Source-specific air pollution in areas and periods with limited data: application to 1940 in the United States

Abstract Studies on the relationship between air pollution and health in the United States primarily use post-1970 data. Estimating earlier exposures enables health and environmental justice research before widespread air quality observations. Approaches for estimating source-specific exposure must balance data availability, computational demand, and the spatial/temporal resolution of health data. We developed methodologies to quantify spatial variability in source-specific air pollution exposure in years 1940 and 2010, the latter with available monitoring data. Total fine particulate matter (PM 2.5 ) and ozone (O 3 ) concentrations were estimated using climate model output from the Coupled Model Intercomparison Project Phase 6 (CMIP6). Annual average emissions were assessed for major sources: automobiles, power plants, oil and gas wells, and residential/commercial coal use. Spatial exposure variability was estimated at the county level using emissions totals or, for elevated sources, inverse distance weighting or dispersion modeling. Pearson correlation coefficients and mean differences compared 2010 estimates against EPA Air Quality System monitor observations. From 1940 to 2010, nationwide average exposure to PM 2.5 , automobile emissions, and residential/commercial coal combustion decreased, while O 3 , power plant emissions, and oil/gas well development increased. Spatial distributions also shifted: in 1940, automobile emissions were concentrated in urban centers, industrial hubs, and coal-burning regions; by 2010, regulatory actions and technological improvements led to more diffuse patterns with fewer extreme hotspots. This ‘flattening’ reflects more uniform pollution levels across regions, driven by reduced localized emissions and pollutant dispersion from mobile and elevated sources. This study demonstrates that reduced-complexity approaches can estimate historical exposure variability in periods with limited data and provides methods for evaluating results using modern observations.