Two Nature papers put numbers on the health cost of a smokier century

Nature published two wildfire papers on the same day this month, and read together they describe the same problem at two different scales [s1][s2]. One asks what smoke will do to mortality in the United States by mid-century. The other asks what it will do worldwide by the century's end, and which countries end up carrying the burden.

The US projection

The first study builds an ensemble of statistical and machine learning models linking climate conditions to wildfire smoke fine particulate matter (PM2.5), then estimates the relationship between smoke PM2.5 and death empirically, using data on all recorded deaths in the United States [s1]. Chaining those two steps together is the hard part — the authors are explicit that understanding of how climate change affects wildfire, smoke and health has remained highly uncertain precisely because that causal chain is difficult to model end to end [s1].

Their central figure: under a high-warming scenario (SSP3-7.0), smoke PM2.5 could produce 71,420 excess deaths per year in the United States by 2050, with a 95% confidence interval running from 34,930 to 98,430 [s1]. That is a 73% increase over the estimated average annual excess deaths from smoke across 2011 to 2020 [s1]. Cumulatively, the study projects 1.9 million excess deaths from smoke PM2.5 between 2026 and 2055 [s1].

The confidence interval is worth sitting with. The lower bound is roughly half the central estimate and the upper bound is nearly 40% above it. That spread is the honest expression of how much uncertainty compounds when a climate model feeds a fire model which feeds an air-quality model which feeds an epidemiological one.

Two secondary findings matter for how smoke is usually discussed. The authors report evidence that the mortality impacts of smoke PM2.5 persist for up to three years after exposure [s1] — meaning the health accounting for a bad fire season does not close when the air clears. And when the projected deaths are monetized, the resulting damages exceed existing estimates of climate-driven damages from all other causes combined in the United States [s1]. That is a strong claim, and it is a claim about the relative size of modelled estimates rather than an observed cost.

The global projection

The companion paper takes a different route: an interpretable machine learning framework that projects global burned area and wildfire emissions, then quantifies the premature deaths and the radiative forcing attributable to fire-derived PM2.5 [s2].

Under SSP2-4.5 — a middle-of-the-road emissions pathway, less severe than the scenario used in the US study — fire carbon emissions are projected to rise 23% from 2010–2014 to 2095–2099 [s2]. Premature deaths from wildfire smoke reach 1.40 million people annually during 2095–2099 (95% confidence interval 0.66 to 2.25 million), roughly six times current levels [s2].

The distributional finding is the one that distinguishes this paper. Africa is projected to see the largest increase in fire-related deaths — an elevenfold rise, driven by changes in emissions and by an ageing population [s2]. Europe and the United States are projected to see a one- to twofold increase under the same pathway, linked to rising fire occurrence in the mid-latitude Northern Hemisphere [s2]. The net effect, the authors argue, is that the health burden becomes more evenly distributed across countries at different development levels than it is today [s2].

"More evenly distributed" is not the same as "fairer." It describes a world where high-income countries acquire a fire-smoke health problem they largely have not had at scale, while the countries already carrying most of it carry considerably more.

The paper also reports a climate-system effect that runs in an uncomfortable direction: increased fire-derived aerosols reduce the 0.06 W m⁻² cooling effect north of 60°N [s2]. Smoke particles scatter sunlight and can cool locally; more fire in the high north erodes some of that.

What these numbers are and are not

Both papers are projections. They describe what models produce when fed emissions scenarios, not what has been counted in death records. The US study's own framing — that uncertainty in this chain has been the central obstacle — applies to its own output as much as to the prior literature it improves on [s1]. The global study's figures depend on assumptions about population ageing and about where people will live, which is why Africa's elevenfold increase is attributed jointly to emissions and demography rather than to fire alone [s2].

They also use different scenarios, which makes direct comparison misleading. The US figure is under a high-warming pathway; the global figure is under a moderate one. A reader who wants a single number for "how bad will this get" will not find one here, and the papers do not offer one.

What to watch

The three-year lagged mortality signal in the US paper [s1] is the finding most likely to change how smoke exposure is measured, because air quality is normally reported and regulated as a short-term daily metric. If cumulative multi-year exposure is what predicts death, the standard reporting framework is measuring the wrong thing.

The second point to watch is whether independent groups reproduce the Africa result [s2]. An elevenfold projected increase in a region where baseline fire-smoke mortality surveillance is thin is the kind of estimate that most needs replication with different modelling choices before it becomes a planning assumption.

Sources

• Wildfire smoke exposure and mortality burden in the USA under climate change — Nature, 2025-09-18
• Global warming amplifies wildfire health burden and reshapes inequality — Nature, 2025-09-18