The primary objective of this Walter A. Rosenblith New Investigator Award study is to create fine-scale daily PM2.5 source impacts from major source and fuel categories from 2011-2020. This objective builds on recent work developing fine-scale daily air pollution exposure products in the United States, which have driven recent epidemiological studies on air pollution health effects.
Poster abstract for HEI Annual Conference 2023
Air pollution source impacts at fine scales for long-term regulatory accountability and environmental justice
Lucas RF Henneman and Ting Zhang
George Mason University, Fairfax, VA
Background. A growing body of evidence elaborates on differing adverse health impacts of source-specific ambient air pollutants. However, spatial and temporal variability in PM2.5 source impacts differs by source and fuel types, warranting the development of consistent approaches to quantify source impacts fields. The study is designed to 1) create fine-scale daily PM2.5 source impacts from major source and fuel categories from 2011-2020, and 2) use the source impacts fields to quantify exposure by population groups, with a particular focus on historically disadvantaged populations.
Methods. The first task has been to clean and prepare observation data for daily PM2.5 source apportionment using positive matrix factorization (PMF) at U.S. monitors. We downloaded observation datasets from two monitoring networks, the Interagency Monitoring of Protected Visual Environments (IMPROVE) and the Chemical Speciation Network (CSN). We assessed the completeness and continuity of downloaded data comprising 311 monitoring sites. We explored the spatiotemporal distribution patterns of PM2.5 species concentrations and reported flags. We compared the flagged and unflagged PM2.5 species concentrations and decided for each flag whether to either 1) keep the flagged data or 2) remove and refill it. For those with unacceptable flags, we evaluated the performance of four interpolation methods. In anticipation of applying multi-site PMF analysis, we clustered monitoring sites separately in the two networks with a random forest-based method. Finally, we quantified the influence of atmospheric dilution on PM2.5 species concentrations with a ventilation coefficient.
Results. We showed summaries of CSN and IMPROVE monitoring location data, including trends over time in concentrations and data flags, spatial coverage, and we discussed the implications of lab protocol changes on the CSN concentrations. In addition, we discussed the implications of filling and clustering the data on the planned PMF applications.