APPLICATION OF AN ARTIFICIAL NEURAL-NETWORK TO IMPROVE SHORT-TERM ROAD ICE FORECASTS

This paper describes how a three-layer artificial neural network (NN) can be used to improve the accuracy of short-term (3-12 hours) automat ic numerical prediction of road surface temperature, in order to cut w inter road maintenance costs, reduce environmental damage from oversal ting and provide safer roads for road users. In this paper, the traini ng of the network is based on historical and preliminary meteorologica l parameters measured at an automatic roadside weather station, and th e target of the training is hourly error of original numerical forecas ts. The generalization of the trained network is then used to adjust t he original model forecast. The effectiveness of the network in improv ing the accuracy of numerical model forecasts was tested at 39 sites i n eight countries. Results of the tests show that the NN technique is able to reduce absolute error and root-mean-square error of temperatur e forecasts by 9.9-29%, and increase the accuracy of frost/ice predict ion. (C) 1998 Elsevier Science Ltd. All rights reserved.