Climate-physiographically differentiated Pan-European landslide susceptibility assessment using spatial multi-criteria evaluation and transnational landslide information

Abstract With the adoption of the EU Thematic Strategy for Soil Protection in 2006, small-scale (1:1 M) assessments of threats affecting soils over Europe received increasing attention. As landslides have been recognized as one of eight threats requiring a Pan-European evaluation, we present an approach for landslide susceptibility evaluation at the continental scale over Europe. Unlike previous continental and global scale landslide susceptibility studies not utilizing spatial information on the events, we collected more than 102,000 landslide locations in 22 European countries. These landslides are heterogeneously distributed over Europe, but are indispensable for the evaluation and classification of Pan-European datasets used as spatial predictors, and the validation of the resulting assessments. For the analysis we subdivided the European territory into seven different climate-physiographical zones by combining morphometric and climatic data for terrain differentiation, and adding a coastal zone defined as a 1 km strip inland from the coastline. Landslide susceptibility modeling was performed for each zone using heuristic spatial multicriteria evaluations supported by analytical hierarchy processes, and validated with the inventory data using the receiver operating characteristics. In contrast to purely data-driven statistical modeling techniques, our semi-quantitative approach is capable to introduce expert knowledge into the analysis, which is indispensable considering quality and resolution of the input data, and incompleteness and bias in the inventory information. The reliability of the resulting susceptibility map ELSUS 1000 Version 1 (1 km resolution) was examined on an administrative terrain unit level in areas with landslide information and through the comparison with available national susceptibility zonations. These evaluations suggest that although the ELSUS 1000 is capable for a correct synoptic prediction of landslide susceptibility in the majority of the area, it needs further improvement in terms of data used.