Europe's warm nights nearly doubled this summer

TL;DR — Warmer 2026 nights in 27 of 28 European locations, median +1.5°C. Warm nights nearly doubled in the sample, 9.2 to 16.4, and came in longer runs. Lisbon went from about one warm night in early August to 15. North America was warmer but more mixed. Selected city points, matched dates, one unfinished season.Interactive dashboard: Were summer nights unusually warm in 2026?(Do you have suggestions on how to make it better, send it my way: joaotome.com)

· Temperature data through 19 August 2026

In a normal early August, Lisbon has one or two nights where the temperature never falls below 20°C. The 2000–2025 average for 1–19 August is 1.4 nights. This year there were 15, and the average daily minimum across those nineteen days reached 20.3°C, which is 1.9°C above the historical average for the same dates and higher than every completed year for the same 1–19 August window since 2000.

Heat coverage runs on afternoon numbers, the 40°C headline and the red map. The part people actually live with is the temperature at 3am with the window open and the flat still holding the day’s heat.

So I went looking for a comparison that did not depend on a few memorable days. The dashboard sets 1 May to 19 August 2026 against the same calendar window in every year from 2000 to 2025, across 28 European locations, 11 of them in Portugal, and 10 in North America.

Two measures do most of the work: daily mean temperature, and the average daily minimum. The second one answers the evening question, because it tells you how much cooling a night actually delivered. One thing to hold onto before the numbers start: this is a selected, unweighted sample of city points, so it describes these locations in this matched window. It is a description of one season, not a climate trend and not an attribution of cause.

The short answer

Were temperatures really higher this year? For this sample and this matched window, yes, and most clearly after sunset.

• 27 of 28 European locations had a warmer average daily minimum than their 2000–2025 baseline. The median difference was +1.5°C.
• 26 of 28 European locations had a warmer daily mean. The median difference was +1.3°C.
• June was the sharpest month. European city averages ran about +2.1°C warmer for the daily mean and +1.8°C warmer for the daily minimum.
• North America was milder and more mixed: 9 of 10 locations had warmer minimums, with a median of +0.6°C.
• All 18 selected offshore points in the European sea-temperature view were warmer than their 2000–2025 average.

Europe is warmer, especially after sunset

Paris had the largest night-time difference at +2.9°C, with Madrid and Bragança just behind at about +2.8°C. For daily mean temperature Paris led again at +3.4°C, followed by Bragança, Madrid, London and Frankfurt.

The season also had, in a way, a shape to it. May stayed close to the historical range, warmer but only by about +0.9°C for the daily mean and +0.7°C for the minimum. June was where it broke open. July and the first nineteen days of August stayed clearly above the historical average.

Anomalies in tenths of a degree are hard to picture, so the threshold counts do a better job. In the simple Europe aggregate, the selected locations averaged 16.4 nights at or above 20°C during the matched window, against a 2000–2025 baseline of 9.2. The longest unbroken run of those warm nights averaged 9.4 days, versus a 6.1-day historical average.

So the warm nights are not only more frequent, they arrive in longer stretches, which is the part that gives a building no chance to catch up. There were also 8.4 days at or above 35°C on average. These are averages across city points, so they count how many warm nights a typical location in the sample had, rather than days when the whole of Europe crossed a threshold. The BBC/AFP analysis uses a different dataset and reports regional population exposure, which is a separate measure and worth keeping separate.

North America is a useful contrast

The North American sample tells a less uniform story. Nine of ten locations had warmer minimum temperatures, but the median difference was only +0.6°C, and eight of ten had warmer daily means with a median of +0.3°C. Chicago had the strongest night-time difference at +0.9°C. Toronto went the other way at −0.3°C for the average daily minimum, the one location in the region that cooled off better than usual. Even inside a broad region, the 2026 signal is uneven.

The sea is carrying the signal too

The sea-surface view points the same way as the city data. All 18 selected offshore points, spread across nine European countries, were warmer than their matched 2000–2025 average. The all-points seasonal average was 20.8°C against 19.4°C in the baseline, a difference of +1.4°C. In the aggregate series, 2026 ranked first for July and for 1–19 August, and second for May and June, among 27 matched years. Barcelona had the largest point-level anomaly at +2.3°C, and Naples the warmest 2026 seasonal average at about 25.0°C. These are open-water grid estimates, so they describe the water offshore rather than the temperature at any particular beach or depth.

Back to Lisbon

Across the full season Lisbon is that remarkable at all next to Paris or Madrid. Its daily mean came in at +0.9°C above the baseline and its average daily minimum at +1.1°C, both modest. The warm-night count tells a different story. Over the whole matched season the dashboard finds 22 nights at or above 20°C, against a historical average of 2.1. That is a city whose seasonal anomaly looks small and whose nights changed almost completely.

On a personal note, why not, that is what I noticed in Lisbon this summer. Some evenings the temperature stayed above 20°C or 21°C for hours, and after several nights of it the flat felt as if it never quite reset. Fifteen warm nights out of nineteen is what that feeling looks like counted, and it tells me more about living through a hot summer than any single 40°C afternoon.

The physics behind that is simple enough. A hot afternoon is survivable if the night releases the heat, because buildings shed what they absorbed and the next morning starts from a normal baseline. When the minimum stays above 20°C, the release never finishes, and the next day can start from a warmer baseline. Hot nights are one route by which heat affects health. This dashboard measures that temperature signal; the mortality section is a separate descriptive check.

There is a useful tension in the Lisbon chart, too. The current window is very warm, especially at night, and yet 2026 is still not the warmest complete matched season in the historical series. The full-period leaders are 2006, 2005 and 2023 for daily mean temperature, and 2006, 2023 and 2005 for the average daily minimum. Using identical dates in every year is what keeps both facts visible at once, so an unfinished season can be compared honestly without being dressed up as a record.

The health impact is becoming harder to ignore

The first national reports for 2026 are arriving now. They cover different periods and use different methods, and they all point to a serious public-health cost.

Germany’s Robert Koch Institute estimates 15,800 heat-related deaths between 6 April and 16 August, its highest figure in at least a decade, with about 9,600 of those linked to the extreme heat of late June. The estimate comes from a statistical model rather than death certificates, and it remains provisional. Read the RKI weekly report.

Spain’s MoMo system estimated 5,030 deaths attributable to high temperatures from 15 May to 24 August, already the highest total since monitoring began in 2015. The previous record was 4,789 in 2022, and the 2026 season is still open. Read the EFE report based on MoMo.

England estimated 2,877 heat-associated deaths during the May and June heat episodes, close to its full-summer 2022 record of 2,985, with a prolonged July episode still excluded from the interim figure. Read the UKHSA report.

France recorded at least 7,007 all-cause excess deaths across two heat-wave periods from 17 June to 22 July. Portugal recorded 680 excess deaths in three periods through 30 July, 555 of them between 2 and 8 July. Both are provisional. Read Santé publique France and the Portuguese report from RTP, based on INSA and DGS data.

These figures should not become a league table. Some estimate deaths statistically attributable to heat and others count all-cause excess deaths during a heat episode, and the reporting windows differ. What matters is the shared signal: several national systems, working independently, detected large mortality increases during the same European heat episodes. The AFP synthesis gathers early figures from eight countries and warns, correctly, that the methods do not line up.

The dashboard I did with Codex keeps a second, historical layer alongside this. In the common 2020–2025 dataset, 2022 was the June–August peak in 18 of 31 countries, and in 2025 there were still 24 of 31 countries with positive summer excess mortality. The relationship with temperature is positive and modest.

Across 96 country-years the descriptive correlation is +0.28 for daily-mean anomalies and +0.24 for minimum-temperature anomalies, which is context for investigation rather than a causal estimate. Age, health, population structure, air pollution and infectious disease all move that number too.

The comparison at a glance

Region Cities above historical minimum baseline Median minimum anomaly Cities above historical daily-mean baseline Median daily-mean anomaly Europe 27 of 28 +1.5°C 26 of 28 +1.3°C North America 9 of 10 +0.6°C 8 of 10 +0.3°C

The baseline is the average for the same dates in 2000–2025. The current observation ends on 19 August 2026, so 2026 remains an in-progress year.

What I take from it

One unfinished season cannot establish a long-term trend by itself, and it cannot show that climate change caused any particular city’s temperature. That needs a longer homogeneous record and formal event attribution, which is a different piece of work than this one.

What it can do is settle a smaller question. The feeling that the house never cools down was not a memory built out of one bad week. In this sample the nights really were warmer, in almost every European location, by a margin large enough to notice. The dashboard is there if you want to check your own city, and I would rather you argue with the data than take my word for it.

Sources and method

The primary source is the NASA POWER Daily API. The temperature fields come from the MERRA-2 reanalysis and are requested in UTC at each city coordinate, so the dashboard uses a gridded estimate consistently across cities rather than a station-observation product.

Each city’s daily mean, minimum and maximum are calculated for every day in the window and compared with the same dates in 2000–2025. A “warm night” means a daily minimum of at least 20°C, and regional summaries are simple averages of the included city series. The data snapshot is dated 19 August 2026, current-year values can change as observations arrive, and the raw snapshot and build scripts ship with the dashboard.

The sea-surface series comes from NOAA NCEI’s Optimum Interpolation Sea Surface Temperature version 2.1. The final daily product covers the historical years and the near-real-time product covers 2026. Each marker is an offshore 0.25° grid cell, and the aggregate averages 18 points.

The historical excess-mortality series uses the EuroMOMO weekly bulletin and Eurostat’s country series. The 2026 national snapshots use the sources linked in the health section. The correlation uses 96 country-years, is exploratory, and is not adjusted for age, population, health conditions or other confounders.

Read the interactive dashboard: Were summer nights unusually warm in 2026?

(Do you have suggestions on how to make this better, send it my way: joaotome.com)

emot | Random Stories is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.