Comparison and Performance Evaluation of Models Used for Vehicular Pollution Prediction

In this paper, an evaluation of four Gaussian dispersion models-GM, CALINE3, PAL2, and ISCST2-commonly used for vehicular pollution prediction is presented. The evaluation has been done using field data monitored at one of the metropolitan cities, Chennai (Madras), India. Various statistical parameters that indicate the modelʹs accuracy in prediction have been calculated. The relative accuracies of the models are compared. Such an analysis will be of use to regulatory agencies in selection of an appropriate model with confidence. As per the analysis conducted in this study, the model GM proved to be the best among the four followed in order by CALINE3, ISCST2, and PAL2. The prediction by all four models was comparable when the windroad angle was perpendicular with wind speed being 2-5 m/s and the atmospheric stability condition being neutral. The prediction was the most accurate for the case of ground level receptor-perpendicular wind direction combination. The model GM provided the best simulation for all conditions tried. The model CALINE3 was the next best but skewed toward underprediction, whereas the models ISCST2 and PAL2 were found to overpredict.