Spatial decomposition analysis of NO2 and PM2.5 air pollution in the United States

Abstract

Background: Length scales for spatial variability of air pollution concentrations depend on the pollutant and the location, from meters to 100s or 1000s of km. Spatial patterns of the concentration at a location could reveal information on pollution sources. Methods: Here, we develop a readily scalable algorithm based on “spatial-increment”, to decompose the air pollution concentration into four spatial components: long-range, mid-range, neighborhood, and near-source. We apply the algorithm to annual-average predicted concentrations of outdoor nitrogen dioxide (NO2) and fine particulate matter (PM2.5) for all (n ≈ 6 million) census blocks in the contiguous US, from national empirical models. We analyze the within-city patterns of the decomposed concentrations and summarize the national decomposition results by state, geographic region, and urban area size. We also calculate how much each empirical regression component contributes to the spatial decomposition results. Results: Our results show that for NO2, “neighborhood” and “mid-range” components dominate both with-city and inter-city concentration differences (both are ~5-fold larger in large urbanized areas than rural areas). Urban area size plays a more important role in the component concentrations than geographic regions. For PM2.5, the “long-range” component dominates; this component varies by region (e.g., three times greater in the Midwest [7 μg/m3] than in the West [2.3 μg/m3]), whereas variation by urban area size is relatively minor. Conclusions: Our study provides the first nation-level fine-scale decomposed pollution surfaces to date, that can be used to estimate, at least to a zeroth order, the contribution of sources at different distances from the receptor to the annual average pollution in the location of interest.