SNOW AVALANCHE SUSCEPTIBILITY MAPPING FOR DAVOS, SWITZERLAND

Abstract. Snow avalanches are among destructive hazards occurring in mountainous regions and spatial distribution (susceptibility) of their occurrences needs to be considered for planning disaster risk mitigation efforts. The susceptibility assessment is the first step avalanche management can carried out using high resolution geospatial data machine learning (ML) algorithms. In this study, we have assessed snow Davos, Switzerland an inventory delineated on satellite imagery a previous study. conditioning factors used include elevation, slope, plan curvature, profile aspect, topographic position index, ruggedness wetness land use cover, lithology, distance road, river. Two ML algorithms, logistic regression (LR) random forest (RF), were comparatively validation split from training (30/70). prediction performances both models based area under receiver operating characteristic curve (ROC-AUC) value. Although AUC value obtained LR method was relatively low (0.74), RF (0.96) demonstrated performance usability approach. results indicate that successfully produce map region, although potential improvements may possible by investigating various input features algorithms as well classifying starting runout zones separately. Furthermore, accuracy expected increase larger dataset.