Abstract
Statistical approaches are routinely applied to analyze landslide processes from several perspectives. During the last decade, the growing accessibility of environmental data as well as an increasing availability of powerful open-source statistical software, such as R, has certainly enhanced the regional scale analysis of landslide phenomena. For large areas, a considerable number of statistics-oriented research focuses on the mapping of landslide-prone zones using binary classification algorithms. The resulting landslide susceptibility maps are commonly regarded as a suitable information source for land management, civil protection, and further statistical analyses. The creation of a statistically-based landslide susceptibility model demands for a multitude of crucial decisions. For instance, besides the selection of suitable explanatory variables or modelling algorithms, also aspects related to the spatial representation of input data determine the final modelling outcome. Opting for one or another option is not always a trivial task, particularly because error-prone models can be associated with high quantitative validation results. This chapter provides an overview on common practices in statistical landslide susceptibility modelling from a broad and critical perspective. Potential pitfalls as well as approaches that reach beyond the pure spatial identification of susceptible terrain are presented and discussed.