You are probably right but when the runs are changing so much, currently a Scandi high setting up on this run

Just for contrasting purposes see below:

06z GFS Op run: - Low pressure for west to east and a westerly wind and mild airflow not a Scandi high in sight 

12z GFS Op run - High pressure to our north east spreading westwards - (easterly wind). with -10c 850's crossing country.

Will be interesting to see where this sits in the ensembles 

The tease continues🙄

Same comment though as with the more atlantic focussed op solutions - will see what the ens say is more likely. 

Interestingly in the medium range of the 12z, the ENS mean is -3.0c @ 850hpa on 18th/19th and the OP run has a mean of +0.5 to 0.7c so the Op is on the milder side in medium range.

Remember one flake of sleet is enough to trigger the snow hatchings with GFS... 😉

I’m sure you’re pleased to see it was not an outlier.

Expect updated thoughts soon 🙂

Indeed. You’d think it would be better than this though. Remember Gooner’s JFF snow charts that never came to fruition for this very reason 

I think it's better for continental locations than the UK. 

Assuming we are looking at London GFS 850 hPa ensembles then it is right in the middle of the average till the 22nd, the only time when it goes cold, then it is a clear outlier, along with possibly one other. 

But that is meaningless by then anyway.

Yes indeed, it's an outlier however post day 12 however that's too far out to worry about. I think all we can say about the 12z is that it hasn't been helpful at all in resolving the uncertainty.

You appear to be confusing the idea of an outlier with simply one of the coldest options. it is not unsupported at any stage of the run, although it is the coldest or one of the coldest in latter stages of the run. 

As has been the case for days now the only clear thing about the outlook is that weather from an easterly quarter continued to look likely in the short to medium term.

So the GFS op isn't at *any* stage an outlier in terms of 850 temps. It does though have very little support in the latter stages of the run, with just one other member bringing similar temps at 850 level. (That's P24 I think and it's a thoroughly entertaining chap). 

In terms of the GEFS I don't think the 12z set brings any clarity or anything different to preceding sets - milder 850s probably more likely than cold but with significant uncertainty for much of the run until deep FI where milder is more likely. A lot of the milder 850 solutions are also still fairly chilly at the surface, at least until 21st Feb anyway. 

The ops are generally suggesting a less cold solution at 850 level is more likely, certainly in the mid term.

Must admit like Ballamar I thought the ECM initially looked a bit more promising than this morning but it flattered to deceive and goes close to full Atlantic in deep FI. 

Edit: What Michael said, really must take less time putting posts together🙃

Thank you for making that point - yet again.

It’s a word that keeps being misused, generally by the same people and with the same purpose.

It’s interesting to look at different parameters, and different locations, and discover a different picture.  I noticed that, for example, for London for the Op the height of the 0c isotherm was at the lower end of the spread, in contrast to the 850hPa picture.  I am assuming that just confirms there’s a continental feed with a cold lower layer - supported in turn by the 2m temperatures.

I don't know if anything has changed in recent years but mathematically an outlier was usually defined as a point in a data set that is more than 2x the standard deviation from the mean given normal distribution. We used to use more sophisticated formula from time to time when analysing critical analytical data however and this isn't classic normal distribution. In the case being discussed it can certainly be more than one point that meets those criteria. I imagine the points in question, though they are well out on their own visually, may not be outliers given the large sd at that point. However, unless somebody has the sd figures for that data set (maybe they do?) neither party can claim whether it is or isn't an outlier. What we do know is that points further from the mean are less likely to occur.

Thanks. I think the point here is that some people use the term ‘outlier’ with the intention of saying “it’s not supported by the ensemble suite and it’s not going to happen”, not in any statistical sense.  I tend to look at the ‘box’ version of the ensembles on Meteociel, which give the median, quartile and upper and lower deciles.

The SD is freely provided for the ECMWF ENS and GEFS.

Have you got a link to that version, it would be useful, thanks. It's not really worth even getting into the statistical semantics at that stage because even if it is an outlier (and some rough maths tells me it probably isn't but on the cusp between 22nd and 23rd) it's too far out for us to judge. It is what it is nevertheless, a very cold point that statistically has less chance of happening than about 95% of the other points on that date! But then I didn't win anything on the premium bonds this month which was about as statistically unlikely so you never know!!