Ensemble forecasts of road surface temperatures

- Ensemble forecast of road surface temperature is presented.
- Probabilistic forecast of road surface temperature is presented.
- Ensemble approach is especially suitable for the prediction of road surface conditions.

This paper describes a new ensemble technique for road surface temperature (RST) forecasting using an energy balance and heat conduction model. Compared to currently used deterministic forecasts, the proposed technique allows the estimation of forecast uncertainty and probabilistic forecasts. The ensemble technique is applied to the METRo-CZ model and stems from error covariance analyses of the forecasted air temperature and humidity 2 m above the ground, wind speed at 10 m and total cloud cover N in octas by the numerical weather prediction (NWP) model. N is used to estimate the shortwave and longwave radiation fluxes. These variables are used to calculate the boundary conditions in the METRo-CZ model. We found that the variable N is crucial for generating the ensembles. Nevertheless, the ensemble spread is too small and underestimates the uncertainty in the RST forecast. One of the reasons is not considering errors in the rain and snow forecast by the NWP model when generating ensembles. Technical issues, such as incorrect sky view factors and the current state of road surface conditions also contribute to errors. Although the ensemble technique underestimates the uncertainty in the RST forecasts, it provides additional information to road authorities who provide winter road maintenance.

The ensemble forecasts of the road surface temperature (RST) in °C for road weather stations K013 (left column) and P004 (right column) as a function of the lead times in h. The red and green curves show the deterministic forecasts and observed data, respectively. The grey areas denote the intervals for probabilities x = 99, 90, 70 and 50% that the RST will be within that interval. The start times of the forecasts are indicated in the titles in a year, month, day and hour (2 digits) format.195