Relating storm and weather factors to dry slab avalanche activity at Alta, Utah, and Mammoth Mountain, California, using classification and regression trees

Abstract

Using classification and regression tree models, we evaluated 31 factors in terms of their importance to explaining avalanche activity indices at two ski areas: Alta, UT and Mammoth Mountain, CA. This study derived new empirical factors that combined wind velocity with new snow amount, air temperatures with time, and total snow depth with time. The analyses created over-fit tree models in exploring structures inherent in the data to obtain the relative ranking and scores of various combinations of the 31 factors. Avalanche activity indices included maximum size, number of releases and sum of sizes of released avalanches. Results showed that time lagged conventional factors describing snowfall and derived wind-drift parameters ranked highest in all tests. Snow drift factors segregated into prominent wind directions showed only moderate importance. Among the non-storm factors, the starting snow depth of a particular year ranked highest showing the importance of interannual variability. This was followed by the accumulated vapor pressure difference, which we formulated to better describe the conditioning of old snow with age. The average snow depth increase and other factors followed in importance.