Large global variations in measured airborne metal concentrations driven by anthropogenic sources

The study investigates the global distribution, sources, and health implications of trace metals in fine particulate matter (PM2.5). While PM2.5 is a well-established risk factor for cardiovascular, respiratory disease, cancer, and diabetes, the specific contribution of PM2.5 components—particularly trace metals—remains insufficiently characterized globally. Metals can drive oxidative potential and oxidative stress due to redox-active species and have been linked to adverse health outcomes (e.g., cardiovascular effects associated with K, Al, Ni, Zn, V; bioaccumulation of As, Pb, Al). Many metals such as As, Cd, and Cr are known human carcinogens, and Pb is associated with impaired cognitive function. Ground-based composition measurements are sparse, with many regions lacking PM2.5 metal data (more often PM10 is measured). Understanding metal distributions can inform exposure assessment, identify emission sources, and improve chemical transport models. The SPARTAN network provides globally consistent PM measurements, enabling assessment of PM2.5 trace metal concentrations and their enrichment relative to crustal sources to evaluate anthropogenic contributions.

The paper synthesizes prior evidence linking PM2.5 composition to health effects, emphasizing the oxidative potential associated with metal content and documented epidemiological associations with specific elements. It reviews known source signatures: K (biomass burning), Zn (traffic/tire wear), V (heavy fuel oil/shipping), coal (Pb, Cr, Mn, As, Se), non‑ferrous metal production (As, Cd, Zn), and vehicle traffic (Ba, Zn, Pb, Fe, Al, Mg, Ti). The authors highlight the scarcity of global PM2.5 metal measurements and the absence of a comparable global network dedicated to PM2.5 trace metals. They also situate findings with respect to European background sites, where heavy metal concentrations are generally lower than in densely populated SPARTAN locations, reinforcing observed enrichments at urban sites.

• Network and site selection: SPARTAN targets densely populated, globally dispersed, under‑sampled regions while favoring representative urban environments (e.g., rooftops) to minimize local anomalies. Nineteen sites across four continents contributed PM samples, with average site sampling duration ~20 months (range 2–50 months). Over 800 PM2.5 filters (as of Oct 2019) were analyzed.
• Sampling instrumentation and protocol: Initially AirPhoton SS4i samplers with two-stage stacked filter units collected PM10 and PM2.5 over nine-day periods on sequential pairs of coarse Nuclepore and fine Teflon filters plus traveling blanks. From late 2017, stations were upgraded to AirPhoton SS5 with cyclone inlets separating PM2.5 and PM10 at 5 and 1.5 L/min, respectively, using eight stretched Teflon filters per cartridge (six PM2.5, one PM10, one blank) pre-assembled centrally. International sites sampled each PM2.5 filter in rotating 3‑h spans across 9 days (24 h total), capturing diurnal variability; PM10 filters sampled 30 min after each PM2.5 sample across 54 days (24 h total). For low-PM MAPLE sites, filters were sampled for 48 h (PM2.5: 6-h periods; PM10: 1-h periods).
• Laboratory analyses: Filters were returned to Dalhousie for gravimetric mass, water-soluble ions (ion chromatography), black carbon (smoke-stain refractometry), and trace metals by ICP‑MS (Thermo Scientific X‑Series 2). For metals, filters received 10–30 µL isopropyl alcohol and were extracted at 97 °C for 2 h in 5% trace metal grade HNO3 (method akin to Fang et al. and Herner et al.). Calibration used 25–500 µg/L standards and internal standards 45Sc, 115In, 159Tb. Field blank concentrations from each cartridge were subtracted from corresponding samples. Extraction efficiencies for some crustal elements (e.g., Fe) can be as low as ~50%; more recent analyses include HCl to improve recovery, with future XRF planned at Washington University.
• Quality assurance and uncertainty: Whole-system uncertainties were estimated via collocated samplers at low (Halifax), moderate (Toronto), and high (Beijing) PM sites over ~3 weeks (24–48 h samples) with results summarized in supplemental materials. A joint campaign with the IMPROVE network was conducted for independent comparison.
• Data analysis: Elements with concentrations above ICP‑MS detection limits in ≥10% of samples were retained (Li, Co, Ag, Ce excluded). Crustal enrichment factors (EFs) were computed relative to continental crust composition (Taylor & McLennan) using Fe as reference: EF_X,2.5 = ([X/Fe]_PM2.5)/([X/Fe]_Taylor. To mitigate potential Fe extraction bias, a complementary EF normalized by estimated coarse PM (PMc*, a soil reconstruction using Al, Si, Ca, Fe, Ti) was computed: EF_X,2.5 = ([X/PMc]_PM2.5)/([X/PMc]_Taylor. Relative abundances (RA) of metals were calculated versus a natural low‑trace metal site (Mammoth Cave): RA_X = [X]_site / [X]_Mammoth Cave.

• PM2.5 mass concentrations: Among 19 sites, Kanpur had the highest mean PM2.5 at 102.8 ± 20.2 µg/m³ (SE), followed by Beijing 58.1 ± 2.2 µg/m³, Dhaka 49.0 ± 3.3 µg/m³, and Hanoi 47.1 ± 7.6 µg/m³. Canadian sites generally had the lowest PM2.5 levels; Mammoth Cave had relatively high PM2.5 due to biogenic/organic aerosol with low trace metal fraction (~4%).
• Crustal enrichment: Metals showed large EF variability by element and location. Strong enrichments (EF > 100) indicative of anthropogenic sources were observed for Pb, As, and Zn, especially at high‑PM2.5 sites. Similar EF patterns were obtained whether normalizing by Fe or PMc*, supporting robustness.
• Health guideline exceedances: • Lead (Pb): US NAAQS 3‑month mean standard 150 ng/m³ was exceeded at Dhaka and Kanpur. Mean Pb in PM2.5: Dhaka 280 ng/m³ vs Kanpur 209 ng/m³; fraction of samples above 150 ng/m³: Dhaka 38% vs Kanpur 30%. PM10 site means were higher than PM2.5 by 41% (Kanpur) and 74% (Dhaka), implying greater TSP Pb. • Arsenic (As): WHO 1:100,000 excess lifetime cancer risk level (6.6 ng/m³) was approached/exceeded at Kanpur, Hanoi, Beijing, and Dhaka. Site mean As in PM2.5: Kanpur 15.3 ng/m³; Hanoi 8.1 ng/m³; Beijing 7.1 ng/m³; Dhaka 6.3 ng/m³.
• Relative abundances (RA) vs Mammoth Cave: High RA (12–311) for anthropogenic elements Zn, As, Pb, Cd at Kanpur, Beijing, Dhaka, Hanoi. Site-specific signatures included: • Beijing: markedly elevated Se (coal-related), with elevated As, Cd, Pb consistent with coal emissions. • Bandung: very high Pb (RA ~38, EF > 700), likely from lead-acid battery recycling/smelting. • Ilorin: elevated Cr (RA ~40, EF > 260), consistent with regional tanning industry emissions. • Singapore: strong V enrichment (RA ~49; EF > 240) linked to shipping/refinery emissions; elevated Zn (RA ~13; EF > 600); minimal enrichment of Pb, As, Cr reflecting limited coal burning. • Several North American sites (Toronto, Bondville, Sherbrooke, Atlanta, Halifax, Kelowna, Lethbridge) exhibited RA < 5 for all elements.
• Element detection and composition: 15 metals above detection thresholds were analyzed (Li, Co, Ag, Ce excluded). Trace metal mass fractions of PM2.5 varied widely across sites; combustion-related elements tended to co-vary, as did crustal components.

The study demonstrates that spatial variability in PM2.5 trace metal concentrations is primarily driven by anthropogenic activities, with the largest enrichments and health-relevant concentrations occurring in densely populated urban centers. High enrichments for Pb, As, Zn, and Cd align with known sources such as coal combustion, non-ferrous metal production, traffic-related wear, shipping fuel oil, and industrial processes (e.g., smelting, tanning). The observed exceedances of Pb and elevated As at several sites underscore potential public health risks, reinforcing the need for targeted mitigation policies addressing specific emission sources (e.g., coal use, battery recycling, shipping fuels). The agreement between Fe‑based and PMc*‑based EFs strengthens the attribution of observed enrichments to anthropogenic sources. Comparison with a natural reference site (Mammoth Cave) via RA clarifies site-specific source influences (e.g., coal in Beijing, shipping/refineries in Singapore, smelting in Bandung, tanning in Ilorin). Findings provide valuable constraints for global chemical transport models and exposure assessments and highlight priority regions for intervention.

Using a consistent global measurement framework (SPARTAN), the study quantifies substantial anthropogenic enrichment of trace metals in PM2.5 across 19 sites, with especially high levels in large, densely populated cities (Beijing, Dhaka, Kanpur, Hanoi). Health-relevant thresholds were exceeded or approached for Pb and As at multiple sites. These results emphasize the global relevance of anthropogenic contributions to fine particulate metals and motivate expanded monitoring in under-sampled regions and rapidly developing economies. The work supports improvements to emission inventories and chemical transport modeling of trace metals. Future work will expand the SPARTAN network coverage, enhance seasonal representativeness, and improve analytical methods (e.g., inclusion of HCl in extraction and XRF analyses) to refine quantification of metal concentrations and sources.

• Analytical extraction bias: Nitric acid extraction efficiencies for some crustal elements (e.g., Fe) can be as low as ~50%, potentially affecting absolute concentrations and EF calculations; complementary normalization to PMc* and planned method improvements (HCl addition, XRF) mitigate but do not eliminate this limitation.
• Temporal representativeness: Some sites had limited sampling duration or seasons, which may reduce annual representativeness despite the rotating sampling design intended to capture long-term averages.
• Spatial coverage: SPARTAN lacks sites in certain regions (e.g., Europe within this study period), limiting global generalizability; future expansion is planned.
• Size fraction comparison: Health guidelines (e.g., Pb NAAQS) are for TSP, while measurements are in PM2.5 and PM10; translation between size fractions introduces uncertainty in direct compliance comparisons.
• Reference composition variability: Use of Fe as a crustal reference assumes relatively consistent soil Fe content; natural variability and anthropogenic Fe could influence EF, though PMc* normalization provides a robustness check.