Road maintenance is one of the main problems Departments of Transportation face during winter time. Anti-icing, i.e. applying chemicals to the road to prevent ice formation, is often used to keep the roads free of ice. Given the preventive nature of anti-icing, accurate predictions of road ice are needed. Currently, anti-icing decisions are usually based on deterministic weather forecasts. However the costs of the two kinds of error are highly asymmetric because the cost of a road closure due to ice is much greater than that of taking anti-icing measures. As a result, probabilistic forecasts are needed to optimize decisionmaking. We propose two methods for forecasting the probability of ice formation. Starting with deterministic numerical weather predictions, they produce a joint predictive probability distribution of temperature and precipitation. This then yields the probability of ice formation, defined here as the occurrence of precipitation when the temperature is below freezing. In the first method, temperature and precipitation at different spatial locations are treated as conditionally independent given the numerical weather predictions. In the second method, spatial dependence between forecast errors at different locations is modeled. The model parameters are estimated using a Bayesian approach via Markov chain Monte Carlo. We evaluated both methods by comparing their probabilistic forecasts with observations of ice formation for Interstate Highway 90 in Washington State for the 2003–2004 and 2004–2005 winter seasons. Results showed that using probabilistic forecasts can save a considerable amount of money compared with deterministic forecasts. The method that takes account of spatial dependence improved the reliability of the forecast, but did not save more money.
Keywords: Numerical weather forecast; Predictive distribution; Spatial dependence; Markov chain Monte Carlo; Latent Gaussian process; Cost-loss ratio.