Developing spatiotemporally resolved air pollution concentrations for epidemiologic study of multiple pollutants in lightly populated low-to-moderate pollution environments

Abstract

Background: Investigation of air pollution health effects in areas with lower populations have historically been hindered by limited availability of monitoring data and/or lack of modeled data. Objective: Develop spatiotemporally resolved air pollution predictions by fusing data available from ambient air monitoring networks with air quality model outputs. Method: Daily air pollution data was obtained from US Environmental Protection Agency (EPA) monitoring networks across the state of South Carolina and from Community Multi-scale Air Quality (CMAQ) outputs for multiple pollutants during 2003 to 2009. Pollutants include carbon monoxide, oxides of nitrogen, ozone, sulfur dioxide, particulate matter (PM2.5), and PM2.5 components: elemental carbon, ammonium, and sulfate. Air pollution predictions are developed using a temporal kriging approach that 'fuses' the data from the ambient monitoring network with CMAQ. Results: Availability of monitoring data varies substantially by day by pollutant with greatest coverage for PM2.5 and O3 and the least coverage for CO, NOx, EC, NH4, and SO4. Examination of modeled surfaces reveals heterogeneous patterns of variation for each pollutant. Preliminary models show prediction accuracy generally reflects data availability and spatial variation of modeled data. Conclusion: Such data improve understanding of the spatial distribution of air pollution and aid epidemiologic study of air pollution.