Probabilistic weather forecasting for winter road maintenance

摘要

Winter road maintenance is a major issue for the Washington State Department of Transportation. Anti-icing is often used to keep the roadways free of ice. Given the preventive nature of anti-icing, accurate predictions of road ice are needed. Currently, the decisions are based on deterministic weather forecasts. The cost of a road closure due to ice is much greater than that of taking anti-icing measures, necessitating probabilistic forecasts to optimize decision making. The study proposes two methods for forecasting the probability of ice formation. Based on deterministic numerical weather predictions, the study models temperature and precipitation using distributions cantered around the bias-corrected forecasts. This produces a joint predictive probability distribution of temperature and precipitation, yielding the probability of ice formation (occurrence of precipitation), when the temperature is below freezing. The first method assumes that temperatures and precipitation are conditionally independent for a set of numerical weather predictions. The second method models the spatial dependence between forecast errors at different locations,estimating the model parameters using a Bayesian approach via Markov chain Monte Carlo. The study evaluates both methods by comparing their probabilistic forecasts with observations of ice formation for Interstate Highway I-90 in Washington state, for the 2003-2004 and 2004-2005 winter seasons. The use of the probabilistic forecasts reduces costs by about 50% when compared to deterministic forecasts. The spatial method improves the reliability of the forecasts, but does not help reduce cost, compared to the first method.