Prediction of snow avalanche occurrence and its run-out distance inherently involves many uncertainties. The main reason for the large uncertainties is that the mechanical behaviour of snow is extremely complicated, and other external factors that affect the problem, such the climate, wind direction, amount of snowfall, loading rate, etc. are all more-or-less random processes. The aim of the work described in this report is to demonstrate potential applications of probabilistic analysis to problems related to snow avalanches. Two problems are looked into. The first problem is evaluating the probability of a slab avalanche. A mechanical model based on traditional geotechnical approach to slope stability was adopted. To account for the uncertainties in the input parameters, the stability model was coupled to reliability analysis software. In addition to the failure probability, the probabilistic analyses provide the most likely combination of parameters at failure and the sensitivity factors that quantify the contribution of each random variable to the total uncertainty. The second problem considered is related to the development of hazard maps for snow avalanche. Current practice is to use a statistical model for run-out distance in areas where topography and climate are favourable for avalanche activity. The procedure may be conservative because it basically assumes that in the area of interest, avalanche activity occurs every year. The actual annual probability of snow avalanche occurrenceat a given area is ignored because it is difficult to estimate this probability from physical models. A procedure is outlined in the report to use historical observations and Bayesian updating to estimate the annual probability of snow avalanche occurrence. The procedureis demonstrated through an example application.

Norges Forskningsråd (NFR)