Large transboundary health impact of Arctic wildfire smoke

The study addresses how wildfires occurring within Arctic Council member states affect regional and transboundary PM2.5 exposure and associated mortality. Against the backdrop of rapid Arctic warming and increased fire activity in Siberia, the paper examines whether changing fire regimes are worsening air quality and health, and how spatial shifts in burning influence populations locally and abroad. The purpose is to quantify the wildfire-attributable contribution to PM2.5 and its chronic mortality burden, thereby informing Arctic environmental health policy and international cooperation on transboundary air pollution.

Prior work shows the Arctic has warmed faster than the global average, with Siberian warming particularly rapid. Although global burned area decreased by ~25% from 1997–2013 due to agricultural expansion and fire suppression, total fire carbon emissions have remained roughly stable (~2 Pg C yr−1) as forests (with higher emission intensity) contributed a greater share of burned area. Climate change is expected to increase boreal forest and peatland fire frequency and emissions in high latitudes (Siberia, Canada), potentially doubling or tripling fire-sourced carbon emissions by 2100. Wildfire smoke degrades air quality locally and remotely via long-range transport, producing primary PM, CO, and PAHs, and contributing precursors for ozone and secondary PM formation. Epidemiological evidence links PM2.5 exposure to elevated risks of cardiovascular and respiratory mortality; wildfire PM2.5 specifically increases all-cause mortality and may be more toxic than PM from other sources. Observational and modeling studies have documented transboundary impacts from Siberian wildfires on downwind regions including Japan, the Arctic, and midlatitudes.

Design: Two global chemistry–climate simulations with CESM v2.2 (CAM6-chem with MOZART-TS1 gas chemistry, MAM4 aerosols, VBS aging) at 0.9°×1.25° resolution, meteorology nudged to MERRA2. Emissions: Biomass burning from QFED v2.5 (FRP-based, AOD-calibrated). The study constructs two scenarios: (1) FIRE ON: all fires globally (wildfires and agricultural). (2) ARCTIC WILDFIRE OFF: wildfire emissions removed within Arctic Council areas (Canada, Greenland (Denmark’s Arctic territory included), Iceland, Norway, Sweden, Finland, Russia, United States (Alaska); contiguous U.S. and Denmark proper excluded), retaining agricultural fires in Arctic and all fires outside the Arctic. Wildfire vs agricultural partitioning: Because QFED does not distinguish fire types, emissions within each grid cell are split using MODIS land cover at 0.01° to estimate the agricultural land fraction, which is used to fractionally allocate QFED emissions between agricultural and non-agricultural (assumed wildfires). QFED was regridded to CESM resolution; MODIS at native resolution was used to compute within-cell fractions. Attributable PM2.5 calculation: To avoid confounding by coupled meteorology and secondary aerosol formation, wildfire-attributable PM2.5 is computed from primary aerosol components (POM + BC) as the difference between FIRE ON and ARCTIC WILDFIRE OFF. Since POM+BC constitute >99.9% of primary aerosol mass from Arctic wildfires, this captures the dominant driver of PM2.5-related health impacts. Significance filtering: A one-sided paired samples t-test (p=0.01) was applied at each grid cell to monthly means (2001–2020) to retain only locations with a statistically significant increase in POM+BC in FIRE ON vs ARCTIC WILDFIRE OFF; otherwise, the fire fraction is set to zero. Bias correction to observations: CESM underestimates PM2.5 vs AIRNOW and PurpleAir in the Arctic (normalized mean bias −0.61 and −0.72, respectively), while the GWR PM2.5 reanalysis has much smaller bias. The FIRE FRACTION is defined as (FIRE ON − ARCTIC WILDFIRE OFF) / FIRE ON using CESM POM+BC. A counterfactual without Arctic wildfires is then estimated as BIAS CORRECTED WILDFIRE OFF = GWR PM2.5 × (1 − FIRE FRACTION). GWR PM2.5 is used up to 68°N; above this, a linear blend with CESM FIRE ON avoids a hard boundary. Health impact assessment: Chronic mortality from long-term PM2.5 exposure is estimated with GEMM (China cohort included), for adults ≥25 years, using a counterfactual concentration of 2.4 µg m−3. Population counts and distributions are from GPW v4 (5-yearly, linearly interpolated). Population age structure and baseline mortality rates are from GBD 2019 (2001–2019; 2020 uses 2019 data). Excess deaths are computed for the GWR PM2.5 scenario and the BIAS CORRECTED WILDFIRE OFF scenario; their difference yields Arctic wildfire-attributed mortality. Trends in PM2.5 are assessed using Theil–Sen estimator, with significance by the Mann–Kendall test.

• Spatial contribution: Within Arctic Council states, wildfires account for an average 21% of annual mean PM2.5 (2001–2020). Russia and Canada have mean wildfire contributions of 22.8% and 27.1%, respectively, while Scandinavian countries are ~2%.
• Hotspots: In Russia near the Chinese and Mongolian borders, Arctic wildfires add >8–10 µg m−3 (40–50%) to annual mean PM2.5; in remote Siberia and Canada, wildfire-attributed PM2.5 exceeds 70% of annual mean PM2.5. Transboundary enhancements are evident in China and Mongolia.
• Population-weighted exposure: Lower than area-weighted due to fires in sparsely populated regions. Canada and Russia have the highest population-weighted wildfire fractions within the Arctic Council (8.7% and 8.2%). Outside the Council, Mongolia has the highest population-weighted wildfire contribution due to proximity to Siberian fire regions.
• Trends in PM2.5: Summer (JJA) PM2.5 increased by up to ~1.5 µg m−3 yr−1 in parts of Siberia from 1998–2020, with regions such as Irkutsk Oblast and Sakha rising from ~10 µg m−3 (1998) to ~40 µg m−3 (2020). The fraction of wildfire-sourced PM2.5 increased by >3% yr−1 in parts of Siberia.
• Latitudinal shift: Wildfire emissions and PM2.5 impacts shifted northward in eastern Russia during 2011–2020 vs 2001–2010, reducing exposure in more densely populated areas near the Mongolia and China borders.
• Mortality burden: On average 21,100 (95% CI: 15,800–27,700) excess deaths per year are attributable to Arctic Council wildfire-sourced PM2.5, with ~8,100 (5,900–10,800) outside the Arctic Council. • Russia: 9,900 (7,300–13,100) deaths yr−1; highest per capita at ~69 (51–92) per million. • Canada: 1,280 (1,051–1,535) deaths yr−1; ~38 (31–46) per million. • China: 4,824 (3,534–6,425) deaths yr−1 despite only ~0.2% of population-weighted PM2.5 from Arctic wildfires, due to dense populations in affected northeastern regions. • Other high per-capita impacts: Mongolia 39 (26–57), Belarus 26 (18–38), Georgia 19 (14–25) per million.
• Temporal change in mortality: Global wildfire-attributed deaths declined from 22,900 (17,300–30,000) in 2001–2010 to 19,200 (14,200–25,400) in 2011–2020, largely due to reduced impacts outside the Arctic Council; within Council states, annual wildfire PM2.5 and related mortalities were relatively stable.
• Population vs area weighting: In Russia, population-weighted wildfire PM2.5 decreased over time while area-weighted increased, indicating rising impacts in sparsely populated regions but reduced exposure in populated areas.

The study demonstrates that Arctic Council wildfires impose substantial transboundary health burdens, with over one-third of attributable mortalities occurring outside the Council. The findings address the core question by quantifying both exposure and mortality attributable to Arctic-origin smoke, revealing that although wildfire-sourced PM2.5 has increased in parts of Siberia, the mortality burden has declined due to a northward shift in burning away from densely populated border regions. Health impacts are amplified when smoke plumes affect large populations (e.g., northeastern China), whereas per-capita risks are highest near the fires in sparsely populated Arctic zones. The results underscore the importance of international cooperation on air quality and wildfire management, and suggest that changes in fire regimes can modulate health risks even amid increasing emissions. The analysis also highlights the need to better characterize wildfire-specific toxicity and the influence of demographic and baseline health variations to refine burden estimates.

This work provides a first comprehensive estimate of the chronic mortality burden attributable to PM2.5 from wildfires occurring within Arctic Council states, quantifying substantial transboundary health impacts and documenting a recent decline in overall mortality due to a northward shift in Siberian fire activity. It shows that wildfires contribute significantly to annual PM2.5 in high-latitude regions and can measurably affect downwind countries. Future research should: (1) improve observational constraints on Arctic PM2.5, including speciated measurements; (2) develop wildfire-specific long-term exposure-response functions and incorporate sub-national demographic and health data; (3) better resolve peat fire emissions and plume dynamics; and (4) investigate the drivers and persistence of the northward shift in fire regimes under climate change and its implications for public health and ecosystems.

• Model bias: CESM underestimates PM2.5 in Arctic regions versus AIRNOW and PurpleAir; although bias-corrected with GWR PM2.5, uncertainties remain, especially where GWR coverage ends at 68°N and blending with CESM is required.
• Fire attribution: QFED does not distinguish agricultural vs wildfires; classification relies on MODIS land cover and fractional allocation, which may misclassify some fires, though cropland area is small in the Arctic Council region.
• Primary aerosol focus: Attributable PM2.5 is derived from POM+BC differences to avoid confounding by secondary formation and meteorological feedbacks, potentially omitting secondary wildfire aerosol contributions.
• Statistical masking: Grid cells without statistically significant POM+BC increases are set to zero fire fraction, which may undercount diffuse or intermittent impacts in populated areas.
• Health function: GEMM represents chronic PM2.5 exposure using general cohorts (including China cohort) and may not capture wildfire-specific toxicity or Arctic-specific exposure patterns and behaviors.
• Demographics: Baseline mortality and age structure are at national scale; sub-national variability is not represented.
• Resolution and plume representation: Model resolution can dilute smoke plumes; satellite detection limits and overpass times may underestimate burned area and emissions, particularly for smoldering peat fires.