This morning's 0z lowers them a little bit, but this would still cause widespread damage across the south:

https://www.theweatheroutlook.com/charts/gfs/00_66_windvector_gust.png 

(There's no 69 hour chart on TWO, but it would show the 70s+ gaining strength as they moved over Kent/Sussex)

The other models continue to show high 50s / low 60s rather than low 70s, so I'm continuing to hope that GFS is just plain wrong. It's the persistency of its forecasts though which has me worried!

The GFS also develops a prolonged cold spell, and even the MetO ends with an ESE'ly across the southern half of the UK as a trough disrupts to the south.

Aside from those winds on GFS, it's a very encouraging start to the day.

Annoyingly the ECM is still pushing the Atlantic through 

The 69 hour chart shows this.

It looks nasty still, but on the plus side: a) it is expected to move away quickly b) other models aren't quite as bad 

Ecm Op is much poorer though mainly because of those pesky Iberian heights.

GEM? I think you mean ECM, as GEM is very similar to GFS in terms of showing a prolonged cold spell with snow for most.

I much prefer GEM's handling of the NYD low, mind you, but I suspect it's a minority option... even the MetO raw now shows winds in the 60s here on NYD.

You must mean ECM is the wobbler,  GEM is great especially for the South.

GEM 0z snow. 

https://www.wetterzentrale.de/en/topkarten.php?map=20&model=gem&var=25&run=0&time=192&lid=OP&h=0&mv=0&tr=12#mapref 

Correct....Still a long way to go on this but a Jan 85 repeat looks the top prize.....meanwhile BBC go rain lows of 6oc lots of rain next week for Kent

There's definitely a chance of a memorable cold spell this morning.  But we are all experienced to know it's so difficult these days to actually get one. 

BBC use ECM so expect it to be milder than others today. 

I know exactly who he is and the term - just that it doesn’t go wrong ever it shows the outcome. Nuance of mine the phrase is annoying

The iPhone weather app, which uses the IBM model, has been flip-flopping a lot recently but this morning has also gone for a colder out look. A full week from Thursday of highs no more than 3C, with some snow around. By next Wednesday it shows a max of 0 here with a low of -6. 

Point taken, and agreed. 

The main encouraging point as Doc made this morning is the date of divergence remaining solid as 5th Jan. Until that pesky low blows through we’ll probably still be none the wiser but at least we’re not in the classic ‘jam tomorrow’ situation (at least so far).

Completely agree. It’s just cringey and utterly stupid. 

Back to the models - still plenty of interest for those that like it colder. Remember the old adage though - until all models are on board, it’s not happening. ECM still a straggler to some extent, and no one should think this is a done deal until it falls in line (if)

WX has firmed up on cold weather, from slightly on the cold side for W Europe in week 1, to week2 in which freezing weather stretches all across to the Atlantic coast, Spain excepted, and Brittany and Cornwall just above freezing. Ever more extremely cold air in Scandinavia and \n Russia. I don't remember seeing such a wide area of W Europe under 0C before. PPtn in week 1 for Scotland and Portugal; in week 2 in an arc from the Pyrenees to the Alps to the Black Sea, dry-ish for Britain so cold but not very snowy.

GFS OP 00z - LPs crossing Britain 995mb Hebrides Tue 31st and 995mb Cornwall Wed 1st, both moving rapidly east to join large depression over Scandinavia and pulling in N-lies behind (gales for Wed 1st). The pressure then rises and wind backs to the NW before the Scandinavian LP bulges W-wards Sun 5th re-establishing N-lies for a couple of days before a re-load of Atlantic HP with NW-lies. From Fri 10th an Atlantic LP runs into Biscay, pressure rises over Iceland; between the two some very cold air floods in from its base in Scandinavia for the beginning of the following week.

JET - always running S of Britain, close but streaky at first; from Sun 12th strong and looping deeply over Iberia

ECM - conflates the LPs for Tue & Wed and takes them as a reduced feature across Scotland so the wind following is never more than a strong NW-ly. On Sun 5th, the LP approaches from the SW, not the NE, and although it combines with the Scandinavian LP to bring in N-lies, these like GFS go back to NW-lies from Wed 8th.

GEM - like ECM suggests a conflated LP but deeper 985mb Scotland with W-ly gales elsewhere. Also shows winds going bac NW-ly but on Sun 5th LP approaches from the Atlantic and resolves into a trough stretching the length of the English Channel Mon 6th dividing col air to the N from milder to the S - a classic recipe for S Coast snowfall. The cold air wins out but there is a rematch on Wed 8th.

GEFS - temp dips into the freezer (5C below norm) Wed 1st and stays ther with the support of most ens members until about Fri 10th, after which although op and control  and several followers remain low, an increasing number break away to milder outcomes and by Mon 13th mean is back to norm. Pptn in the S (rain most likely esp towards SW)  Wed 1st and in most runs for a few days after Sun 5th, then frequent but not in every run. Temp profile in N similar, but pptn (most likely as snow) Tue 31st and Wed 1st, less after Sun 5th but still a cluster and mostly dry after that.

https://www.wetterzentrale.de/maps/AIFSOPEU00_168_1.png 

https://www.meteociel.fr/cartes_obs/gens_display.php?x=282&y=103&type=2 

ECM refusing to see eye to eye with the idea of notable cold. 

Even by day 12 it is still cold on the latest run: 

https://www.wetterzentrale.de/maps/AIFSOPEU00_282_1.png 

One thing that's interesting about the AIFS/pattern-matching service is that it was never keen on that New Year's Day low, instead keeping it as a wave or a very shallow feature as it crossed the UK. Most of the other models, including the op ECM, have now followed suit. It's GFS which is the big hold-out, for now at least.

I'm awaiting the verification with interest, and not just because I fervently hope it's right (as it means lower wind-speeds). IF it managed to get the low right several days before the other models, that's worth taking note of.

We'll know in a couple of days!

But is its AI going out and 'scraping' the other models rather than producing a forecast in its own right?

No, it's just fancy pattern-matching from a database of old charts.

In theory it's a good idea, as unlike us humans it won't have a bias to only focus on the "good" or interesting bits of pattern matching. And in practice? This NYD low will be an excellent test.

No, as far I’m aware it’s both (and may be the highest skilled model there is as a result based on what I have read).

As far as I'm aware it takes the ECM's starting data, then maps it to previous similar outcomes from the ERA (reanalysis) and ECM archive database, blending the charts as necessary if there's not an exact match. It's a sophisticated algorithm, to be sure, and as you say it's consistently beating the traditional models, including ECM, in terms of verification.

I quite like the MC description of it!

These data are provided experimentally as a test of forecasts made by artificial intelligence/machine learning. The calculations for the AI model are completed in one minute! The model was trained with ERA5 reanalyses and old ECMWF high-resolution runs.

No, it’s built using its historical dataset. IIRC it’s 30 years up to ca 2018 plus an overlay of more recent data, all of which is fed into some form of machine learning system.

Brian posted a link to a detailed explanation.