Keywords: atmospheric dispersion models, model evaluation, natural variability, air quality models, model comparison, air pollution, environmental pollution, modelling
Comparative evaluation of two air quality models: within-regime evaluation statistic
At the Workshop on Operational Short-Range Atmospheric Dispersion Models for Environmental Impact Assessment in Europe, Lee and Irwin (1994) presented a methodology for conducting an objective comparative evaluation of two or more air quality models. The methodology involved the use of bootstrap resampling to allow computation of confidence bounds on model comparison measures, and an inverse weighting to accommodate pooling of data from several field experiments. Mention was made that it would be useful to explore alternative choices for the primary within-regime statistic selected for defining the differences to be seen between the modelled and observed concentration values. In this discussion, we explore the differences to be seen in using the Robust Highest Concentration versus the median of the upper quartile of values as the primary statistic for defining differences between modelled and observed concentration values, and assess which might be less sensitive to natural variability (or uncertainty).