Months with maximum temperature records set in the 21st century

[Metwatch]

Saw this post the other week by Aidan Mcgivern, one of the Met Office meteorologists, and we have 7 out of 12 months where their maximum records have all been recorded this century, with all 3 winter months included in the last 5 years. All the spring records are from last century which are still yet to be broken this century along with June and September.

However, I researched what the warmest temperatures of those remaining 5 months were but for this century alone and this is what I found:

June: 2017, 34.5°C on the 21st, in Heathrow, London

March: 2021, 24.5°C on the 30th, in Kew Gardens, London.

April: 2018, 29.1°C on the 19th, in St James Park, London.

May: 2005, 31.4°C on the 27th, in Gravesend / Herne Bay, Kent.

September: 2016, 34.4°C on the 13th, in Gravesend Kent.

All of those months above are all with 1.5°C of their all time monthly records, and with april, 0.3°C below!

So it begs the question when will the remaining 5 months see their records broken this century, perhaps by 2030?

(P.s this would be of most interest to @CryoraptorA303  @WYorksWeather and @Earthshine I think!)

 

[BruenSryan]

Also worth mentioning that the July record was beaten 3 times in succession this century before 2022.

2006 36.5C

2015 36.7C

2019 38.7C

And if 2019 didn't happen, 31st July 2020 would have also been a record with 37.8C. 

[CryoraptorA303]

 Metwatch Thanks for tagging me with this.

Some things to note:

I have previously derided the September 35.6°C record to be quite intuitively inaccurate. However, Harwarden Bridge reached 32.3°C on September 1st 1906, and this is a well-known hotspot. Harwarden also reached 37.1°C on July 18th 2022, and the next day saw the infamous 40.3°C at Coningsby. If we apply the same disparity between Harwaden and the final record in 1906, we get a value of 35.5°C, which is very close to the quoted value. There is still the argument to be made that being in September and with the Sun being lower, the actual gradient of disparity between Harwarden and the final record would be lower, however I am now quite satisfied to accept that September 2nd 1906 was extremely hot and did indeed see temperatures well in excess of 34°C and probably in excess of 35°C. Such a September heatwave repeating today would likely see temperatures in the region of 37-38°C at Coningsby, Gringley-on-the-Hill or Normanby Hall, or perhaps somewhere a bit further south like Cambridge if the wind wasn't as strong. This heatwave repeating after a particularly dry and/or hot period with extremely high uppers could even see us challenging 39°C in September, which would be one hell of a sight to see. I'm not sure if 40°C or an outright all-time record break would be possible with the Sun this low in the scenario of an extremely dry and hot period, however I would be inclined to say it's not absolutely physically impossible if the air is the hottest we've ever seen over the UK, the kind of air that would see something in the region of 45°C in July or early August.

The April 1949 record is complete bunk and out of line with other London stations, which all recorded 27.8°C or less on that day. Marham also recorded 27.8°C on that day but keep in mind, 1. this is rounded Fahrenheit so we don't have precise values, and 2. the same urban heat island that London sees today didn't exist in the same form then. There is no explanation for the disparity. 29.1°C on April 19th 2018 is the reliable record for April. Just for comparison, the rest of London reached at least 28.4°C on that day, and the next highest was Northolt at 28.9°C, so this disparity really has no justification for existing in 1949.

The quoted June 1976 record is total bunk and completely out of line with local stations. The next highest temperature on that day was 34.5°C at Cheltenham and another station in Southampton literally up the road from Mayflower Park recorded 34.4°C on that day, a far cry from 35.6°C. Other stations on the south coast didn't exceed 34.0°C on that day. The previous day, June 27th, with 35.5°C at Mayflower Park, is much more reliable and should be the quoted June record. I don't count June 1957 in this as that 35.6°C record was set by Camden Square, the same station that recorded the bunk April record and has far too many consecutive annual maxes to be legitimate. This was also recorded in Fahrenheit and so is a less precise measurement than the June 1976 35.5°C anyway.

The March and May 1944 (not 1922 as that was set by the same Camden Square and has an unreasonable disparity with the rest of London) records are reliable enough with either multiple accounts of extremely high temperatures or many modern, well-known stations reaching very high temperatures along with the record setter.

If I'm being honest, I would not count pre-Celsius records to begin with unless the quoted figures are exceptionally backed up as is the case with May 1944 and September 1906. The celsius transition roughly coincides with when data can be assumed to be reliable without needing to be proven anyway. I suspect a major reason behind not scrubbing the bunk records would be deniers accusing the Met Office of hiding the existence of previous records. There is already a certain cretin on the internet who counted Glaisher screen measurements made in the 1911 and 1932 heatwaves and is genuinely perplexed as to why these records aren't counted

If I had to guess why the remaining records haven't been broken yet, I would mostly say it's simply because we haven't had enough time to break them all, and because of luck. A bunch of the hottest days on record in this century as pointed out by @BruenSryan all happened to fall in July. Had all of these record-breaking heatwave synoptics fallen in the other months without a 21st century record and left July with the 2006 record then we would only have a single month without a new record, and that would probably be September simply owing to how extreme that heatwave was. As mentioned that was analogous to potentially reaching as high as 38°C or maybe even higher today.

Were I to put a date on when these records will fall, I would suggest the March, May and June records will imminently fall in the next few years, and the September record will be surpassed by the mid-2030s. Last year already seriously threatened the September record until the hottest air coincided with the easterly wind and by the time the necessary southerly and westerly winds arrived the hottest air had already left.

[WYorksWeather]

Cryoraptor has covered the details of timing above. Just worth diving a bit more into the statistics, to assess what they tell us.

The way to do this is through hypothesis testing. The null hypothesis would be that temperatures have not increased over the period we have reliable measurements (which is of course false, but bear with me). So, if that were true, we would expect that the probability of a new monthly record should be roughly 1/n, where n = the number of years since reliable records. If we take the period of reliable recording to be roughly since 1880, then we have roughly 1/120 or lower, or just under 1%, as the expected probability of a record in a given year. This can be seen to be trivially true by considering the case of a small number of years. The probability of the first year in any record being a new record under the null hypothesis is of course 100% (no previous record). It then follows that the second year has a 50% probability of breaking that record, the third year a 33% chance, etc.

So, what does this tell us in reality? It means that in the 21st century, we should expect, if there were no warming trend, that the probability of a new monthly record for a particular month is just under 1%. Therefore, the expected number of records since 2000 is around three. The true number of records is of course at least seven, but then we also have to add interim records, which are at least the three from July (2006, 2015, 2019), which makes ten by this definition (there may be one or two more I'm not aware of), and it also depends on whether you include the tie from January 2003. But let's go with ten for now unless anyone makes me aware of any more. That means that at an absolute conservative minimum, the rate of new monthly records is at least three to four times the expected rate.

In terms of timescale, I'd expect a few of these records to survive longer just by sheer chance. The September one has the best chance since it was so exceptional for the time. I'm going to say that I actually think it will take until just before 2050 for all of the records to go. By chance I expect one of them will survive, barely hanging on through a succession of freak events by chance. Especially since these are spring and autumn ones, they are more susceptible to timing issues. E.g. a notable heatwave in early March won't be a record-breaker, neither will one in late September.

If we assume that the record setting rate averages a 10% chance for each month per year (which would be unbelievably high, higher even than the last few years, and more than ten times background rate), then this holds true. Of the five remaining monthly records, each one would have a chance of 0.9 ^ 25, or 7.2%, of surviving for 25 years. The combined odds of none of the remaining records surviving under that scenario works out at 69%. If you do the same maths for 2040, then that probability decreases to 32%.

It is of course possible to construct higher probability scenarios, but a 10% probability of a new record for each month is an extremely high rate, so I find it hard to imagine we'll go above that.

Apologies for the maths! Comments welcome on this @CryoraptorA303 , @Metwatch or anyone else - happy to explain this or continue to discuss this subject.

[Earthshine]

 Metwatch Very interesting!  Going by highest temperatures on record for each country is also pretty fascinating:

England: July 2022

Wales: July 2022

Scotland: July 2022

Northern Ireland: July 2021

Record high minimums:

England: July 2022

Wales: July 2022

Scotland: July 2019

Northern Ireland: August 2001

[dryfie]

I note that the list given by McGiven suggests a downgrade in the Jan max recorded just a couple of weeks ago.  His list shows it as being 19.2 (provisional but a couple of weeks ago).  There were reports higher than that at at lest two sites, one automatic and the other manual (the 19.9 previously reported).  Have these now been ruled out?

[CryoraptorA303]

 WYorksWeather That's a great breakdown of it, bravo.

The reason I say I expect them to all be gone by the mid-2030s is because the odds of exceeding these temperatures in these months will exponentially increase year on year.

July has likely seen the most new records simply as a consequence of the same maths. If we say from the period June 20th-August 20th, the chances of seeing record breaking temperatures is roughly the same due to a relatively constant sunlight-airmass warmth relationship, that leaves July with all 31 days, August with 20 days and June with just 10. July saw five instances of temperatures above historical extremes, August saw one and June has so far seen none, which would appear to back up this timeframe.

In March every single day counts and on each day the chance of record heat is higher than the last. Not only does the airmass need to be warm enough and be blowing in the right direction for hotspots to get hot, the timing is really restricted to the last 10 days of March at best if you want 25.6°C to be exceeded, although the window for this temperature to be exceeded will increase year on year.

April is the fastest-warming spring month and it makes a lot of sense why. April already sees a return to August-strength insolation in the second half and so the solar potential is there for hot weather. The deciding factor in April then is airmass warmth, and of course the airmasses have only begun warming up in the last month or so. However, because of climate change, the troposphere is absorbing more heat in the heat-gathering six months of the year and is losing less of it in the heat-losing six months of the year, so year-on-year we will see higher and higher temperatures being retained into March and a more intense heat-gain once the heat-gathering period begins. This will translate to summer temperatures being reached much earlier, and the most profound effect will be in the middle spring month which is highly dependent on airmass temperature. We would also have to factor in polar air temperature at this point, and we know that this is also undergoing a positive feedback loop due to climate change as more and more ice is lost each year, decreasing polar albedo and ensuring more and more heat is gained in polar airmasses. This will already be having an effect in winter, but by April this will be an absolute game-changer and will severely limit how low temperatures can get. The faster spring warm-up will also encourage more polar heat transfer to occur, transferring more early heat polewards to high-midlatitude areas like the UK.

September is also time-dependent as the Sun is rapidly receding, however the September Sun is still quite strong, and it managed to get to 32.2°C on September 19th 1926, which likely means that temperatures in excess of 34°C are still possible this late in the year in the modern day with hot enough air. We also of course reached 34.4°C on September 13th 2016, and 34.6°C on September 8th 1911. Remember what I said about the troposphere gaining more heat in summer and losing less of it in winter year-on-year? September at this latitude is as warm as it is because of the residual heat left over from summer. The same effect of more heat being gained and less being lost will have profound consequences for September as hotter and hotter air will be lingering around for longer each year. This effect is even more profound for northern Africa and Iberia due to their insolation being higher anyway, which are the two places we get our extreme heat from when the air there decides to have a temper tantrum. I would suggest that in a worst case "Coningsby" scenario, the 35.6°C record can be well exceeded up to the solstice, and in the first 10 days of September a temperature in excess of 38°C would still be possible. Of course by the equinox Solar potential is being lost at peak rate, so the hottest potential temperature at this point in the year is likely dropping by around 0.5°C per day from the equinox. Assuming 36.0°C is the absolute maximum possible on the equinox in the current year were a worst case scenario to occur, this would leave us with around 32°C the highest possible temperature achievable on October 1st, which sounds about right.

This then leaves May as the outlier. On May 1st the hottest stations in the country have a daily average around 16°C. If we apply the highest anomaly recorded during Coningsby, which is +17.6°C at Nottingham, to these stations, we get 33.6°C, well in excess of the 32.8°C record, suggesting that we can go above 32.8°C anytime during May in theory if the right heatwave materialises. By the end of May the very hottest stations are averaging at 19-20°C daily max, suggesting on May 31st temperatures well in excess of 36°C could be possible. As this is around 4°C below the all-time record of 40.3°C, and the 32.8°C record was also around 4°C below the 1911 all-time record of 36.7°C, this would appear to be a sound line of logic. So why hasn't May seen a new record yet? I would put it down to sheer luck. It just hasn't materialised yet.

The same can be said for June. The maths demontrated for May clearly show that 35.5°C can be way exceeded at any point in June. So again, the answer for why June hasn't seen a new record is simply luck. This is not for a lack of trying though; both 2019 and 2022 really wanted to break the record, but failed to do so.

It should be noted that May 2005 had a serious chance of really smashing the 32.8°C record to pieces, but the synoptics didn't line up. So there has been at least one serious attempt in this century.

For the reasons stated here, I believe the March record will actually have the highest chance of surviving the longest, purely due to the need for timing, wind direction and high uppers to exactly line up. With the other months here, there is far more legroom for impreciseness.

 dryfie That was the initial reading by Kinlochewe before it reached 19.6°C, and then Achfary reported in the next morning with 19.9°C.

[Metwatch]

Some interesting discussions!

The airmass is definitely the limiting factor I noticed. Below a chart from the very end of February 1960 seeing an extraordinary warm airmass advected from the south, but then with March and April I cannot find any examples of an airmass this warm reaching the UK, and even into early May, something similar I can see the earliest is May 1992.

I think that is down to airmass transport not being as strong as we head through spring as the PV weakens, and things get sluggish / slower moving, so we have to wait until May for a particularly warm to hot airmass to make its way here, that is my theory but not overly sure how true that holds. Solar insolation by April is very strong as mentioned and even with a not really noteworthy warm airmass and high pressure just to our east, temperatures respond very well, heading into the low 20s to even mid 20s easily; early April 1946 or mid April 2003 a good example. March 2021 and April 2018 they didn't require overly warm airmasses to see surface temperatures respond well either, same in the case of the March 1968 and April 1949 due to that stronger solar insolation.

So I think if we can get that warmer airmass with 850s of above 15C, with good enough mixing to remove any inversions with a southerly breeze we could see those spring records fall, it'a only a matter of time.

September and October have seen record breaking hot airmasses such as last autumn, so that is not a problem following the excessive residual warmth in summer, just need that southerly component to the winds and no inversions higher up / good mixing of the airmass lower down to allow the highest possible temperatures, and they can still get higher than current records as well!

Edited February 12 by Metwatch

[BruenSryan]

 dryfie The McGivern tweet was before the 19.9C figure came in.

[BruenSryan]

 Metwatch Also related to this is how the last week of April seems to lack warmth and the well-known April quirk where April's record warmth has tended to be biased towards mid-month. I myself cannot recall much in the way of April warmth in my lifetime coinciding with the final week - the closest being 2011 where 27.8C was achieved on the 23rd but that's just before.

Get a good southerly and April 30C is easily achieved in my view. I wouldn't even totally rule out a 31C provided all variables are in order and timing to the end of the month.

[CryoraptorA303]

1 hour ago, Metwatch said:

I think that is down to airmass transport not being as strong as we head through spring as the PV weakens, and things get sluggish / slower moving, so we have to wait until May for a particularly warm to hot airmass to make its way here, that is my theory but not overly sure how true that holds.

That is actually not a bad shout, the atmosphere tends to be the least active in spring, as evidenced by spring being the driest season on average. I would guess it's mostly to do with the oceans and airmasses both being near their coolest points. This year the atmosphere and oceans are at record warmth though, which would appear to support that this year in particular could see some spring records fall. The tendency of having HP to the south of us atm will also make the odds of a spring temper tantrum more likely.

1 hour ago, BruenSryan said:

Also related to this is how the last week of April seems to lack warmth and the well-known April quirk where April's record warmth has tended to be biased towards mid-month. I myself cannot recall much in the way of April warmth in my lifetime coinciding with the final week - the closest being 2011 where 27.8C was achieved on the 23rd but that's just before.

There are a few examples. 2010 seems to have had its warmest April temps at the end of the month, as did 2013. I think this is related to the concept of climate singularities, which is really just a matter of timing related to the predictable Solar year.

1 hour ago, BruenSryan said:

Get a good southerly and April 30C is easily achieved in my view. I wouldn't even totally rule out a 31C provided all variables are in order and timing to the end of the month.

Using the anomaly method, even 32°C is theoretically possible in late April. I doubt we'll see anything of that magnitude for a while, but it's possible, and I'm in agreement that the next big April heatwave will see us go above 30°C. Bonus points if it's for more than one day.

What would be a sight to behold would be a severe late May heatwave that sees us go above the present June record, perhaps peaking at 35.8-36.2°C. This year will coincidentally be the 80th anniversay of the 32.8°C record, so it would be quite fitting for us to break the record this May. Even better if it's on the 29th as that will be exactly 80 years to the day. If I could chose the stations that set the record, I would suggest that it be a joint record between East Malling, Wiggonholt and St James's Park, as those are the three contempary stations closest to the 1944 record setters (Tunbridge Wells, Horsham and Regent's Park respectively), however as that was recorded in rounded Fahrenheit, we'd be highly unlikely to see a three-way record today. If we couldn't do that then I'd suggest Wiggonholt as the third hottest with 35.8°C, St James's Park in second with 35.9°C and East Malling being the record holder with 36.2°C.

I have no idea how likely a severe May heatwave is this year; most indications I'm seeing would rather point to serious heat later in summer instead. We could see a 1989 pulled; February 1989 is a close relative of February 2024 so far, so that's definitely not off the table. February's other close relatives are February 1995, February 2000, and of course, it's closest relative of all is February 2022, according to Starling's Roost. Along with January 2024's closest relative being January 2003 along with 1971 and 1936, this would seem to suggest we're in for a very hot year, so any heat record of any description is on the table.

[A Face like Thunder]

1968 was a strange year, with a record March temperature, a one-day summer (90+ on 1st July), severe flooding in July and September, and only an 'Indian summer' post-September floods to wipe away the memories of a 'garbage' summer, temperature-wise.   

[Harry's House]

 A Face like Thunder Also a spell of unusually thundery weather for so early in the year in April and a shockingly dull early August in southern England.

[WYorksWeather]

Worth mentioning as well that the number of opportunities for heat is probably lower in the spring. There's actually more of a tendency towards blocking, and northerlies and easterlies. One of the other issues is that the still cold SSTs also tend to mean that you get nagging onshore breezes that mean that many stations are knocked out of contention.

The key I think would be to get a 15C isotherm for the March record, and in the second half of the month, preferably as part of a longer period of dry weather. I think it would be very hard to break that record this year, for example, since we've had such a wet few months, unless March were extremely dry and settled, leading up to a plume at the end.

So it's also whether the whole period leading up to the month in question is suitable for a record. I think September 2023 may have broken the record, for example, if not for the exceptionally wet July.