Using Chemical Transport Model Predictions to Improve Exposure Assessment of PM_2.5 Constituents

Air pollution health-effect studies commonly use central monitor concentrations (CMCs) of airborne fine particulate matter (PM_2.5) to represent population exposure near the monitoring sites. The spatial distribution of PM_2.5 constituents is presumed to be the same and is well-represented by the CMC. Here we apply chemical transport models in California and show that the population-weighted concentrations (PWCs) of secondary PM_2.5 constituents within the 12 km buffer zone are within ±20% of the respective CMC values, but the PWCs of primary PM_2.5 constituents differ from the CMC values by -40 to +60%. The appropriate CMC representative distance varies significantly in different cities due to the unique combination of population distribution, emissions patterns, and meteorology at each location. We conclude that exposure misclassification can be significant if the same representative distance is assumed for multiple CMC PM_2.5 constituents across all sites in a single air pollution epidemiology study that has a large spatial and temporal range. This misclassification will increase the variance around the effect estimate and therefore reduce the likelihood of finding a statistically significant effect.