Abstract

Avalanche forecasting at the slope scale is an essential skill for traveling safely in avalanche terrain. Recent studies indicate that this skill is difficult for recreationists to master, and is subject to a number of persistent biases that can lead to avalanche accidents and their associated fatalities. In an effort to address this problem, researchers have developed a number of algorithmic, rule-based decision aids that simplify slope-scale avalanche prediction. Many of these decision aids are taught and used widely but to date, there has been no comprehensive comparison of their effectiveness. In this paper, we describe five of these decision aids (Reduction Method, Elementary Reduction Method, NivoTest, SnowCard, Obvious Clues Method) and analyze how they would have performed had they been used by accident parties in 751 historical US avalanche accidents. Using a principally nonparametric analysis, we found that these decision aids would have prevented between 60% and 92% of historical accidents, with the actual proportion of accidents prevented varying significantly among different decision aids. We also found that the preventive values of decision aids were statistically invariant to the level of training of the accident party, the type of recreational activity, and the type of slab avalanche released. In contrast, we found that preventive values differed significantly by snow climate, with four of the five decision aids being more conservative in maritime mountain ranges. All of the decision aids exhibited the highest preventive values during periods of considerable and high avalanche danger, but only one of them (the Obvious Clues Method) performed well during times of moderate and low danger. We also examined the mobility afforded by each method using a simple comparative metric based on slope angle, and found that two of the decision aids (NivoTest and Obvious Clues) allowed more latitude for travel in timbered areas adjacent to hazardous slopes. Using a Monte Carlo simulation to assess the effectiveness of the various decision aids across many virtual users, we found that a decision aid based on a simple checklist of obvious clues most frequently provided the optimal balance between accident prevention, mobility and ease-of-use.