STARTER: a statistical GIS-based model for the prediction of snow avalanche susceptibility using terrain features—application to Alta Val Badia, Italian Dolomites

Abstract

This article describes a methodology to localise areas with high potential towards natural snowpack instability under particular meteorological conditions, at the scale of an Alpine valley. Localisation is based on statistically relating known release areas of past avalanche events to maps of: (1) slope inclination, (2) slope orientation (aspect), (3) elevation, (4) distance from crest lines, (5) terrain roughness and (6) concavities/convexities. The maps have been built using two different GIS softwares while the statistical analyses have been performed with a specific software handling also Fuzzy Set theory algorithms. The results of the statistical analyses have been verified on test release—areas which have not been used as input data for the statistical analyses. Verification allowed to quantify how reliably the susceptibility values were calculated, to compare the values obtained using different combinations of terrain features and to finally decide on the most efficient combination. The susceptibility maps were calculated and verified for three different meteorological scenarios (given by three classes of snow depth). Verification has shown that the accuracy of the susceptibility maps was between 67% and 82%. The three susceptibility maps show a remarkable difference in the spatial pattern of the highest susceptibility pixels suggesting that for different meteorological scenarios different classes of terrain features need to be considered. The possibility to make combinations of terrain features and to assess and verify their statistical relationship with release areas of past avalanche events is the major original step made by STARTER. Linking those release areas to meteorological scenarios is an attempt to include in the analysis the combined influence of terrain features and meteorological conditions towards snowpack instability.