Air quality model validation: application to the San Francisco Bay Area and St. Louis

Complex urban-scale photochemical models are increasingly being used to predict air quality and the impacts of alternative emission control strategies. Yet confusion about acceptable levels of performance for these models still exists. A workshop was convened by the American Meteorological Society, for example, for the purpose of reviewing and recommending procedures for evaluating air quality models. They recommended a set of statistical measures for testing the accuracy of models but did not specify the accuracy required for acceptance or validation. This paper addresses the question of what quality of model performance is necessary for model validation. To study this question, we applied our urban-scale photochemical model to both the San Francisco Bay Area and St. Louis. While model performance is judged to be adequate, more work is needed to fully validate the model. This is because model performance is limited by the accuracy and representativeness of the input data.