https://www.wetterzentrale.de/en/show_diagrams.php?model=ecm&lid=ENS&var=2&bw=1&geoid=50625 

As Darren pointed out, even the smallest errors in the initial data can quickly amplify, resulting in significant changes to the weather several days later. This is why ensemble runs use slightly tweaked starting conditions to better capture the range of possible outcomes.

Even here might get an inch of snow from that 😂 if only it would happen 

To me, real life is too random to be narrowed down to a scientific formula. I'm sure if you got a ball and a robot leg kicking it with the same speed and force each and every time, that the ball would not stop in the same single spot each time. I'm surprised that the models are not capable of accounting for such randomness from the same starting point with each run. 

Isn't that the ensembles set in a nutshell?

And that's why I'm indifferent to easterlies.

I feel I'm outside the No-Homers club.

Because none of the ens know the exact starting point including the opp

Your assumption of the ball is correct it would not stop exactly in the same place each time slight changes in vacuum and external calculations to name a few including unknown ones unknown calculations also exist everywhere including the models butterfly effect etc because detailed physics can’t calculate the variables they don’t know about that exist in any one exact moment of time which of course errors become greater with time.

Think it as a perfectly tuned engine, it is only as perfect as the computer says it is because it has not considered factors it does not know about.

Hope that makes sense.

https://www.theweatheroutlook.com/charts/gefs/gefsens850london0.png 

The op very much towards the top of the pack on tonight's late run, in stark contrast to this afternoons full fat Beasterly. 

There are some VERY cold members still showing their hand, with two late run clusters at or around - 10 and - 15. They are however still very much outnumbered by less cold solutions, but at least the hope lives on for another day! 

There's nothing really to buy into. The possibility of a potent beast is there in the ens, though it's currently the minority solution (maybe at best a 1 in 4 chance). 

It'll be several days yet before we have any idea what to expect in the final third of January.

The GEFS has had 9 or 10 members showing -10 for London over its past 4 runs, so around a 1-in-3 chance. What's more interesting though is the really deep cold, i.e. the -15s, as there were three in last night's 18z - the most for a long time. 

Yes - that’s my count and score on the chance. Its still reasonable odds against past cold feed signals, but we must be realistic that the 60% chance of a sustained westerly feed is reality. More runs needed and upgrades quite possible as the data becomes clearer.  It is after all, Normal for the models to play this out in a rollercoaster fashion., but when you are partial to one outcome over another - as I am, it is a bumpy ride. 

Clearly didn't explain myself properly. In real life, the same single starting point would yeild multiple different results in any test. So why 'tweek' the starting conditions of each run to achieve the same? to appease an unthinking computer program that is incapable of dealing with natural chaos?  Surely mulitple runs based on the same actual raw data (rather than slightly tweeked each time) would result in a better ensemble set? Of course, until a program is capable of that, then it won't happen.

It'd be pointless, because 2+2 always equals 4. All the traditional models do is apply maths to the starting data, so there's only one possible result. The difference between the control and the op is purely down to differing resolutions.

(And it'll be why the MetO is looking to phase out the op - if your ensembles have the same resolution as the op, you may as well just use the control run instead and save yourself some computation time).

Indeed.  You’ve highlighted the issues:  we will never have a 100% accurate starting position and the models are only an approximation of the physics involved.  As the models improve and the resolution gets higher the accuracy improves but there’s a limit.

Even if by some miracle the models could be fed a 100% accurate starting position I doubt the models can ever represent the actual physics precisely.  

I don’t think its ’errors’ so much as as incomplete data, although that could be what you mean in a broad sense.   There are grid points for which there is no, or incomplete data, so these will have ‘educated guesses’ added to fill in the blanks.  If one grid point says -9c and -12c dew point and the one two points across has -11c and -13c what goes in the missing one?  Suppose the reality is materially different?

I didn't phrase it well. By 'errors' I mean that the initiation data can not possibly be an absolutely correct representation of the current state of the atmosphere, because of the lack of data. 

I’d like to see the MOGREPS suite respond a bit so on that basis I’d say an easterly of any sort is 50%, an easterly of substance around 20%. 

GFS Op - the above contest continues to Wed 21st with S-lies for Britain. At that point the HP is winning with the aid of LP in the Med, but collapses Sun 25th as LP forces its way into the N Sea. This area of LP moves S-wards and the battle resumes by Fri 30th, though further north with W-lies for Britain. Hopes of a dramatic E-ly from yesterday have disappeared. (GEM keeps the contest going to Sun 25th, though)

ECM - agrees with GFS to about Wed 21st but instead of a separate LP in the Med the Greenland LP extends a major trough to SW Britain and France by Sat 24th which resolves into LP sitting over Wales 980mb Wed 28th, 

The models, not unexpectedly, have difficulties with the boundary between HP in northern Europe and LP in N Atlantic. Expect further changes!

AIFS - London, maxima 10C now, dropping to 3-6C from Mon 19th and a further cooling after Mon 26th, mostly dry for the weekend, small amounts of rain at other times. Winds S-ly with bursts of E-ly at teh weekend and later. Edinburgh, maxima close to 5C (but with a dip this weekend), light rain at first, heavy and frequent after FRi 23rd.

GEFS - in the S heavy rain now, mainly dry until occasional rain after Sun 25th (but remaining damp in the SW), mean temp near norm for the coming week, then a very wide spread of ens members with a slight balance in favour of colder ones. In the N, both N England and Scotland, not much rain at any time, perhaps a little later on (different from AIFS). Snow row figures higher in a fortnight's time, but not dramatically so.

There were some issues between P1 and P16. I'm confident the downloads on TWO completed fully. It's possible some plots on other sites / apps were not the full set - it depends on how they schedule chart generation.