Just checked the Pert’s for the GFS pub run, and 11 of the 22 members - 50% on the nose - show the -5 or lower isotherm within Mid East Anglia (MBY) and all points north, by T+192 - which is pretty much the same as this morning - despite the dodgy Op (Dodgy for cold weather fans)!

https://www.wetterzentrale.de/en/topkarten.php?map=1&model=gfs&var=2&run=18&time=186&lid=C00&h=0&mv=0&tr=6#mapref 

But most of the ensembles are crap

However when it comes to these synoptic patterns, the extra data is crucial and so more emphasis should be placed on the op and control

I was waiting for the ECM op before posting something similar (from a southern point of view), but the fact that the 0z ECM op is 7C warmer at 850 here by 192 shows how much uncertainty remains in the outlook. The ECM ensembles have been trending milder here and don't show a median frost at night any more - and the daytime median in the potentially deep cold part of the forecast has gone up from 3C to 6C, so not much of interest there. 

That's the thing with something which is roughly a 30% chance (going by GEFS up 'til yesterday's 18z) - it usually won't happen. 

Of the ops this morning only the GFS brings in anything reasonably cold before 240 here, going as low as -8 at 850. The MetO, ECM and GEM struggle to reach -4C at 850! The "chasing a rainbow" theme continues.

Looks like we can add AIFS to models who get fooled by into mis forecasting them. Odd, annoying and roll on Spring. 

It show how poor all the models and more inaccurate than before. Many need to be scrapped as well had more accurate non models in the past when an Easterly arrive without hitch.  Seen the app temps really poor with average to less cold temps.  We need to worry more about less cold temps than what the models showing.  Was surprised yesterday when GFS was so good but ensembles very poor with less cold temps and rain  from the slider as uppers mean far too high. 

Of the GFS 0z - 13 of the pert’s past T+216 hours being in the Atlantic - of 22 inc control  and Op. So it is a downgrade- but not a capitulation. 

I'm surprised you ever gave it that much credence, TBH. For one, it's only pattern-matching the past climate, not the current one. Two, no two scenarios are the same - and there's a *very* limited pool of potential matches to draw on. Three, the "AI" models won't be able to pick up on small-scale features at a considerable distance, consider how "clean" they look compared to the physics-based runs. And four, though they verify better than the traditional models it's a bit misleading - as that's on a hemispheric basis, not a tiny (in global terms) set of islands at the edge of a massive ocean.

The thing I've taken from this is that the pattern-matchers showed there was a high chance of a snowy spell if we'd been in the 80s or 90s. Sadly we're not, we've had over a degree of warming since then and it's made it all the harder to get a textbook Scandinavian High in midwinter. 

None of this rules out a potential cold spell, but it remains around a 30% chance at this morning down here, the odds may vary elsewhere in the UK.

Fair summary Retron - it’s only ever  been a 50% chance at best. Now 30% may be more realistic looking at all  the models - although the GFS op is a peach this morning: proper eye candy. Unlikely but fun to watch it through.

Obviously it still might happen but it's very unlikely now. And it's not just AIFS that's got it wrong ECM and GFS have as they always seem to. Why?

Do ECM and GFS also pattern match ? Genuine question I dont know .

https://www.wetterzentrale.de/en/show_diagrams.php?geoid=50893&model=aifs&var=2&run=0&lid=ENS&bw=1 

It's no less likely than it's been for the past week, really - it's been hovering at 30% (in terms of -10s for London in the latter half of the GEFS) for days on end. It's been as low as 15% and as high as 55%, but more often than not in the 30 -35% range. For really deep cold, i.e. ice days, it's been 15% at best, usually below 10%. 

ECM and GFS ops just run on physics, so aside from perhaps a bit of "bias correction" it's just the usual struggles - the butterfly effect, I think of it being, where early errors magnify. It's more noticeable in these types of setups as frankly nobody cares if it's 6mm or 8mm of rain in a SW'ly, or whether it's 12C or 14C on a March day. 

The new GEM, incidentally, is still mostly physics-based but uses "AI", i.e. pattern-matching, to tweak its temperature fields. 

Very interesting, it is a underrated model.

That's certainly what it seems to be. The start date for a cold easterly was initially flagged as 21st/22nd which is now only a couple of days away, and we seem to be no closer than we were 10 days ago!

The GFS op run was the first one I looked at this morning and gave a bit of false hope, but it's not representative of the ensemble, and the other models seem to be variations on the theme of cold rain with a bit of sleet and maybe temporary snow for the south.

Really? I pointed out two or three days ago that it was consistently being shown as from the 25th to the 27th (i.e. the last week of the month)? I don’t remember it ever being before the weekend except in one or two GFS operational runs. The overall easterly influence remains but any “real” cold has always been more of a coin toss and to me that’s the way it remains. The ECM 00z op run seems to either “gone off on one” or picked up a new trend and seems to be arpa odds with its ensembles in the crucial medium term.

1. The medium range outlook remains very uncertain and things haven't significantly changed in the last 72 hours

2. The most likely period for a cold easterly flow remains the end of Jan and the first half of Feb, as I've consistently said

GFS Op - less detail than FAX, LPs often deeper, but then continuing 995mb Sun, 1010mb Tue 27th, 990mb Wed 28th (this gets monotonous!) but then the Scandi HP asserts itself 1035mb and pushes some E-lies across Britain  Fri 30th, rather weak in the west. On Sun 1st an area of LP runs SE-wards across Britain into France and by Tue 3rd this rather than the HP is the driver of stronger and colder E-lies. The HP by then has shifted from Scandi to become a ridge Iceland - S Russia.

ECM - also keeps LP trundling up and down the west coast to Tue 27th but then develops a deep LP 960mb crossing the Scottish Borders and bringing colder N-lies behind esp for the E coast (No E-lies in this model). Then Britain is in a col between the departing LP and a new one mid-north-Atlantic, like GFS with its final ridge but less definite. 

GEM is like GFS with LP off the SW at first, but from Wed 28th, the LP deepens 975mb  English Channel bringing in rather cold air from the NW (cold rain?) and on into Europe, threatening E-lies Thu 29th but these less certain than GFS

AIFS - London, maxima 8C now, down to 2C Mon 26th (no ice days), back up again Sat 31st, rain in the milder periods winds S-ly except a day of N-lies  on the 26th. . Edinburgh, temp profile similar though only down to 4C on the 26th, rain likely at any time, winds entirely E-ly. (as in models above, winds imply circulation around LP sitting to the SW)

GEFS - temps sliding from near norm now to a period from Mon 26th when mean 2-3C below norm - considerable ens variation but the majority of runs stick with the norm, (op much colder in S, 10C below norm Feb 2nd), small amounts of pptn (heavier in southwest),  in one run or other most days.  Snow row figures suggest snow unlikely in far S, possible for E Coast of England, quite likely in Scotland

I saved the ens graphs and iPhone app screenshots from a week ago today and they showed the dip starting on Wednesday (21st). The app showed a max of 3C for Wednesday which fitted with the ensemble mean. It now shows a max of 9C so there has most certainly been a delay.

Incidentally last night it was showing a max of 0C for next Tuesday and this morning that has risen to 6C again. Any proper cold from the east has consistently been pushed back, always staying at least 8 or 9 days out.

The ECM meteograms IMBY suggest similar, a decent chance of temps less than 6c on their scale, with a low chance of cold and a low or non existent chance of very mild. 

Whether this morning's 0z signal is maintained and the ens scatter continues to reduce is of course TBC. So to summarise, a chilly spell still looks likely from around 25th or 26th, colder as you head north and that's where the greatest chance of snow will be. It's generally unsettled so precip at times and transient snow pretty much anywhere can't be ruled out, depending on usual track, timing etc. of encroaching LPs. Stepping back I don't think the bigger picture has changed that much from recent days but we might be getting a bit better agreement that the atlantic will have a bit too much energy for the HP to have enough influence for a cold spell IMBY but not so much influence to bring much in the way of mild or even average weather and with the HP just to the north or north east that's where you're more likely to be under colder air. 

I still believe the models are only good enough for trends , as we see on a regular basis ,we see extremes for both options cold/warm, it invariably ends up a compromise between the two. The beast always seems to stay at 300 hours plus on the charts ! 😃

Hope it stays there , still plenty of snow here , and some sharp frosts every night since the cold spell “ ended”