UK September Weather Forecast 2026: Heatwave Risk, Rain Outlook and Temperatures

The Met Office’s forecast for September 2026 runs out on the 16th — and it has just taken a little heat out of it.

When we read its long-range page, it carried exactly two blocks: one for 24 August to 2 September, and one for 2 to 16 September. For 17 to 30 September there was nothing — not a hedge, not a probability, nothing at all.

The summary is undramatic and has cooled twice over: an unsettled first half with rain heaviest in the north and west, drier spells developing after mid-month, temperatures near average rather than above it, and warm spells possible in the south — but no mention of a heatwave anywhere on the page.

What does the Met Office actually say about September?

Here are both blocks, verbatim, as we retrieved them on 2 September 2026. First, “Sunday 6 Sep - Tuesday 15 Sep”:

“An unsettled period is likely, with many places seeing showers or longer spells of rain at times. Rain will likely be most frequent across northern and western areas, while the south and east are more likely to see some drier interludes. There is also the potential for some strong and gusty winds, particularly in the northwest. Temperatures are likely to be near average, although brief warm or very warm spells are possible, especially in the south.”

And “Wednesday 16 Sep - Wednesday 30 Sep”:

“Confidence is very low from mid-September onwards. A changeable pattern, with the wettest conditions in the northwest, will likely transition towards more in the way of settled or dry spells between weather systems. The distribution of rainfall is uncertain at this range, but drier conditions than earlier in the month are likely for most. Temperatures are likely to be around average overall, although some warm spells are possible, particularly in the south.”

Read the hedges, because they are the content. “It looks like”. “Signals … soon become weak”. “Becoming most likely”. “Will probably”. “A chance that … hot spells or heatwaves develop”. This is a forecast written by people who are telling you how much they do not know.

Two things have dropped out since mid-August, and they are the ones worth noticing. The temperature sentence has gone from “above average overall” (18 August) to “a little above average overall” (19 August) to “near average” now. And the heatwave sentence — “a chance that late season hot spells or heatwaves develop” — has gone from the page entirely.

Why did it stop on 16 September — and why doesn’t it now?

Because the window moves with you. The long-range forecast looks out about 30 days from the day you read it. On 19 August, 30 days out was the middle of September; on 2 September it reaches 30 September, which is why the whole month is now inside the window. Nothing was added to the forecast — the calendar caught up with it.

Chart must be regenerated: the September timeline should now show 2–5 September under the standard forecast, 6–15 September and 16–30 September as the two long-range blocks, and the 16–30 band flagged “confidence very low” rather than “no forecast published”. Alt text to be rewritten to match.

It is not rewritten once a day, either. We read the page twice on 19 August, thirty minutes apart. The September block was identical both times. The first block was not: at 11:52 BST it was headed “Sunday 23 Aug - Tuesday 1 Sep” and described a period starting “largely dry and settled”; at 13:22 BST it was headed “Monday 24 Aug - Wednesday 2 Sep” with different prose. If you are checking this page against a news story, check the date range in the heading first.

The Met Office is candid about why it will not go further, and about why even the part it does publish is written for the whole country rather than your town:

“when looking at forecasts beyond five days into the future the chaotic nature of the atmosphere starts to come into play … whilst we can still forecast the general feel of the weather to a relatively high level of accuracy using our ensemble models, it becomes harder to offer local detail”

So what about the second half of September?

It is now covered — and the Met Office’s first words about it are “confidence is very low”.

What the Met Office publishes for that fortnight is one paragraph, and it hedges in every clause: a changeable pattern “will likely transition” towards drier spells, the rainfall distribution is “uncertain at this range”, temperatures “likely to be around average overall”.

So there is now a forecast for late September — but it is a direction of travel for the UK as a whole, with the Met Office’s own confidence rating attached. Any article giving a temperature, a date or an “exact day” for late September still has nothing published behind it. That is either a single model run — which changes every six hours and is not a forecast — or it is invention.

Does the three-month outlook say more?

It says something, but it is not a September forecast, and it is easy to misread in two separate ways.

The Met Office publishes a 3-month Outlook for contingency planners. The August–October issue this article originally covered has been replaced: the current one covers September, October and November 2026 and carries an issue date of 01.09.2026 and a printed next issue date of 28.09.2026. It is a three-month average, so September is still not broken out on its own — the outlook’s numbers describe the autumn, not the month.

Chart must be regenerated with the September–November 2026 figures: temperature 5 / 55 / 40 (cool / near average / warm), precipitation 15 / 55 / 30 (dry / near average / wet), wind 15 / 70 / 15 (calm / near average / windy), against the 20/60/20 normal baseline. Alt text to be rewritten to match.

Normal is 20% for each outer band and 60% for near average, so:

The first trap: these are not thirds. The Met Office’s own normal likelihoods are 20 / 60 / 20, not 33 / 33 / 33. If you assume thirds, 40% looks like a mild 1.2×. Against the real baseline it is the 2.0× the Met Office prints — and the outlook says so in words: the next three months are “around twice as likely to be warm compared to the 1991-2020 reference period”.

The second trap: the signal is back-loaded now, not front-loaded. The new outlook keeps an “increased chance of hot spells during the early autumn”, but its summary says “the latter part of the period is the most likely to be wet” and that “the chances of impacts from rain and wind become higher than normal later in the period” — that is late October and November, not September. It also names a driver the August issue did not lean on: “a very strong El Niño event, that is likely to be historic within the modern recorded era”. Read as a September document, it says less than it appears to.

“Close to the climatological norm” — what is normal?

A normal UK September is 12.93°C, with 90.9mm of rain and 127.5 hours of sunshine, on the 1991–2020 baseline.

The forecast’s other structural claim is a “broad northwest-southeast split”. That phrase sounds like a forecast of something unusual. It is not — it is a description of what September in the UK is like every year.

Scotland’s average September is 1.8 times wetter than England’s and gets about a quarter less sunshine. So when the Met Office says both “close to the climatological norm” and “a northwest-southeast split”, it is not saying two things. It is saying the same thing twice.

“Above average” temperatures — above which average?

This is where a real number can mislead without anyone lying. The Met Office now says “a little above average overall”, and does not say which average. The 1991–2020 UK September mean is 12.93°C. The 1961–1990 mean is 12.20°C. The most recent thirty years, 1996–2025, average 13.22°C.

A September that comes in at 13.0°C is “above average” against two of those three and below the third. When you read “above average”, the useful question is not how much — it is against what.

For context on the range: the warmest UK Septembers on record are 2006 and 2023, tied at 15.2°C, and the coldest is 1952 at 9.9°C. Five of the ten warmest have occurred since 2000.

What does it say about rain?

More than it did, and it has moved wetter. The first half of the month now leads with rain rather than with a split:

“An unsettled period is likely, with many places seeing showers or longer spells of rain at times. Rain will likely be most frequent across northern and western areas, while the south and east are more likely to see some drier interludes.”

Watch how far that has travelled. On 16 August the block said only that “drier, more settled weather is slightly more likely towards the south and east”; by 18 August it named “the lion’s share of the rain and any strong winds” for the north and northwest; it now opens by calling the period unsettled outright. The geography has been stable throughout — north and west wet, south and east drier. What has moved is how confidently the Met Office says it will rain.

Two things to take from it, and one not to:

• It is a split, not a total. The forecast says where rain is more likely relative to elsewhere. It does not say how much, on which days, or in which towns — and by design it never will at this range.
• The autumn as a whole now leans wet — but the wet signal sits in late autumn, not September. The September–November outlook now leans wet (30% against a normal 20%), and the Met Office addresses the question directly: “it is too early to say whether rainfall this autumn will provide meaningful drought relief in the areas that have experienced the driest conditions”.
• It is not a rainfall figure. A normal September delivers 90.9mm across the UK on the 1991–2020 baseline, but that is climatology, not this forecast. Nothing published says September 2026 will be wetter or drier than that.

⚠️ The wording has now moved several times: the rain sentence appeared on 18 August, the temperature sentence was softened on 19 August, and by 2 September both blocks had been replaced outright — the heatwave sentence is gone and temperatures read “near average”. If you are reading this well after 2 September, expect it to have moved again. The page carries no issue date and is redrafted more than once a day.

Does any of this mean another heatwave?

No — and the Met Office has now dropped the word from the page. The heatwave sentence that stood in mid-August is gone; the strongest language left is “brief warm or very warm spells are possible, especially in the south”. Its three-month outlook is explicit that it cannot do better: it “is unable to predict specific heatwaves or prolonged periods of hot weather and cooler periods are also likely”.

That is worth holding onto given the summer behind it. As we covered when the August heatwave broke , 2026 had already produced six heatwaves and records at several thresholds before September began. A warm September would not be surprising. But “would not be surprising” is not a forecast, and the Met Office has not issued one.

What about the drought?

The forecast will not answer that question, and this is the part that actually affects people.

By mid-August, all of Wales and around half of England were in official drought status, with hosepipe restrictions covering millions of households. Those restrictions come off on the state of reservoirs and rivers, not on a weather outlook. The Met Office now expects an unsettled first half of September and leans wet for the autumn as a whole — but it is blunt about what that is worth: “even a wet month may only partially alleviate water pressures in parts of the UK, following the prolonged dry weather through the summer”.

If you want to know whether a ban applies to you, your water company’s own page is the authority. We keep a guide to the 2026 hosepipe bans and the areas they cover .

What to check, and when

• For the next week, use the location forecast, not the long-range page. The long-range text is written for the UK as a whole by design and will never mention your town.
• Read the date headings before the words. Both the wording and the period boundaries move, and they move more than once a day — we watched the first block change period and prose inside half an hour.
• Watch for the September–November issue in early September. It is the first product with a September-specific section. The Met Office states only that the Outlook is published monthly and dated the current one 2 August; it does not announce the next date, so this is a cadence inference, not a diary entry.
• Treat anything about 17–30 September as noise for now. No published forecast covers it.
• When you see “above average”, check the baseline. Against 1961–1990 it means something different from against 1991–2020.
• Warnings only run seven days out. There is no such thing as a September weather warning today.

Sources

Checked 2 September 2026. No affiliate links, and no payment was received for any link on this page.