AbstractA numerical road‐ice prediction model is further developed to predict surface temperature and surface state for porous asphalt pavement, which is significantly colder at night and warmer at daytime than conventional asphalt pavement. The model is validated against temperature measurements on both porous and conventional asphalt pavements, using actual input data for the complete winter 1992/93. The validation of model rum at noon shows that, on average, the model has a near‐zero bias for minimum, maximum and overall hourly temperature forecasts. The model has the smallest standard deviation (0.78 °C) for minimum temperature forecasts and the largest standard deviation (1.92 °C) for maximum temperature forecasts. Nearly 90% of frost days are predicted. Model runs with different start times at 1200 1500, 1900 and 0000 show consistently good re