower displacement rates. Time-series obtained over the three main landslides provide interesting results of their long-term kinematics, primarily driven by precipitation. We propose simple statistical hydro-kinematic models, relating yearly motion to seasonal rainfall, to explain the observed time-series. We found that annual precipitation is controlling the landslide displacements after a certain rainfall threshold is reached. Besides this control, we show the possible impact of a local Mw 5.4 earthquake in 1991 on the kinematics of the Maca landslide. Our results suggest that the earthquake accelerated the landslide and has an effect during several years on the precipitation threshold required for triggering a motion. These results suggest that the rainfall threshold can vary in time following strong earthquakes shaking.