The strength and weight of evidence in backcountry avalanche forecasting

Abstract

In North America and Europe, most fatalities due to snow avalanches occur in the backcountry during recreational pursuits. Of these, more than 90% of the fatal avalanches are triggered by the victims themselves. This pattern suggests that the primary cause of avalanche fatalities for human-triggered avalanches is a failure in human perception. For the latter, people thought that the state of stability or instability of the snow cover was different than it actually was. In this paper, the strength and weight of evidence used to make decisions in backcountry travel are discussed from: (a) the perspective of the favored hypothesis to proceed for good recreational enjoyment based on stability evaluation and (b) the null hypothesis based on an assessment of instability. Based on the facts about snow slab avalanche release, it is argued that instability analysis is the best framework for avalanche forecasting, whereas human action is most closely related to the favored hypothesis (stability evaluation). Using scaling laws derived from: (a) fracture mechanics about the size of imperfections causing avalanches and (b) avalanche dimensions, it is suggested that a snow slab could show stability over more than 99% of the total area. From the concepts of Bayesian probability, it is shown that overconfidence about stability can arise when the weight of the likelihood is high and the weight of prior is low. Similarly, underconfidence (excessive conservatism) often results when the weight of the prior is high with little regard for the likelihood, which may be low. Overconfidence about stability is considered to be a prime source of accidents.