Human-Caused Warming Now Drives Two-Thirds of Europe’s Urban Heat Deaths

More than half a million people live in the crosshairs of a silent killer that has been growing steadily stronger for a quarter of a century, and a landmark new analysis has now put a number on how much of that danger is our own doing. In the most comprehensive heat mortality attribution study ever conducted, researchers estimated that 430,365 heat-related deaths occurred across 854 European cities between the summers of 2000 and 2025, and that 233,626 of those deaths would not have happened in a world without human-induced climate change. Even more striking is the trajectory: the share of heat deaths attributable to human activity climbed from 39 percent in 2000 to 68 percent in 2025, meaning that by the most recent summer, roughly two out of every three heat-related deaths in European cities were caused by warming that humanity itself set in motion.

The study, published in Nature Health by a team led by Clair Barnes and Garyfallos Konstantinoudis of Imperial College London, together with colleagues at the London School of Hygiene and Tropical Medicine and the University of Bern, represents a step change in the science of event attribution. Previous analyses had quantified the fingerprint of climate change on individual disasters, such as the devastating 2003 European heatwave or the record-breaking summer of 2022, but no one had systematically tracked how the attributable mortality burden evolved across an entire continent over 26 consecutive summers. By combining probabilistic climate attribution with city- and age-specific temperature-mortality relationships, the team produced an unprecedented spatial, temporal and demographic portrait of a crisis that is accelerating in plain sight.

The technical architecture of the analysis rests on a deceptively simple counterfactual. The researchers first assembled factual daily mean temperature series for each of the 854 cities, drawn from the ERA5-Land reanalysis dataset, which blends historical observations with sophisticated data assimilation at a resolution of roughly 11 kilometres. They then constructed a parallel set of counterfactual temperature series representing what each city’s summers would have looked like in a preindustrial climate, by subtracting the local warming attributable to the approximately 1.3 degrees Celsius of human-caused global warming that had accumulated by 2025. That local warming signal was itself synthesized from two independent lines of evidence: observational products including ERA5-Land and the E-OBS station-based dataset, and a 36-model ensemble of coupled global circulation models from the CMIP6 archive, each regressed against global mean surface temperature.

With both climate scenarios in hand, the team applied exposure-response functions that describe how mortality risk rises as temperatures climb above each city’s optimal range. These functions, derived from the Multi-country Multi-city Collaborative Research Network using quasi-Poisson regression and distributed lag nonlinear models fitted to daily mortality records from 232 European cities, were extended through multivariate meta-regression to predict risk curves for all 854 cities and five age groups. The difference between the deaths expected under factual temperatures and those expected under counterfactual temperatures yields the mortality attributable to climate change, with uncertainty propagated through 1,000 bootstrap samples spanning both the climate and epidemiological components of the pipeline.

The results reveal a burden that is both enormous and strikingly uneven. The 854 cities studied are home to nearly 160 million people, and across the study period the annual estimated heat death toll rose from 9,720 in 2000 to 24,572 in 2025. The deadliest single summer was 2022, with an estimated 29,480 deaths, followed closely by the infamous summer of 2003, which claimed an estimated 28,462 lives. The heaviest burden fell on the oldest residents: people aged 85 and over accounted for 165,992 of the estimated deaths, and they also showed the highest attributable fraction, at 0.58. When standardized per million population, Croatia, Malta and Romania emerged as the hardest-hit countries, with 327, 226 and 223 estimated heat deaths per million respectively, while among capital cities, Zagreb, Bucharest and Rome topped the rankings.

Perhaps the most sobering comparison in the study is between the summers of 2003 and 2025. In absolute terms, 2003 remains the second deadliest summer in the record, and its mortality signal was concentrated in a brutal early-August heatwave that produced sharp spikes in daily deaths across western and southern Europe. Summer 2025 followed a different rhythm, with temperature spikes in June and early July followed by a shorter heatwave in mid-August, yet its standardized mortality footprint traced a strikingly similar geography, concentrated in eastern and southern Europe. The crucial difference lies in attribution: only 42 percent of the 2003 deaths were linked to human-induced warming, whereas 68 percent of the 2025 deaths were. Summers as deadly as 2003, once considered extraordinary anomalies, are becoming an increasingly regular feature of European life, and a growing share of each new catastrophe is directly traceable to fossil fuel combustion.

The geography of attribution also tells a subtle story. The highest levels of human-induced summer warming, exceeding 2.5 degrees Celsius, were found in northern Italy, central Spain, southern France and parts of Eastern Europe. Yet the highest attributable proportions were estimated for Ireland, Austria and Czechia, at 0.67, 0.64 and 0.63 respectively, while coastal cities along the Atlantic and North Sea coastlines tended to show lower fractions. This apparent paradox reflects the nonlinear shape of the exposure-response curve: in cooler northern cities, even modest amounts of warming push summer temperatures more decisively past the threshold where mortality risk begins to climb steeply, whereas in already-hot southern cities the burden is larger in absolute terms but a somewhat smaller share of it is newly attributable. In every one of the 30 countries studied, however, the attributable proportion increased throughout the study period.

The authors are careful to spell out the assumptions and limitations that frame these estimates. Baseline mortality rates and exposure-response functions were held constant, isolating the contribution of temperature change, though sensitivity analyses using variable populations and mortality rates produced broadly consistent attributable proportions. Because the analysis relies on area-averaged daily mean temperatures, it cannot capture within-city variation from urban heat islands, green space, housing quality or air conditioning, nor does it account for humidity, nighttime heat or heatwave duration, all of which suggest the true burden is likely underestimated. Climate models also tend to underestimate long-term warming trends, making the attributable warming a conservative estimate. And because the exposure-response relationships were fitted to data from 2000 to 2019, any adaptation that has since reduced vulnerability, or any maladaptation and demographic ageing that has increased it, remains invisible to the model, with uncertain effects on the final numbers.

Comparisons with official statistics underline that different methods answer different questions. For summer 2025, the UK Health Security Agency estimated 317 heat-related deaths in London across all age groups, compared with 416 in this analysis, while Euromomo reported 451 heat-attributable deaths in Madrid against 544 here. The study’s authors argue these approaches are complementary rather than competing: excess-mortality methods excel at describing extreme events using directly observed deaths, while the counterfactual attribution framework enables real-time forecasting and rapid health attribution. The findings also align well with earlier work, including a global study covering 1991 to 2018 that attributed 37 percent of warm-season heat deaths to climate change, a figure this new analysis matches for the early years of its own record before the proportion surged upward.

The policy implications are as clear as they are uncomfortable. The pronounced spatial variation in risk demands locally tailored adaptation planning, particularly for older and vulnerable populations in southern and southeastern Europe, where the absolute burden is greatest. But adaptation alone cannot neutralize a threat that is growing because the planet itself is warming: a substantial share of expected heat-related mortality is linked to anthropogenic warming and could in principle be reduced under stronger mitigation. Without deep and sustained emissions cuts, the analysis suggests, summers with death tolls rivaling 2003 will keep arriving, and the proportion of those deaths written in humanity’s own signature will only continue to climb.

Subject of Research: Attribution of heat-related urban mortality in Europe to human-induced climate change from 2000 to 2025

Article Title: Heat-related urban mortality attributable to human-induced climate change in Europe from 2000 to 2025

Article References: Barnes, C., Masselot, P., Xi, D., Clarke, B., Rayner, A., Chen, X., Wang, R., Raju, E., Theokritoff, E., Mistry, M., Gasparrini, A., Otto, F., Vicedo-Cabrera, A. M., & Konstantinoudis, G. (2026). Heat-related urban mortality attributable to human-induced climate change in Europe from 2000 to 2025. Nature Health. https://doi.org/10.1038/s44360-026-00209-8

Image Credits: AI Generated

DOI: 10.1038/s44360-026-00209-8

Keywords: heat mortality, climate change attribution, European cities, extreme heat, public health, urban heat, CMIP6, exposure-response functions, counterfactual climate, ageing population, adaptation, emissions reduction