Spatiotemporal modeling of PM2.5 oxidative potential using source impact and model fusion techniques

Abstract

Exposure to elevated levels of air pollution can lead to cardiorespiratory disease, birth defects, and cancer. However, observational air quality data are spatially and temporally sparse due to high cost of monitors, limiting the scope of epidemiologic analyses and introducing error in exposure assessments. This dissertation presents the development, evaluation, and applications of multiple mathematical and computational modeling approaches for estimating spatiotemporal trends in air pollutant concentrations where and when data is not available for use in health studies. Specifically, source apportionment techniques with multivariate regression analyses are used to estimate long-term (years 1998—2010) and large-scale (eastern US) spatiotemporal trends in a novel pollutant metric, fine particulate matter (PM2.5) oxidative potential measured with a dithiothreitol assay (OP_DTT). OP_DTT measures a particle’s ability to catalytically generate reactive oxygen species while simultaneously depleting a body’s antioxidant defenses, leading to oxidative stress and, in turn, inflammation in the respiratory tract and cardiovascular system. Results show that biomass burning and vehicle sources are significant contributors to OP_DTT and that OP_DTT exposure presents higher risk ratios for asthma/wheezing and congestive heart failure emergency department visits than PM2.5 mass. Additionally, statistical downscaling techniques and model fusion approaches are developed to simulate fine-scale spatiotemporal trends (250m resolution) in air pollutant concentrations (OP_DTT, PM2.5, carbon monoxide, and nitrogen oxides) in Atlanta, GA. These methods estimate steep spatial gradients in pollutant concentrations near roadways that monitors and regional air quality models with coarse grid resolutions do not capture. The models developed in this dissertation can estimate concentration fields of air pollutants, including the novel pollutant metric OP_DTT, at regional and local scales, making them valuable tools for current and future epidemiologic and environmental justice research.