Apocalyptic weather

I'm on your mums side with this one. We've spent all winter begging for warm sunshine, when it finally shows I'm inviting it in 😎

It's certainly something worth exploring. There are some forecasters who use pattern matching to try and produce medium to longer range forecasts, but the problem is that no one global setup is like any other that has happened previously.
As a scientific type you'll appreciate this one: when working on the earliest models back in the 1960s, one of the meteorologists found the computations were running too slow, so he rounded the data from 4 decimal places to 3 to speed things up. The results were markedly different from a few hours in, and totally different from 12 hours onwards.
And that's the problem with pattern matching; a seemingly small difference in one of the input variables cascades through the results and causes huge changes just a bit further into the forecast. That's why they run each model 50 or more times with slight tweaks to input data to allow for imperfections in the source data (ensemble forecasting). Generally the results of each run are close out to about 3 days, often to 5 days, and sometimes out to 7 days, with increased scatter the further you go. However on other occasions, like leading up to this spell, the forecast temperatures have differed by up to 5 degrees only 12 hours ahead.
Anyway, AI will be able to pattern match far better than humans, and i think they have been doing a lot of back testing to try and train AI to pick up on differences and adjust for the most probable outcomes.
When it comes to longer range forecasts, it's purely about the overall setup, the long wave patterns. Conditions higher up, around the 500 to 850mb level, are easier to forecast as they are not impeded by topography. It is easier to forecast where upper level troughs and ridges will be placed, and that drives weather closer to the surface. That can give us some confidence as to the overall trend, but not the details. So, longer range forecasts will focus on the likelihood of above/below average temps and rainfall, but not day to day forecasts. AI could certainly help with these forecasts. That doesn't even begin to take account of the various teleconnections which also influence patterns (ENSO state, i.e. El Niño/La Niña being one of the major ones).
FWIW they do run deterministic models out to 3 months, but you may as well toss a coin because the accuracy approaches 10% outside of two weeks. It's one of the reasons I obsess over it; I love the complexity, the science and the maths involved.

Anyway, Cardiff provisionally set a new record for the warmest June night, coming in with a minimum of 22.5, which is above the average maximum for June in Cardiff.
Of course, the June maximum was also broken yesterday with 36.1 recorded in Gosport. There is a decent probability that this will be exceeded today, probably somewhere between West Sussex across to Gloucestershire and down south of Bristol which is where the hottest air currently lies. Yeovilton and Shoreham are already reporting 33 at 11am.

37 recorded in Yeovilton this afternoon, beating yesterday's record by 0.9, and the 1976 record by 1.4. This is a very large margin in weather terms; in a stable climate you'd expect maybe 0.2-0.3 above a previous record.
Jersey hit 39.1, which is the first time they've broken the 100F barrier.
Temperatures are forecast to fall to around average by Monday.

The record high for June has been beaten for the third straight day. 36.9 observed in Suffolk, with a couple of hours heating to come. Hitting 100F seems quite a high probability at the moment. That would be over two degrees higher than the highest temperature recorded in 1976, which is the benchmark summer for the nostalgia merchants claiming that it's just summer and perfectly normal.

Heading to Siena and Tuscany for a week tomorrow. Forecast for Saturday and Sunday is 98 (37) & 99 (36) degrees, perfect conditions for an afternoon stroll.... Scorcio as a Harry Enfield sketch may say...

Scochio was a Fast Show sketch which, whilst sharing actors with Harry Enfield's Television Programme, was not Enfield. Not many people know that.

Anyone who watched The Fast Show knows that. Try harder.

Anyone who watched the fast show would know a half arsed Michael Kane impression when they saw one as well.

If Curbishley isn't involved in this rapidly evolving weather based technological ecosystem, innovation matrix and synergistic data-enabled future landscape, I'm not interested .