Detailed traffic data collected from seven major roads in the city of Athens, Greece are presented and analysed in this study. Vehicles are split into seven categories while vehicle speed is also recorded. Based on these data the emissions of five major pollutants (CO, Benzene, NO(X), PM(10) and VOCs) were calculated with the aid of the COPERT methodology and, based on these results, an Artificial Neural Network was also developed. The results of the two methodologies were compared and it was found that the differences were very small. The ANN model seems to be a reliable alternative to calculate road traffic emissions in a busy road environment. The results reflect the spatial and temporal distribution of the concentrations of the pollutants examined. Alternative "what if" scenarios of the fleet distribution were also applied by means of environmental policy. Since Athens experiences low air quality conditions the correct estimation of traffic emissions is crucial since they play a significant role in the design of an environmental abatement strategy.