The Preliminary Applications of Numerical Weather Prediction and Big-data Machine Learning Methods in Winter Road Weather Conditions Prediction System

The weather related accidents account for 30% of the road accidents in China, especially at the winter seasons. The road authorities used to make the decisions based on experiences and the happened weather conditions. The Beijing Piesat Co., Ltd. is developing a road weather prediction and decision support system to help road authorities to improve the road maintenance efficiency and to decrease the cost. The core part of this system is the road weather condition prediction system, which is consisted of data assimilation, high-resolution numerical weather prediction, and the big-data machine learning post-processing system. The system is able to predict the road surface temperature, road slippery index, road icing. The road surface temperature is the basic information for predicting the icing, slippery, etc., It is shown that with more accurate representation of the low road surface temperature could be more accurately predicted for the road conditions. The tests also demonstrate the important role the big data machine learning post processing played in improving the road temperature accuracy and suggest that the proper machine learning post-processing method is critical in the winter road prediction systems. The system will be used in this coming winter season for one busy high-speed way authority in China. The presentation/poster will explain how the results are produced and will show the selected verifications.