Model output discussion - 1st December onwards

[Paul_1978]

1 minute ago, northwestsnow said:

How on earth do i stop this from happening!!!! Every time i want to post on the thread it keeps making me quote a post from days ago arrrghghgh

Ask in the Help, Support and Feedback forum. 

[northwestsnow]

Thanks Paul  

 

[Daniel*]

59 minutes ago, JOPRO said:

One mans Outlier is another mans trend setter...! Am sure the 12hrs will be different again, this time showing a beasterly........

Well I urge folk to not lose their heads over an individual run through consistency and narrowing timeframe can we start to assert a prognosis as of yet this is nonexistent with wild swings in the model output. Patience and temper the emotion with the more obvious.

Edited December 8, 2015 by Daniel*

[sparky1972]

good afternoon everyone, i have been enjoying reading the posts on here regarding the search for cold and the lack of it showing on the models, i for one am a coldie and got a little excited after the weather outlook 16 day forecast for my postcode has us down for 6 days of snow from the 18th to the 23rd, that forecast used the GFS06z data, so i thought i would come here and see what other peoples views are and if it stands a chance a comming off, so far i would say 80-20 against, keep up the good work everyone and may the search soon be over, lol

[Methuselah]

11 minutes ago, Paul_1978 said:

Ask in the Help, Support and Feedback forum. 

It's a bug that we are all being annoyed with...

It's not just the usual run-to-run fluctuations that we need deal with, just now: the models' ensemble members are also going off on one post around 7 days...So, at the moment, I'd suggest that once the chaos sets in, the anomaly charts are probably as accurate as anyone can get?

So don't rule anything out.

[johnholmes]

Grief I make a post based on 500mb anomaly charts and all hell breaks loose.

For those not used to my posts in the 6-15 day time frame they give a good indication of MAJOR changes or NO MAJOR change. The do NOT and never have, in my use of them, given any detail in what 500mb charts or surface charts may shows within that 6-15 day time frame in detail, note the word detail. It is rare, less than 30-35% when they show no major change, be that a major difference to current charts or a marked change, that they are wrong.

This post perhaps sums up better than I have about them

John gave a fair summary of what the anomaly charts show but others have misinterpreted this. It is frustrating because time and again it has been said what anomaly charts are good for and what they are not so good for. They are good for giving an overall indication of the likelihood of a blocking pattern setting up but poor for picking out details within the longwave pattern - especially in a situation as complex as this where there is wide divergence between ensembles and often the Op and control runs are at odds at anything beyond day 5-7. 

I continue to post that there is little probability of anything more than 24-48 hours of deep cold air penetrating into all parts, into northern most areas more likely than southern areas. I could be wrong, for the majority seeking cold AND snow wherever they live, I hope I am. If I am I will as I always do be on here to hold my hands up and say I was wrong. I doubt I will be far out in the next 2 weeks. Beyond that is not my province to even speculate. I leave that to others.

So long as ALL of us feel confident in putting our views in print, preferably with charts to back this up, that others do not mis read or knock the posts then this forum gives everyone the chance to air those views. Simply respect every poster unless they are obviously trying a wind up. If you think they are don't react just press the report button and tell the team who will deal with it.

Happy viewing everyone.

 

Edited December 8, 2015 by johnholmes

[Paul]

36 minutes ago, northwestsnow said:

How on earth do i stop this from happening!!!! Every time i want to post on the thread it keeps making me quote a post from days ago arrrghghgh

As has already been said, please ask this sort of thing in the dedicated thread over in the help section (this being the model thread!) - you're way more likely to get your question answered over there, as that's what it's setup for, and better still - in many cases, a quick read of the thread or search for the answer will probably provide it.

See this post on how to remove the old quote, among other things.

https://forum.netweather.tv/topic/84553-forum-upgrade-info-bugs-support/?do=findComment&comment=3297018

[jvenge]

19 minutes ago, sparky1972 said:

good afternoon everyone, i have been enjoying reading the posts on here regarding the search for cold and the lack of it showing on the models, i for one am a coldie and got a little excited after the weather outlook 16 day forecast for my postcode has us down for 6 days of snow from the 18th to the 23rd, that forecast used the GFS06z data, so i thought i would come here and see what other peoples views are and if it stands a chance a comming off, so far i would say 80-20 against, keep up the good work everyone and may the search soon be over, lol

It is quite unlikely that the exact run will verify and it is much less than a 20% chance ;-) The GFS has been changing almost every run, even for my part of the World. Don't put too much stock into anything past 6 days at the moment :-)

Having said that, it is certainly possible, otherwise it wouldn't have shown ;-)

 

 

[mulzy]

1 hour ago, nick sussex said:

You lost me at the Gaussian distribution! lol But have looked at that now and yes I agree. A mean by itself is pretty useless, especially because it can mute out what the different clustering is suggesting.  We often see this divergence between operational run and ensembles when theres a change in pattern, generally what happens is one backs down, so either the ensembles will move towards the operationals or the latter will move the other way. In strongly zonal patterns a slight tweak in starting conditions within the ensembles often makes little difference however in complex patterns things can go wrong.

We often see easterlies poorly modelled by the ensembles and these will be slow to come on board, every synoptic pattern needs to be viewed in a different way.

Generally if cold spells are suggested I view things as what can go wrong to decide on how many hurdles need to be jumped. For this reason any type of Scandi high scenario has a code red warning!

 

Hi Nick - we have always had the odd operational 'going off on one' so I think it's equally dangerous just to rely on a few operational runs.  The background signals all point one way -  the fact is that there is unlikely to be any significant cold weather or even cool weather in the near term for the vast majority.  The Xmas period is up for grabs but here in London we're getting 15C on a regular basis (just checked that the CET is at an incredible 10.6C after the first 7 days) - all this talk of potential cold and yet at the same time we're slowly sleep-walking into the warmest December ever recorded...

 

 

Edited December 8, 2015 by mulzy

[johnholmes]

one way of checking how consistent any model is in the far reaches of its run, F1 as it is termed as for you to check each run each day for the same time of issue and see if it shows a similar pattern for the same day say initially at 10 or 12 days ahead. Then see how it compares as you count down the days. Obviously by the time, usually (!), we get to T+48 all the models at any time of issue will be converging on a similar solution, usually (!). It will work at long time scales better than checking each run against the last AT LONG TIME intervals NOT below about 120-144 hours though.

Give it a try if you have not before, save each run and compare, this site allows you to do it easily for the current and last run, but not I don't think for anything further back. Try this web site in that case

http://www.weatheronline.co.uk/cgi-bin/expertcharts?LANG=en&CONT=euro&MODELL=gfs&VAR=prec

[chrisbell-nottheweatherman]

45 minutes ago, johnholmes said:

Grief I make a post based on 500mb anomaly charts and all hell breaks loose.

For those not used to my posts in the 6-15 day time frame they give a good indication of MAJOR changes or NO MAJOR change. The do NOT and never have, in my use of them, given any detail in what 500mb charts or surface charts may shows within that 6-15 day time frame in detail, note the word detail. It is rare, less than 30-35% when they show no major change, be that a major difference to current charts or a marked change, that they are wrong.

This post perhaps sums up better than I have about them

John gave a fair summary of what the anomaly charts show but others have misinterpreted this. It is frustrating because time and again it has been said what anomaly charts are good for and what they are not so good for. They are good for giving an overall indication of the likelihood of a blocking pattern setting up but poor for picking out details within the longwave pattern - especially in a situation as complex as this where there is wide divergence between ensembles and often the Op and control runs are at odds at anything beyond day 5-7. 

I continue to post that there is little probability of anything more than 24-48 hours of deep cold air penetrating into all parts, into northern most areas more likely than southern areas. I could be wrong, for the majority seeking cold AND snow wherever they live, I hope I am. If I am I will as I always do be on here to hold my hands up and say I was wrong. I doubt I will be far out in the next 2 weeks. Beyond that is not my province to even speculate. I leave that to others.

So long as ALL of us feel confident in putting our views in print, preferably with charts to back this up, that others do not mis read or knock the posts then this forum gives everyone the chance to air those views. Simply respect every poster unless they are obviously trying a wind up. If you think they are don't react just press the report button and tell the team who will deal with it.

Happy viewing everyone.

 

 

I honestly think this needs to be quoted whenever Mods open a new Model Discussion thread.  I for one as an inexperienced model followr who relies on knowlegeable people's analyses respect John's posts and try to view them for what they are.  It seems at present, based on his analysis that the anomalies are showing no movement towards prolonged cold, but there is still every chance of a short-lived colder spell between lows, with the trend being for a more southerly mean jet location.

Edited December 8, 2015 by chrisbell-nottheweatherman

[Rocheydub]

4 minutes ago, knocker said:

Given, looking at various posts this today, many seem to think the EC32 summary more than useless I shall desist in the future. I can then concentrate on rambling about the current ops output.

Can I ask please, that you do continue. I for one, find the insight very interesting. It also does sometime, pick up signals for changes in weather patterns, and keeps my mind sharp looking for these changes as they near the short and medium range NWP. I certainly value this input Knocker.

[Iceberg]

Talk of sleep waking into the mildest Dec ever is silly IMHO 

interesting 06z ens

[Iceberg]

Indeed the EC32 has its place I commented that "Re anom and EC32, you have to remember they arn't really picking up any signal, so its normalised, confirmation bias. i.e temps around average with a few milder and colder bits."

its very useful to see the ec32 summary, but has more use when it does pick up a strong signal (same with CFS but to a lesser degree imho)

[knocker]

There was at least one quote but then we had the ever returning subject regarding the means which I think must be scheduled on a repeat loop. The summary I use is from the ECMWF EPS monthly ens means

Edited December 8, 2015 by knocker

[Nouska]

34 minutes ago, mulzy said:

Hi Nick - we have always had the odd operational 'going off on one' so I think it's equally dangerous just to rely on a few operational runs.  The background signals all point one way -  the fact is that there is unlikely to be any significant cold weather or even cool weather in the near term for the vast majority.  The Xmas period is up for grabs but here in London we're getting 15C on a regular basis (just checked that the CET is at an incredible 10.6C after the first 7 days) - all this talk of potential cold and yet at the same time we're slowly sleep-walking into the warmest December ever recorded...

 

 

The background signals will be almost identical at the start of the run but what a model then does is based on how it interprets these background signals - heavily skewed to tropical forcing at this time!

GFS is looking at a very strong Nino response (neutral MJO) through the tropics whereas the ECM and UKMO see a signal for amplified progression into phase 4 - this translates to a signal to build heights through Scandinavia and have a trough digging into SW Europe.

The extended modelling from both the ECM and GFS is typical of the tropical forcing they anticipate - they cannot both be right - need time to let them settle on a consensus before we see any stability in what lies ahead towards Christmas.

Lot of discussion on the mean but funnily enough, so far this late autumn/early winter, the ECM extended mean (10 to 15 days) is actually pretty close to what we actually see when verification time arrives.

[chrisbell-nottheweatherman]

9 minutes ago, knocker said:

There was at least one quote but then we had the ever returning subject regarding the means which I think must be scheduled tin a repeat loop. The summary I use is from the ECMWF EPS monthly ens means

 

If you're referring to the conversation I was involved in, I think all we were saying is that a mean average is only as useful as the data from which it emerges allows it to be.  Combining my generic understanding from three years of undergrad biology, plus some practical experience in a masters' programme with the specific meteorological knowledge from others, I get the understanding that periods of widely-diverging ensemble members when the current pattern seems uncertain in terms of whether it will continue as it current is presents typically-non-Gaussian (i.e. unequally-distributed) data.  Such distribution makes mean averages without standard deviation information less powerful as it will tend to show a result between clusters of data as opposed to representing the concentrations of individual data points created by the clustering of the data.

Edited December 8, 2015 by chrisbell-nottheweatherman

[knocker]

19 minutes ago, chrisbell-nottheweatherman said:

 

If you're referring to the conversation I was involved in, I think all we were saying is that a mean average is only as useful as the data from which it emerges allows it to be.  Combining my generic understanding from three years of undergrad biology, plus some practical experience in a masters' programme with the specific meteorological knowledge from others, I get the understanding that periods of widely-diverging ensemble members when the current pattern seems uncertain in terms of whether it will continue as it current is presents typically-non-Gaussian (i.e. unequally-distributed) data.  Such distribution makes mean averages without standard deviation information less powerful as it will tend to show a result between clusters of data as opposed to representing the concentrations of individual data points created by the clustering of the data.

Only partly as it's a well rehearsed subject.

I do emphasise that the summary is just a broad overview of the EC32 output as I certainly do not have access to the full suite which obviously will give much more detail. That being said I still think it gives a good idea of the trend for the coming month. Two things I take as read. One is that variations around the theme will obviously occur due to the day to day variations in atmospheric dynamics so it will encompass Pm incursions and periods of more settled weather within which there will be temperature swings. Two, it is not written in stone and pattern changes are still possible.

That being said using it in conjunction with other indicators that are available is always a good idea.

 

Edited December 8, 2015 by knocker

[chrisbell-nottheweatherman]

46 minutes ago, knocker said:

Only partly as it's a well rehearsed subject.

I do emphasise that the summary is just a broad overview of the EC32 output as I certainly do not have access to the full suite which obviously will give much more detail. That being said I still think it gives a good idea of the trend for the coming month. Two things I take as read. One is that variations around the theme will obviously occur due to the day to day variations in atmospheric dynamics so it will encompass Pm incursions and periods of more settled weather within which there will be temperature swings. Two, it is not written in stone and pattern changes are still possible.

That being said using it in conjunction with other indicators that are available is always a good idea.

Fair points.  You're much more knowledgeable than I am on the meteorological side of things, so I'm happy to defer to you on that.  My point is that there's a difference between calculating a mean on the one hand, and using it as a trend guide on the other.  If we had a group of people at a conference or public lecture, for example, we could calculate their mean height.  If the distribution of their heights was a standard continuous variation, i.e. a normal or Gaussian distribution, the curve, where heights on the x-axis are plotted against number of persons on the y-axis, would be something like this:

 

In this case, we might expect the mean value to give us an idea of the trend regarding the height of attendees.  However, if the curve is closer to this:

then we have a situation where the mean will lie between the two clusters (in my example, it might suggest that the trend is for attendees to be around 5 feet 9 inches in height, where, in actual fact, most are between either 5 ft 2 and 5 ft 6 or 5 ft 11 and 6 ft 3).

Applying this to the current model situation, my understanding (and please correct me if I haven't grasped this properly) is that the high levels of Shannon entropy as a result of changing underlying patterns and drivers means that the ensembles are tending to fall into one or two (or more?) clusters in terms of output.  As a result, while an ensemble mean is still an ensemble mean, the trend it suggests my lie between clusters of members and may therefore infer a trend which doesn't exist (perhaps showing a solution which is neither one thing nor the other when the individual ensemble members tend to fall into one definite trend or the other).

Apologies Mods for the off-topic diversion.

Edited December 8, 2015 by chrisbell-nottheweatherman

[Ali1977]

well the next few frames will tell us if the 12z will follow the 06z, already looking iffy but we'll see!!

[Mucka]

One thing that has been fairly consistently modeled and pretty well documented by some here is the very high risk of more flooding from the stalling low into next weekend and early next week

 

[Iceberg]

Jet stream going to struggle to get to Europe at this rate. 

That's all I want to see at t144 so i am happy now regardless 

[Gavin Hannah]

At 1.135. It does seem that HP off the west is more pronounced than in the 06Z and occupying more space inbetween the lows.

t.150 and that low in the Atlantic is centered much further south than in the 06Z. Could be a key factor in the FI portion of the run.

Edited December 8, 2015 by Gavin Hannah

[radiohead]

Staggering difference between the UKMO and GFS over Canada at T144

[Mucka]

Nice looking UKMO this afternoon at 144h but it has performed poorly at that range in recent times, at least that is my perception.

 

 

We would likely get a good Atlantic ridge out of that and possibly a small cut off high somewhere N of the UK

Any blocking that did take hold looks likely to be reinforced by the upstream pattern - if only that were close to the actual...

 

Nice looking GFS and pretty good consistency today with the overall pattern considering what has gone before, it is just playing with the Eastward progression and amplification.

12z/06z/00z

 

Edited December 8, 2015 by Mucka