Investigating the transferability of statistical disposition models for slope-type debris flows

Abstract

Statistical disposition models are used both in geomorphology and in natural hazards research to spatially predict the occurrence of mass movements, other geomorphic processes or landforms. This is achieved by establishing some statistical relationship between an inventory of mapped occurrences and relevant geofactor maps. Statistical approaches rely, among others, on the assumption that future events will take place under the same conditions as they did in the past. In the present paper, the Certainty Factor (one of the favourability functions) is used to quantify the association of debris flow initiation sites with the geofactors landcover, slope and CIT index in three alpine study areas. The transferability of the models, i.e. their applicability to other study areas, is investigated using a comparison of model results (CF parameters) from different study areas and an extensive cross-validation. It is shown that, in similar study areas, the model parameters are generally very similar, indicating that the geofactors represent well enough the relevant conditions and processes that lead to debris flow initiation. The cross-validation is generally successful, independent of the similarity of the area for which the model has been established and the target area. Therefore, it can be concluded that statistical disposition models such as the CF model are, in principle, transferable between study areas. Nevertheless, some critical issues such as data quality have to be kept in mind when trying to apply a model to another study area.