40°C in the UK?

Currently, the highest officially recorded temperature in the UK is 38.7°C, recorded at Cambridge Botanic Garden on 25 July 2019. Prior to that, the record was 38.5°C (10 August 2003 in Faversham, Kent), which beat the record of 37.1°C set on 3 August 1990 in Cheltenham, Gloucestershire. Before the 1990 heatwave, the record had stood for 79 years: 36.7°C on 9 August 1911. It is also worth mentioning that 31 July 2020 saw the temperature reach 37.8°C at London Heathrow airport, which became the UK’s third hottest day behind 25 July 2019 and 10 August 2003 (and thus would have itself been a record prior to 2003).

In our rapidly warming world, it is unsurprising that we have observed more extreme heat in recent years, and the 36.7°C record from 1911 seems almost cold in comparison to what we have seen so far in the 21st century.

After the 2003 heatwave saw the UK eclipse 100°F for the first time, attention turned to 40°C. It’s a seemingly unthinkable temperature for a country with an aging population which does not have widespread residential air conditioning. Only in parts of the London Underground is 40°C widely experienced in the UK, as John Hammond demonstrated during the July 2018 heatwave.

But with anthropogenic global warming, 40°C is becoming more likely. Eunice Lo and Dann Mitchell provide a good overview of how climate change is affecting UK heatwaves in this open-access Weather briefing paper which formed part of the October 2021 COP26 Special Issue. More specifically, Christidis et al. (2020) from the Met Office Hadley Centre wrote in Nature Communications:

“The probability of recording 40 °C, or above, in the UK is now rapidly accelerating and begins to rise clearly above the range of the natural climate. The return time for the 40 °C threshold is reduced from 100–1000s of years in the natural climate to 100–300 years in the present climate and to only about 15 years by 2100 under the medium-emissions scenario (RCP 4.5) and 3.5 years under the high-emissions scenario (RCP 8.5).”

Christidis, N., McCarthy, M. & Stott, P.A. The increasing likelihood of temperatures above 30 to 40 °C in the United Kingdom. Nat Commun 11, 3093 (2020).

40°C in the UK still seems hard to comprehend, even with robust science to support its increased likelihood due to greenhouse gas emissions. After all, the UK has thus far only managed to get past 100°F on three individual days, and typically at very isolated locations. But then came June 2021, when the Pacific Northwest heatwave obliterated local temperature records in Canada by as much as 4.6°C to reach a simply mind-bending record of 49.6°C in British Columbia. To eclipse a prior record by such a margin shocked meteorologists and climate scientists alike, and showed what was possible in today’s world when all the favourable factors come together. And it makes 40°C in the UK seem all the more plausible.

Everything changed with the 12Z initialisation of the NCEP GEFS model on 30 June 2022 (Figure 1). As far as anyone I have spoken to can recall, this was the first time 40°C appeared in a major medium-range global forecast system for the UK. It wasn’t even just the mere appearance of 40°C that was remarkable — this forecast showed a huge area of southeast England eclipsing 39°C. Given that the UK’s previous hottest days had only seen 38°C exceeded very locally, this was unlike anything anyone had ever seen before. Furthermore, site maxima are usually higher than what global models predict due to their relatively coarse resolution — so seeing a global model explicitly produce conditions that pulverised the existing records was not something I think anyone expected to see.

But that was just a single member of one ensemble forecast system. Ensembles are designed to capture a range of possible outcomes, and it’s common to see the odd member do something wild — especially beyond 10 days. At long lead-times, errors in the model can, given the right circumstances, occasionally lead to rapid error growth and produce unphysical outcomes. A well-known example of this is the runaway deepening of tropical cyclones in the 3km NAM model. There can also be situations where the model has a lead-time dependent bias, and thus drifts to a similar outcome at a fixed lead-time rather than a fixed valid date (the ‘chasing the rainbow’ scenario).

However, in the limit of no unphysical model error or drift, the individual ensemble members can be considered as alternate realisations of the real world. Parallel universes, so-to-speak. We need these because weather consists of a predictable and an unpredictable component. (As an aside, I’m not sure how well this is appreciated outside of those with an interest in weather or physical science knowledge — there are fundamental limits on day-to-day ‘weather’ predictability.)

Thus, the appearance of 40°C in that single ensemble member suggested it was a possible evolution over the next 2 weeks — but extremely unlikely.

I expected that this would be the only time we would see such an extreme in the forecast, and it would join the hall of fame of crazy GEFS forecasts. But it wasn’t. As subsequent initialisations of the GEFS came in, more ensemble members continued to show 40°C — or even more broadly, more members started to show >35°C, which in a global model at that sort of lead-time would be suggestive of the potential for a new UK maximum temperature record. Then came the forecast for 16 July from the 18Z initialisation of the deterministic GFS model on 2 July:

Historic, unbelievable chart.

GFS 18Z with widespread temperatures of at least 41°C in England on 16 July. The current UK record is 38.7°C.

It is possible that the model has runaway warming in its surface layer. It is also possible that an unprecedented heatwave may develop. pic.twitter.com/REfK6z1qEO

— Dr Simon Lee (@SimonLeeWx) July 3, 2022

As I noted in the tweet, it was possible the model was excessively drying its surface layer — soil moisture analyses suggested this might be the case — leading to over-zealous surface temperature rises. But note that the valid time was similar to that first GEFS member (Fig. 1). We were no longer talking about something at a fixed lead-time, but a relatively fixed valid date. And the lead-time was shrinking. Furthermore, it’s not immediately clear to me what would constitute unphysical surface drying in southern England over the next 2 weeks, given forecasts show little precipitation and the region has experienced the driest opening 6 months of a year since 1976.

It’s now a full six days since that first extreme GEFS member, and the GFS/GEFS continues to throw intermittent support to record-breaking heat for the UK somewhere around 15-17 July. For example, look at today’s 12Z GFS for 16 July (Figure 2), the same date as that single GEFS member from 6 days ago. Not just 38°C, but 40°C once again:

We are now also within the range of medium-range forecasts from ECMWF, which are also suggesting the potential for extreme heat at the same time as the GFS/GEFS. Figure 3 shows a 15-day ensemble forecast for London alongside the model climatology (based on reforecasts of the past 20 years), to give an idea of how unusual this forecast is. On 17 July, more than 25% of the forecast ensemble members exceed the 99th percentile of the model climatology, and more than 10% of the ensemble members exceed 35°C. That’s a big shift in the distribution versus what might normally be expected.

However, all this said, truly extreme heat in the UK is not yet the most likely outcome, even if the risk is likely the highest it has ever been. Taking the ECMWF ensemble forecast for 17 July as an example, the median sits just below 30°C. One forecast member is down in the low 20s — near the model average. A large part of this uncertainty remains because a specific evolution of the large-scale weather across western Europe is required in order to pump the warmest air from Africa, through Spain and France, and then into the UK. This involves the development of a low pressure system west of Iberia and its subsequent northward drift during next week, with warm air advection on its eastern flank. This effectively has to happen perfectly for the UK to attain 40°C.

What is remarkable and concerning is that it is increasingly apparent that should that large-scale evolution occur, then it is increasingly likely that the UK will see unprecedented heat. Put another way, there seems to be relatively high confidence in the temperature forecast for that given synoptic scenario — but the scenario uncertainty dominates.

Extreme events are inherently unlikely, and are thus unlikely to dominate predictions at lead-times beyond 10 days or so. We see this with split-type sudden stratospheric warmings, for example, which are generally not predicted more than 2 weeks ahead of time. You wouldn’t get anywhere trusting a single ensemble member, and so forecasters have to emphasise the ensemble mean — even though the ensemble mean isn’t actually a specific realisation of the real world, merely an attempt at diagnosing the ‘predictable component’.

We have to now sit and watch what happens. Regardless of whether 40°C or even a new UK maximum temperature record is set this month, we have entered uncharted territory in forecast-world. Eventually, this will translate to the real world, and with the global climate continuing to heat up, we can expect to see more situations where 40°C appears in forecasts for the UK in the coming years and decades.