EU gas storage will be about 77% full when the heating season opens on November 1, 12 points below the fifteen-year norm and short of the 90% target. What happens after that will dependend heavily on the weather. Is a repeat of the 2022 gas crisis on the cards? Not necessarily. An IntelliNews Lambda analysis shows, the problem is not the empty tanks, that will not push up gas prices by itself, but the fact that Europe has so little wiggle room because it has too small a gas cushion underneath it.

That is not a disaster and it's not so unusual either. Comparing this expected result with the last fifteen Augusts and tanks storage lands between 72% and 80% half the time, between 69% and 82% four times in five.

But it does mean the probability of hitting the EU’s mandatory 90% target by November 1 this year is zero. The probability of starting this year’s winter with tanks less than 80% full is 75%, according to IntelliNews Lambda 15-year baseline calculation.

As IntelliNews Lambda reported, most of the tanks in Europe are on target, at or close to, their historical averages. Ukraine in particular is actually well ahead of the average in anticipation of a brutal Russian bombing this winter. However, Germany and the Netherlands are the problem kids.

Most of the hole that can’t be filled before the heating starts is due to the extremely low levels of gas storage in Germany, which has the biggest banks in Europe: currently only half full, if imports continue at the traditional pace, Germany’s tanks will only be 67% full on November 1, with a one-in-two chance of landing between 60% and 72%, and a two-thirds chance of being below 70%.

Its tanks currently hold 71 TWh less than a normal late August, which is 39% of the entire EU shortfall on 22% of the EU's storage capacity. Germany is where the missing gas is.

And this late in the summer there is not much anyone can do about the shortfall; it is close to locked in already, because the gap between where storage is and where it should be barely moves between late August and the start of winter, according to IntelliNews Lambda’s analysis of the historical data. As IntelliNews reported, the EU has been importing every single drop of LNG that Russia is producing for months and closed the gap in many countries, but it has not been enough and the EU will start winter with a deficit.

That makes the severity of the winter a key factor going forward. If it is mild, Europe will muddle through. If it is cold, that will be a problem.

IntelliNews Lambda forecasts of storage volumes and price set against the weather. Find an intereactive version of these charts here.

The current gas storage is missing a January

EU storage stood at 62.32% full on August 21, holding 704.3 TWh. The fifteen-year seasonal curve for that date is a lot higher - 78.29%.

The shortfall is 15.97 percentage points, or about 180 TWh, which is roughly what the EU withdraws in on average for the whole month of January, the coldest month in the calendar.

And this summer’s extreme weather that produced the hottest summer on record means the temperature curve is not a simple average of past years.

IntelliNews Lambda uses a four-harmonic Fourier fit to day-of-year averages for 2011 to 2025 (with the crisis years of 2022 and 2023 stripped out) so the shape of a normal year is not distorted by the two years when there was no normal.

On that measure Europe came out of last winter at 27.62% full on March 31, the sixth-lowest trough of the sixteen winters on record, and it has gotten worse since then: the gap was 9.4 points in April; it is 16 points now.

Despite the unseemly scramble to buy Russian gas, despite the rhetoric of ending imports completely, injections are still running behind. Over the 30 days to August 21 the EU added 7.68 points of fill against a seasonal norm of 9.17 - 87 TWh where a normal August delivers 104 TWh. Europe was not closing the gap in the weeks when closing it should have been easiest.

Germany is carrying 40% of the hole

German storage was 50.43% full on August 21, holding 124.3 TWh against a seasonal norm of 79.26%. The gap is a gapping 28.83 points, nearly twice the EU's average, and it has been the defining feature of this restocking season since the spring, when German tanks bottomed out at 20.46% on February 25 - the lowest reading in the German series since records began in 2011.

Germany holds 247 TWh of working capacity, a fifth (22%) of the EU total. It is short 71 TWh. The EU as a whole is short 180 TWh. So two-fifths of Europe's missing gas is missing from German caverns that have not been refilled. That is why the country slipped into a gas crisis months before the rest of the bloc felt it, why the pain from this restocking season has fallen so unevenly, and why gas prices in Germany are twice that of the rest of the Continent.

At the current pace of re-stocking, Germany is on track to have its tanks 67.3% full on November 1, or 166 TWh. The 80% confidence band runs from 55% to 75%. Even the optimistic end of that range would leave Germany 15 points below its own seasonal norm on the day the heating season starts.

 

Weather does not hit the price, it empties the tank

That short fall is going to pump up gas prices in Europe. According to IntelliNews Lambda calculations, every percentage point of storage below the seasonal baseline is worth about 1.6% on the gas price. The current 16-point hole is worth roughly an increase in prices of 27% against an otherwise identical month with normal tanks.

Regressing the log gas price on the storage gap, Brent and coal across 163 months from 2011 to 2026 and the fit explains 77% of the variation in prices.

The relationship survives the tests that kill most macro correlations. It holds in sub-periods - minus 0.67 in 2016-20, minus 0.52 in 2024-26. It is held in the heating season alone. It holds in first differences, at minus 0.27, weaker but real, where the storage-to-industry correlations in the same data collapse to nothing.

It also peaks at exactly zero lag. Storage does not lead the price and the price does not lead storage; they move together within the month, because both are readings of the same balance. Anyone waiting for the storage number to tell them where the price is going has the causality backwards.

Cold weather has almost no direct relationship with the gas price, according to IntelliNews Lambda’s analysis. Across the 163 months in the data set , the correlation between a cold month and the seasonally adjusted price never gets above 0.19 at any lag, and once the storage gap is in the regression the weather term is statistically indistinguishable from zero, at a t-statistic of −0.6.

There is no interaction term linking price and storage. The markets appear to be remarkably efficient and often anticipate problems. The statistical variance on prices during a double whammy of a cold snap while tank volumes are low is also close to zero the data shows across all months, and effectively zero in the heating season alone.

What the cold does do is empty the tanks fast. Across 67 heating-season months, each month that runs one standard deviation colder than normal deepens the storage gap by 1.71 percentage points (with a t-statistic of −7.4) the strongest relationship anywhere in this data.

Outside the heating season the same coefficient is 0.41, four times weaker, which is what you would expect if heating demand is doing the work. Low storage does not make each cold day hurt more. It means every cold day will lower the storage base even further.

So, the chain runs cold weather, then a deeper hole, then a higher price of 1.6% for each percentage point the storage falls. The double whammy is real but sequential rather than multiplicative, and Europe starts this winter with less room underneath it than at any point in the fifteen-year record outside the war in Ukraine.

Looking at the actual numbers compared to the 15-year baseline and the rule-of-thumb is one standard deviation - a sigma, which is a measure of the distance from the mean – is not a problem as two thirds of a normal distribution fall within this range. But any result over two standard deviations from the mean are a problem, as they are outside 95% of the normal results. Above four sigmas, and you are in a full blown, once in a century crisis.

In volume, a winter running one standard deviation colder than normal draws an extra 97 TWh out of EU storage across November to March, and about 21 TWh out of Germany's. Two standard deviations colder doubles it to 193 TWh.

This summer was a crisis, and it is not over yet

This summer’s heat that stopped Europe refilling was extraordinary. Measured against the same fifteen-year baseline, June ran 2.39 standard deviations above normal, July 2.23 and August 3.3. In the week to August 4 the ten-city, demand-weighted average ran 4.75 sigma above the norm for the date; in mid-July it touched 5.91 – off the scale stuff.

 

On that scale this summer spent weeks well beyond crisis territory, and the June-to-August anomaly of 3.14°C is double the previous record in the series, 1.55°C.

Temperatures have come down since. The seven days to August 21 averaged 24.1°C against a seasonal norm of 21.6°C - an anomaly of 2.5°C, or 2.23 sigma. That is well off the peak, but it is still outside the two-sigma band of normality. Europe is cooling towards normal without having got there yet.

The winter that has just passed offers no warning at all. Every month from November 2025 to March 2026 sat inside one standard deviation off its norm, and the November 1 seasonal norm is 10.3°C. There is no useful signal running from a hot summer into the following winter: the correlation between the June-August anomaly and the following November is 0.10, and 0.26 for November to January. Summer tells you nothing about the winter that follows.

The El Nino question

Will this year’s super El Niño make a difference? It is already running hotter than any El Niño on record on the back of record high ocean temperatures in the Pacific, and will continue throughout the European winter. Does that extra heat mean Europe can look forward to a milder than normal winter? Actually, the past suggests a hotter El Niño usually means a colder winter in Europe.

The US Climate Prediction Center issued an El Nino Advisory on August 13, with the Nino-3.4 sea surface temperature anomaly at 1.4°C in July and a greater than 90% chance of a “very strong event” through the northern hemisphere autumn and winter. There is a 69% chance the anomaly exceeds 2.5°C between October and December, which would put it among the largest on record.

The common assumption is that this is good news for European gas, on the grounds that El Niño means a mild winter. But that is not how it works. The European Centre for Medium-Range Weather Forecasts finds "cold conditions over Northern Europe during El Niño", with the signal "strongest in late winter". NOAA's own analysis describes the mechanism: El Niño shifts atmospheric wave patterns so they "strengthen into the stratosphere, causing a weaker stratospheric polar vortex in late winter and more frequent major disruptions", and cold outbreaks over Europe happen when the vortex weakens, not when it holds. Last winter suffered from a big freeze after the polar vortex collapsed, bringing record snowfalls to places like Russia and Arctic conditions in the US. The vortex is a circulating current of air over the Arctic that traps its cold air on top of the planet. When the vortex breaks down, that cold air mass spills out over northern America and Europe, bringing freezing temperatures. If it happens again this year, Europe is on course for another freezing winter.

But another big freeze is not a given. Two caveats cut the other way. The ECMWF work finds the relationship is non-linear by event size - "only the moderate to weak events show the canonical negative NAO pattern while the strongest events show a rather different pattern" – so the very strong El Niño will not necessarily have the same impact on the polar vortex. This is a case where the textbook signal is least reliable.

In the fifteen European winters since 2011, the five El Niño winters averaged 0.84°C warmer than normal against 0.01°C colder for La Niña winters, a correlation of 0.52 that strengthens rather than weakens when both series are detrended. The two strong El Niño winters in that window, 2015/16 and 2023/24, were both mild.

Fifteen winters is not enough to overturn the modelling, and the modelling is not confident enough to overturn fifteen winters. The defensible position is that El Niño gives Europe no reliable steer either way, which means the storage number is the thing to plan around, because it is the part that is already known.

Three winter scenarios

Given the uncertainties there are three likely scenarios for getting through this winter. The insufficient November 1 tank levels are already a given. The rest of the season is like to play on in one of three ways. The most important variable will be the weather.

 

And that will make a big difference to the price of gas. Averaging over every winter the distribution allows, rather than picking one, January TTF has a median of €67/MWh, a 50% band of €57 to €78 and an 80% band of €49 to €97.

That carries a 51% chance of January’s prices averaging above today's level, a 23% chance of clearing €80 and a 9% chance of exceeding €100. Set against a fifteen-year January average of €26.10/MWh, the central case is 155% above normal and even the bottom of the 80% band is nearly double it.

The heating season ends when the tanks stop emptying, and that date is more consistent than the history of how much gas is left then. EU storage has bottomed on April 1 on average across the last thirteen non-crisis years, with a standard deviation of eight days.

This winter, IntelliNews Lambda’s Fourier baseline suggests the winter bottoms out on March 30 at 37.6% full. Germany bottoms slightly earlier, on March 29, but with a much wider scatter of +/- 14 days.

Running the same arithmetic forward to that date the average EU storage will fall to a low of 24.4% full at the end of the heating season (about 276 TWh) against a seasonal norm of 37.7% (426 TWh). Germany's trough comes out at 23.9% (59 TWh) – half of its norm of 42.3%.

The bottom line is the uncertainty this year is huge. The bands are enormous. The EU's 80% band runs from 5% to 42.5% full; Germany's runs from empty to 55%. Where November 1 is already close to being determined, the end of the season is almost entirely unknown and will depend almost entirely on the weather.

The distribution puts a 35% probability on EU storage finishing below its 17.81% record low, set on March 30, 2018, and a 38% probability on Germany finishing below its own 14.13% record.

Those tails should be read as a warning about the model, not only about the gas. Storage cannot physically go to zero – technical gas minimums demand that storage cannot fall below 10% and withdrawal-rate limits bite long before that - so what the low end of the distribution is really describing is a winter in which the price rises far enough, far enough in advance, to destroy the demand that would otherwise have emptied the tanks. The rationing happens in the price, not in the caverns.

TTF in April, on the same distribution, has a median of €58/MWh and an 80% band of €34 to €102 - the widest of any month, because by then the whole winter has either happened or it has not.

What the market already knows

One number should temper all of this. The storage-and-seasonality model takes July's average price forward to about $18.94/mmbtu by late August. TTF actually settled at €66.00/MWh on August 21, equivalent to $22.59. The market is running 19% above what storage and the calendar together explain so the higher price limiter that tempers gas demand and will cushion a shortfall crisis has already kicked in.

Across thirteen complete gas years, the storage gap on November 1 has no statistically significant power to predict that winter's price relative to trend (t-statistic of −0.25). Winter cold does, at about 11% per standard deviation. By the time the heating season opens, the tank level is public information and it is in the forward curve. What moves the price after that is the weather surprise.

That has a direct consequence for how these forecasts should be used. They say where the price lands if a given winter arrives. They do not say the market has failed to notice the hole, because the evidence is that it has already priced a good deal of it. The 19% premium is the market's own view of winter risk, arrived at without any of this arithmetic.

Europe's position going into this winter is worse than at any point since the invasion year, its largest storage market is carrying two-fifths of the shortfall, and the price has already moved to reflect it. What is left to be decided is the weather, and on that the honest answer is that nobody knows, including the people who have spent the summer explaining why El Niño means it will be mild.