Edenfel DCNN 9437 – The mysterious case of artificial intelligence not knowing it – possibly for good reason.

54.59346 -7.28254 Met Office CIMO Assessed CLASS 5S Installed 1/1/1865 Archived temperature records from 1/1/1981.

The Centre for Environmental Analysis (CEDA) archives show a weather station known as “Edenfel” established in County Tyrone over one hundred and sixty years ago. CEDA holds digitally archived rainfall data back to 1872 though temperature records for the site only start from 1/1/1981. In starting my site review I located the station and then tried to find imagery using the Google search engine. This brought about some very weird findings.

Firstly I input the weather station name as it appears on the Met Office full listing of Synoptic and Climate stations here https://www.metoffice.gov.uk/research/climate/maps-and-data/uk-synoptic-and-climate-stations. Below is the screen shot of the surprising response claiming Edenfel is “not a real weather station”

This rather confused me, it certainly was not a typo. I opted to try “X” and their “GROK” AI.

GROK similarly claimed no such station existed. This is exceptionally strange given the longevity of Edenfel making it one of the UK’s longest running sites. Now incredibly confused and also wary of the Met Office’s proven ability to keep “Zombie” stations alive I opted to double check their stations map.

So far so good, it must exist because the Met Office clearly shows it. So I double checked their Location Specific Long Term Averages site for data – after all it is an impressively old site.

What? “Sorry we don’t have any climate stations at or nearby that location” So what is all this data below about?

I decided to revert to BING for any images and their “Bird’s Eye” view dutifully supplied the imagery I sought and proved beyond doubt (well maybe) the site’s existence.

That is very definitely a Stevenson screen and it is in the location specified by the WGS84 coordinates supplied on the CEDA archive page. It also confirms that this is an unenclosed station (so much for the Met Office’s claims that they should always be is a secure enclosed area) and it is in the back garden of a property set in a woodland clearing. Bear in mind that Class 5 states no regulatory requirements at all simply “2.6 Class 5 (additional estimated uncertainty added by siting up to 5 °C) Site not meeting the requirements of class 4.” So Edenfel is the lowest of the low and even then the Met Office adds that “S” indicating it is subject additional shading effects. In other words so unreliable it really should not be used for climate reporting – but of course they do.

Quite how the Met office manages to use the data from Edenfel is also something of a mystery. Studying the downloadable temperature data from this manually reporting station reveals the following. 2016 – just 305 days readings, 2017 – NONE AT ALL , 2018 – 275 days readings, 2019 – 338 days readings, 2020 – just 306 days, 2021 – 350 days, 2022 – 334 and the last archived year of 2023 shows just 332 days readings. Given that this is almost certainly an “adopted” amateur site in a private garden, such lack of diligent readings (with an entire unrecorded year) seems rather disappointing. I do wonder quite how anyone could use such partial data to form daily, monthly or annual averages but no doubt the Met Office finds a way. Perhaps they use the same formulae as used to divine annualised readings at places like Hartpury and Mickleham amongst very many others.

My final research was to try and establish the provenance of this site – was there any incriminating evidence of the site being poorly kept for example? Well no, not really, largely because as I went back in historic imagery the Screen simply …….disappeared!

Search high and low, around and about (and I am unfortunately all too experienced at having to do that) and I cannot locate where this screen was from before that year of no readings in 2017. The image above shows no vestige of a Stevenson Screen and there are no archived notes of a relocation. There are, however, notes relating to the “Gauge” (as in Rain) location giving (Irish) grid reference numbers 2466E, 3719N which derive a location in a walled garden (not unusual at all) about 300 metres away. If anyone can track it down I would welcome their information.

In conclusion, perhaps AI was correct, there may, or possibly may not, be an “Edenfel” weather station and the Met Office certainly does not want to show any “climate averages” for it. I seriously doubt it really makes any difference though as it is/was a junk site with a poor readings record – par for the Met Office course.

Source: https://tallbloke.wordpress.com/2025/05/05/edenfel-dcnn-9437-the-mysterious-case-of-artificial-intelligence-not-knowing-it-possibly-for-good-reason/