Topography-driven satellite imagery analysis for landslide mapping

We describe a semi-automatic procedure for the classification of satellite imagery into landslide or no landslide categories, aimed at preparing event landslide inventory maps. The two-steps procedure requires knowledge of the occurrence of a landslide event, availability of a pre- and post- event pseudo-stereo pair and a digital elevation model. The first step consists in the evaluation of a discriminant function, applied to a combination of well-known change detection indices tuned on landslide spectral response. The second step is devoted to discriminant function classification, aimed at distinguishing the only landslide class, through an improvement of the usual thresholding' method. We devised a multi-threshold classification, in which thresholding is applied separately in small subsets of the scene. We show that using slope units as topographic-aware subsets produces best classification performance when compared to the ground truth of a landslide inventory prepared by visual interpretation. The method proved to be superior to the use of a single threshold and to any multi-threshold procedure based on topography-blind subdivisions of the scene, especially in the validation stage. We argue that the improved classification performance and limited training requirements represent a step forward towards an automatic, real-time landslide mapping from satellite imagery.