Data from centrally-sited monitors are often used to characterize the exposure of humans to air pollution in urban areas, but ambient concentrations of many air pollutants can vary widely. Because many important spatial details at fine scales (especially in hot spot regions and near major roadways) cannot be derived from monitoring data, they can be augmented with air quality modeling data.

In this paper, we explore a method for for improving exposure assessments of air pollution with Eulerian grid scale models, such as the Community Multiscale Air Quality (CMAQ) modeling system. Such models provide deterministic outcomes of concentration as single-values for each grid cell across a modeling domain, and the spatial resolution is dependent on grid size. Several major limitations apply to applications of such models. First, modeled concentration details at a resolution smaller than that of the finest model grid is unavailable; and second, the magnitudes and spatial gradients of each modeled pollutant species is dependent on the size of the grid. Of practical significance, these limitations affect the so-called “change of support” problem (statistical terminology for inferences of non-representativeness of point measurements to grid values) in model evaluation studies, and they can constrain the use of air quality models for regulatory implementation and exposure assessments. A third limitation is that direct model simulations at fine grid sizes is currently too computer resource intensive for practical assessment analyses in urban areas.

To address these concerns, we consider the value of describing and then introducing parameterized sub-grid scale concentration variability (SGV). Our approach is to derive and parameterize pollutant SGVs as sub-grid concentration distribution functions (SCDFs) from a combination of fine-scale CMAQ and hybrid modeling results (based on combined local-scale dispersion model and CMAQ outputs) approaches for each grid cell. We then introduce a method for incorporating these SCDFs with operational CMAQ simulations. For short and long-term population exposure assessments, we briefly review and discuss SCDF parameterizations for diurnal distributions on a daily and seasonal basis. Our preliminary results suggest that the Weibull distribution provides a reasonable framework for describing and parameterizing SCDFs to CMAQ on a grid-by-grid basis.