Atmospheric pollution models are generally evaluated using measures of the difference between predicted and observed concentrations. These statistics can be inadequate as they tend to be sensitive to small errors in the timing of peak events that are generally of little consequence from a health point of view. This paper presents a technique for evaluating atmospheric carbon monoxide models based on the statistical properties of the uptake of carbon monoxide, rather than relying on traditional model evaluation measures. A simple semi-empirical atmospheric carbon monoxide model is used as an example to illustrate the approach.
