A 26-year attribution study puts Europe's climate-driven heat deaths at about 252,000

Attribution studies of individual heatwaves have become routine — a few weeks after a severe event, a team publishes an estimate of how much of it was made worse by climate change. A preprint posted this month attempts the longer arithmetic: not one summer, but every summer from 2000 to 2025, across 854 European cities [s1].

The study has not yet been peer reviewed, and that qualification should travel with every number in it.

The method

The authors combined probabilistic climate attribution with city-specific and age-specific temperature–mortality relationships, then estimated heat-related mortality under two conditions: the factual climate, meaning the one that actually occurred, and a counterfactual climate, meaning the one that would have existed without human-induced warming [s1].

The difference between the two is the attributable burden. This is the standard architecture of attribution work, and its credibility rests almost entirely on two inputs: how well the counterfactual climate is simulated, and how accurately temperature–mortality curves are estimated for each city and age band. Neither is directly observable, and both carry substantial uncertainty that compounds when multiplied across 26 summers.

The numbers

Across the 854 cities and the 26 summers, the authors estimate 463,730 heat-related deaths [s1]. Of those, 251,956 are attributed to anthropogenic climate change — just over half of the observed burden [s1].

The trend inside that total is arguably the more important finding. The proportion of estimated heat deaths attributable to human-induced climate change rose from 45% in 2000 to 64% in 2025 [s1].

That rising fraction is what distinguishes this from a study about how hot Europe has been. As baseline warming accumulates, a larger share of any given summer's heat mortality becomes something that would not have happened in the counterfactual world. The absolute burden and the attributable fraction are climbing together.

Who and where

The largest burdens fell on older age groups, and on cities in southern and south-eastern Europe [s1].

Neither is surprising, and both matter for policy. Age-specific temperature–mortality curves steepen sharply in the elderly, and southern and south-eastern European cities combine high summer temperatures with older populations and, in many cases, less residential air conditioning and weaker heat-health infrastructure than northern and western Europe. The geography of the burden runs roughly opposite to the geography of adaptation capacity.

The 2003 comparison

One result works against the simplest reading of the data. Summer 2003 — the European heatwave that killed tens of thousands and reshaped the continent's heat-health planning — remained the deadliest summer in absolute terms across Europe as a whole, over the entire 26-year period [s1].

The authors' framing is that summers with similarly large mortality burdens have become more frequent in recent years [s1]. That is the meaningful shift: not that any single summer has surpassed 2003, but that 2003-scale summers have stopped being anomalous.

It is also a reminder that 2003 prompted real adaptation. Heat-health action plans, alert systems and care home protocols were built across Europe in its aftermath. Some portion of why recent summers have not exceeded it, despite hotter conditions, is likely that response — though this study is not designed to measure adaptation, and does not attempt to separate it from the temperature signal.

What the design cannot do

Attribution studies estimate deaths; they do not count them. The 463,730 figure is not a tally of death certificates. It is the output of a statistical model applying exposure-response relationships to temperature data, and its precision is illusory — a number given to five significant figures represents a central estimate within a range the model itself defines.

Heat-related mortality is also mostly indirect. People die of cardiovascular events, respiratory failure and kidney injury during heat, and these are recorded as such. That is why the epidemiological approach used here — modelling excess deaths against temperature — captures more than official heat-death counts, and also why it depends entirely on the model being right.

The counterfactual climate is a simulation. Different climate model ensembles produce somewhat different counterfactuals, and the attributable fraction moves with that choice.

Finally, this is a preprint [s1]. Peer review can change numbers, methods and conclusions, and studies of this scale often do change between posting and publication.

Why it matters anyway

For European cities, the operational implication does not depend on the exact figure. If roughly half of the heat mortality accumulated since 2000 is attributable to warming, and that share is rising toward two-thirds [s1], then heat planning built around historical baselines is planning for a climate that no longer exists. The alert thresholds, hospital surge assumptions and care protocols set after 2003 were calibrated to a distribution of summers that has shifted underneath them.

The authors' own conclusion is that anthropogenic climate change has already greatly increased heat-related mortality across European cities, and that both rapid emissions reductions and targeted local adaptation are needed [s1]. The second half of that is the part European health systems can act on in the next four years.

Sources

• [s1] Attributing heat-related mortality to human-induced climate change in 854 European cities during the summers of 2000-2025 — Research Square preprint, posted 4 August 2026. Not yet peer reviewed. https://doi.org/10.21203/rs.3.rs-10552432/v1