The Cambridge Convolution – Part Two

Image: Courtesy of https://www.botanic.cam.ac.uk/cambridge-university-botanic-garden-records-highest-ever-uk-temperature/

This post is part two and follows on from my recent post on the discrepancies between the two nearby weather stations in Cambridge which operate to different historic temperature recordings protocols. I asserted that these differences artificially create the impression of warming where none may actually exist by a crude convolution. I follow this up with more data to demonstrate this point.

A brief recap. The two Cambridge sites are at opposite ends of the CIMO weather station assessment scale for temperature recording. They are 6.29 km (3.89 miles) apart as shown below.

Cambridge University Botanic Gardens (CUBG) is rated as very poor CIMO “Class 5 (additional estimated uncertainty added by siting up to 5 °C)”. It was removed from the Central England Temperature series in 1931 due to suffering from identified Urban Heat Island Effects (UHI). The site quality has dramatically worsened in this respect over the intervening 95 years. I personally find it unacceptable that the Met Office feels it can ratify national highest temperature records there as it did in 2019 – a private weather station in such circumstances would unquestionably be rejected as unsuitable. A prime noted effect of UHI is that night time temperatures are elevated significantly above those of rural areas – differences of +6°C higher are considered routine.

This site continues to take once daily manually observed readings at 09:00. The headline image from 2019 shows the irony of this “manual observation”. Minimum readings are noted (by hand) from the horizontal minimum Liquid in Glass thermometer (LIGT) whilst maximums are read from the electronic data logger which uses a Platinum Resistance Thermometer (PRT). The PRT takes readings every 15 seconds with each 4 averaged per minute to provide 1,440 timed readings throughout the entire day. Thus the minimum LIGT is actually redundant. If this site were to be converted to electronically transmitted readings (a small task) the LIGT would not be used – this makes a crucial difference as detailed by Stephen Connolly here .

Cambridge NIAB Is one of the very few CIMO Class 1 stations the Met Office operates. It lies in a rural location outside the city within the controlled areas of the agricultural research institute. It would be expected to record lower temperatures (particularly overnight) than the artificial and heavily modified urban climate of the Botanic Gardens site. The weather station has been in its present location since 2009 when it was also converted to automatic operation solely dependent on a PRT/Data logger with no LIGT – again a crucial point to note.

My initial “one off” comparison of the temperatures recorded on 1st February 2019 demonstrated the reverse effect to UHI in that the CUBG had an archived minimum recorded temperature of -5.8°C (from the LIGT) whilst the NIAB only showed +0.2°C (from the PRT/data logger) making a 6°C discrepancy. This was clearly wrong due to the identified over recording of nighttime temperatures of “traditional” Met Office systems that are now, realistically, unfit for purpose. This warranted further research.

Dave Woolcock has been especially busy in analysing data from those weather stations such as Cassley where the changeover period readings from both manual to automatic stations within the same screen were available. I will be further reporting the dramatic findings on these sites shortly, however, to test some of my hypotheses Dave then analysed the two Cambridge sites to give a longer and more recent time frame. {N.B. It would be expected that minima recorded at CUBG would normally be much warmer than NIAB due to strong UHI effects at the former. }

The graph below is the basic indication of variances in recorded daily minima. There was no period data from most of 2017 from CUBG, whilst the PRT was fitted and screen design changed, hence the gap period. Data covers daily minimums variance from end 2009 to end of 2023.

As an explanation all the dots above the zero line indicate the scale of difference where CUBG is recording colder than NIAB with extreme variances reaching 9°C. This is the direct reverse of what would be anticipated and indeed is really happening. The dots below the zero line indicate the anticipated lower readings from NIAB.

To further indicate what is happening here Dave then added a 90 day rolling trend to indicate the seasonality of variation.

I feel it is reasonable to deduce that winter nighttime minimums are being recorded regularly colder at the CUBG than they actually are. This is not because of any real differentials in like for like readings but solely due to the differences between the two recording protocols.

The major concern is that when the CUBG transfers to automated recordings (along with up to 40% of all current UK manual weather stations already fitted with PRT/Data loggers) the change will eliminate the long term historic cooler bias and portray a warming trend. This would obviously show as a “step change” for any individual site but when sites are progressively added into the total stations mix over time will only indicate a continuing warming trend.

There is a lot more analysis to do and tidying up of presentation. The full methodology is being logged and will also be published. However, all the first indications are demonstrating the real reasons for the likes of the graph below showing modern phases of warming. There was a spate of conversions in the early 1990s, a short hiatus of about 10 years with few conversions and then a major programme between 2009 to 2012. These changes are shown in the Met Office presentation. Such significant winter/cooler seasonal changes will move annual averages upwards as well.

The future Talkshop historic temperature reconstruction will follow the principles identified in my report on Wye and also take changing readings protocols into account by setting a before and after reset point for term comparisons where readings protocol/instrument changes are noted.

The next post regarding Dunstaffnage in Scotland will progress this issue and indicate the preposterous nature of Met Office comparisons and an absurdity of “Well Correlated” stations determining data infilling.

Source: https://tallbloke.wordpress.com/2025/11/26/the-cambridge-convolution-part-two/