March wet avalanche prediction at Bridger Bowl ski area, Montana
Wet avalanches are a safety concern for all ski areas because they are difficult to control artificially and the shift from safe to dangerous wet snow conditions can happen very quickly. Forecasting for wet avalanche conditions in intermountain ski areas, such as Bridger Bowl, Montana, can be especially difficult because intermountain snow climates can exhibit wet avalanche characteristics of either maritime or continental snow climates. Various statistical models have been developed for avalanche prediction; however, most are tailored specifically for dry avalanche forecasting. Archived meteorological, snowpack and avalanche data for the month of March from 1968 to 2001 (1996 data unavailable) were used to develop 68 possible predictor variables related to temperature, snowpack settlement, and precipitation characteristics. The original Bridger Bowl dataset was divided into a ♯new snowα and an ♯old snowα dataset. A ♯new snowα day has newly fallen snow that is less than 48 hours old; an ♯old snowα day has newly fallen snow that is more than 48 hours old. The two datasets were used to determine whether the factors that influence ♯old snowα and ♯new snowα wet avalanche occurrence differ. Hypotheses were developed and tested to determine which ♯old snowα and ♯new snowα variables behaved significantly different on days with wet avalanches compared to days with no wet avalanches. The 33 ♯old snowα significant variables and the 22 ♯new snowα significant variables were analyzed with binomial logistic regression to produce one prediction model for ♯old snowα wet avalanche conditions and another prediction model for ♯new snowα wet avalanche conditions. The ♯old snowα model uses the prediction day minimum temperature and the two day change in total snow depth as predictor variables. This model has a 75% success rate for calculating accurate wet avalanche probabilities for ♯old snowα days. The ♯new snowα model uses the prediction day minimum temperature as well as the three day cumulative new snow water equivalent as predictor variables. This model has a 72% success rate for calculating accurate wet avalanche probabilities for ♯new snowα days.