An Automated Nowcasting Model of Road Surface Temperature and State for Winter Road Maintenance

Abstract

Winter road maintenance is an important application field of meteorology in western and northern Europe, North America, and many other parts of the world. In order to provide timely short-period high-accuracy forecasts of road surface temperature and state (dry, wet, frost, or ice), an automated road ice prediction model is developed for the purpose of nowcasting (up to 3 h ahead). The model was validated against observations from 41 road surface sensors in seven countries. As far as the authors are aware, this model is unique in that it is the only fully automated physical road ice prediction model that requires no external meteorological input data other than automatically collected sensor measurements of surface temperature, air temperature, dewpoint, and wind speed from the forecast site. The results show that (a) the projection of primary meteorological parameters by the model itself as input is acceptable for this purpose and (b) model performance becomes poorer as nowcast period gets longer. However, all nowcasts of surface temperature have a near-zero bias and their weight-averaged root-mean-square errors are less than 1.1°, 1.6°, and 2.0°C for minimum, overall (for every hour in all days), and maximum temperatures, respectively. Also shown is that (c) over 95% of minimum temperature forecasts are within an absolute 2°C error band for up to 3 h ahead and (d) over 92% of frost (minimum temperature at or below 0°C) and no-frost nights are successfully predicted. Mean error in predicting time of frost is less than 22 min for all sites and up to 3 h ahead, although there is large difference for individual stations. (c) The accuracy of the prediction of dry and ice states is over 82%, and more than 72% of wet states are predicted accurately in most countries.