That's what my Dad used to say and he should have known with his job. But it's a big IF  to get a good feed from a frozen watery Baltic continent this year. Maybe later this month but I wouldn't put money on it.

I seen the ensembles and same story no break down to real cold or Spring alike mild dry weather that we badly needed.  Those silly -10C uppers always at day 10 onward for a month now.  A break down are much needed, a zonality to return, a northerly, or Azores HP to move in dry out and bring more milder sunnier days.  This time any break down we get won't be worst than what we got now which is the worst non-weather and non-seasonal.

https://www.wetterzentrale.de/en/topkarten.php?map=1&model=ecm&var=1&run=12&time=216&lid=OP&h=0&mv=0&tr=24#mapref 

The next one is interesting for coldies.

Every winter. As I posted the other day. At least the last two winters have seen the model discussion be about potential at day ten. 

GFS Op - like FAX this week (LP a bit further N so SE-lies rather than E-lies). Next week's LPs start in mid-Atlantic and fill as they drift past Cornwall into France Mon 9th and Wed 11th, the latter connecting with cold air over Scandinavia and bringing a NE-ly blast to the SE. This doesn't last as pressure rises generally over Britain Sat 14th (still N-ly for E coast). Then mobile pattern from the west, LP 1000mb England Wed 18th, HP 1030mb Ireland Fri 20th.

ECM - agrees generally with GFS 

AIFS - London maxima around 8C and frequent rain (colder and drier briefly Sun 15th). Edinburgh, maxima around 5C, a slower decline to 2C Sun 15th then a quick recovery to 8C. Rain every day.

GEFS - in the S, mean temp and ens members near norm to Wed 11th, the sudden wide divergence of ensemble and mean drops to a little below norm. Some rain for the coming week, rather more prominent in ens members after 11th, heavy in SW, some chance of snow in the east by then. In the N, mean and ens members again near norm to 11th, then all rather colder (mean 2-3C below norm) Less pptn than in south but never really dry, 50% chance of snow in lowland stations after 11th, obviously more on mountains

For me, the models generally seem to have had a really hard time of it ever since during the festive season, with the edge of the reliable timeframe often less than a week ahead.

As well as the easterly spells that have been suggested at times over the past fortnight or so, I also recall that just before New Year, the models teased us with a strong rise in pressure over Greenland which would have considerably prolonged the post-New Year cold spell had it materialised. It seems that we have been stuck in the rut we are in at the moment ever since then. 

When the atlantic has full control of our weather at this time of year, the models seem to have little if any difficulty. However, it is often a different story whenever they have to deal with any kind of blocked set-up, like what we have just now.

If those models cannot cope with different weather pattern it need to be fully scrapped otherwise a massive overhaul to upgrade it to ban the fake/silly 10 days nonsense, no push backs and any weather pattern MUST count down without hitch. 

The models are just based on the classical engineering formulae connected with fluid dynamics as applied to the atmospheric circulation. The problem is that the lack of precise initial data and the density of the measurements means that the forward projection of the outcomes is overtaken by the complexity of the many feedback loops after about 7 days in all but the most stable atmospheric conditions. So the models after this date (often termed FI in here) are basic subsumed by chaos after that date and effectively increasingly turn into random number generators.

The only real solution is to provide increasingly more precise and denser initial data to overcome this limitation

As you say, NWP uses historical data to make a prediction on what will happen next at every stage of the output, using current scientific understanding based on that historical data. If just one of those predictions/assumptions is even slightly incorrect, the impact obviously gets increasingly amplified as the run progresses.

Until every area of the planet has accurate feedback of measurements and 'the computers' have a sufficiently enormous dataset to 'know' what would happen next in every circumstance/global synoptic set-up (which would take eons to gather), then there can never be full accuracy (which, as we know, amplifies through the rest of the run)

As an added complexity, climate change is constantly changing the 'rules'.

Is that actually true? Or do they just take the starting point and apply pure physics to see what happens to the atmosphere? I have wondered this for years and never got a straight answer. Several people have suggested that the models somehow use past weather patterns to make their predictions, but I am not sure that is true, is it? I assumed it is a purely mathematical process.

I thought that was how the AI ones worked, while the normal models use pure physics.

I mean you can summarise what every model does in one equation

Du/Dt = (1/rho)(DelSigma) + f

u is the three dimensional flow, so the left hand side is how the atmospheric winds change over time. 

f is external forces acting on each bit of air (gravity, friction as well as the effect of perceived force from the earths rotation such as Coriolis and euler). 

The other term on the right hand side is linked to internal forces of each little bit of air. So for example at the surface the friction is stronger on the bottom bit than the top bit so the air is 'deformed'

And that's it. Simple right? Except these forces all go down to a molecular level so you have to do something called parametizations to understand what is going on at a bulk level (since even if you could mesure the initial conditions of every particle, its not deterministic anyway). Parametizations also allow you to relate stuff like flow and pressure to other bulk properties like temperature.

In a sense any difference in any model will be linked to how these parametizations (approximations for bulk quantities) are done. Better models solve that equation with better precision.

Perhaps I worded it clumsily, but 'physics' is merely the understanding of how 'things' work. And we understand how 'things' work by observing how they work.

The weather computers 'applying physics' is merely applying what are understood to be the 'laws of physics' through what we have observed them to be by figuring (to put it very simply) "If x is applied to 'A', then the end result will be 'B'". And we test that theory by repeating the process in exactly identical conditions.

Take a golf shot. If you play a shot then place a ball in the exact same position, do the exact same swing, and the atmospheric conditions are exactly the same, 'physics' dictates that the second shot will finish where the first did.

We understand that the laws of physics apply uniformly - and we gain our understanding by observing how 'things' react in a variety of situations.

Sticking with the golf shot analogy, we know that if one even tiny factor is different for the second shot - our grip is slightly different/we stand half a cm further away from the ball/the wind strength changes/etc - the ball will not finish in the same position. 

Expanding on that further in the context of NWP models, if you're just making a short putt of, say, 20cm, the second ball wouldn't finish in exactly the same position, but it'll be close as dammit because it's not travelled that far. If the putt was 20m, the end position of the ball with even a tiny initial deviation will end up in potentially a visibly different position. That's like the difference in looking at NWP at t+24 and t+384.

(it's a much easier to have the nebulous concept in your head than it is to explain! 😕)

.

Looks such a promising chart and then you look further into it!!

A real saving grace is just how deeply embedded the cold is over Scandi. Would be interesting to see stats on that, is scandi having its coldest winter in many years? Anyway, keep the persistent cold there for long enough, we have to get lucky eventually.

Problem being it not being a true Greenland high! Not saying that can’t change