Cassley Addendum – A demonstration of how changing observation artificially creates warming and the implications for the future.

In my report on why winter nighttime minimums are increasing, I introduced the concept of traditional and historic observation timings over-recording colder minimums due to observations being taken often at the coldest time of day. Expanding on this idea regarding the long term record at Rothamsted I demonstrated how the transfer to more frequent observations on station automation effectively eliminated this prior over cold reporting and indicated higher nighttime temperatures {more to be revealed later}. This change created the statistical impression of warming where none actually existed.

Most weather stations have transitioned from once daily manual observations to automatic multiple readings with no visible overlap period to run comparisons, however, some did such as Rothamsted and also Cassley among a few others. Studying this transition period for Cassley is incredibly revealing.

Cassley weather station in Sutherland operated from 1/1/1974 to 28/9/1999 as a manually observed station with readings taken once daily at 09:00GMT. On 29/9/1999 the site was automated and twice daily readings were archived (09:00 and 21:00 ) in addition to the once daily manual observations. The manual observations stopped 1/6/2002 with solely automatic readings continuing from 2/6/2002. This gives a 976 day comparison period.

To briefly recap, manual observations are taken at 09:00 which is frequently (especially in winter) very close to, or actually at, the coldest time of day which is typically around 30 minutes after sunrise. For reference sunrise on the 1st December at Cassley is 08:25 and does not get earlier than that until the last week in January. This means a cold night’s reading may not effectively be reset at any interim time so that no matter how much the temperature actually rises, the next days observation will still show that minimum and get recorded the following day. To further clarify this point from the Met Office’s own “Daily Extremes” page which only reports data from automatic weather stations, – no manual reporting ones are used.

A specific example was the 4th March 2001 when a low of minus 11.3 °C was recorded by the manual observer which matched the automatic 09:00 reading exactly. The following day on the 5th March 2001 the 09:00 minimum manually observed was minus 8.5 °C. And that was the figure that historically would have been solely entered into the historic archive for manual stations – 2 consecutive very cold days. However, the automatic recorder effectively reset the minimum recording at the 21:00 reading on the 4th and thus recorded the real lowest temperature at 09:00 5/3/2001 at plus 3.6°C………..a 12.1°C differential between the two different reading protocols.

This is what those figures look like in the CEDA Archives. AWSDLY is the 12 hourly automatic weather station data, whilst DLY3208 is the manual recording. Column I is the maximum reading and column J the minimum.

If such a large daily variance seems improbable, it is worth remembering that Cassley is very prone to regular Foehn effects.

“A record temperature of 16.8C was verified for Cassley in Sutherland. It was recorded at 03:00 GMT on Sunday 29 December. It is the highest on record for 29, 30 or 31 December.”

So have I simply exposed an occasional freak phenomenon or some occasional quirk? Or is there a systemic problem in making this changeover from manual to automatic recording? Enter the statistical and analytical skills of Dave Woolcock who has diligently downloaded the entirety of the CEDA historic temperature data for every UK station on the data base. His methodology was simply to compare the AWSDLY and DLY3208 reading for each of those 976 days plotting the variances between the two simultaneously running but separate systems.

Although I deliberately highlighted the extreme case, it is worth emphasizing the two thermometers (Liquid in Glass for manual readings and Platinum Resistance for automatic) will have insignificant variance in overall accuracy ignoring their differing response times. There should be no differences in the readings provided they are reading the same thing but, unfortunately, they are not.

The timeline scale (X axis) clearly shows the variation seasonally focusing on the colder months (but not solely winter) with the Y axis having to extend to double digit degrees to accommodate the variations. The remarkable scale and frequency of deviations indicates this observational frequency change is causing a very large distorting effect. Whilst the auto system is correcting a prior system deficiency it is also proving that there is much more to comparing historic records than would normally “meet the eye”. The result is that a continuous comparison timeline of weather station readings from before and after automation is not practical. Whilst many “climate scientists” may claim “adjustments” (that they are so fond of) can be made, the obviously stochastic nature makes any such manipulations grossly unreliable.

This situation is then further complicated by the lack of transparency in this conversion process. Whilst Dave has been able to identify a few stations with visible overlap data (more will soon be detailed notably Rothamsted itself) the vast majority do NOT show such comparative data. Below is typical of more recent data. In this case for Alice Holt Lodge which converted from manual to automatic reporting on 29/5/2009. No overlap comparative data is available and this is what has happened with almost every conversion – the likes of Cassley are very much the exception not the rule. Note the change from single 24 hourly reading to twice daily 12 hour reading in Columns D and F

Dave is in the throes of producing more comparatives for the few other stations where overlap is available which I will publish shortly. More detail regarding the numbers of variances and the effect of daily averages these have will be produced. However, all the preliminary indicators show this change has a major effect – after all, just 1 days double digit discrepancy will noticeably affect even monthly averages.

To put this change over from manual reporting to automatic reporting into context requires an indication of the scale of the change over time. Given the Met Office is not inclined to offer me any assistance at all, I am having to compile my findings rather laboriously by checking each individual site’s records. I have so far scanned all the English weather stations and the numbers are highly significant.

ENGLAND: Total current operating weather stations ………….212

Of above total, Automatic Reporting………………………………….149

Manual reporting Stations…………………………………………………63

Of the 149 automatic reporting stations currently operating, 38 converted from manual to automatic in the four year period from 2009 to 2012. In addition a further 5 converted after 2012 plus an additional 7 are new automatic stations replacing closed down manual reporting stations in nearby locations.

The end result is that 50 (34%) of all those currently automatic reporting weather stations in England alone have effectively change reading protocol over the last 15 years. Alternatively put, of the total English currently operational sites over 24% (so far) have been subjected to an observational change which artificially generates an appearance of warming where none exists. {i will shortly be publishing a list of changeovers by country/date to cover all the UK when time permits}

I must stress again that it was the old reporting system at fault for over recording cold events, however that is not relevant to the fact that this changeover effect is known to the Met Office who conveniently overlook the impact. Spot anything significant about the average winter minimum recorded temperature in the graph below starting around that 2009 to 2012 period from a previously flat 10 years? Or is it my “wishful thinking”?

Those who have studied Dave’s variances graph will also have noticed another oddity – some of the discrepancies are negative with the automatic weather station (PRT) actually recording a colder temperature than the manual LIGT – and surprisingly often in the summer months. That discrepancy cannot be explained by time of reading protocol and neither can it be by variances in instrument accuracy. They should both read the same temperature when read……..provided it is at the same time.

This next part of this post is potentially very subjective on my part and unquestionably requires further detailed investigation which I would welcome any meteorologists or observers viewpoints on.

Again I will refer to the real world observational practice as shown in the video below taken by the Floors Castle observer. If not viewed before it is a must watch as an important reference to view real world practise.

Floors Castle is a manually reporting station, however, it is important to note that, following World Meteorological Organisations recommendations, all Met Office weather stations are now equipped with Platinum Resistance Thermometers. This manual site changeover to PRTs was completed 2017

The wire coil running to the left of the image connects the PRT probe to its Data Logger visible central screen. {An automatic reporting weather station would not require this type of data logger as it transmits electronically and does not require observer reading}. The relevance of this is to point out that both PRT and LIGT were mounted within the same screen during these earlier days of changeovers. In fact in all current manual stations the minimum reading LIGT is still retained and used as the observer demonstrates above. Note that important point, the minimum LIGT is effectively redundant and would NOT function following conversion to automatic reporting – so why retain it and specifically use it only at manual stations?

With both PRT and LIGT mounted within the same screen and used simultaneously this requires the screen door to be opened daily – this would again not happen at a solely automatic site which is only opened on servicing visits. Now what is it that the Floors observer says?

“That’s our current temperature outside…..sometimes it changes as we open the box, erm…. obviously ‘cos once you let a normal bit of air inside its cosy little box it changes” …

The response time of PRTs is vastly quicker than that of an LIGT as I noted in my report on Lentran. 

“PRTs are much faster reacting with the Met Office standard platinum foil Vaisala PTU series being quoted at approximately 20 seconds for 63% in air at the same velocity as above { 3 metres/second}. 95% of final value is often achieved well under 40 seconds.” 

Conversely consider the LIGT used by a traditional triage nurse who shakes down the thermometer first and pops it in your mouth under your tongue into a much faster responding solid/liquid medium. The nurse will then take your pulse rate viewing that upside down watch they wear on their lapel for a full minute, check a few other physical factors (eyes etc) before removing the thermometer to only then take the reading. A reminder

The point I am demonstrating is that a weather observer confronted with both LIGT and PRT at a dual recording screen will take the LIGT readings for maximum and minimum immediately on opening the screen as they are trained to and then go onto other obs functions. This leaves the screen door open for a long enough period for the PRT to “adjust” to the outside air temperature it is suddenly now exposed to. Of course if it does change that demonstrates the inadequacy of the screen itself……it should expressly not be “in its own cosy little box” it should be accurately recording the outside air temperature anyway.

This is where the known failings of naturally aspirated/ventilated screens are exposed. In general terms wind speeds fall overnight in the cooler air. In low wind speeds of less than 1 metre/second, Stevenson Screens are well documented to retain warmed daytime air inside the screen and in this insulated environment not fall as low as the outside temperature. The reverse problem is also noted in still but hot and sunny conditions leading to solar radiance warming the static entrapped air and over recording temperatures. This latter effect was well noted in the alleged record breaking 2022 heatwave where very poorly sited and sheltered screens recorded improbably higher than they should.

I suggest these lower PRT than LIGT readings may be a strong indication of Aitken effect that is an additional distorting factor particularly when either screens are poorly sited or compromised by subsequent development or unmaintained vegetation in their vicinity.

In summary this is the first of a series of reports on this ongoing compromising feature both reading times protocol and how it can artificially create the appearance of warming. Consider that 30% (63 out of 212) of all weather stations in England are currently once daily manual reporting sites BUT everyone of them is already fitted with a PRT and recording mechanism that can readily be automated. In doing so the problem of artificially induced global warming by data representation alone will be massively increased.

I welcome comments and advice on this issue that I consider to be just one of many adverse but also well known issues that are adversely affecting the accuracy of the historic temperature record.

Source: https://tallbloke.wordpress.com/2025/11/20/cassley-addendum-a-demonstration-of-how-changing-observation-artificially-creates-warming-and-the-implications-for-the-future/