Automated snow avalanche release area delineation – validation of existing algorithms and proposition of a new object-based approach for large-scale hazard indication mapping

摘要

Snow avalanche hazard is threatening people and infrastructure in all alpineregions with seasonal or permanent snow cover around the globe. Coping withthis hazard is a big challenge and during the past centuries, differentstrategies were developed. Today, in Switzerland, experienced avalancheengineers produce hazard maps with a very high reliability based on avalanchedatabase information, terrain analysis, climatological data sets andnumerical modeling of the flow dynamics for selected avalanche tracks thatmight affect settlements. However, for regions outside the consideredsettlement areas such area-wide hazard maps are not available mainly becauseof the too high cost, in Switzerland and in most mountain regions around theworld. Therefore, hazard indication maps, even though they are less reliableand less detailed, are often the only spatial planning tool available. Toproduce meaningful and cost-effective avalanche hazard indication maps overlarge regions (regional to national scale), automated release areadelineation has to be combined with volume estimations and state-of-the-artnumerical avalanche simulations.In this paper we validate existing potential release area (PRA) delineationalgorithms, published in peer-reviewed journals, that are based on digitalterrain models and their derivatives such as slope angle, aspect, roughnessand curvature. For validation, we apply avalanche data from threedifferent ski resorts in the vicinity of Davos, Switzerland, whereexperienced ski-patrol staff have mapped most avalanches in detail for manyyears. After calculating the best fit input parameters for every testedalgorithm, we compare their performance based on the reference data sets.Because all tested algorithms do not provide meaningful delineation betweenindividual PRAs, we propose a new algorithm basedon object-based image analysis (OBIA). In combination with an automaticprocedure to estimate the average release depth (d0), defining the avalancherelease volume, this algorithm enables the numerical simulation of thousandsof avalanches over large regions applying the well-established avalanchedynamics model RAMMS. We demonstrate this for the region of Davos for twohazard scenarios, frequent (10–30-year return period) and extreme (100–300-yearreturn period). This approach opens the door for large-scale avalanchehazard indication mapping in all regions where high-quality and high-resolutiondigital terrain models and snow data are available.