Doughnut Summer 2023: Still On!

Posted on by

The September 2023 heatwave in England was, for the time of year, unprecedented in duration (7 successive days when 30C was exceeded somewhere), and apparently for the mean temperature on at least one day, though not for the maximum temperature (which was higher in 2016). Overall, the mean temperature for September 2023 is on track to exceed that in August, which would make June and September the two hottest months of the year for the first time in the entire Central England Temperature record (back to 1659)! Even if the month ends up being not quite as hot as August, it is still likely to be be one of the warmest Septembers in the CET record.

Here’s something you don’t see very often:

Fig 1 CET data summary for 2023, year to date, from Met Office

In my previous post on 5th September, I wrote: “I’d guess that the mean CET for the first third of September is likely to be 5 or 6C, maybe more, above the 1961-90 mean for the month of 13.5C” and that is exactly what has come to pass. Yes, the mean CET for the first 10 days of September have been 6.8C higher than the monthly mean (over the period 1961-90)! Astonishing.

Various records have been broken, but the one I’m most interested in is whether we’re experiencing the first true “Doughnut Summer” (in the CET) since at least 1659. I’m defining a Doughnut Summer as one where June and September are the two hottest months – usually they’re July and August. For that to happen the September mean temperature would need to end up as higher than August’s 16.4C, or around 3C higher than the baseline.

With the first third of September 6.8C higher, the rest of September only needs to be 1C warmer than the September mean. That’s not as easy as it sounds because temperatures (usually) cool during September so the final two-thirds of September needs to be more than 1C warmer than the mean for the final two-thirds of 1961-90 Septembers. On the other hand, due to you-know-what, Septembers are typically somewhat warmer now than they were from 1961-90:

Fig 2 September mean CET temperatures, graph by the author, from Met Office data

Specifically, looking at the medium-range forecasts, I think it’s actually quite likely that the rest of September will be warmer than average:

Fig 3 Latest 14-day forecast for London, from Meteoblue

Note that it’s still a bit sticky as I write on Tuesday 12th September and was even warmer yesterday. Next weekend, 16th-17th, is also expected to be warm, and there’s no sign of colder weather. The picture from TheWeatherOutlook is broadly similar:

Fig 4 Latest 14 day forecast for London from The WeatherOutlook

What of the heatwave that’s made all this record-breaking possible?

Well, as has been widely reported, we saw 30C broken somewhere in the UK (south or south-east, in practice) for 7 consecutive days, when the previous record was 3! Very unprecedented.

I suggested last time that “[i]t’s not expected to be incredibly hot”, and it’s fair to say that daily records were only broken on the last two of those 7 days – 33.2C at Kew Gardens on Saturday 9th and 32.5C in Cambridge on Sunday 10th (beating records set in 1898 and 1891 respectively!).

On the other hand, that 33.2C (as well as the 32.5C, incidentally) would have been a date record – the hottest temperature recorded so late in the year – had 34.4C not been recorded at Gravesend on 13th September 2016.

And here’s a little detail I noticed. See the graph at the top right of Fig 1? Here’s a bigger version:

Fig 5 From CET data summary for 2023, year to date, from Met Office

That blue line seems to me to indicate a mean CET temperature date record and quite a significant one at that. I haven’t downloaded daily CET data (just monthly – as shown, for example, in Fig 2), but from that graph, the mean CET temperature for (I think) Thursday 7th September 2023 seems to have been higher than for any date after mid-August in any other year in the dataset. Wow!

Finally, I was reading recently about climate attribution studies. Now, I’ve long had a degree of scepticism about quantifying the increased risk due to global warming of extreme climate events, for the simple reason that the risk in the models is simply not the same as in the real world, an assumption that really should be made a bit more explicit.

But sometimes the increased risk can simply be assessed empirically. Referring to Fig 2, I note that the mean CET for September was 16C or more 4 times from 1659 through 2005. It’s really quite likely that mark will be met or exceeded this month, for the 4th time in 18 years. So, very roughly, the chance of what we might call historically a “very hot” September has increased by a factor of around 20 (and rising, since the last was in 2021 and the one before that in 2016).

Of course, we can’t say what the probability before global warming was of 7 consecutive days with a temperature of above 30C somewhere in the UK, because it previously didn’t occur at all, at least back as far as the daily data goes. My favourite book, The Wrong Kind of Snow, quotes this as 1875, so maybe we can suggest that we’ve just experienced a heatwave, the like of which – before global warming – occurred less than once in 150 years.