Classification of Landslide Susceptibility in the Development of Early Warning Systems

Statistical classification techniques complemented by GIS yield good results in predicting landslide hazard/ susceptibility. In this work, several well-known classification methods are applied to data from distinct alpine areas in Vorarlberg, Austria. It is shown that kernel methods (Support Vector Machines – SVM – and Gaussian Processes) outperform classic techniques for this task. As a further result, hazard maps for the study areas are generated, which can be used as input for early warning systems focussing on landslide hazard.