Development of an artificial neural network to predict benzene concentrations in a street canyon

Nowadays, the prediction of atmospheric pollutant concentrations in street canyons' environment is of great importance. To achieve this, many kinds of modeling techniques were proposed. One of the most promising techniques is Artificial Neural Networks (ANNs). In this study, an ANN was developed to predict benzene concentrations in a heavily trafficted street canyon. It also evaluates the importance of the variables determining these concentrations. The training procedure was developed based on data collected by an annual measurement's campaign, performed in a specific street canyon. The data include benzene concentration, traffic flow and speed, vehicle's type distribution, wind speed and direction. The results from the simulations indicate that ANN is a promising technique for predicting benzene in an urban environment, and can be used for environmental management purposes. Fresenius Environmental Bulletin