The physical assumptions in various models for the diffusion of inert substances in urban areas are critically reviewed. It is suggested that there are no significant differences among the predictions of the models, and that often simple integral models have more valid physical assumptions than do complex computer models. Correlations between observed and predicted concentrations of carbon monoxide (CO), sulfur dioxide, and suspended particles are typically in the range of 0.4 to 0.8. The magnitude of the prediction can be expected to be within a factor of two of the observation.

Mixing depth and pollutant concentration are not well correlated (R ≃ -0.2) because local pollutant plumes do not as a rule reach the capping inversion. Inverse wind speed and pollutant concentration are better correlated (R ≃ 0.6). A method for accounting for stability in the simple Atmospheric Turbulence and Diffusion Laboratory model is suggested, based on observations of CO emissions and concentrations and of wind speeds at numerous stations.