Abstract Mathematical models can play an important role in assessing the air quality implications of urban planning by linking the emissions that result from specific land uses to ambient concentrations of those pollutants. Where motor vehicles are a dominant source of air pollutant concentrations, it is necessary to predict the dispersal of air pollutants from road networks under average emission and meteorological conditions. It can also be important to identify extremes or worst case conditions that may occur in any given year, particularly where these extremes relate to health guidelines used for urban planning. This paper examines an approach that can satisfy these requirements for providing air quality advice to the planning of urban areas. A hybrid approach is outlined which combines a deterministic model with statistical techniques for estimating the frequency distribution of concentrations from the model output and the analysis of historical pollutant concentrations. The particular method described is a very simple but useful implementation of the general hybrid approach. The deterministic model component requires only basic traffic and meteorological measurements to infer estimates of mean annual concentrations. The statistical component allows inference of the frequency distribution of concentrations about this mean, including extreme values such as the maximum. Results are illustrated for 24-h average measurements of carbon monoxide over annual periods in Canberra, although the method can be applied to predict seasonal values of shorter term concentrations. The hybrid approach can also be applied with more detailed deterministic models, should there be adequate meteorological and other information to drive them.