The impact of human-induced perturbations on the chemical state of the atmosphere has received a great deal attention during the last two decades, from acid deposition, to elevated tropospheric ozone, to the impact of greenhouse gases.  Scientific efforts to understand these atmospheric conditions have involved a combination of laboratory, field and numerical modeling experiments.  The complexity of physical and chemical atmospheric processes, combined with the enormity of the atmosphere, make results obtained from laboratory and field experiments difficult to interpret without a clear conceptual model of the workings of the atmosphere.  Atmospheric models are thus used to integrate and synthesize our evolving knowledge of various atmospheric processes.  Modern atmospheric air quality models, such as EPA's Community Multiscale Air Quality Modeling system, are essential tools used to explore the future state of the atmosphere and formulate effective abatement strategies.  Over the past decade, an increase in the number and duration of air quality model simulations has led to an increase in the time required to thoroughly evaluate each simulations.  A wide variety of graphical, exploratory and statistical modeling techniques are used to assess model performance in order to identify needed improvements in the model parameterizations and inputs, as well as to quantify the level of confidence in the model-predicted values.  For a statistician, the task of model evaluation on this scale requires interacting with scientists from a wide variety of scientific fields, becoming familiar with the various components and inputs of the numerical model and the observational data used in an evaluation, and the ability to communicate statistical ideas in way that is relevant to the atmospheric modeling community.