Unequal airborne exposure to toxic metals associated with race, ethnicity, and segregation in the USA

The study addresses whether exposure to toxic fine particulate matter (PM2.5) metals varies systematically with racial residential segregation (RRS) in the United States and how these disparities compare to disparities in total PM2.5 mass. PM2.5 exposure is a major contributor to disease burden, and prior work shows higher total PM2.5 in communities with higher proportions of persons of color and lower socioeconomic status. PM2.5 is chemically heterogeneous, and certain constituents (notably trace metals) have stronger toxicity profiles. However, most environmental justice studies have focused on total PM2.5 mass rather than toxic components, and often on neighborhood racial composition rather than segregation metrics that capture broader urban patterns. The authors focus on the dissimilarity index (DI) quantifying segregation between non-Hispanic Black (NHB) and non-Hispanic White (NHW) residents, hypothesizing that higher DI is associated with higher concentrations and mass proportions of toxic trace metals in PM2.5, and that disparities are larger for metals from anthropogenic sources than for naturally derived metals.

Prior research documents disproportionate exposure to higher total PM2.5 concentrations in communities with higher percentages of racial/ethnic minorities and those of lower socioeconomic status, with links to adverse outcomes such as cancer, asthma, cardiovascular disease, and mortality. Racial residential segregation (RRS) has been associated with increased risks for infant and all-cause mortality, cardiovascular disease, COVID-19 mortality, and pregnancy complications. Some recent studies link RRS with higher total PM2.5 exposure and increased health risks, but evidence connecting RRS to specific toxic PM2.5 components is limited. Trace metals in PM2.5 can drive oxidative stress via reactive oxygen species generation and are linked to respiratory and cardiovascular morbidity and hospitalizations; several (e.g., Pb, Ni, Cr, V, Ti) are known or suspected carcinogens, and others (e.g., Al, Pb, Mn, Fe, Cu) have neurotoxic potential. Most environmental injustice studies to date have relied on total PM2.5 mass rather than its toxic components, leaving a gap this study aims to address.

- Data sources: Obtained PM2.5 chemical speciation data (trace metals: Cu, Zn, Ni, Cr, Pb, V, Fe, Mn, Ti) and total PM2.5 from the US EPA Chemical Speciation Network (CSN) and IMPROVE networks for 2000–2019. Data were downloaded from the Federal Environmental Manager Environmental Database. Only valid samples were retained; monitors reporting >50% of expected every-third-day measurements (~121/year) were included. Measurements below the minimum detectable limit (MDL) were retained; sensitivity analysis imputed MDL/√2 for values below MDL. - Metals grouping: Classified Cu, Zn, Ni, Cr, Pb, V as predominantly anthropogenic and Fe, Mn, Ti as predominantly natural tracers based on literature source apportionment. - Spatial and urban/rural context: Monitors were classified as urban or rural using US Census designations. For counties with multiple monitors, annual values were averaged. - Racial residential segregation (RRS): Computed the dissimilarity index (DI; range 0–1) using American Community Survey 5-year data at census-tract (unit) within county (reference) scales. Primary analysis used NHB (minority) vs NHW (reference); supplemental analyses considered Hispanic, Asian, and Native American vs NHW. Counties with only one census tract were excluded. - RRS categories: Counties were categorized as well-integrated (DI 0–0.3), moderately segregated (0.3–0.6), and highly segregated (0.6–1). - Outcomes: Annual mean concentrations and PM2.5 mass proportions of each metal. Population-weighted means computed across counties within each RRS category using county population weights. Relative disparities across RRS categories quantified via coefficient of variation of population-weighted means, enabling comparison independent of absolute magnitude. - Statistical analyses: Univariate linear regression with log-transformed concentrations tested associations between DI and pollutants, expressed as percent change in concentration per 10% increase in DI. Models were stratified by urban/rural status and adjusted for geographic region fixed effects. Extended models added county percent NHB (or percent NHW) to assess joint associations. Statistical significance assessed at p<0.05. Analyses conducted in Python (Pandas, Statsmodels).

- Overall disparities: Residents of racially segregated communities are exposed to PM2.5 with over three times higher mass proportions of known toxic/carcinogenic metals. Total PM2.5 concentrations are roughly two times higher in racially segregated communities, while concentrations of metals from anthropogenic sources are nearly ten times higher. - Geographic patterns: Pb (anthropogenic) shows strong spatial clustering with elevated concentrations in the industrial Midwest/Ohio River Valley and lower levels in Western/Mountain/Border states. Fe (natural) exhibits weaker spatial dependence, reflecting significant natural dust contributions, especially in the Southwest. - Association with segregation (DI): Across all sites, a 10% increase in DI is associated with larger increases in metals from anthropogenic sources (about 9–16% higher concentrations) than in metals from natural sources (about 4–7% higher) and total PM2.5 (about 5% higher). For Pb specifically, concentrations increase by about 9% per 10% DI increase. These associations persist after controlling for geographic region and when restricted to urban sites: in urban counties, a 10% DI increase is associated with about +5% Pb, +10% Zn, and +10% Cr; Fe shows a smaller positive association and Ti shows no increase. - Race/ethnicity composition: Counties with above-average NHB population share have consistently elevated fine particulate metal concentrations compared with counties with higher NHW or Native American shares. In models including both DI and population composition, anthropogenic metals increase by roughly 4–8% per 10% DI increase and 4–6% per 10% increase in NHB population; in contrast, they decrease by about 4–10% per 10% increase in NHW population (not always statistically significant), indicating stronger burdens in high-DI, high-NHB counties. - RRS category contrasts: Metal concentrations are elevated in highly segregated counties relative to moderately segregated and well-integrated counties. Lead and iron concentrations are approximately twice as high in highly segregated vs well-integrated counties. Mass proportions of anthropogenic metals are markedly higher in highly segregated counties: on average 3–12 times higher than well-integrated and 1.5–18 times higher than moderately segregated counties (with some CIs overlapping 1 for individual metals). Mass proportions of natural-source metals are more similar across categories; Fe mass proportion is about twice as high in highly segregated vs well-integrated counties, while Mn and Ti show weaker differences. - Relative disparities: Population-weighted relative disparity across RRS categories is higher for metals associated with anthropogenic emissions than for total PM2.5. For total PM2.5, relative disparity is about 0.33 (95% CI ~0.32–0.47), whereas anthropogenic metals (e.g., Pb) exhibit substantially larger disparities. Disparity patterns have been broadly stable over the past decade for most metals. - Vanadium case study and policy signal: Following regulations limiting sulfur in marine fuel oil (2010–2015), V concentrations declined substantially, especially in coastal cities, with the greatest decreases in highly segregated counties. In 2010, V in highly segregated counties was about 2–9 times higher than in well-integrated and about 2 times higher than in moderately segregated counties. By 2019, highly and moderately segregated counties had similar V levels, and highly segregated counties were only about 3 times higher than well-integrated (with wide CIs). The relative disparity in V declined from about 0.60 (2010) to 0.40 (2019), indicating targeted emission reductions can reduce segregation-related disparities.

The findings demonstrate that racial residential segregation is strongly associated with elevated exposure to toxic trace metals in PM2.5 beyond the well-documented disparities in total PM2.5 mass. Because anthropogenic metal sources (industrial/metallurgical processes, traffic-related emissions, heavy fuel oil combustion) are concentrated in urban cores with historically high segregation, populations in highly segregated counties experience both higher total PM2.5 and a version of PM2.5 that is enriched in toxic metals, amplifying potential health risks. The associations persist when adjusting for geographic region and when considering only urban sites, and are exacerbated in counties with higher NHB population shares. The vanadium case illustrates that targeted emission controls can simultaneously reduce overall concentrations and narrow segregation-linked disparities, underscoring the potential of source-focused policies to mitigate environmental injustice related to PM2.5 metal exposure.

This study contributes evidence that racial residential segregation is associated with disproportionately higher concentrations and mass proportions of toxic, anthropogenically derived metals in PM2.5, with disparities exceeding those for total PM2.5 mass. The burden is greatest in highly segregated counties and in counties with larger NHB populations. Temporal analysis of vanadium suggests that targeted emissions policies can reduce both overall exposure and segregation-related disparities. Future work should expand monitoring coverage (especially in rural and low-DI areas), improve within-county spatial resolution, examine additional toxic components and sources, evaluate causal pathways linking source-specific reductions to disparity changes, and assess health outcome implications of component-specific disparities.

- Potential urban bias: Anthropogenic metal sources are more prevalent in urban areas, which also tend to have higher segregation, potentially confounding results; analyses restricted to urban sites and models controlling for region mitigate but may not eliminate this. - Monitoring network coverage: Limited spatial coverage and typically one monitor per county hinder assessment of within-county variability; counties with monitors have higher average DI than all US counties, under-representing low-DI areas. - Detection limits: Many trace metal annual means are near or below MDL; sensitivity analyses using MDL/√2 imputations did not materially change conclusions, but uncertainty remains. - Population weighting scope: Population-weighted means reflect only populations in counties with CSN/IMPROVE monitors and assume county monitor values represent county-wide exposure. - Segregation metric constraints: DI is invariant to tract spatial arrangement and sensitive to spatial scale; it captures residence only and not mobility or time spent in other neighborhoods. Focus on NHB vs NHW in main analyses may not capture all dimensions of multi-ethnic segregation, though supplemental analyses suggest similar patterns for other groups.