A EUMETSAT analysis finds El Niño summers in western and central Europe warmer, sunnier and drier than normal. We checked its claims, maps and event list against NOAA's data, and kept a historical pattern, a mechanism and the cause of 2026's heat apart.
EUMETSAT's case study Impact of El Niño Southern Oscillation on European climate: A satellite view (5 October 2026, Rob Roebeling and co-authors) compares July to September weather in six El Niño years with the 1991 to 2020 normal and with six La Niña years. It finds El Niño summers warmer, sunnier and drier than normal across much of western and central Europe, with the opposite signal over parts of the eastern Mediterranean and southeastern Europe. Its authors call the link a probabilistic attribution and not proof of a direct cause, and they say human-caused warming remains the dominant driver of extreme European summers. This piece goes through the study's claims one at a time, reads what its maps show and how large the signals are, checks its event list against NOAA's ONI and RONI files, compares 2026 with the composite, and sets out what kind of evidence would be needed to say El Niño mattered in 2026.
Update, 7 October 2026. The first version of this piece rested on a secondary report because the EUMETSAT page would not load for us. We have now read the original case study, including its figures, from a saved copy, and this version replaces the secondary report with it. It adds a claim-by-claim review, a check of the study's event list against NOAA's data, a comparison of 2026 with the composite, and the study's winter and spring 2027 analogue. Where the original changes what we said, the text says so.
The short answer is that a pattern across past summers, a physical mechanism and the cause of one summer's heat are three separate claims. The EUMETSAT study speaks mainly to the first, and its authors say so. Europe's 2026 heat had other documented ingredients, and those observations describe the regional conditions without establishing how much El Niño influenced them.
The study takes July to September anomalies against the 1991 to 2020 normal for four variables: 2 m air temperature (ERA5-Land, labeled ERA-5 on its maps), precipitation (GPCP), root zone soil moisture (EUMETSAT's H SAF scatterometer record, defined as the layer between 28 and 100 cm) and downward surface solar radiation (CM SAF). It averages them over six El Niño years (1991, 1997, 2009, 2015, 2018, 2023) and, separately, six La Niña years (1998, 1999, 2007, 2010, 2020, 2021). The authors say such events typically begin around July and persist until the following June, so these are developing-year summers, which is the right comparison for 2026. They describe El Niño summers as generally warmer, sunnier and drier than normal and La Niña summers as cooler, cloudier and wetter, although the La Niña maps are not shown.
To test whether this exceeds chance, the authors compute z-scores for the El Niño versus La Niña contrast relative to the variability in non-ENSO years. The authors report that several regions exceed plus or minus 2 and interpret that as roughly the two-tailed 95% level under a Gaussian assumption. That is their interpretation. Dividing a contrast by background variability does not by itself give a calibrated significance test for a difference between two six-event means, and the published method does not say enough (how the background spread and the contrast's own sampling variability were estimated, or how many degrees of freedom apply) for us to verify the inference independently. They call the result a probabilistic attribution rather than definitive proof of a direct causal relationship, and they say human-caused warming remains the dominant driver of extreme European summers whatever the ENSO phase, while allowing that the teleconnections may have amplified 2026's warmth and drought.
What the article does not give. There is no table of magnitudes or uncertainty ranges, so the size of the effect can only be read off the maps. No leave-one-out test is described, so we cannot tell whether one or two events drive the pattern. With six events per group and many grid points, some regions could exceed any such threshold by chance, and the text describes no adjustment for that. The contrast also treats El Niño and La Niña as opposite responses, although the authors note earlier that the two phases are asymmetric, with La Niña impacts weaker and shifted westward.
Three questions, three levels of evidence. A pattern across past summers is an association. An explanation of how tropical Pacific warmth reaches Europe is a mechanism. A statement about how much El Niño changed one particular summer is attribution, and it needs its own counterfactual analysis: model ensembles with and without El Niño's tropical Pacific temperature pattern, with the assumptions stated. Removing human-caused warming answers a separate attribution question. A historical comparison, however carefully built, speaks to the first question only.
Here is each claim in the case study, what the article offers for it, and what this site was able to check.
| Claim | What the article shows | What this site could check |
|---|---|---|
| El Niño summers in Europe are warmer, sunnier and drier than normal. | Composite maps of six July to September seasons (Fig. 3), with scales to plus or minus 1.5 K, 18 W/m², 10% soil moisture and 1 mm/day. | The pattern is real in the figure, but it is a regional dipole, and the event-selection rule is not specified in enough detail to reproduce the list (see the event table below). |
| The El Niño versus La Niña contrast exceeds background variability in several regions. | Z-score maps (Fig. 4) shaded only above about 1.5 and reaching 4.5. | Not reproducible: no values, no field-significance adjustment, no leave-one-out test, and too little method detail to verify the 95% reading. |
| The warm, dry summers may be partly attributed to teleconnections. | The authors' own wording: a probabilistic attribution, not definitive proof. | Consistent with an association. The authors allow that 2026's warmth and drought may have been amplified, but they present no event-specific analysis and call human-caused warming the dominant driver. |
| Summer 2026 was broadly consistent with some El Niño patterns but not a textbook response. | 2026 June to August maps (Fig. 5) and the statement that eastern Europe and Scandinavia were cooler than the composite. | The maps use much wider color scales than the composite, and the 2026 event is already stronger than the composite average (see below). |
| Teleconnections are strongest in late winter, about six months after onset. | Text citing Stockdale et al. (2017) and Brönnimann (2007); schematic in Fig. 2. | Not tested in the article. The timing agrees with the ONI file, where all six composite El Niño events peak in November to January (five) or December to February (one). |
| Winter 2026/27 and spring 2027 are likely warmer and wetter, with a cloudier spring. | Composite maps (Figs. 6 and 7), called a statistical analogue and not a dynamical forecast. | Not testable yet. Checkpoint 3 below sets a date and a failure condition. |
Every size below is read by eye from the article's color scales, since it gives no numbers. The July to September temperature composite (Fig. 3) has its darkest shading, near +1 K or a little above, over France, the Alps and southern Germany, a warm band running toward Poland and the Baltic, and weaker anomalies (roughly 0.2 to 0.5 K) over the United Kingdom. Over the Black Sea, Turkey and the eastern Mediterranean the composite is near zero or slightly cool. The solar radiation composite is brighter by roughly 6 to 14 W/m² over Germany, the Alps, Poland and the Baltic and dimmer over Greece and the eastern Mediterranean. Soil moisture is lower by roughly 2 to 5 on the map's percent scale across France, Germany and Poland and higher over the Balkans and around the Black Sea. Precipitation is below normal along a band from the Alps to Poland, by a few tenths of a millimeter per day at the darkest spots, and above normal by up to roughly 1 mm/day over the Balkans and Greece.
So the headline that El Niño summers are warmer and drier describes western and central Europe. Over southeastern Europe and the eastern Mediterranean the same composite shows the opposite signs for temperature, rain, soil moisture and sunshine. The z-score maps (Fig. 4) agree: warm, sunny and dry significance over France, Germany, the Alps and the Baltic, and significance of the opposite sign over the Balkans, the eastern Mediterranean and North Africa. The shading starts at about plus or minus 1.5, so the plus or minus 2 threshold the authors cite cannot be read from the maps. The sign convention is a question for the authors. The text defines the contrast as the La Niña mean minus the El Niño mean, yet in the temperature and radiation panels positive values appear where the El Niño composite is positive. Without the La Niña maps or the calculation, we cannot tell whether the definition, the plotting convention or the wording is at fault.
Only some of the places the authors name as especially dry in 2026 (France, Spain, Germany, Austria, Hungary, the United Kingdom) line up with the strongest composite signal. By eye, the strongest dry and warm shading sits over France, Germany and the Alps, and it is weaker or patchier over the United Kingdom and Iberia. That fits the authors' description of a partial match and does not show that El Niño drove the 2026 pattern in any place.
The article says its El Niño years had ONI above +0.5°C and its La Niña years below −0.5°C. This table puts those years against NOAA's ONI file and RONI file (both read 7 October 2026). The peak column is the largest three-month ONI value from July to September of the year through April to June of the next. The ONI is not the relative index; the RONI subtracts the tropical-mean warming.
| Year | Status | JAS ONI | JAS RONI | ONI peak |
|---|---|---|---|---|
| 1991 | El Niño, in study | +0.67 | +0.83 | +1.54 |
| 1997 | El Niño, in study | +1.79 | +1.84 | +2.37 |
| 2009 | El Niño, in study | +0.56 | +0.42 | +1.50 |
| 2015 | El Niño, in study | +1.73 | +1.54 | +2.59 |
| 2018 | El Niño, in study | +0.30 | +0.16 | +1.05 |
| 2023 | El Niño, in study | +1.25 | +0.79 | +1.99 |
| 1998 | La Niña, in study | −0.83 | −1.14 | −1.51 |
| 1999 | La Niña, in study | −0.94 | −0.86 | −1.53 |
| 2007 | La Niña, in study | −0.77 | −0.79 | −1.76 |
| 2010 | La Niña, in study | −1.24 | −1.39 | −1.57 |
| 2020 | La Niña, in study | −0.43 | −0.85 | −1.11 |
| 2021 | La Niña, in study | −0.46 | −0.76 | −0.91 |
| 2002 | Beyond +0.5 at JAS, onset year, not in study | +0.63 | +0.80 | +1.19 |
| 2004 | Beyond +0.5 at JAS, onset year, not in study | +0.54 | +0.69 | +0.71 |
| 1995 | Beyond −0.5 at JAS, onset year, not in study | −0.55 | −0.57 | −1.05 |
| 2022 | Beyond −0.5 at JAS, continuing event, not in study | −0.78 | −0.94 | −0.89 |
| 2026 | Current event, provisional | +2.16 | +1.69 | not yet |
What the table shows. The article says its events had an ONI above +0.5°C or below −0.5°C, but it does not say whether that applies to the July to September value, to some other seasonal or annual mean, or to the event as a whole. The sub-threshold years are therefore not shown to violate its rule. Three of the twelve listed years (2018, 2020 and 2021) have a July to September ONI inside plus or minus 0.5°C. Each belongs to an event that went beyond that level later (2018's ONI first reached +0.5°C in August to October, and the peaks in the table are +1.05°C, −1.11°C and −0.91°C), so an event-based rule could explain them.
The omissions differ in kind. By a simple test, 1995, 2002 and 2004 are onset years: their ONI had not reached the threshold by April to June (+0.03°C, +0.30°C and +0.12°C) and had by July to September. 2022 is a continuing event: its ONI was already −0.83°C in April to June and is −0.78°C in July to September. An onset-only rule would not explain the list either, because 1999 (April to June ONI −0.79°C) is also a continuing year and is included, and the article does not say how it treats multi-year events. We therefore read the selection rule as insufficiently specified, not as demonstrably broken. The article also calls the events strong, although +0.5°C is CPC's threshold for a weak one, and the weakest listed events (2018, 2020 and 2021) peaked at only 0.9 to 1.1°C in magnitude.
The index changes the sample. The study cites the ONI, and this site tracks the RONI for its checklist. Applying the same plus or minus 0.5°C rule to July to September RONI values would pick a different El Niño set: 2009 (+0.42°C) and 2018 (+0.16°C) would drop out, and 1994, 2002 and 2004 would come in, for 1991, 1994, 1997, 2002, 2004, 2015 and 2023. The La Niña set would grow to 11 years. The article's reference list includes a WMO document on ONI versus RONI, but its text names only the ONI. That does not make either index wrong, but it shows that the composite depends on how an event is defined, and the article gives one definition without testing the other.
A trend check, with limits. The composites are anomalies against a 1991 to 2020 normal whose midpoint is 2005.5, so a warming trend would add a warm offset to any composite from later years. The study's El Niño years average 2008.8 and its La Niña years 2009.2, almost the same, so a common linear trend would approximately cancel in the El Niño versus La Niña contrast the authors test. Similar mean years do not show how large the trend contribution is in either composite, and it does not cancel in the maps of El Niño years alone (Figs. 3, 6 and 7), which carry a positive trend component of unknown size that the article does not remove and that we did not calculate. This is our own arithmetic from the year lists, not something the authors report.
Copernicus reports its figures in its summer review and PDF, using ERA5 and a 1991 to 2020 reference. Over European land (25°W to 40°E, 34°N to 72°N), summer was 1.17°C above that average, the third warmest, behind 2024 (+1.55°C) and 2022 (+1.34°C). Over western Europe (11°W to 15°E, 37°N to 55°N), it averaged 21.7°C, or 2.5°C above, the warmest on record and ahead of 2003. Those are two different regions and two different rankings, and they should not be merged.
Copernicus's news release also reports that 52% of European land experienced at least very strong heat stress at some point during the summer, a cumulative exposure and not the share affected on any one day. The review describes repeated blocking high-pressure systems over western Europe from May to September, and record sea surface temperatures with marine heatwaves around it. Those observations describe the regional conditions; they do not establish how much El Niño influenced them.
The EUMETSAT authors compare 2026 with their composite and stop short of a match. Their Figure 5 shows June to August 2026 anomalies for temperature, root zone soil moisture and solar radiation, and they call the summer broadly consistent with some El Niño patterns but not a complete textbook response: eastern Europe and Scandinavia were cooler than the composite, where it would have been warmer and drier. They add that teleconnections are weaker and more variable early in an event, that blocking, marine heatwaves and soil moisture deficits can reinforce or cancel the signal, and that the dry soils and very sunny weather raised wildfire risk.
The scale of 2026 against the composite. The 2026 maps are drawn on much wider color scales than the composite maps: about plus or minus 7°C for temperature against 1.5 K, plus or minus 20% for soil moisture against 10%, and plus or minus 50 W/m² for radiation against 18. A single year will always show larger extremes than a six-year mean, so this does not argue against El Niño by itself, and the two sets of maps cover different windows (June to August against July to September). The two sets of maps differ in season, averaging and data source, so they support a comparison of scale but not a quantitative estimate of how much of 2026's anomaly a composite-sized signal would explain.
The event was building through the summer. CPC's relative Oceanic Niño Index (RONI), a three-month running mean, read +0.49°C for April to June, +0.97°C for May to July and +1.36°C for June to August. The latest season in the file, July to September, is +1.69°C. The chart shows the trend and not the official ENSO status, which CPC declares in its monthly discussion and not from one seasonal value.
Not by strength. The July to September ONI for 2026 is +2.16°C, against a mean of +1.05°C for the study's six El Niño years, about twice as large, and above every one of those six (the highest is 1997 at +1.79°C). It is also the highest July to September ONI in NOAA's file, which starts in 1950, ahead of 1997, 2015 (+1.73°C), 1987 (+1.46°C) and 1965 (+1.27°C). On the relative index the picture is less extreme: 2026's +1.69°C ranks second to 1997 (+1.84°C) for that season, with 2015 at +1.54°C, because the RONI removes the tropical-mean warming that the ONI includes. The 2026 value is the latest season in the file, so it is provisional and can be revised.
The composite events reached a mean ONI peak of +1.84°C, so 2026 is already above the average peak of the six events, and the authors, citing a WMO release, expect the 2026 peak around the end of the year. The RONI also rose quickly, from +0.49°C for April to June to +1.69°C for July to September. There are two ways to read that. The composite may understate what a stronger event can do. Or the strength of the response may not scale with the index, since the authors themselves say teleconnections depend on the background flow and differ from event to event. The article does not separate the two, and with six events it could not. A sensitivity test would recompute the composite from 1997 and 2015 alone, the two cases closest to 2026 in strength, while stating plainly that two events cannot establish a pattern.
El Niño shifts tropical rainfall and heating, which can move jet streams and storm tracks, and the North Atlantic is the route to Europe. The EUMETSAT article lays out the usual pathway, and its Figure 2 is a schematic of it: a Walker circulation over the equatorial Pacific, Hadley cells on either side, and Rossby wave trains arcing across the Northern Hemisphere. During El Niño the Walker circulation shifts east, ascending air moves to the eastern Pacific and descending air toward the tropical Atlantic, which strengthens the Hadley circulation and can excite Rossby waves that travel poleward and eastward. Those can alter the North Atlantic jet, shift storm tracks and affect the North Atlantic Oscillation. The figure is an illustration, not an analysis of any one summer.
The article also says these teleconnections are strongest in late winter, that their impacts depend on the season and on the event, and that La Niña impacts are generally weaker and shifted westward. So its summer result is best read as consistent with that literature and not as a demonstrated summer mechanism. ECMWF's review of ENSO and European climate likewise describes responses that vary by region, season and event strength, including cold over northern Europe in late winter. We found no verified summer mechanism, so the summer association remains a statistical result with a plausible but untested pathway.
The article treats the mean anomalies of the same six events as an empirical analogue for January to March 2027 (it calls this winter) and April to June 2027, on the assumption that the event persists to June and behaves like those six. It concludes that winter is likely warmer and wetter than in neutral or La Niña winters, that these conditions persist into spring, and that spring adds more cloud. It notes that the analogue resembles ECMWF's seasonal forecast for winter 2026/27, as accessed on 28 September 2026, and stresses that it is uncertain how strongly the teleconnections will emerge. We do not repeat it as a forecast.
The maps are less uniform than that summary. In the winter composite (Fig. 6), by eye, the temperature signal is strongest over central Europe and the Balkans, at about 1.2 to 1.5 K at the darkest spots, with cooler anomalies over the northern North Atlantic, northern Scandinavia and northern Russia. Soil moisture is higher across central and eastern Europe, rainfall is higher over Iberia, the adjacent Atlantic and parts of the Alps and central Europe, and rainfall is lower over Greece, Turkey and the eastern Mediterranean. The solar radiation signal is weak. In the spring composite (Fig. 7), most of Europe is warmer by roughly 0.2 to 0.8 K, with cooler anomalies over the Balkans and around the Black Sea. The cloudier signal, roughly 6 to 14 W/m² less sunshine, is concentrated over France, the Alps, Italy and the Balkans, while southern Scandinavia and the Baltic are sunnier by a similar amount. Rain is higher across Iberia, the Alps, central Europe and the Balkans, and slightly lower over Scandinavia and the British Isles. So warmer and wetter is a description of the central and southern parts of the continent.
One labeling point matters for reading these panels. The spring panel lists 1992, 1998, 2010, 2016, 2019 and 2024, the calendar years after each onset year, while the winter panel's title lists the onset years themselves, 1991 to 2023. January to March of an onset year comes before the event develops, so we read the winter title as shorthand for the following year, but the article does not say so.
These are points where the case study is unclear or inconsistent with itself. None overturns its main result, and each is a question for the authors.
Is the historical association robust? The article gives the pattern and its significance maps, not the numbers needed to judge it. A fuller test would state the event-selection rule, report the size of the El Niño versus La Niña difference with an uncertainty range, and recompute leaving out each event in turn. It would also repeat the composite with an event list built from the RONI, and with the omitted years added, since the table above shows the sample depends on the index and the rule. A difference that stays warm, with an interval that excludes zero and a sign that survives dropping any single event and changing the list, would support a real association. A difference that shrinks toward zero or leans on one or two events would weaken it. The article reports none of these checks, and the maps cannot supply them.
How much did El Niño contribute to 2026? None of those checks can say. They test whether a pattern is dependable, not what it did in one year. Quantifying El Niño's contribution would require model ensembles with and without its tropical Pacific temperature pattern, with the experiment's assumptions stated explicitly. Blocking and ocean conditions could be part of the response being tested, so a design should say whether they are left free or held fixed. Removing human-caused warming answers a separate attribution question. We did not find such an analysis for European summer 2026, and the EUMETSAT article does not claim to provide one.
The evidence here does not support either extreme: that El Niño caused Europe's heat, or that it has no effect there. It supports a narrower statement. In a small sample of past developing El Niño summers whose selection rule is not fully specified, western and central Europe tended to be warmer and drier than in La Niña summers, southeastern Europe tended to show the opposite, and the authors themselves rank human-caused warming first among the drivers of extreme summers. Regions and months can differ, and historical averages are not a forecast, so this piece makes none for winter 2026 to 2027. Related coverage: why El Niño did not cause a US heat wave and the Met Office's Northwest Europe expectations.