Demise of the European summer: the challenge of unpredictable extremes

There used to be a season called European summer – one that was well-behaved, predictable and easy to plan around. It arrived in June, warmed the rivers, drew down reservoirs a little, and left in September. Farmers irrigated on schedule. Rhine barges could usually load fully. Alpine snowmelt replaced much of what evaporation took. Harvests, holidays and hydropower budgets were built on the assumption that one year would look broadly like the last.

That assumption no longer holds. Since 2018, Europe has experienced a multi-year drought without precedent in more than 250 years of reconstructed records, one of the deadliest floods in modern German and Belgian history, and, this year, a combination of heat and drought that pushed several of the continent's largest rivers to their lowest levels on record. The problem is not only that summers are hotter, it is more so the whiplash – the swings between extremes are larger and come closer together, and that institutions tend to prepare for whichever disaster happened most recently.

Rivers are where these pressures meet. They supply drinking water, irrigation and cooling water, carry freight and generate electricity, and they cross national borders. When they flood or run dry, the consequences reach well beyond hydrology.

The drought that reset the baseline

The current sequence is usually dated to 2018. That spring, large parts of central, western and northern Europe entered a drought that did not fully end until 2020. A study by researchers at the Helmholtz Centre for Environmental Research (UFZ), which reconstructed soil-moisture droughts back to 1766, found the 2018–2020 event to be the most intense in that record [1]. On average it covered about 36% of Europe's land area and lasted just over 12 months at any given location; at its peak, half the continent was affected. Only the drought of 1857–1860 lasted longer.

What set 2018–2020 apart was temperature rather than rainfall. The precipitation deficit, around 20% below the long-term mean, was comparable to earlier major droughts. The air-temperature anomaly over the affected area, about +2.8 °C, was not. The impact on farming was substantial. After adjusting for long-term yield gains, grain maize production fell by 20–40% in Belgium, the Netherlands, Luxembourg, Germany and France, wheat by up to 17.5% in Germany, and barley by around 10% across most of the continent [1].

The same study ran the analysis forward using climate-model projections. Future droughts that exceed the 2018–2020 benchmark are projected to be of similar intensity and somewhat larger in area. They are also projected to last much longer: about 40 months on average under a moderate-emissions scenario, and potentially several times longer under high emissions (Figure 1).

By 2020, scarcity dominated European water management. Reservoir operators, agriculture ministries and river authorities developed irrigation restrictions, low-flow protocols, drought contingency plans and early-warning thresholds. Most of that planning assumed the next danger would resemble this one: a slow-building deficit, visible months ahead, manageable through rationing.

Then the water came all at once

In mid-July 2021, a slow-moving low-pressure system stalled over the Eifel and Ardennes uplands straddling Germany, Belgium and the Netherlands. More than 90 mm of rain fell in a single day around the Ahr and Erft rivers, far above previous records [2]. The Ahr, normally a shallow river with a mean flow of around 7 cubic metres per second, rose by several metres within hours.

Most gauges on the Ahr were damaged or destroyed before the flood peaked, so its size has had to be reconstructed. The Rhineland-Palatinate hydrological service estimates the peak at Altenahr at about 850 m³/s. Published estimates for the valley range from roughly 400 to 1,300 m³/s depending on location and method [3]. Either way, the flood was several times larger than anything recorded since systematic gauging began after the Second World War.

It was not, however, unprecedented. Historical records and flood marks show that a comparable flood struck the valley in 1804, and a large one in 1910 [3, 4]. Those events were excluded from the statistics used to draw official hazard maps, because the maps relied only on the gauged series from 1947 onwards. As a result, the 100-year flood for the Ahr had been estimated at just 241 m³/s [5], a fraction of what arrived in 2021 (Figure 2).

At least 184 people died in Germany and 38 in Belgium. The Ahr valley alone accounted for 134 German deaths, more than half of them people over 60. Thousands of buildings and more than a hundred bridges were destroyed or badly damaged. Measured by loss of life, it was Germany's worst natural disaster since the North Sea flood of 1962.

The flood also showed how far preparedness had drifted from actual risk. The real flood exceeded the "extreme" hazard scenario mapped for the valley. Warnings reached some communities only shortly before midnight. Post-event surveys found that about a third of affected residents in the worst-hit German states said they had not been warned, or had not understood the danger in time.

A rapid attribution study by the World Weather Attribution (WWA) group concluded that climate change had made heavy rainfall of this kind in the region between 1.2 and 9 times more likely, and 3–19% more intense [2].

When flood protection becomes the whole agenda

The policy response was swift and large. Germany's federal and state governments set up a reconstruction fund of roughly €30 billion. Forecasting agencies reviewed warning chains, siren networks and the gap between meteorological forecasts and local decisions. In Wallonia, municipalities began drawing up buyback and relocation schemes for the most exposed riverside properties. Across the EU, adaptation funding and hazard-mapping programmes shifted towards flood risk. For several years, "river resilience" in policy discussions largely meant resilience to flooding.

That was a reasonable response to a disaster of that scale. But it reflects a familiar pattern: attention and money follow the most recent catastrophe, and the most recent catastrophe is a poor guide to the next one when the underlying hazard runs in both directions. A higher dike does little for a river that, a few summers later, is carrying a fraction of its normal flow.

The drought came back, and worse

It returned in 2022, a summer that a senior researcher at the European Commission's Joint Research Centre described at the time as possibly the driest in 500 years. On the Rhine at Kaub, the shallowest point on the river's main shipping route, the water level fell to 30 cm in mid-August, close to the record of 25 cm set in October 2018 [7].

This year went further. Between April and June 2026, much of Europe, from Portugal to southern Finland, received below-average rainfall, and successive heatwaves deepened the deficit [8]. On 5 August, the Kaub gauge fell below its 2018 record [9]. By mid-August it had dropped into single digits for the first time since records began [10]. On 1 October it read −6 cm [7]. The gauge reading is not the depth of the water: at a reading of zero, the navigation channel at Kaub still holds a little over a metre. But freight traffic becomes practically impossible below about 40 cm, and a large container vessel can carry only about a quarter of its capacity at 75 cm (Figure 3).

The low water was not confined to the Rhine. At Lobith, where the river enters the Netherlands, the water level fell to a record 6.47 m above Amsterdam datum on 30 July, just below the previous low set in August 2022 [13]. Modelled discharge there averaged 850 m³/s in July, 37% of the 2019–2025 norm. That monthly average was lower than the lowest single July day in the preceding seven years [14]. In France, the Garonne at Tonneins carried about 5% of its normal July flow and the Loire at Saumur about 13%. On the Danube, flow through the Iron Gates fell to 48% of normal (Figure 4a,b).

In France, the environment ministry reported that nearly 70% of the country was under water-use restrictions on 9 August, with 67 departments at the highest crisis level [15]. Later in the month, reports put the share above 80% [16]. The WWA group noted that France was heading for its smallest maize harvest in 50 years, and that Romania had lost more than a million hectares of maize [8].

Heat, not rain

The scientific explanation for 2026 differs in an important way from 2018–2020. WWA's attribution study, published on 23 July, separated the drought into its components: low rainfall, high evaporative demand, and the soil-moisture deficit that results from both [8].

In western Europe, neither observations nor climate models show a long-term change in low spring rainfall. The rain shortfall of April–June 2026, roughly a one-in-25-year event, does not appear to have been made more likely by climate change. Evaporative demand is a different matter. This is the atmosphere's capacity to draw water out of soils, plants and open water, and it rises steeply with temperature. The researchers found that the highly evaporative conditions of April–June 2026 had become about 80 times more likely, equivalent to about 7% more intense, than in a climate 1.4 °C cooler. The resulting soil-moisture drought in the west was about five times more likely, although the uncertainty range is wide. In eastern Europe, where dry conditions had persisted since 2025, the January–June soil-moisture deficit was about 11 times more likely and the associated evaporative demand about 40 times more likely (Figure 5).

The distinction matters for planning. A drought driven by missing rain can, at least in principle, be anticipated from rainfall deficits. A drought driven by heat can develop in years when rainfall is close to normal, and it intensifies with every increment of warming. As the WWA authors put it, rising evaporative demand has shifted drought thresholds: rainfall deficits that would not previously have caused drought, now do so.

Fire

Heat and dry vegetation also fed the fire season. By late July, more than 116,000 hectares had burned in Spain and 42,000 in France [8]. Across the EU, data from the European Forest Fire Information System show that by 5 August about 505,000 hectares had burned in 2026. That is two and a half times the 2006–2025 average for that date, and close to the highest value recorded by that point in any year of the period (Figure 6).

The timing is telling. From March to June, the area burned in 2026 ran about 40% above average, an early start but not an exceptional one. The step change came in July, as the heat intensified and river flows collapsed. In the single week to 22 July, about 186,000 hectares burned, roughly seven times the area that burns in an average week at that time of year.

Why this is a political problem, not just a hydrological one

None of the above is only an engineering story. Rivers in Europe sit inside a dense web of competing claims — drinking water, irrigation, industrial cooling, hydropower, freight, tourism, ecosystems — and a transboundary one at that, since the Rhine, the Meuse, and the Danube all cross multiple national jurisdictions with their own institutions, priorities, and electorates. When water is plentiful, of course, all is well. When it is scarce, the administrative details being to come to the forefront. A government deciding whether to keep cooling water flowing to a power station, protect irrigation for a weakened harvest, or maintain minimum ecological flows is making energy, agricultural and environmental policy at once, often within days.

The energy angle is especially sharp as an example because Europe's electricity system leans on rivers in two structurally different ways at the same time: hydropower needs the water to move, and thermal and nuclear plants need it to be cool and abundant enough to absorb waste heat. A drought that lowers river levels and raises river temperatures attacks both simultaneously.

That is what happened along the drought corridor from the Alps to the lower Danube this summer. Hydropower output in July was 53% below normal in Hungary, 38% in Italy and around a third in Austria, Serbia and Romania. The shortfall across the corridor from April to July was about 13 TWh [14] (Figure 4c). On the Danube, nuclear plants also reduced output as river flows fell to their lowest in seven years. At the end of July, daily generation at Hungary's Paks plant fell by 45% in three days. The analysis found that these cuts coincided more closely with river conditions than with peaks in air temperature [14].

The effect was uneven. Portugal's hydropower output in July was more than double its norm after a wet first half of the year, and Spanish reservoirs were near the top of their historical range. That surplus offset about 70% of the corridor's losses at continental scale, but not where they occurred.

Because European electricity markets are interconnected, the shortfall showed up in prices well beyond the worst-affected rivers. July day-ahead prices averaged €122 per megawatt-hour in Hungary, compared with €103 a year earlier. Austria was at €117, up from €88, and France at €95, up from €58 (Figure 4d). Prices rose in two phases. A heat-driven spike in demand pushed Hungarian, Romanian and Serbian daily averages above €225/MWh on 1 July. A slower climb followed as rivers emptied, with Swiss prices averaging €134/MWh over the last ten days of the month [14].

Two hazards, one problem

The political difficulty is that flood defence and drought resilience do not always pull infrastructure investment in the same direction, and governments that spent heavily and sincerely on the former after 2021 are now being asked, mid-decade, to pivot toward the latter without abandoning what they just built. The more encouraging strand of thinking — increasingly visible in EU adaptation policy and in river basin management plans from the Rhine to smaller cross-border catchments — treats the two hazards as symptoms of the same underlying problem rather than as competitors for budget. Reconnecting rivers to their historic floodplains, restoring wetlands, and slowing water down across a catchment rather than channelling it away as fast as possible can blunt a flood peak and hold moisture in the landscape for the dry months that follow, which is the basic logic behind the nature-based approaches now being piloted in flood-prone, drought-prone reaches across the continent. It is a harder sell politically than a single-purpose dike or reservoir, because its benefits are diffuse and slow to show up in a press release. But they fit a climate that can deliver both extremes within a few years.

WWA's own policy assessment points to a gap here. Drought management plans are not mandatory under the EU Water Framework Directive, so preparedness varies widely between member states [8].

An unreliable season

The old European summer was trustworthy because it was boring: predictable enough that entire economies, from Alpine hydropower schedules to Mediterranean tourism, could be built on its rhythm without much hedging. The pattern that has replaced it does not vary gently around the old average. It moves from one extreme to the other within a few years, and each extreme tends to crowd out preparation for the next.

The record so far suggests both extremes will recur. The Ahr flood was matched in 1804 and will be matched again. The 2018 low-water record at Kaub stood for less than eight years before it was broken by more than 30 cm. Rivers will continue to rise and fall as rainfall and evaporation dictate. The question for the communities along them is whether planning can be built for both at once.

---

Notes on the figures

All figures were produced for this article from published data. Values are transcribed from the sources listed below, except Figure 6, which uses the EFFIS cumulative burnt-area export directly. Where sources report only ranges or central estimates, the figures show only what was reported. Values for 2026 are provisional and may be revised.

Sources

1. Rakovec, O. et al. (2022). The 2018–2020 multi-year drought sets a new benchmark in Europe. Earth's Future 10, e2021EF002394. https://doi.org/10.1029/2021EF002394
2. World Meteorological Organization (2021). Climate change made floods in Western Europe more likely (summary of the World Weather Attribution study by Kreienkamp et al.). https://wmo.int/node/14172
3. Fluvial deposits of the Ahr river (western Germany) reveal recurring high-magnitude flood events over the last 1,500 years. Earth Surface Processes and Landforms (2025), Table 4. https://onlinelibrary.wiley.com/doi/10.1002/esp.70219
4. Roggenkamp, T. & Herget, J. (2014). Reconstructing peak discharges of historic floods of the river Ahr, Germany. Erdkunde 68, 49–59.
5. CEDIM Forensic Disaster Analysis Group (2021). Hochwasser Mitteleuropa, Juli 2021 (Deutschland). Karlsruhe Institute of Technology.
6. Vorogushyn, S., Apel, H., Kemter, M. & Thieken, A. H. (2022). Analyse der Hochwassergefährdung im Ahrtal unter Berücksichtigung historischer Hochwasser. Hydrologie und Wasserbewirtschaftung 66(5). https://gfzpublic.gfz-potsdam.de/pubman/item/item_5013380_4/component/file_5013767/5013380.pdf
7. Pegel Kaub. Wikipedia (German), accessed October 2026; wasserpegel.com, Pegel Kaub (annual minima 2018 and 2022). https://de.wikipedia.org/wiki/Pegel_Kaub
8. World Weather Attribution (2026). Increasingly hot Europe faces more severe droughts and growing challenges for water and land management. 23 July 2026. https://www.worldweatherattribution.org/increasingly-hot-europe-faces-more-severe-droughts-and-growing-challenges-for-water-and-land-management/
9. Argus Media (2026). Historischer Tiefstand beim Pegel Kaub. https://www.argusmedia.com/de/news-and-insights/latest-market-news/2861391-historischer-tiefstand-beim-pegel-kaub
10. Kaub gauging station. Wikipedia, accessed October 2026 (citing Contargo, Low Water, 2017, and ELWIS). https://en.wikipedia.org/wiki/Kaub_gauging_station
11. ELWIS, German Federal Waterways and Shipping Administration. Pegel KAUB: gauge reference values. https://www.elwis.de/DE/dynamisch/Wasserstaende/Pegeleinzeln:KAUB
12. KuLaDig. Pegelturm an der Rheinuferstraße in Kaub. https://www.kuladig.de/Objektansicht/KLD-352119
13. NL Times (2026). Rhine reaches lowest water level ever recorded at Dutch border. 30 July 2026. https://nltimes.nl/2026/07/30/rhine-reaches-lowest-water-level-ever-recorded-dutch-border
14. Sen, S. (WSP) (2026). When the rivers stop flowing: Europe's drought summer measured. pv magazine, 10 August 2026. https://www.pv-magazine.com/2026/08/10/when-the-rivers-stop-flowing-europes-drought-summer-measured/
15. Euronews (2026). Drought forces water restrictions across nearly 70% of France, government says. 10 August 2026. https://www.euronews.com/my-europe/2026/08/10/drought-forces-water-restrictions-across-nearly-70-of-france-government-says
16. Le Journal de la Météo de l'Ain (2026). Sécheresse : 80 % du territoire sous restrictions d'eau. 17 August 2026. https://www.lejma.fr/2026/08/17/secheresse-80-du-territoire-sous-restrictions-deau/
17. European Forest Fire Information System (EFFIS), Copernicus Emergency Management Service. Cumulative burnt areas in EU27 countries, 2006–2026 (data export). https://forest-fire.emergency.copernicus.eu/