The CALMET meteorological model and its puff dispersion model CALPUFF were used to predict dispersion of the sulfur dioxide emissions from industrial and domestic heating sources in Izmir, the third biggest province in Turkey. The modeling domain covered an area of 80 x 100 km centered at the metropolitan area of Izmir with grid spacing of 1000m. Statistical analyses were carried out to evaluate the model performance by comparing the predicted and measured time series of sulfur dioxide concentrations at four monitoring stations using two main methods: root of the mean square error (RMSE) and an index of agreement (d). The index of agreement varied from 0.51 to 0.77 at four monitoring stations and the total RMSE ranged from 0.36 to 0.66 for the year 2000. The overall model performance for four monitoring stations was found good with an accuracy of about 68%. The agreement of model predictions and measurements was better for two urban monitoring stations (Karsiyaka and Bornova), compared with the other urban stations (Alsancak and Konak). (C) 2003 Elsevier Science Ltd. All rights reserved.