I think this is the way to go, because I’ve spent another few hours behind the scenes trying to obtain a positive result for heatwaves at the national population level, and I’ve come away empty-handed. Aside from my rather frail ARIMA modelling result back in part 3, nothing is coming along and singing a little song of death during heatwaves for the nations of England & Wales. Interesting is it not, that the only result I can weakly muster is one that arose from modelling and not empirical evidence?

Mrs Dee is not surprised, and points out our gardener and his family have recently returned from Sardinia where the temperature didn’t drop below 30°C, even at night, and where they roasted in 40°C heat during the day. Nobody in the hotel died, ambulances were not blaring away, and everyone went about their business as usual. But, as knowledgeable professors like to point out, heatwaves tend to bring deaths forward in the sick and the elderly and, because of this, a better measure of their true impact on society would be years of life lost.

Larder Raid

I’m not in a position to go calculating years of life lost so I’ll stick to plain old mortality (deaths/100k). The good news is that this is no longer crude mortality because it is calculated for quinary age bands rather than the population lumped together as a whole, the available bands being: under 5 years, 5-14, 15-19, 20-24, 25-29, 30-34, 35-39, 40-44, 45-49, 50-54, 55-59, 60-64, 65-69, 70-74, 75-79, 80-84, 85+. That’s a lot of bands to analyse, so I’m going to focus on the five oldest bands.

I shall start straight in with five basic plots of mortality for England & Wales over the period January 1970 – December 2019:

There you go - please pay attention to the scaling of the y-axis! If I bung them all together in one colourful slide this is what we get:

A few months stand out as being particularly vicious across all five age groups, with November of 1989 leading the way. You’ll see some periods affecting the older subpopulations but not so much the younger, as may be expected, and this sort of thing makes me think I ought to model age bands separately when it comes to dialling-in temperature data.

Getting Stuck Right In

I’m going to get stuck right in by looking for any sort of correlation between mortality for the older groups and my collection of heat-related measures derived from the Central England Temperature Record (HadCET). Here goes…

I’ve shaded the 25 cells of interest, and you’ll see a fair few statistically significant correlations appearing. There are two notable features we should note and these are:

1. My superhot day count for the number of days reaching or exceeding 30°C has failed to elicit a single correlation of note, whereas my standard hot day count of days reaching or exceeding 26°C has managed to elicit statistically significant correlations with all five age groups.

2. Of the 20 statistically significant correlations on show, no less than all 20 possess negative signs.

What does this tell us?

It tells us that it is cold weather, and not warm weather, that dominates mortality for the older age groups. Anyone working in a UK hospital, hospice, care home or residential home will tell you the same. The cold kills big time in the UK, and yet the Met Office and the media invariably focus on heatwaves.

This also tells me that I ought to concentrate on the summer months alone if I am to get anywhere, so here’s that table again but for summer months over the period 1970 - 2019:

Again, there are two notable features:

1. Of 25 possible correlations only 7 made it to statistical significance, and all of these exhibited negative coefficients, indicating that mortality among the elderly of England & Wales is elevated during cooler summers.

2. The hot day and daily excess heat variables that, arguably, truly mark out the heat that is felt during heatwaves fell from grace, leaving only the two measures of maximum daily temperature.

I’m hoping that readers are beginning to see why Imperial College and chums have to resort to statistical modelling, theoretical considerations and buckets of assumption in order to squeeze out a result. But, let me just trundle on and produce another correlation matrix for the summer months of June, July and August alone:

Nope, nothing there whatsoever! Apart from one measly negative correlation all has evaporated into nothingness at a time when we should be smacked about the head with a clear connection between summer heat and elevated mortality in the older age groups.

I shall remind you that we’re now dealing with genuine and actual date of death data, so there’s no messy delays to registration and/or processing to confound matters. Since I’m looking at the entirety of summer en bloc then it’s also less likely for there to be any awkward business arising from extended stays in ICU and the like (though there will no doubt be something of this for folk succumbing in mid to late August). In plain English there’s nothing to see when there should be no doubt!

After yet another boring matrix I suspect we need to see a colourful plot revealing summer mortality in England & Wales for the period 1970 – 2019. Feast on this…

Note how smooth and stable those curves are. Apart from a bit of a bump for summer 2003 for the 85+ year group we don’t see much by way of outlier and variation over a 50-year period. Many of us recall the blistering and prolonged scorching summer of 1976, yet it doesn’t even register as a blip on this slide.

So… I’ve come away empty-handed after trying a whole bunch of statistical tricks, including my favoured ARIMA intervention approach. Here’s a table of outlying summers (in terms of mortality) that were identified across all five age groups following an initial ARIMA run. On the heatwave front the evidence we should be looking for are positive value estimates (elevated mortality) that also happen to correspond to particularly hot summers. We should also see the same (hot) summers cropping up across all age groups. Except we don’t:

There’s not a single repeated summer (hot or otherwise), and just three estimates that run to positive values, with nothing at all to show for the 80 – 84 year group, nothing to show for 1976, nothing to show for 1995, and nothing to show for 2018. Keen readers may wish to eyeball the following slide that reveals total excess daily heat for each summer period, this being my best stab at capturing sustained hotness:

So… there you go. Trying to pin elevated mortality in the elderly on hot summers in the UK using empirical data and standard statistical methods isn’t easy, despite what the media and certain experts may claim. Instead, we have to resort to modelling if we wish to scare people.

Kettle On!

The overall mean of this series is 20.3°C; so if I go all nominal on everybody and select 22.6°C for my upper ‘hot summer’ threshold, and 18.0°C for my lower ‘cool summer’ threshold, then I hope you can see what happens as a result. It transpires we have a decent distribution of hot summers over the 54-year period, but the cool summers are bunched-up down prior to 1994. This is going to introduce a serious amount of bias given the crude mortality rate decline that we clocked back in part 4… we are going to end-up comparing apples with pears and calling them lemons.

Here’s the crude mortality rate plotted out for June, July and August of each year to match the above slide:

There are a number of ways I can deal with this, and the first idea that springs to mind is a first order differential for crude mortality. Here’s what this looks like:

We now get to see when the death rate suddenly rockets more clearly, with December 1989 leading the way with a change of +52.7 deaths/100k, beating COVID into second place with a change of +47.3 deaths/100k in April 2020.

With yer differential series we get to identify problem months easily, and we (largely) don’t have all the bother of bias introduced by declining mortality rates over time. I say ‘largely’ because there is a modest correlation between crude mortality and its differential (r = 0.336, p<0.001, n=650) that points to a tendency for greater month-on-month differences during periods of elevated mortality, as we may expect. Not perfect, but it’s a start.

Seeking Perfection

Something else that isn’t as perfect as it should be is use of indicator variables to signal hot days, hot weeks or hot months. We know that heatwaves can drag on, and my ancient mother complains of heat when the maximum temperature hits just 24°C if it has been really hot (+30°C) for a couple of days before. This feeling the heat more over time – an incremental effect, if you will – will never get addressed by indicator variables flipping in and out of existence. Neither will it get addressed by considering the daily absolute maximum temperature, for in statistical analysis each day, each week or each month is treated as a series of independent measures. If we are not careful with our number crunching, we’ll miss the impact of building heat on folk. I thus stopped to cogitate on how best to capture this notion, and came up with the idea of excess daily heat.

The HadCET daily data record for January 1970 – July 2026 tells me that the average daily maximum temperature for the summer months of June, July and August is 20.4°C, with a median of 20.0°C and mode of 19.1°C. If I take 20°C as my climatological normal value, then a day upon which the bulb outside hits 32°C will return an excess over normal of +12°C. If I then derive excess daily values for all temperatures above 20°C, and zero anything falling below this, then I end-up with a coherent daily time series that can be summed by week, month or year to reveal what I am calling excess daily heat.

Let’s have a look at how this measure stacks up by month of year:

Now this is interesting because a few moments ago Mrs Dee stated that August is the hottest month of the year. In terms of peak temperatures attained this may well be the case, but when we look at the measure I’m calling excess daily heat, then it is July that delivers. We also see a flutter for May and September, which makes sense to us mightily confused Britlanders.

So let us now look at excess daily summer heat when crunched into annual mean values:

Fascinating indeed!

I’m not quite sure where global warming went, but in terms of a rough and ready proxy for excess heat delivered each day above the 20°C ‘summer norm’, then we don’t see any evidence of a warming trend using our eyeballs. But I’ll do the decent thing for alarmists everywhere and try and tease a trend out by boiling these figures down into year by year totals for summers over the period 1970 – 2025:

A trend emerges… but don’t start applauding and screaming for a ban on air travel just yet, because if we run a linear regression we discover that trend is statistically insignificant (p=0.124). Statistically insignificant in this context means the positive trend you see cannot be considered to be any different from a horizontal line owing to the vagaries of variance and sample error; that is, it could have come about by chance.

But… if we extend this series then things may well change, so I guess I better do just that before I get accused of cherry-picking by the cherry pickers:

OK, so you keen activists out there can start banning air travel right now and closing down the entire economy for that trend now becomes super significant (p<0.001).

One little thing I perhaps ought to mention at this point is that I’m fully expecting the planet to warm gently as it continues to emerge from the last-ice age1, and that such warming generates atmospheric CO2 from a multitude of natural sources, as well as increase the amount of our primary greenhouse gas (water vapour). What we’re looking at with a time series that stretches all the way back to 1878 is the faint signature of our inevitable move toward an interglacial period. To blame every scrap of warming on fossil fuel use since 1878 is a political stance and not terribly scientific. If haters gotta hate, then activists gotta scare!

Now What?

I rather like my quirky little measure I’m calling excess daily heat but Mrs Dee wasn’t at all happy with my long-winded explanation over wine last night. Not a lover of hot weather, I ended by asking her whether she’d prefer a week of 29°C, or week of 24°C that ended with a record-breaking 38°C. She plonked unequivocally for the latter, appreciating that such a scenario would be likely to conjure blaring headline news and cause a stir among alarmists. “There you go”, I said, “it isn’t the height of the mercury column, but the sustained heat of day that surely matters to us humans”. She agreed, and so we opened another bottle.

With the 5-year historic baseline approach truly b*ggered for assessing the impact of the summer heat of 2026 by a step-like and somewhat inexplicable leap in deaths between 2021 and 2022, I realised the safest route forward at this juncture was to focus on the period 1970 – 2019, and look for a relationship between excess daily heat and mortality for the summer months of June, July and August using the somewhat exciting date of occurrence death data that was obtained under FOI.

Total counts of summer deaths by year seemed a bit weird to fiddle with at first, but this handy series lent itself to an additional measure, and that was to derive the ratio of summer deaths to total deaths for each year. I quite liked this for ratios always eliminate a host of tricky factors, and this ratio would account for population changes, changes in healthcare, changes in disease profiles, changes in age profiles, changes in legislation and the like. It should also be a measure that is sensitive to excess deaths arising from particularly hot summers. With that all said here’s a plot of the new kid on the block:

Now that is utterly fascinating.

Of course, we have to realise that the rising ratio could be an upswing in summer deaths as well as a downswing in non-summer deaths (or both), so we need to keep an open mind. Whether this pattern correlates nicely with my newly forged excess daily heat is the bit of cooking I’ve got lined-up next, but for the time being (and since I’ve reached the Substack email limit) here’s a taster by way of one of those Pearson bivariate correlation matrices:

Oh dear, yet another null result, with none of my temperature-based measures for summers over the period 1970 - 2019 coming anywhere near some sort of correlation with my freshly baked ratio.

I guess I’ll just forge ahead with ARIMA modelling of crude summer month mortality and see what this brings to the table... and, of course, I can always conjure a heavily biased 5-year baseline from which to derive excess summer deaths. I suggest we don’t hold our breath, for here’s a sneak preview of crude summer month mortality in action:

Is anyone else getting the creeping feeling that excess summer death and heatwaves aren’t really going to be a thing when it comes to looking at deaths on a small, damp island bathed by the North Sea? Mrs Dee would like to point out her colleagues always zip off to somewhere rather warm during the school holidays; and they don’t fear death, only poor service.

Kettle On!

Thus, even with a period of reasonably stable mortality, we cannot detect any relationship between summer heat and all cause death. Given the lion’s share of all cause mortality is generated by the older age groups, then we are effectively looking at zero correlation between heat and death for the most vulnerable members of society. No wonder Imperial College took-up modelling - there’s no empirical evidence to be seen!

Seasonal Excess

So… here’s another idea… how about I turn the pages back and derive estimates for seasonal excess all cause death? The reason for doing this at this stage is to find the weeniest scrap of empirical evidence that the heatwave of 2026 is doing something that is of sufficient wallop to make its way into national figures. Plus, time has moved on and the ONS have released a bit more data for us to chew on.

Seasonal excess in this context means summer excess but the summer isn’t over and, at the time of writing, the weekly death registrations are only up to week 31 (w/e 31^st July), and I have hitherto assumed that summer is quintessentially week 27 – 39. If I move the goalposts and set the study period to start at week 21 (w/e 22^nd May), then we get to include that rather warm period that popped up at the end of May. This gives us an 11-week window we might tentatively call ‘pseudo-summer’.

With that all done I felt I ought to look at some raw data before rolling my sleeves up, and so proceeded to derive counts of pseudo-summer season all cause deaths for England & Wales, which were also converted to crude mortality. I decided to crayon a couple of quick slides to get a feel for how the numbers would eventually shape into some sort of baseline, and this is when I noticed something very odd indeed. Take a look for yourself:

What on Earth happened during 2022 to cause that sudden surge in deaths and crude mortality? The word is once again WERGH, and this takes us into a decent-sized rabbit hole. There’s no way I can derive a sensible 5-year baseline, and so seasonal excess death has to be shelved as an idea for the time being. This also wades deep into John Dee’s Almanac territory and I feel I ought to take up further investigation over in that publication stream. It’s never bloomin’ straightforward!

Getting Granular

In analytical circles it’s fashionable to go granular these days. By that I mean drilling down into small subgroups and finer definitions. It is often the case that large-scale effects, on the whole, are weak or non-existent, and this sort of thing doesn’t please paymasters (been there, done that). Today I’m going to go all granular for two very good reasons. Firstly, deaths by date of registration isn’t the best variable in the world owing to the vagaries of notification and processing; secondly, all cause death might be handy to fiddle with, but it’s not the same as a physician certifying that heat-related conditions were a secondary cause of death. In sum, I’d like to go brutally granular and look for daily deaths by date of occurrence and secondary cause. Whilst this should sort the wheat from the chaff, it does mean I’m going to have to forget chasing 2026 since the necessary files won’t be available until next year.

Sound good? Sounds fruity? Let’s bake!

The Nitty Gritty

I thought I’d start straight in with this granular mode malarky by looking at secondary cause of death. As I’ve already mentioned, secondary cause is where we need to look to find mention of heat-related deaths (ICD10 code T67). The only file available to the public on the ONS website that is of any use in this matter is this one dated 27^th May 2026, which takes us up to December 2025 for England & Wales. If you flip to table 12 in this file you’ll see a monster entry that is 239,405 rows long. Despite this the datafile only covers death registrations for the period 2021 – 2025, which isn’t much at all for anyone trying to analyse the impact of heatwaves. If I boil the many options down into all people of all ages for the umbrella ICD10 code of T67; that is to say, the topline figures, then I end-up with this puny little table:

That’s the nitty gritty. If I want more then I’m going to have to stump up a great deal of cash and wait several weeks whilst the ONS process my FOI request. Sorry, dear readers but no can do! Even though it feels somewhat pointless, I’m going to bolt on some temperature data and call this the nutty grotty analysis. That being said, this nutty grotty analysis tells us one important thing right from the get-go, and that is physician-certified heat-related deaths in England & Wales are very rare. Imperial College and their chums might turn the handle on fancy epidemiological models that estimate thousands of deaths but real physicians in real hospitals tending to real people tell a very different story.

What I shall do now is turn back to the trusty HadCET dataset and crunch the absolute daily maximum temperature, the mean maximum daily temperature and count the number of hot days (daily maximum >=25°C) for the summer period (week 27 – 39) for the years 2021 – 2025, and bolt these figures on to the table. Try this:

Now this is darn interesting. We see 2022 maxxing out on heat-related registered deaths, absolute maximum temperature and mean maximum temperature, but not quite bagging the hot day top count into the bargain – though it’s close. A tiny table it may be but this nutty grotty is providing evidence of excessive heat = death. It’s a shame the ONS restrict the file to 2021 – 2025, but I guess it’s the sensible thing to do given the sheer size of table 12 at 239,405 rows long. I don’t want waste these tasty scraps, so here’s a paltry Pearson correlation matrix based on just 5 observations:

Result! Though we can’t expect much from this, we do have two max temp variables surpassing the 95% level of confidence in their correlation with heat-related deaths (ICD10 T67). The hot day count falls short at p=0.091, but this doesn’t surprise me since this crude indicator bins days at 25°C with days at 35°C and so smears any impact of truly excessive heat.

BUT…

We need to realise what we’ve gone and assumed before we get carried away. We’ve gone and assumed those deaths marked with ICD10 T67 as a secondary cause of death occurred during the hottest part of the summer. That ain’t necessarily so. Alarmists would have you believe that people keel over and die on the hottest days, with heat as the primary cause right then and there, on the spot. The reality is that healthcare services get stretched on the hottest days, with anything from insect bites to sunburn to drownings to pedal cycle accidents to BBQ burns and beyond. Ambulance service delays rise and so vulnerable people, as well as those exposed to truly excessive heat, can easily deteriorate due to lack of available care.

When it comes to analysis there are further tricky issues, for heat-related deaths are rarely instantaneous in the UK, and may well occur several days, if not weeks after exposure as Intensive Care Units battle to save lives. We need to bear all this in mind because climate alarmists would have you believe otherwise. We must remember that alarmists are selling an agenda rather than truth. But, let me see if I can squeeze anything else from these 5 data points.

Nutty Grotty Models For Heatwave Death

Try these for size…

Nice! I wasn’t expecting much but these are rather encouraging in that they’re pointing in the direction of linear or near-linear relationships. If I reach for my spanner that is generalised linear modelling (GLM) and assume a Poisson error structure then I get the following outputs:

These translate into the following equations:

• Annual T67 deaths = 1.579 x absolute maximum temperature – 41.525

• Annual T67 deaths = 5.959 x mean maximum temperature – 118.762

Interestingly enough, these predict zero ICD10 T67 heat-related deaths at 26.5°C for absolute maximum temperature, and at 19.9°C for mean maximum temperature, with the first T67 death predicted to occur at 26.9°C for absolute maximum temperature, and at 20.1°C for mean maximum temperature. Crude, yes, but I’m altogether terribly joyous at this little nutty grotty result. If I had the time and cash I’d now go and scour the literature to see how my fiddling squares with the big guns, and go buy a load of relevant data under FOI.

What I have got time for is to pull down the latest HadCET temperature data and find the hottest day in England this year so far was 26^th June, for which an absolute maximum value of 32.8°C was recorded. If we bung this into the equation for absolute maximum temperature we arrive at an estimate of just 10.3 ICD10 T67 secondary cause deaths for the entirety of 2026; a far cry from the thousands estimated by Imperial College.

With tons more data I’d be happy to squeeze more out of this but I think there’s little point in chasing this further, apart from one last table that sums the T67 deaths by age band so we might see who suffered the most over the 5-year period:

For sure this is rough and ready, and makes no effort to consider the size of each sub-population, but that’s not the point. All I’m doing here is confirming what we should already know, and that is the older age groups tend to suffer most, with the 75 – 89 year group accounting for 53% of all ICD10 T67 deaths over the period 2021 – 2025.

Going BIG

OK, so that’s going granular for you. Itty bitty little pieces but, surprisingly, we seem to have a result worth writing home about, and a possible model upon which to hang our grant application. I can’t take physician-certified secondary cause of death due to heat any further, but I can go granular in other ways, and that is to look at all cause death not by date of registration but by date of death. Yes indeed, we can go down to monthly, weekly or even down to daily counts for our time series and see how this squares with monthly, weekly, or even daily temperature fluctuations.

Sound good? Hop in…

Kettle On!

Another problem we face is that people in the summer do not act like people in the winter. Behaviours change, attitudes to risk-taking change, environmental factors change, the weather (in general) changes, activities change, diet changes, social interaction changes, disease profiles change, and so on. The modelling strategy I have used implicitly assum…

Read more

Intervention analysis (also known as interrupted time-series analysis) is a quasi-experimental statistical method used to estimate the effect of an exogenous intervention—such as a policy change, law, or external shock—on a time series. It tests the causal hypothesis that observations after a treatment differ significantly in level or slope from those before the intervention, accounting for the autocorrelated structure of the data.

The technique is typically conducted using the Box and Jenkins ARIMA framewor…

Read more

I also couldn’t see a mortality signal reflecting claims of a climate crisis, climate catastrophe, climate collapse, climate breakdown or climate apocalypse. Interestingly enough, neither will you find any of these terms used in the IPCC AR6 working Group 1 Report, this being the climate status ‘Bible’ for all climatologists everywhere. Read that weighty and most sober technical tome and you’ll wonder what brand of tea politicians, activists and eco-journalists are all drinking.

But, hey, I promised a slide of the 5-year crude mortality baseline so we could see what was being subtracted from weekly figures to arrive at weekly excess all cause crude mortality…

Read more

More than 2,700 people died from heat-related causes during May and June’s record hot spells, as climate change makes heatwaves more deadly, experts estimate. Researchers from Imperial College London, the Met Office, and the London School of Hygiene and Tropical Medicine (LSHTM) estimated that there were 2,700 excess deaths in England and Wales over the two heatwaves.

And they found that climate change significantly drove up the death toll, as more than two fifths (42%) died as a result of the extra heat caused by human activity – mostly burning fossil fuels.

On closer reading he discovered this wasn’t 2,700 actual people but an estimate furnished through statistical modelling… by Imperial College… Yes indeed, THAT Imperial college who, back in 2020, told us a zillion people would die of COVID. Except they didn’t. Dig around the internet and you’ll find Imperial College guessing numb…

Read more

A quick word about my climate-related Substack notes… I use them now and then to pop out rapid bake analyses, so please do check on them from time to time. Given the amount of excrement being shovelled on heatwave deaths I’ll be running a free-to-view series on this starting Wednesday 22nd July. And now back to the studio…

Disappointment is the word.

There you are extolling the virtues of the world’s longest running temperature record only to find it has more holes than my garden sieve!

We’ve got a shortlist of eight stations that are supposed to make the HadCET data record what it is (Radcliffe, Rothamsted, Ross-on-Wye, Stonyhurst, Malvern, Squires Gate, Ringway, Pershore College), so it occurred to me to count how many of these eight were operative on any one day, then convert these daily counts into percentages that could be averaged over the year to get a feel for overall completeness of the data record.

What came out of this shocked me somewhat, for the maximum daily count only managed to reach four operative stations out of the eight at best! TBH I had to cross-check my files a few times to ensure I hadn’t made a blooper. I hadn’t, and 50% was thus the best capture rate we could expect.

Herewith a handy little time series plot that reveals how the annual mean capture rate varies from just 12.5% (just one station operative) to 50% of stations operative (largely driven by Rothamsted, Malvern, Squires Gate and Ringway):

Read more

I’m very fond of the Central England Temperature record. For one thing you can pull down the monthly mean temperature for as far back as January 1659 (3°C). That’s kinda freaky: we’re talking 368 years ago, with Richard Cromwell on the throne serving as Lord Protector. His dad was the rabid puritan Oliver Cromwell (of roundheads fame), and one of the blokes who decided to bring the axe down on King Charles I. Nasty business!

Now, the first device capable of detecting temperature changes - known as a thermoscope - was developed by Galileo Galilei around 1592–1593. However, this instrument lacked a numerical scale and was affected by air pressure. The first true thermometer (defined as a device with a calibrated scale to measure specific temperatures) was created by Santorio Santorio in 1612 (with clear diagrams published in 1625).

In 1629, Joseph Solomon Delmedigo, a student of Galileo and Santorio in Padua, published what is possibly the first description and illustration of a sealed liquid-in-glass thermometer. This was described as having a bulb at the bottom of a sealed tube partially filled with brandy – now that’s my kind of thermometer! Interestingly enough, Delmedigo did not claim to have invented this instrument; he simply described what he had seen somewhere. But then, in 1654 or thereabouts, Ferdinando II de’ Medici (Grand Duke of Tuscany) went and actually manufactured the first sealed liquid-in-glass thermometer using alcohol. The key here is that seal, which renders the instrument independent of changes in atmospheric pressure. Good one Ferdinando!

The use of two references for graduating a thermometer is said to have been introduced by Joachim Dalence in 1668, although Christiaan Huygens in 1665 had already suggested the use of graduations based on the melting and boiling points of water as standards. In 1694, Carlo Rinaldini proposed using them as fixed points along a universal scale divided into degrees. Now, this is interesting coz it means that that first monthly mean temperature back in January of 1659 would have been recorded using a sealed liquid-in-glass jobby (I like the thought of it being brandy), but with possibly only the freezing point of water used as a single reference mark and a nominal scale of sorts. Quite so, for our modern-day use of the Fahrenheit scale didn’t come about until Daniel Gabriel Fahrenheit got going in 1724, with Anders Celsius adding to the pot in 1742. Quite how Hadley Centre bods have reconciled those very early records dating back to 1659 is beyond my ken, but maybe some keen reader might dig out a dusty paper or two.

So there you go, but since I tend to use HadCET daily data dating back to just 1878, then I’m well clear of early teething troubles and into the precision gauge era with mercury and whatnot. This leaves the tricky question of where you stick your thermometer. Out in the blaze of direct sunshine is not a good idea, hence the use of the protected enclosure designed by Thomas Stevenson (father of author Robert Louis Stevenson). The development of a screen with double-louvered walls on all sides and no floor was reported in 1864. Modifications by Edward Mawley of the Royal Meteorological Society in 1884 included a double roof, a floor with slanted boards, and a modification of the double louvers; this being the standard design adopted by the British Meteorological Office. This means those early daily HadCET records dating back to 1878 were likely taken using a basic form of Stevenson screen, with the full Monty arriving within a decade.

This leaves us with the fundamental issue of screen location, and this is the bit I want to get my teeth into today. As one sharp-as-razors reader has already commented, the weather station sample going to make-up the HadCET daily record has flipped around a bit since its inception. And there are other problems, besides. Those who are super-keen ought to read this paper by Parker et al detailing the development of the HadCET dataset.

Back to those stations…

You can pull down an ASCII file called cet_station_active_flags.txt from here, and if you do you’ll find a simple log of binary codes for the following eight stations:

• Radcliffe

• Rothamsted

• Ross-on-Wye

• Stonyhurst

• Malvern

• Squires Gate

• Ringway

• Pershore College

If I had time I’d run a little report on each station with a little piccy summary collage showing you where they have shoved their central England thermometers. But I ain’t got that time – business before pleasure, as they say - and these days I have to earn my crust elsewhere before I mess around with Substack owing to the microscopic following of paying subscribers. Such are the lives of statisticians.

The data comes with an explanatory note that we need to be aware of…

Note: Where a station observation is unavailable but the station is active, a backup station is used.

…OK, so they’re happy to flip around flippantly, and I guess I better flag those occasions when an unknown backup station would have entered the frame. The fly in the ointment here is that the datafile doesn’t distinguish between those occasions when a station is active but with data unavailable, and those occasions when a station is inactive with data unavailable. This means we’ll need to assume that backup stations were used each time the unavailable flag shows its face. When it comes to climate science it’s never straightforward!

And now for some basic plots…

Read more

Well, it sure has been hot down here in the South West. So hot, in fact, that we broke the England temperature record twice on one day in the land of trakkers and cider… or so they have claimed. I’ve been jotting down a few notes exposing the farce, and you can clock them all again under the following hyper-headings:

Smashing Records

Beers & Beef

Summer 2026 vs. 1976 I

QUESTIONS, QUESTIONS!

According to the BBC...

Back Garden Bias

A Bit Of Common…

Summer 2026 vs. 1976 II

Please do keep an eye on my notes section coz this is where I’m likely to dump a quick and dirty freebie analysis any time of the day or night, with and without warning, and with or without biscuits.

Apart from the Met Office leading the gullible deeper into a globalist narrative, one other thing that gets my goat is when folk say “I don’t ever remember it being this hot”. I’m never quite sure what they mean by this, for when closely quizzed it usually becomes clear that they’re also never quite sure what they mean by this.

If I’m being particularly cheeky, I’ll fire-off a barrage of questions, usually starting with “what do you mean by hot?”, followed by “what is the temperature today, and how does it compare with temperatures of fifty years ago?”. It rapidly becomes clear that the general public haven’t the first clue about anything other than them feeling hot and uncomfortable in the present moment.

The impact of humidity never seems to enter their mind, and neither does the impact of wind, trees, clothing, buildings, ventilation, hydration levels, age, well-being, technology and all the rest. I generally get a blank stare if I mention that the maximum temperature measured this very day is actually less than it was on this day five decades ago. It’s as if some folk have made their mind-up that climate change, as promulgated by the BBC, is pushing us all toward a doom-laden scenario that humans (being evil) fully deserve. Occasionally I’ll get, “was it really hotter then? I guess I’ve forgotten”, and this pleases me no end. Some, when prodded to actually remember rather than regurgitate propaganda, will start telling tales of how they engaged in water fights and had to sleep in the garden etc etc etc.

Then, of course, there are the perennial Met Office shenanigans with the reporting of mean values as opposed to maximum values; and the churning of the many definitions we could apply to generate media headlines mentioning ‘heatwave’, ‘record’ and/or ‘deadly’; that sort of thing. These days we’re not allowed to enjoy a pleasant summer’s day without a stiff dose of fear porn or a weather chart showing very scary blood red and black shading. My smartphone used to frighten me several times a day with dire weather warnings until I turned the blasted thing off, and now there’s a local neighbourhood keep-you-safe group emailing us all with terrifying warnings that the impending heatwave will be deadly. TBH, and speaking as a biological entity, I’d rather stand naked in the garden on a blistering June day than a freezing cold February day if I want to survive. It wouldn’t take long for hypothermia and death to set in, but alarmists don’t seem to care about genuinely deadly cold weather these days. This tells me they’re not at all interested in saving vulnerable others, but they are somewhat interested in saving their very political cause.

Heatwaves – the definition thereof – provide a rich source of fun and games with numbers each summer, and the Met Office are particularly keen to play this game; presumably because they’re paid to do exactly this. If they can’t declare a record-breaking heatwave each year just watch them jiggle the numbers about until they get one. If they can’t declare a record-breaking maximum temperature they’ll flip to counting days where a nominal threshold was exceeded; and that threshold can flip from maximum values to mean values if necessary. Then there’s the ‘three or more consecutive days’ clause they can dial-in; and if this doesn’t work, they can rely on using the phrase ‘sustained’, which can mean anything they want it to mean. We live in the strangest of times when numerical garbage is swallowed by the public wholesale without question, and especially so if somebody on the small screen delivers the info in a posh voice, suit and (blue) tie.

Dee’s Hot Day Count

So, without further ado, let me generate my own heatwave indicator and be as transparent about it as possible. What I shall do right now is take the HadCET maximum daily temperature for the whole year period 1878 – 2025, and flag those days where the value reached or exceeded 25°C. This is a nominal threshold that I can easily change, but I wish to settle on this because it is my favourite summer temperature: beyond 25°C and I start to get sticky, flushed and in constant need of cold beer. It’s also a handy threshold to adopt because it generates 1,170 hot days amongst the 54,215 days in the sample (2.2%). This means we’ll get a graph that is neither laden with excessive counts nor suffering from anaemia.

As Yoda would say that graph now is…

Read more

So far in this mini-series we have looked at historic data for the mean, mean maximum, and absolute maximum May temperature and have discovered that May 2026 was warm but not exceptional. What I’d like to consider today is the cool side of things and I shall start out by plotting the mean maximum and mean minimum temperature series for the month of May on one slide. I’m hoping that readers will realise that the mean maximum temperature is, by and large, the mean daytime temperature, and that the mean minimum temperature is, by and large, the mean nighttime temperature. I have to declare “by and large” because it is quite possible to experience warmer nights and cooler days under certain conditions.

With our cuppas in hand, let us look how these two series stack up over the period 1878 - 2026:

First impressions are that both series are climbing at the same modest rate over time. A quick and dirty linear regression confirmed this is indeed the case with the slope rate for the mean maximum temperature series fetching-up at +0.9°C per century (p=0.001), and the slope rate for the mean minimum temperature series fetching-up at +0.8°C per century (p<0.001). Don’t hold your breath. If this is what global warming looks like, and those globalists are telling the total and utter truth about atmospheric CO2, then I suggest we don’t panic.

One little feature of May 2026 that I ought to point out is that the mean minimum series doesn’t offer anything exceptional (night time temperatures were fairly typical), with all the hard work being done by the mean maximum series (day time temperatures were exceptionally high). Now, I don’t know about you guys, but I’m not keen on pinning the blame on cow farts, SUVs, eating meat, and air travel. Neither am I keen on pinning the blame on atmospheric CO2, but that’s another story for another day. What I am keen on is figuring what the sun was doing in terms of hours of uninterrupted solar radiation, and this means we have to sit and figure what was happening to cloud cover. We also ought to figure what was happening to barometric pressure, weather fronts, wind and jet streams, for the UK tends to bake when hot Spanish or North African air starts flowing over these soggy islands. I guess you can’t tax or legislate for weather fronts and the like, so our poor farmers are going to be clobbered in a bid to prop-up a failed economy.

A Broader View

Read more

Yes, it was hot over here in the UK last May; and it was particularly hot down here in the South West where we make cider and drive around in tractors (pronounced ‘trakkers’)1. I’ve no idea what the record-breaking record for the UK as a whole was and, TBH, I’m not bothered about this. I’m not bothered for two reasons, these being: 1). the reliance by the UK Met Office on weather station data of such a lowly grade that we may as well throw it in the trash, and 2.) the sheer nonsense we find ourselves bathing in on a regular basis. I recently typed out a short note on this that I’ll reproduce here for convenience. A few days ago, I said this…

Whenever the Met Office announces another record-breaking temperature we ought to stop and consider what it means. If the previous record was set 100+ years ago, well before we invented the SUV, and when atmospheric CO2…

Read more

Dry as a bone, that. The most exciting part of this is the zero mean for the raw residual series, which is telling us that the trendline went smack down the centre leaving no significant drift or bias. That’s a start, but we need to see how the residuals rise and fall over time. Here we go:

Read more

I often resort to reckless cheeky-mongering with CO2 because it can help drive the point home that Earth’s climate is far more complex than some experts make out, and I can think of no better slap in the face than to home-in on the substance some believe to be the root of all evil. We may laugh, but I was first introduced to climate alarmism by a man who quite seriously declared atmospheric CO2 to be toxic to the poin…

Read more

Enjoyment with biscuits

What I’m primarily doing with my Substacks is enjoying myself numerically on a daily basis. I sit down with a cuppa and poke about until I find something that grabs my attention. Yep, this is big time data mining – and with biscuits! I’ll then type thing…

Read more

I thus turned to the world authority on sunspot activity over at the Royal Observatory of Belgium in Brussels, and pulled down their daily datafile from this webpage. First up – and arguably the most logical approach - was crunching the annual mean daily sunspot number (SSN); then I had an idea of counting the number of days when the sun went berserk. To define ‘berserk’ I settled on the 95^th percentile value attained for the whole SILSO data series from January 1818 through to December 2015, and this fetched up at a count of 233. This gave me the base threshold from which I could flag (and count) seriously wild days each and every year.

So… let us begin at the beginning with a quick look at the logical approach, this entailing running our five (annual) variables of interest through a Pearson bivariate matrix:

Read more

Mean sea ice extent dominates the statistics you’ll see bandied about, but this is only half the story, and we really ought to be considering variance in ice extent. Variance is something that alarmists tend to avoid, for whenever we look at variance we discover Earth’s climate is far less predictable than claimed. By definition, the catchphrase climate change invokes the notion of variability, and thus leads to the formal statistical construct we call variance. I tackled this issue in a fourteen-part series that kicked-off with an article entitled Climate Change & Climate Variability (part 1) back in June 2024. It’s worth digesting this because I arrive at the conclusion that the climate is likely not getting more variable over time.

So, then, variance is once again le plat du jour, and I shall now furnish two choice plots that got me thinking a wee bit more over a rather nice mug of Grumpy Mule. First up is the (annual) standard deviation of Arctic daily sea ice extent over time:

Read more

We shall start with the North Pole, and I shall colour-code the 566 data points that take us from November 1978 through to December 2025 according to the seasons of the northern hemisphere. Here we go…

What a banana!

If we take the blue winter blobs we see lots of ice extent and little variation in this, as we may expect. Spring (green) is very similar in both mean and variability about that mean such that there’s very little difference between spring and winter ice dy…

Read more

Antarctic Sea Ice Index (parts 1 – 10)

Antarctic Sea Surface Temperature (parts 1 – 2)

Antarctic Land Surface Temperature (parts 1 – 8)

Is Antarctica Boiling Over With Heatwaves?

My fingers and thumbs make that 21 articles and, being terribly cynical for a wee moment, I presume this is proportionate to the sheer number of bollards talked-up in the popular press when it comes to the continent few can actually go and visit.

By strange and wonderful coincidence, I was one of the scientists invited to spend some time there as part of a British Expedition back in 1981, but I declined owing to romantic entanglements of a deep and tender nature. Yes indeed, that was me doing some old-fashioned courting that ended with betrothal - and marriage - to Mrs Dee. A good decision, and a warmer decision!

And now for some gentle entertainment by way of slides characterising what is going on down …

Read more

There it is in a fetching ice blue looking for all the world like a set of stalagmites. Back at school our geography teacher offered a priceless mnemonic for distinguishing stalagmites from stalactites, and I shall share it with you today. He simply said, “tights come down”. And that was it: a class full of sniggering students for the rest of the term! Very Finnbar Saunders. Speaking of which, here’s a suggestion I may well submit to Viz Top Tips…

Save money on botox by wiping your lips with a Scotch Bonnet - Colonel John ‘Debris’ Dee of Colchester (ret.)

Getting back to serious business I need t…

Read more