The application of artificial neural networks in prediction of

Air pollution is one of the main problems in metropolitan areas. Therefore, the prediction of air pollution can be regarded as one of the important issues of air quality research in urban areas. There are many methods for the prediction of air pollutant concentration. In recent years, there has been considerable progress in the development of neural network models for air quality prediction. In this paper, artificial neural network has been used for the prediction of CO concentration in Tabriz. Hourly correlation between the concentration of CO and metrological variables was calculated by multilayer perceptron for the months of December and January 2003. Multilayer perceptron performance was compared with traditional methods such as auto-regressive integrated moving average (ARIMA). The results indicated that there was a high non-linear correlation between CO concentration and metrological variables such as: speed and direction of wind; and relative humidity. The location of large industries in the west and southwest of Tabriz were found to be the most effective elements in air pollution. Blowing of winds from west and southwest directions, on the months of December and January have caused pollution are transferred to the inner parts of Tabriz.