Abstract:In order to solve the problem of meteorological service for transportation operation in areas with a few observation stations, three typical regions in northern China (Beijing-Tianjin, Shaanxi-Gansu and Qinghai) are selected. The data used include the transportation weather observation data, the surface data from China Meteorological Administration Land Data Assimilation System (CLDAS V2.0) and Multi-source Precipitation Analysis System (CMPAS), as well as surface short wave and long wave radiation retrieval products from FY-4A. The change characteristics of road surface temperature and its relationship with environmental meteorological factors are analyzed. Three methods (linear regression, random forest and deep neural network) are employed to construct a 1 h updated real-time road surface temperature model, and the effects of model results from different methods and different data combinations are tested. Besides, the spatial generalization ability of the model is also explored. The results show that the road surface temperature is significantly correlated to the environmental meteorological factors, but show different feactures in different regions, seasons and time periods. The independent test shows that there is no much difference among the model results based on different methods. All the model results can well reproduce the daily changes of road surface high temperature in summer and road surface low temperature in winter. The mean error in winter is significantly lower than that in summer. The application of satellite radiation products has a significant improvement effect on the road surface high temperature observation model results in summer. The model has good spatial adaptability. However, compared with the model constructed according to local observations, the error of the model using data from nearby traffic stations in the same climate region shows an increasing trend at different degrees, of which the error increasing in Beijing-Tianjin region is the smallest.