The 71% Jump in Europe's Heat Deaths Is About 54% Like for Like

A green LED cross outside a pharmacy on rue Belgrand in Paris read 44 °C on the afternoon of June 26. It was a street display, not a weather station, and the study released overnight into Friday, Oct. 23, would never have seen it. The model registers one average temperature per city per day, on a grid about 11 kilometers across.

The estimate, from researchers at Imperial College London and the London School of Hygiene & Tropical Medicine (LSHTM), is 50,391 heat-related excess deaths in 854 European cities from June 1 to Aug. 31. The 95% interval runs from 46,026 to 53,878. The authors attribute 65% of the total, 32,619 deaths, to human-caused warming.

That is strong enough to act on, and weaker than the headlines make it sound. It is not a body count but a model of cities only, built from daily mean temperatures and temperature-death curves fitted to 2000-2019. The most repeated comparison, that the toll is 71% higher than in 2022, fails a like-for-like test: on the report's own rerun, the gap is about 54%.

The cities hold almost 173 million residents in 30 countries, almost 40% of those countries' population. The report from Imperial's Grantham Institute and LSHTM says the Europe-wide burden "will be larger, although the magnitude cannot be inferred directly from our results." It stops at the end of August; a further 1,508 deaths for May 21-31 were reported separately. The report is accepted by Nature Health; the companion paper in Nature Health, covering 2000 to 2025, was published Oct. 14.

Two curves and a subtraction produce the number

Step one is temperature. For each city, the team takes observed daily mean temperatures from the ERA5-Land reanalysis, then builds a no-warming series by subtracting the local human-caused warming, estimated from observations and 36 climate models. Averaged over the cities, that warming is 2.5 C. The summer ran around 2.4 C hotter than the 1991-2020 climatology.

Step two is health. Temperature-death curves (exposure-response functions) were fitted on real daily deaths from 2000 to 2019 in 232 European cities, then extended by regression to all 854 cities and five age groups. The model runs twice, on real temperatures and on the no-warming series, and the difference is the climate-attributable part. Deaths are counted only above each city's minimum mortality temperature, the point where risk is lowest. Repeating everything across 1,000 simulated draws produces the interval.

Why a model at all? Heat rarely appears on a death certificate. AP, citing Imperial's Garyfallos Konstantinoudis, reported that heat deaths "are underreported and often not listed in death certificates." Malcolm Mistry, an LSHTM co-author, said, as quoted by Euronews, "The true burden of mortality, in fact, due to heat is not directly observed or measured." Andrew Hartle, an intensive-care consultant, put it more bluntly at a press briefing, as Bloomberg reported:

"A postmortem won't tell you whether somebody was hot or not."

Andrew Hartle, intensive-care consultant, Imperial College Healthcare NHS Trust

The companion paper is explicit that its figures "do not represent observed counts but modelled estimates." It lists what the method cannot see, including urban heat islands, housing, air conditioning and "daytime extremes, nighttime heat, humidity, apparent temperature or heatwave duration." Those gaps "may further contribute to an underestimation of the absolute heat burden." The report says the late-June event combined extremely high humidity with unusually high night-time temperatures. The curves are also frozen at 2000-2019, and the intervals "do not capture all potential sources of uncertainty."

"Heat-related" is the 50,391. "Attributable to climate change" is the 32,619 that vanish in the no-warming run. The difference leaves about 17,800 deaths in a summer without human-caused warming (approximate, because medians do not subtract exactly).

The share is not a dial that only turns up. The paper puts it at 39% in 2000, 42% in 2003 and 68% in 2025; this year's 65% is three points lower, with heavily overlapping intervals. In the worst 15 days, June 17 to July 1, it was 57%. The report explains: in more extreme heat the proportion falls, "because the high temperatures reached would have been dangerously hot even in a preindustrial climate." At the other end, Estonia's cities show 81%, but of only 13 modeled deaths.

The record holds, but the 71% is not a like-for-like gap

The record is the sturdy part. In the same framework and cities, 2022 comes to 29,480 and 2003 to 28,462, and 2025 to 24,572, so this summer is almost double. Clair Barnes, the lead author, told CNN: I think we can be quite confident in stating that the summer of 2026 is likely to have been the worst on record for heat-related mortality across much of Europe.

The 71% is where the arithmetic slips. AP reported 50,391 as 71% higher than 2022's 29,480; RTE said the toll was more than 70% above both 2003 and 2022. Straight division does give 70.9%, but the benchmark years were run on the mean population and annual mortality rates of 2000-2019, while 2026 was run on its own population, almost 173 million people, and its own rates.

The report reran 2026 on the same populations and rates as the earlier years. That gives 45,492 deaths (interval 41,429 to 48,719), which the authors call "around 50% higher than in either 2022 or 2003." Computed directly, that is 54.3% over 2022 and 59.8% over 2003. The attributable share stays stable, and the 32,619 climate-attributable deaths alone, as published, exceed the entire modeled 2022 summer.

"Record" is also relative to a fixed set of cities. A 2023 Nature Medicine study led by Ballester estimated 61,672 heat-related deaths in Europe from May 30 to Sept. 4, 2022, across 823 regions and over 543 million people. That is a larger number on a different footing, not a contradiction and not a like-for-like record.

People 85 and over account for 22,502 of the modeled deaths, about 45%; roughly 6,000 were under 65.

June 17 to July 1 holds 14,539 of the modeled deaths, 29% of the total in 16% of the days. That is about 969 modeled deaths a day against about 466 across the other 77. Barnes told Grist: Any one of those heat waves would have been deadly, but we've had five.

The peak followed our earlier piece on France's record-hot night. Germany logged a provisional 41.3 C on Friday, June 26, a mark provisional data put at 41.5 C the next day. In Paris, Frédéric Adnet, head of the Samu emergency medical service, said workload rose "by around 50 percent," according to a European Broadcasting Union analysis.

In Naples, the hardest-hit city, the Guardian reported 815 deaths per million (age-standardized). Dr. Mario Guarino, an emergency doctor there, described a terminal cancer patient who lived in a hot attic because his family could not find a company to install air conditioning.

Official counts agree in places; the €17 billion answers a different question

The report checks itself against official figures in five places. For Europe in June 22-28, a rapid report counted 20,390 deaths against the model's 9,585, with the study covering about 40% of the population. France's Santé publique France reported 5,764 all-cause excess deaths for June 17-July 2 against the model's 2,305, though the agency stresses that is not yet an estimate of deaths directly attributable to heat. Germany's Robert Koch Institute reported 16,000 heat deaths for April-early September against the model's 6,477 for May-August.

The authors call the results "broadly compatible": scaling the European figure by 40% gives 8,156, "only slightly lower than our estimate." A rough coverage-scaling, which assumes deaths scale with population, puts the model 18% to 35% above the scaled official figures for Europe, France, the UK and Germany. Spain is the outlier: its city-only model figure, 6,824, exceeds the national 5,123 from the MoMo monitoring system, the reverse of what a cities-only model should show.

David García García, a MoMo researcher and not an author, gave the best-qualified criticism:

"However, it is important to bear in mind that these are estimates based on observed temperatures and epidemiological models, which do not incorporate actual data on deaths that occurred in the summer of 2026. Furthermore, the methodology bases the attribution to heat on the temperature associated with minimum mortality, which may result in high estimates."

David García García, researcher, Mortality Monitoring System, Carlos III Health Institute

His concern runs opposite to the authors' own list, which points toward undercounting. Lasse Vestergaard of EuroMOMO told CNN to treat the estimates with caution "as they involve several uncertainties and assumptions," though they are "overall broadly in line" with EuroMOMO's own summer reports.

Beware the other number in circulation. A European Broadcasting Union analysis of EuroMOMO data found over 36,000 more deaths than expected in weeks 26-35. That is net all-cause excess for whole countries, not an estimate of heat deaths, so it cannot be subtracted from 50,391. Official figures also move: Germany's Robert Koch Institute first estimated about 5,100 late-June heat deaths, then revised to about 9,600.

The economic number is a third kind of thing. Bloomberg bundled the study with a separate one, from Bloomberg Economics, saying heat wiped roughly 0.17% from European Union GDP from April through September, close to €17 billion. Bloomberg prints both figures, but they do not reconcile: 0.17% of an EU economy near €18 trillion would be about €30 billion. Ulrich Eberle, its author, put the second-quarter hit at 0.2% of French GDP, 0.1% of Germany's and 0.05% of Spain's.

The printed amounts differ mostly by currency: the $18.9 billion ("nearly $19 billion") and the $24.3 billion in The Straits Times headline (S$24.36 billion) are the same figure as the €17 billion. The €180 billion ($208 billion per CNN) is a different estimate from Triodos Bank, about 1% of EU GDP across 2026. It covers all of 2026 and four channels, from labor productivity to transport, about 10.6 times the Bloomberg Economics figure, and is not a competing answer to the same question.

Lower tolls track cooler summers, and the sources do not show what worked

Rates follow heat exposure. Finland comes in at 16, Estonia 26 and Norway 28, against 561 for Italy. Malta's 716 is the highest, but its attributable share has a very wide interval, so it is not tabled. In the north, the report says, "only the most extreme heat events are associated with high mortality rates."

None of that ranks who adapted best, and by construction it cannot: with the curves fixed at 2000-2019, the model cannot register adaptation. The outside evidence is anecdotal. Wallonia's mortality ran about 77% above normal against 31% in Flanders, RTBF reported, in a European Broadcasting Union review, with no explanation. Geneva opened pools and cinemas to people over 65, and Bern, Lucerne and Zurich ran contact or help programs, yet Switzerland's modeled rate was 386 per million in a summer its cities ran 4.0 C above the 1991-2020 mean. That shows neither failure nor success. Adnet said France had fewer deaths than in 2003, particularly in care homes, because "the public was warned" and staff knew how to treat heatstroke.

Barnes framed the stakes in degrees: For many people, it's certainly the case that that extra two-and-a-half degrees (Celsius) of warming is what's going to tip those temperatures from uncomfortable to dangerous and potentially deadly.

The model finds about 17,800 deaths even without human-caused warming, which adaptation can still work on, and 32,619 that emissions cuts could have prevented.