Castlemartin Warren – Another new hotspot requiring a detailed inspection of the issues around sample point distribution. {A long post}

Castlemartin Warren (above) is NOT listed by the Met Office as either a climate or synoptic weather station on their current listing. It does appear in the CEDA archives but solely as a wind measurement point with no historic records since its installation in 2014. As far as I am aware this is the first time it has appeared as a regional “extreme”. Two issues are raised here: new sites appearing at notably warmer recording locations and, more importantly, the very significant modification new sites are making to the distribution of climate data sampling points and subsequent “averages”.

The CEDA archive shows a “Start Date” for Castlemartin as 22/7/2014. The below image of the site is from 2011 clearly indicating the wind mast predates the claimed date.

The November 2024 image of the site still does not indicate a Stevenson Screen, however, it may be possible that it is a non standard sensor and pole mounted as at Aonach Mor .

What is perhaps most significant of the very recent inclusion of temperature readings from Castlemartin is its location. It lies under 4 miles from the existing long term site at Milford Haven opened 1969 and just 6 miles from the Dale Fort site, opened 1959. Why include data from yet another weather station so close to other existing sites, especially one also on the coast?

In reality it gets much more unusual regarding Met Office claims of site spacing and distribution.

“What is a weather station? 

A weather station is a site equipped with instruments that measure a range of meteorological parameters. The Met Office operates more than 350 weather stations across the UK, each recording a selection of data including air temperature, atmospheric pressure, rainfall, wind speed and direction, humidity, sunshine, cloud height and visibility. However, not every station records all meteorological parameters. These stations are, on average, spaced about 40 kilometres apart, allowing meteorologists to monitor current weather systems and the UK’s ever-changing climate over a longer period.”

This “average” 40 km spacing seems somewhat improbable given the existing proximity of Milford Haven and Dale Fort with the inclusion of Castlemartin almost equidistant between the two. Rather than start comparing reified “averages” or even 40 km separation how about just a much smaller 40 km circled area like this below?

How many official Met Office weather stations are there likely to be in the described area above? Firstly we now know of three coastal located sites at Milford Haven and Dale Fort plus now the very recent addition of coastal Castlemartin. Anymore? Well how about seaside Tenby…..the clue might be in the name of Pembrey Sands…….”Aber” indicates the sea end of a river as in Aberporth……..Saron is inland by just 10 miles…….Scolton Country Park is just 8 miles inland……..Whitechurch is 7 miles from the nearest sea…….and then the almost equally very modern Whitesands is well under 1/4 mile from its self named bay.

In the 40 km described area above there are no fewer than TEN official Met Office weather stations. Were the two “new” ones at Whitesands and Castlemartin really an imperative for the benefit of the climate record of the UK ? Do additional sites in very mild locations only help to distort historic readings ever upward?

In my review of Buxton I noted there were only 3 sites in the entire inland cooler county of Derbyshire, whilst in the very much smaller area of “hot recording” west London there were eleven operational sites with many newer ones. In my report of Baintown I noted just 3 weather stations in all of the cooler east coast Scotland county of Fife, whilst at Exeter I could not help but notice the increasingly large number of sites in balmy Devon. In Dorset not only were station numbers increasing, the Met Office was still resorting to the inexcusable practise of opening new stations in walled gardens such as Kingston Maurward in 2017.

So does this seemingly selective site opening to replace (or often add to) existing locations make much difference to historic records? Well, all those long deceased stations do require their numbers to be fabricated from “somewhere”. Those “gridded cells” still require “somewhere” to derive their numbers from so all the better if those modern “somewheres” are warmer than their predecessors. {If only the Met Office would own up to where those “somewheres” actually are!}

As I noted in my report on Elmstone in my home county of Kent, large numbers of very well located inland rural sites have closed over the last 25 years with the remainder of their modern replacements highly compromised.

Below is the 40 km circled area of Kent comparable to south west Wales highlighted above.

How many official Met Office sites remain within this described area?

RAF Manston lies just 2 miles from the English Channel on the isle of Thant, whilst Faversham is under 3 miles from the Thames Estuary. At Langdon Bay (Heights) it is inadvisable to wander around in the dark lest a stray step results in a 100 metre rapid descent down the “White Cliffs” into the Straits of Dover. This only leaves 1 other station within that radius – more on that in a while.

Whilst a 40 km radius in mild south West Wales can support 10 stations and increasing, east Kent warrants just 4 – with a major reduction in recent times.

An observation….whilst Castlemartin Warren was reporting 12.4 °C, that 4th Kent station at Frittenden recorded that same day what was actually the national lowest temperature at -3.7 °C.

This differential is, in itself, nothing of note, a significant east England /west Wales variation is quite normal. What is important though is that with only one inland recording point in east Kent, the area average for the likes of non existent sites at Dungeness, Folkestone and Dover, is dominated by the remaining 3 coastal recording points.

In Frittenden there was the above figure of -3.7 °C, in my own back garden near Canterbury I recorded – 4 °C, the Private Weather Station (PWS) in a nearby vineyard recorded – 4.2°C , an also nearby private airstrip recorded -4.1°C. In Ashford a PWS ( by its own admission subject to Urban Heat Island effects) which reports to the Met Office WOW site recorded an urban -2.2°C as shown below.

“My station is in a moderately built up part of Ashford, about 20 minute walk east of town centre. I am using a WH1080 weather station and a custom built Stevenson Screen.“

However, none of these lows will be adequately reflected in the “infilling” data used for the like of those non existent “Climate Averages” stations because that data will be fabricated from the likes of coastal Manston only simultaneously reading +0.8°C as a minimum.

Or coastal Langdon Bay only reading a minimum of +2.2°C

To complete the hat-trick, as a manual, once daily observed site, the official Met Office Faversham station does not openly report its data. However, a nearby PWS likely to record the same as Faversham only dropped to +0.6°C as below

The point I am demonstrating here is that the ongoing changes to operational sites will inevitably affect “averages” drawn from the changed cohort. If the members of that cohort are selectively changed to warmer (maritime influenced) locations over time, the appearance of warming can be created.

Below is the list of all post 1950s sites in the similar 40 km Kent area.

The removal of inland sites such as Elmstone, Anvil Green, Wye, Throwley, Charing, Ulcombe, Peckham and Tunstall has succeeded in moving the averages up by the elimination of cooler recording sites in Kent. It should be remembered that meteorological averaging is highly responsive to extremes at either end of the warm/cold scale. Notable lows rapidly wipe out any highs when averages are calculated as demonstrated in the absurdity of May 2024 being rated much warmer than May 2025. It is much more effective to adjust averages upwardly by elimination of intense and/or frequent lows . Risible Class 4 Faversham regularly records the nation’s highest annual temperature (notably again in 2025) and is not having its overall area averages “diluted” with the likes of Elmstone.

The same is happening in south west Wales as demonstrated by the new inclusion of the likes of coastal Whitesands and now Castlemartin in favour of now closed cooler reporting sites such as Lampeter and Llandeilo and others.

Progressing this point of “These stations are, on average, spaced about 40 kilometres apart” consider that “average” is a term that seems to do a lot of “heavy lifting” for the Met Office. There are 10 operational weather stations in the above milder south west wales peninsula. How many weather stations should be expected in this same sized area of the Welsh/ English borders covering parts of the counties of Powys and Shropshire?

Within this delineated area, the weather stations at Hope closed in 1978, Forden closed in 1987, Welshpool closed in 1994, Newtown closed in 1996 and Llanfair Caerinion closed in 2006 leaving precisely ZERO operating stations.

The Met Office offers no “Climate Averages” for anywhere within this delineated zone.

What the Met Office does offer are Climate Extremes by country, county and by month. Looking at the lowest minimums for Wales.

Several things are evident.

1. Powys records 8 out of the 12 month’s coldest temperatures.

2. Long closed Welshpool itself (almost central to the delineated area) is an all time monthly record holder.

3. Only the Class 5 sites at Llysdinam and Alwen still exist.

Referring to the same data for England is equally revealing.

1. The coldest December and January temperatures at Newport (Shropshire) and Shawbury are towards the eastern border of the delineated area.
2. All the other sites except Kielder Castle no longer exist. {The Santon Downham site in Norfolk was known as “Grimes Graves” and is not the same as the current Santon Downham site in Suffolk}

Again I refer readers to Darrell Huff’s seminal work “How to Lie with Statisitcs” which starts with the first chapter “The sample with the built in bias” . In the case of Met Office weather stations, for whatever reasons (be they intent or chance) the claimed representation of the natural climate by geographic spread is quite obviously not impartial.

A cynic may suggest that the system is open to being “played” to manipulate data in an exceptionally surreptitious manner almost invisible to an untrained observer – a form of difficult to spot and passive “error by omission” rather than an active “error of commission”. Others may claim other system requirements have inadvertently modified how representative the data actually has become. Whichever side of this potential debate the answer may lie, I suggest the changing weather station network is an important factor (amongst many others) distorting the temperature record.

It will be with a uniform geographic spread in mind that any future Talkshop temperature reconstruction will operate.

Source: https://tallbloke.wordpress.com/2025/12/07/castlemartin-warren-another-new-hotspot-requiring-a-detailed-inspection-of-the-issues-around-sample-point-distribution-a-long-post/