Air pollution models help us to understand the way air pollutants behave in the environment. In principle, a perfect model would enable the spatial and temporal variations in pollutant concentration to be predicted to sufficient accuracy for all practical purposes, reducing the need for measurements. It is noteworthy, however, that a model is no use unless it has been validated to show that it works. Hence, the process of model development goes hand in hand with developments in measurement. There are many reasons for using models, such as for regulatory purposes in establishing which sources are responsible for what proportion of concentration at any receptor, estimating population exposure on a higher spatial or temporal resolution than is practicable by measurement, targeting emission reductions on the highest contributors, predicting concentration changes over time. There are four main families of models:

Dispersion models, which are based on a detailed understanding of physical, chemical and fluid dynamical processes in the atmosphere. They enable the concentration at any place and time to be predicted if the emissions and other controlling parameters are known.

Receptor models, which are based on the relationships between a data set of measured concentrations at the receptor and a data set of emissions that might affect those concentrations.

Stochastic models, which are based on semi-empirical statistical relationships between the pollutant concentrations and any factors that might affect them, regardless of the atmospheric physical processes.

Compartment or box models, in which inputs to, and outputs from, a defined volume of the atmosphere are used to calculate the mean concentration within that volume.