SomeLikeItHot

Yes I agree. Personally I think the 6z and 18z should be regarded more as ensemble members rather than individual runs.

We don't mathematically and scientifically prove the best model by looking at what happened in a measure over a few days, but by long term stats, and the ECM consistently has beaten GFS over the years. What about the 5 & 6 days stats though, which is probably where there are real and significant divergences at the moment (and have vastly better confidence) here the ECM is better.

19th of Feb is of course well past the end of the ECM 240 hour run. And yes there is huge uncertainties out past 15days. The main differences at this point between GFS and ECM are +120 hours. Relevant part of the UKMO guidance with regard this is UK Outlook for Saturday 9 Feb 2013 to Monday 18 Feb 2013: Most areas remaining on the cold side this weekend. Some bright spells for many areas, but also some wintry showers in some North Sea coastal areas. Meanwhile parts of western and southwestern UK are likely to be cloudier with some rain, sleet and hill snow, especially later this weekend. Then, next week western parts of the UK are likely to continue to see spells of cloudier weather with rain, hill snow and, perhaps, some snow temporarily to lower levels. Elsewhere conditions will tend to be colder and brighter, though some parts may catch a few wintry showers. Looking ahead to the weekend after next and the start of the following week, conditions most likely to be a little colder than average, with northwestern parts most at risk of unsettled weather.

Much to like about the last 100 hours of GFS FI.

I'd be surprised if they didn't look at it even if its a secondary consideration, (edit: ah I see Gavins post that seems to be exactly what they do) but my point wasn't about the gefs in particular, but what ensembles are likely to be showing in general with the gefs as an example that I can actually see. - a majority of cold members but a decent minority of mild ones meaning that if you had to nail your opinion to a post as one estimate you say something like likely below average rather than likely wintery and be totally wrong if it turns out to be mild.

With regards to the Meto update for next week if you look at Ensembles then there is something like 5 vs 12 split for mild versus slightly wintery next week with a couple in the middle. Given they give just a single point outlook (ie below average) obviously they are hedging their bets when there is still a range of outcomes. It looks something like 25% chance of mild, 60% cold and 15% of cool. I'd prefer if they came out and said something like those probability estimates but they don't, so you just get wishy washy "cooler than average with mild outbreaks" unless there is a really strong signal.

Not sure who said don't look, but surely the point is don't believe everything you see uncritically and exercise some judgement about whether the pattern would have to change much to have a big effect on experienced weather. We've seen models recently do big changes in 72-96hrs, at least in terms to how they effect the UK. But other features like the band of rain and warm air I experienced today/ last night were well flagged a week in advance, even if the exact details wasn't nailed down. If multiple models and multiple runs, ensembles, etc all agree on a feature at 168hrs and its not marginal whether it will effect the UK or not then I think you can be fairly certain something like it will occur at that range. Other times there isn't agreement at that range or its marginal so you pull in your confidence to a shorter time frame.

GFS 6z seems to end nicely doesn't it, highs in Greenland, mid atlantic, and HP pushing down into Scandi with northerlies looking to turn to easterlies as the low moves away from scandi. Gets there for the last few frames also, not just in the last one. FI but nice to see its possible.

AS I understand it the storm doesn't show on all ensembles or models (didn't show on the 0Z) and the GFS Op is towards the most extreme end of models when it does show so they are probably waiting to get a better fix on it.

Yes the mean is certainly above the majority of members due to a couple of very high near +10 solutions. The mean can be a poor measure if you have a bi-modal distribution. When people are saying there is no cold they are talking about getting the synoptics for prolonged cold not 1-3 days shots of cold interspersed with mild temps which seems to be mostly what is showing. I think some ensembles and the OP got to a blocked synoptic in the end so there is some hope for all. Personally I am happy with a bit of cold and maybe a snow shower or two, but then again I lived most of my life in Sydney so I get excited at a light frost. Just so long as we don't get 13 degrees and raining, as I could have done that back home.

Amazing difference between the GFS 0Z and 6Z at 90hrs with that low being non existent on one and intense on the other.

Quite so. The outcome over the last few weeks is regularly well outside the range of Ensembles once we go into FI. Ensembles are great for range of variation in close but at 240 hours? What they can tell you is how sensitive your outcome is to variations in initial conditions, but they will reflect all the other short comings as the Op runs. Eg. if the Op run doesn't accurately take into account some background influence be it SSW or hyperactive butterflies then this also won't be reflected in the ensembles.

Some excellent posts there SB. A couple of points, the term "Chaotic" in mathematics has a fairly clear meaning and doesn't mean that there is no discernible pattern, but rather that the dynamical system is highly sensitive to initial conditions - exactly what we see when ensembles diverge rapidly after a few days from the initial conditions. As for the question of a ceiling, it seems rather than a ceiling that progress will become asymptotically more difficult as time goes on. There are effectively mathematical limits to how far forward you can meaningfully model non-linear dynamical systems in a truly deterministic sense, but that doesn't mean that we can't obtain useful statistical information beyond this point which is of course entirely the point of the ensembles. I would not be surprised if greater knowledge of various dynamics and linkages results in better statistical forecasts in the med-long term, while not greatly increasing the range of accuracy of deterministic short range forecasts.

Again it looks as if the 06Z GFS pushes the block breaking later. Its been consistent showing the block breaking down 5-7 days later than the run

GFS is nothing if not consistent. Consistently saying for the last week that the block will last for another 120-144 hours and then collapse. It will eventually be right I am sure, but the question is when?

Been varying between snow/sleet/rain light-moderate in City of London now for about 40 minutes. currently moderate snow.