Evaluating regional-scale air quality models

Numerical air quality models are being used to understand the complex interplay among emission loading, meteorology, and atmospheric chemistry leading to the formation and accumulation of pollutants in the atmosphere. A model evaluation framework is presented here that considers several types of approaches, referred to here as the operational evaluation, diagnostic evaluation, dynamic evaluation, and probabilistic evaluation. The operational evaluation is used to reveal the overall performance of the model, and diagnostic evaluation approaches are then used to identify what processes and/or inputs significantly influence the predictted concentrations and whether they are simulated correctly. Dynamic evaluation entails assessing a model's ability to reproduce observed changes in pollutant concentrations stemming from changes in weather and emissions. Probabilistic evaluation approaches will provide the confidence that can be placed on model results for air quality management or forecasting applications. Here, we present example results from several different model evaluation studies that consider questions related to the operational, diagnostic, and dynamic evaluation of a model, and discuss their complementary goals toward model improvements and characterization of model performance.