Air Pollution and Mortality
Air pollution is the largest environmental risk factor in global health statistics. The estimates that travel as headlines — millions of deaths a year — rest on a specific chain of inference: ambient particulate matter, household fuel use, exposure-response curves, and counterfactuals. This page unpacks each link.
Last reviewed on 2026-04-27.
What “air pollution” means in health statistics
PM2.5 — the fine fraction
Most quantitative work on air pollution and health centres on PM2.5: airborne particles smaller than 2.5 micrometres. Their importance comes from their size — small enough to penetrate deep into the lung and to enter the bloodstream — and from the fact that they are produced by almost every relevant source: combustion of coal and oil, biomass burning, vehicle exhaust, secondary chemistry from agricultural ammonia and industrial sulphur, and natural sources such as dust and wildfires. PM2.5 is reported as a mass concentration in micrograms per cubic metre of air (µg/m³). The 2021 WHO guideline tightened the recommended annual mean to 5 µg/m³; many cities sit several times higher.
Ozone, NO₂ and others
Ground-level ozone (O₃) is treated separately because it is a secondary pollutant, formed by sunlight acting on emissions of nitrogen oxides and volatile organic compounds, and because its health effects are seasonal. Nitrogen dioxide (NO₂) is the workhorse traffic indicator in European cities. Sulphur dioxide and carbon monoxide were major historical pollutants and remain locally important. Most cross-country mortality estimates aggregate PM2.5 and O₃; NO₂ is more often used at the city level.
Ambient versus household
Ambient (outdoor) air pollution affects everyone who breathes the same air shed. Household air pollution is the result of cooking and heating with solid fuels — wood, charcoal, dung, coal — usually in poorly ventilated kitchens. The two overlap: household smoke leaks outside and contributes to ambient PM2.5, and ambient PM2.5 leaks inside. WHO and IHME report them separately because the affected populations and the mitigation policies are different. Household air pollution is concentrated in low-income rural settings; ambient PM2.5 is everywhere but worst in dense, fossil-fuel-intensive urban regions.
Where the data comes from
Ground monitors are the gold standard but cover a small share of the world’s land area, and almost none of low-income rural Africa. Satellite retrievals — chiefly NASA MODIS and ESA Sentinel — fill the gap by inferring surface PM2.5 from aerosol optical depth, calibrated against the monitor network. WHO publishes the consolidated Ambient Air Quality Database; the Atmospheric Composition Analysis Group at Washington University in St Louis publishes one of the most widely cited global gridded PM2.5 products. These satellite-blended grids are what most country-level “PM2.5” figures actually represent.
From a concentration to an attributable death
Exposure-response functions
The link between PM2.5 exposure and disease is built from cohort studies — large, long-running studies in which the air quality where people live has been measured and their health has been followed for years or decades. These cohorts produce relative-risk curves showing how mortality rises with PM2.5 exposure for specific causes: ischaemic heart disease, stroke, chronic obstructive pulmonary disease, lung cancer, and lower respiratory infections. The Global Burden of Disease project at IHME stitches these curves into a single integrated function, and applies it to the global PM2.5 grid. The 2019 GBD added type 2 diabetes; later vintages have added chronic kidney disease and a handful of other endpoints.
The counterfactual
An attributable death is not a death caused by air pollution in a forensic sense; it is the difference between observed mortality and the mortality expected under a counterfactual exposure. The choice of counterfactual matters a great deal. WHO has historically used the “air quality guideline” level as the counterfactual; IHME uses an empirical “theoretical minimum risk” level fitted to the data. The two yield different totals — IHME’s headline numbers tend to be larger — even when applied to the same underlying PM2.5 grid. When two reports give very different totals for the same year, this difference is usually most of the explanation.
Population versus individual risk
Attributable-death numbers are population statistics, not individual claims. A country’s “X attributable deaths from PM2.5 in 2019” means that, applied across a population of millions, that is the predicted reduction in mortality if exposure had been at the counterfactual level. No specific individual death is being assigned to air pollution. The same logic governs DALYs (disability-adjusted life years), which add years of disease-related disability to the years of life lost.
Why estimates evolve
Air-pollution mortality figures change between report vintages for at least four reasons: PM2.5 grids are revised as the satellite chain is recalibrated; cohort studies add years of follow-up and refine the exposure-response curve; new endpoints are added; and the choice of counterfactual is updated. A rising headline total from one year to the next does not necessarily mean air quality has worsened — it may mean more diseases have been linked to PM2.5.
A short reading checklist
Pin down the indicator
Is the figure annual-mean PM2.5, 24-hour PM2.5, ozone, or a composite air-quality index? Mixing them is the most common chart-reading error. WHO daily and annual guidelines are not interchangeable.
Pin down the source
A monitor reading from one urban site is a different object from a satellite-blended grid average across a whole country. Reports usually mix the two; the better ones say which.
Pin down the year
Air quality varies year-to-year with weather and economic activity. A single-year ranking is sensitive to wildfires, sandstorms, lockdowns and stagnation events. Multi-year means are more stable.
Separate ambient and household
A country can have moderate ambient PM2.5 and severe household exposure, or vice versa. The two come from different sources, affect different populations, and respond to different policies. Treat them as separate facts unless you specifically want the combined total.
Sources
The standard references on this topic are the WHO Ambient Air Quality Database, the WHO Global Health Observatory air-pollution indicators, the Health Effects Institute’s State of Global Air, the Institute for Health Metrics and Evaluation’s Global Burden of Disease results, and the U.S. Environmental Protection Agency’s integrated science assessments for criteria pollutants. Country and city air-quality agencies (EEA in Europe, EPA in the United States, MoEFCC in India, MEE in China) publish the underlying monitor data. Each chart on this site cites a specific publisher and vintage; follow that citation for the methodology details that drive the headline number.
Air pollution sits at the intersection of fossil-fuel use, sectoral emissions and population health. Reading those pages alongside this one helps explain why air quality and climate policy share so many of the same levers.