Western Europe averaged 20.74°C in June — a record, and 3.06°C above normal. For the single week the heat peaked, Europe has two figures for the dead: 10,650 and 20,390. One starts from registered deaths. The other never looks at one.
June 2026 was western Europe’s warmest June on record, the Copernicus Climate Change Service reports: an average of 20.74°C, 3.06°C above the 1991–2020 average for the month, beating June 2025.1 That is a measurement, reported to a hundredth of a degree.
It is also a sentence with two dials on it. The first is the box. “Western Europe” here means 11°W–15°E, 37°–55°N — roughly Spain and Britain east to Italy and Germany.1 Widen to all European land (25°W–40°E, 34°N–72°N) and the same month, in the same dataset, comes in at 19.14°C, 1.78°C above average — and second-warmest, behind June 2019.1 The box alone moves the anomaly by 1.28°C, and the record from first to second.
The second dial is the calendar. May was violent: an early heatwave from about 21 to 30 May broke May records in France, the UK, Ireland and Portugal.3 Yet May 2026 ranks only seventh-warmest for Europe, at 13.87°C — because, as Copernicus states plainly, a colder-than-average first half balanced the monthly mean.3 Move the month boundary and the same air produces a record or an also-ran.
Between 22 and 28 June the heat peaked. Europe now has two figures for what it cost, and they are not two attempts at the same sum.
The first is 10,650: EuroMOMO’s pooled estimate of excess all-cause deaths that week across 27 European countries or subnational regions. The network is hosted by Denmark’s Statens Serum Institut and backed by the ECDC and the WHO.4 It starts from deaths that were registered, and subtracts the number expected. More than 9,000 were aged 65 or over — and over the preceding eight weeks the same countries had been running about 500 deaths a week below expected.9
The second is 20,390: a preprint by Christopher Callahan of Indiana University Bloomington, posted on 30 June, which applies previously developed heat–mortality response functions to the same week’s temperatures.5 It is not a count and does not claim to be. It infers deaths from how hot it was, using a relationship fitted to earlier events, and it has not been peer-reviewed.
So the ratio — 1.91 times, a gap of 9,740 — is not a scandal. It is what happens when one instrument counts bodies and subtracts an expectation, and another reads a thermometer and consults a curve. Neither is the toll.
France is the cleanest test: the country June hit hardest, and the one with the most-cited number. Santé publique France recorded 8,973 deaths from all causes in the week of 22–28 June, against 6,948 the week before — 2,025 more, a rise of 29.1%.6 That has been widely reported as excess deaths. It is not. It is a comparison with the previous week, not with an expected baseline; France’s actual excess-mortality estimate appears three weeks after a heatwave ends, once about 95% of deaths are consolidated.6
It is also built from electronic death certificates, which the agency’s own epidemiologist notes cover roughly 60% of deaths nationally — and unevenly: about 25% of deaths at home, 45% in nursing homes, close to 80% in hospitals.10 The most striking line in the bulletin is a 91% rise in deaths at home.6 That is the stratum the system sees least well.
Callahan’s model puts France at 5,210 for the same seven days.5 Same country, same week, 2.6 times apart — again two measures, not one disagreement. The desk’s reading, offered as analysis: no published figure today is a count of people killed by heat in France, and none claims to be.
No death certificate says “heatwave”. Excess mortality is a residual against a modelled expectation, not a cause of death — so even the registered figures here are partly modelled, at the baseline. What this piece could not establish: (a) the Callahan preprint’s 95% confidence interval — the PDF could not be machine-retrieved and the interval circulating online traces only to a tertiary aggregator, so it is omitted rather than printed; the country split comes from consistent reporting of the preprint by three outlets.11 (b) A second estimate of about 15,000, attributed to a researcher at Poznań University of Medical Sciences, could not be traced to any primary or press source, and is not used. (c) EuroMOMO publishes no country breakdown and flags that recent weeks may move. (d) The UK Health Security Agency’s registered estimate for England and Wales is still to come, and will test the modelled 2,700.8 (e) Whether Europe’s heat-health warnings cut this event’s toll below what a function fitted on past heatwaves predicts — possibly part of the gap; nobody has shown it either way.
One question here is genuinely open among credible researchers, and it decides what any of these numbers mean: whether heat deaths are lives cut short by years, or by days. The effect is called mortality displacement, or harvesting. Peer-reviewed estimates of its size differ across most of the available range.12
Heat kills disproportionately among the very frail, and a spike is often followed by a dip as deaths are pulled forward by days or weeks. A study of 15 European cities put harvesting at around 70% of heat-associated deaths; a European analysis found heat impacts cut by about 75% once displacement was allowed for. On this reading a weekly excess figure badly overstates life-years lost — and EuroMOMO’s eight preceding weeks running ~500 below expected is the pattern that invites the question.129
Analyses of the 2003 French heatwave put displacement under 10% of deaths. Daily time-series methods can only detect harvesting at short lags; studies using annual windows find temperature effects operating at much longer lags than daily studies can see, which suggests the deaths are not merely rescheduled. On this reading the excess is close to real loss.12
Evidence can narrow this: it turns on mortality in the months after June 2026 — measurable, not yet measured. Noodle News states no bottom line. Note the scale: the range in the literature, roughly 10% to 70%, is wider than the gap between the two headline death figures above.
For long pieces, end on an auditable dump of the data behind the exhibits — the reader can check the work.
| Exhibit | Measure | Value | Note |
|---|---|---|---|
| Exhibit 01 | W Europe Jun (1st) | +3.06°C | Measured ERA5 anomaly vs 1991–2020 |
| Exhibit 01 | Europe Jun (2nd) | +1.78°C | Measured ERA5 anomaly vs 1991–2020 |
| Exhibit 01 | Europe May (7th) | +0.60°C | Measured ERA5 anomaly vs 1991–2020 |
| Exhibit 01 | Globe Jun (2nd) | +0.56°C | Measured ERA5 anomaly vs 1991–2020 |
| Exhibit 02 | Registered excess | 10,650 | 27 countries and regions, 22–28 June |
| Exhibit 02 | Modelled estimate | 20,390 | Model output (Callahan), 22–28 June |