This project analyzes the management of Utah Highway 210 with respect to avalanche danger. The occurrence of an avalanche crossing the road is a low-probability, high-risk event. Avalanche hazard is managed both by active control (explosives) and by passive control (road closures). The study emphasizes the road closure decision. The probability of an avalanche crossing the road is modeled through logistic regression. The decision to close the road, however, reflects a highly asymmetric loss function. Closing the road when no avalanche occurs is very costly, but does not compare to the effects of failing to close the road when an avalanche does in fact occur. This feature of the problem is incorporated into the analysis using an asymmetric loss function. Decisions to close the road are scored against actual avalanche occurrence by means of a proper scoring rule and this performance is contrasted with that of professional avalanche forecasters currently managing the road. The possibility of augmenting the current practice with statistical analysis is evaluated. The results of the project are generalized to other decision making problems involving quantal forecasting under asymmetric loss. Asymmetric loss is a common problem with low-probability, high risk events.
