Model Output Discussion - cold spell to end Autumn

[chrisbell-nottheweatherman]

4 minutes ago, northwestsnow said:

Not sure its unlikely chris- infact i'd say its a toss of a coin.

Apologies - I was referring to the 6th and 7th in the short ensembles.  Thereafter the op and control are back to being well within the ensemble scatter.

[bluearmy]

11 minutes ago, Ed Stone said:

Thanks for pointing that out, FG; it explains why I seldom if ever pay any attention to ensemble means...in many cases, they simply don't mean anything. Pardon the pun!

This is where I get confused

looking at the ens mean and op for day 8 and the upper pattern looks pretty much the same. If you are going to use ens forecasting for placement of highs and lows then you are going to struggle. clearly on that mean we could be the warm side or cold side of the flow but that's what the mean shows you. The number of clusters and spread tells you the probability (which will change, run to run),  but to say the mean is of no use is surely wrong ???

     

[Day 10]

19 minutes ago, chrisbell-nottheweatherman said:

Op and control are outliers which suggests that the second northerly is unlikely.  Given the low ppn. signal perhaps a UK or Euro high is the more likely set-up?

Are they? They look to have support to me, however it's 7 days away that...

[chrisbell-nottheweatherman]

2 minutes ago, Day 10 said:

Are they? They look to have support to me, however it's 7 days away that...

https://www.netweather.tv/forum/topic/88883-model-output-discussion-cold-spell-to-end-autumn/?do=findComment&comment=3659224

Edited November 29, 2017 by chrisbell-nottheweatherman

[Methuselah]

2 minutes ago, bluearmy said:

This is where I get confused

looking at the ens mean and op for day 8 and the upper pattern looks pretty much the same. If you are going to use ens forecasting for placement of highs and lows then you are going to struggle. clearly on that mean we could be the warm side or cold side of the flow but that's what the mean shows you. The number of clusters and spread tells you the probability (which will change, run to run),  but to say the mean is of no use is surely wrong ???

     

But I haven't said that it's 'no use' BA...just that it is, in some cases, wholly unrepresentative of what's being suggested. Should a run consist of only two distinct clusters (one at, say, -10C; the other at +10C, for example) what does the mean (0C) really tell you about upcoming events?

[karlos1983]

some crazy cold runs in there!!!

And some utterly disgusting mild ones it has to be said. Majority at D15 (yes I know, miles away) are sub 5c 2m Temps

ECM loses the plot around 7th Dec, Basically when it blows up that low to our west. Good news. 

[Gavin Hannah]

Just now, Ed Stone said:

But I haven't said that it's 'no use' BA...just that it is, in some cases, wholly unrepresentative of what's being suggested. Should a run consist of only two distinct clusters (one at, say, -10C; the other at +10C, for example) what does the mean (0C) really tell you about upcoming events?

In short, it's a Best Guess scenario.  If the Op and Control are on the same / similar page, would one not prefer to follow the op / control rather than any of the members?

[knocker]

18 minutes ago, Ed Stone said:

But I haven't said that it's 'no use' BA...just that it is, in some cases, wholly unrepresentative of what's being suggested. Should a run consist of only two distinct clusters (one at, say, -10C; the other at +10C, for example) what does the mean (0C) really tell you about upcoming events?

That the clusters are outliers and should be ignored Way to go Sidney

Edited November 29, 2017 by knocker

[Methuselah]

3 minutes ago, knocker said:

That the clusters are outliers and should be ignored Way to go Sidney

Well the milder of two obviously is!

[Man With Beard]

Closer to the now, Thursday needs keeping an eye on for many. Latest EURO4 has snow coming in quite some way from the east coast.

      

[knocker]

4 minutes ago, Ed Stone said:

Well the milder of two obviously is!

I'm afraid I have no wish to discuss this with you further since you have gone over to the dark side and joined the forces of evil.

Edited November 29, 2017 by knocker

[jvenge]

29 minutes ago, Ed Stone said:

But I haven't said that it's 'no use' BA...just that it is, in some cases, wholly unrepresentative of what's being suggested. Should a run consist of only two distinct clusters (one at, say, -10C; the other at +10C, for example) what does the mean (0C) really tell you about upcoming events?

I'm generally of the opinion that means are useless unless you can see how they were made up. We have posters giving more insights into ECM members than we had in previous years.

How I personally use the ECM:

1. Check Operational.

2. Drink a nice green tea if its the 12z, or a coffee if its the 0z.

3. Check the ECM Control. 

4. Check the ECM EPS Means and try to figure out where the Op and Control sit, while trying to then guess what is a majority solution showing.

The further you go out, the more pointless the means become. This is especially evident on the EPS 46 after day 15. Prior to then, though, you can sometimes figure stuff out from them. 

I don't think you can forecast from ensembles or means, but you can use them to favour a certain evolution and then if you are lucky, use the operations or controls to forecast, if you really had the desire to even try :-)

Edited November 29, 2017 by jvenge

[bluearmy]

34 minutes ago, Ed Stone said:

But I haven't said that it's 'no use' BA...just that it is, in some cases, wholly unrepresentative of what's being suggested. Should a run consist of only two distinct clusters (one at, say, -10C; the other at +10C, for example) what does the mean (0C) really tell you about upcoming events?

Those charts are for the timescale where the bi polar clusters on uppers exist. I would say the mean chart is a pretty good guide of where we are at that point - uncertain !!!!!!

and I would add that at some point beteeen east and west the upper temp  has to be what the mean is showing. The current distribution of members had two distinct clusters but that doesn’t mean that we don’t end up verifying inbetween which would be close to the mean. 

Would be interesting if we could get gefs clusters on Meteociel using their colour refs so we could compare in these instances.

100% agree with post above re timescales - without anomolies the mean charts generally become of little visual use the further out they go as they tend to revert to climatology. Of course where we see distinct mean features beyond day 8 on the mean output we should take notice of them !!!

Edited November 29, 2017 by bluearmy

[Methuselah]

1 minute ago, bluearmy said:

Those charts are for the timescale where the bi polar clusters on uppers exist. I would say the mean chart is a pretty good guide of where we are at that point - uncertain !!!!!!

and I would add that at some point beteeen east and west the upper temp  has to be what the mean is showing. The current distribution of members had two distinct clusters but that doesn’t mean that we don’t end up verifying inbetween which would be close to the mean. 

Would be interesting if we could get gefs clusters on Meteociel using their colour refs so we could compare in these instances.

What charts? I was merely inventing an hypothetical situation so as to make my point clear(er)...

Another miserable failure on my part.

 

[Man With Beard]

2 minutes ago, bluearmy said:

Those charts are for the timescale where the bi polar clusters on uppers exist. I would say the mean chart is a pretty good guide of where we are at that point - uncertain !!!!!!

and I would add that at some point beteeen east and west the upper temp  has to be what the mean is showing. The current distribution of members had two distinct clusters but that doesn’t mean that we don’t end up verifying inbetween which would be close to the mean. 

Would be interesting if we could get gefs clusters on Meteociel using their colour refs so we could compare in these instances.

Not quite "Meteociel"d, but GEFS clusters available on Weatheronline:

http://www.weatheronline.co.uk/cgi-bin/expertcharts?LANG=en&CONT=euro&MODELL=gefs&MODELLTYP=2&VAR=cpre

I would post the individual clusters but unfortunately I'm not techy enough

[ArHu3]

22 minutes ago, jvenge said:

I'm generally of the opinion that means are useless unless you can see how they were made up. We have posters giving more insights into ECM members than we had in previous years.

 

Means are not completely useless, when it sees something it will probably be there but when it doesn't see anything it doesn't mean there isn't anything there

[bluearmy]

16 minutes ago, Ed Stone said:

What charts? I was merely inventing an hypothetical situation so as to make my point clear(er)...

The ones I posted to illustrate that the mean at the time of heightened polarity is pretty retesentitive of the op which was in the cold cluster 

[phil nw.]

Clearly as most of us know by now the split jet coming off NA in the next few days is creating differing outcomes wrt Atlantic ridging and the subsequent pattern for next week.

Day 4

We have seen a lot of this in the last few weeks because with ridging- as opposed to an independent high to the north of the jet- some energy goes over the top.

While we are in this repeat pattern of the ebbing and flowing of the Atlantic ridge with +ve ht anomalies over the Arctic then this set up will continue teasing us for a while yet as we wait to see if we get enough Atlantic amplification to bring us another northerly.

A look at the 06Z gef clusters for this time next week show a pretty even split for cold v mild

http://www.weatheronline.co.uk/cgi-bin/expertcharts?LANG=en&MENU=0000000000&CONT=euro&MODELL=gefs&MODELLTYP=2&BASE=-&VAR=cpre&HH=180&ZOOM=0&ARCHIV=0&RES=0&WMO=&PERIOD=

There are 4 different clusters in the 06Z Gefs and by this time next week they are split pretty evenly cold v mild.Looks to me like 2 are for the northerly(40%)1 goes mild retaining the high to the south with a south westerly flow (30%)and 1 cluster with a battleground setup(30%).

It's all very finely balanced as the cliche goes but if we are to get the northerly reload next week we don't want to see any further dampening down of that regressing ridging around days 4/5  as this would inevitably mean more jet energy leaking north.

Edited November 29, 2017 by phil nw.

[Seasonality]

In the past there were one or two folk on here who used the 384h GFS mean to show what would be happening in a fortnight's time and usually to support their own weather preference. Problem is that at that range the spread is usually such that the mean charts are useless, a muted representation of a whole suite of broadly varying 384h outcomes, and we know how accurate any model is at that range (example from today's 06z below).Glad to see it isn't being done this year and that more considered analysis is being undertaken. I prefer to look for clustering in the more reliable time frames and the position of the op and the control within the suite to make an analysis. I also like the function on weathercast that shows a nice little graph of ECM and GFS side by side.

 

[londonblizzard]

Can someone explain to me (with my very limited knowledge) why the GFS op and control have a higher degree of importance than the any of the individual GEFS members? I.e. why is the op or control not just another pertubation from this set with no more significance than any of the others? 
I know this will probably sound like a stupid question to most of you. 

[knocker]

Interesting that Michael Ventrice tweeted the EPS 11-15 mean earlier in order to make this point

Quote

An impressive tropospheric ridge is forecast to over take the North Pole during mid-December. This will shove a lobe of the Polar Vortex o/ E. Asia and the eastern U.S.. Expecting a high wavenumber pattern to emerge around the N. Hemisphere. Encourages extreme weather worldwide.

Thus demonstrating that respected professional meteorologists do use the mean as one of the useful tools in the box. If it is good enough for them then it is certainly good enough for me, Unfortunately he only posted the anomaly so I will add the GEFs which does not show some ridging in mid Atlantic as does the EPS although the latter does lose it at the end. What I will say is there does seem a consensus on this quite extreme upstream amplification

Edited November 29, 2017 by knocker

[❄️❄️❄️❄️❄️❄️❄️❄️❄️]

Examples and steps are lower than winter first ... So I think there is a chance to have another goal during the month. Gefs and Eps are updating the stratosphere related to the second part of November 3 and good ideas from Wave 1 are rising briefly in the next few days. This underscores weakened weaknesses and causes stratosphere near Alaska. These include sunflower model and red polar. At the same time, we have a large hole in the North Pacific Ocean and north of the North Atlantic to prepare. Miracles of miraculous use are the finest appearance found in late September and October (ie restoration) and slow movement (second sign, big sign). It is important to look at the amount of cold fever that can be shattered in the northwest.

Edited November 29, 2017 by RAIN RAIN RAIN

[Cloud 10]

6 minutes ago, londonblizzard said:

Can someone explain to me (with my very limited knowledge) why the GFS op and control have a higher degree of importance than the any of the individual GEFS members? I.e. why is the op or control not just another pertubation from this set with no more significance than any of the others? 
I know this will probably sound like a stupid question to most of you. 

 

The GFS op and control run are run at a higher resolution than the other GEFS members,so in theory they should verify better.

[Man With Beard]

1 minute ago, knocker said:

Interesting that Michael Ventrice tweeted the EPS 11-15 mean earlier in order to make this point

Thus demonstrating that respected professional meteorologist do use the mean as one of the useful tools in the box. If it is good enough for them then it is certainly good enough for me, Unfortunately he only posted the anomaly so I will add the GEFs which does not show some ridging in mid Atlantic as does the EPS although the latter does lose it at the end. What I will say is there does seem a consensus on this quite extreme upstream amplification

Michael's tweet in no way contradicts my take on the means:

When there is a large anomoly on a mean chart, it is very useful. It could indicate an even more extreme median, since any outlier ensemble members closer to the norm will take the edge off the anomaly

When anomalies are small, it is less useful as ensembles could be clustering more tightly, or there could be a large split either way.

[winterof79]

8 minutes ago, RAIN RAIN RAIN said:

Examples and steps are lower than winter first ... So I think there is a chance to have another goal during the month. Gefs and Eps are updating the stratosphere related to the second part of November 3 and good ideas from Wave 1 are rising briefly in the next few days. This underscores weakened weaknesses and causes stratosphere near Alaska. These include sunflower model and red polar. At the same time, we have a large hole in the North Pacific Ocean and north of the North Atlantic to prepare. Miracles of miraculous use are the finest appearance found in late September and October (ie restoration) and slow movement (second sign, big sign). It is important to look at the amount of cold fever that can be shattered in the northwest.

Perhaps another visit to the "edit" function may be an option