Jump to content
Sign in to follow this  
Captain Shortwave

Model Output Discussion - The final stretch to Christmas

Recommended Posts

Yes we will agree to disagree but looking at the D13 means, GFS P -v- GEFS it really does show how out of kilter the P is:

 

attachicon.gifgfs-0-312.png attachicon.gifgens-21-1-312.png

 

Is there a reason you post means at such a long range? It's just that I find them useless and I was wondering what I should be looking for in them. As far as I can see, any isolated long range mean compared to an actual chart, does is flatten out what might exist towards the norm.

 

If you showed the mean moving from day to day then you can see movement but I'm at a loss as to what there is of worth that you want me to see.

 

Could you possibly explain as to what I'm missing?

Share this post


Link to post
Share on other sites

Anyone got a link for ECM De Bilt esembles?

 

Nice GFS parallel this morning, but looks rather on it's own. ECM not great, neither is the GFS (OP) and the ukmo is so so...

Share this post


Link to post
Share on other sites

Hi yes I agree that the trend has been zonal for the GEFS for a few runs, but there were still a cluster showing potential in the Pacific region. This morning that is now noise with that baton seemingly dropped. Of course it could all change later and we need 2-3 runs with that signal before we get too worried. but with the CFS churning out consistent zonal means for January it is worth noting.

As for cold without an SSW, that is possible, but a sustained block without one would be a long shot.

 

For balance: both GFSP and JMA continue to model a significant Alaskan high ridging right through to the pole.

 

JN192-21.GIF?21-0

 

gfsnh-0-192.png?6

 

 

and this sits rather well with strat forecasts that - this morning at least - are sending most of the vortex energy back over to Siberia.

 

NH_HGT_10mb_192.gif

 

This is all good news really, though not as instant or as dramatic as was hoped for. Longer term strat modelling is suggesting an elongation of the vortex as it is hit with a sustained top down wave 1 attack. A long way to go before we know exactly the axis of that elongation and whether it might encourage height rises over greenie or scandy... but not all is doom and gloom this morning, that's for sure.

Edited by Catacol_Highlander

Share this post


Link to post
Share on other sites

The GFS parallel "official" release has been put back another week. Hardly surprising!

Share this post


Link to post
Share on other sites

Is there a reason you post means at such a long range? It's just that I find them useless and I was wondering what I should be looking for in them. As far as I can see, any isolated long range mean compared to an actual chart, does is flatten out what might exist towards the norm.

 

If you showed the mean moving from day to day then you can see movement but I'm at a loss as to what there is of worth that you want me to see.

 

Could you possibly explain as to what I'm missing?

 

The ops are the best for up to D6 IMO. But always worth comparing to the mean at that time as well. You can see the mean of 2m temps for this weekend and compare it to the GFS P:

 

post-14819-0-07461600-1419249759_thumb.g

 

P around 7-8c higher than the mean where as the OP and Control are close to the mean. So in all probability P is outside the SD.

 

The verification of means -v- op is between 30-50% better from around D10 so if you are going to forecast from that period over the long run you will be more successful using the means. I use them with clusterings, 8-10 and 10-14 day anomalies and try to get a pattern and trend. They are not perfect but better than using the ops for sure.

Share this post


Link to post
Share on other sites

Can anyone call anything after +120 though, I've never seen such chaos at such short timescales. There's some conflicting signal that is causing such madness within the output, just what though I've no idea as it can't just be down to the track of the low on the 27th as we've seen plenty of those features over the years.

It's the movement and form of the polar vortex that is the root cause of all the model mayhem and flip flopping.

Watch the progress of the PV and it's lobes up to 96 hours and you can get a better read on where the models will go.

As we stand the PV is situated at the Eurasian side of NH.

Models keep trying to find the right placement hence the variation on the runs

Edited by KyleHenry

Share this post


Link to post
Share on other sites

It will be interesting to see how the GFS compares to the GFSP in verification of accuracy with some wild deviation in  the operational beyond T+180.  ECM spread is still significant for 27th-29th cyclogenisis with many possibilities in tracking, complex baroclinic secondary development, cylcone phasing and attendant cold pooling.  The last two GFS runs have been similar with a less intense cyclogenisis across the United Kingdom, peak cyclogenisis occurring North Sea/Norway.  Note very significant pressure gradient increasing upwelling in the North Sea around 27th - 29th with the risk of coastal flooding across East Anglia and Eastern England during the period.

Edited by Thermohaline Conveyor

Share this post


Link to post
Share on other sites

The ops are the best for up to D6 IMO. But always worth comparing to the mean at that time as well. You can see the mean of 2m temps for this weekend and compare it to the GFS P:

 

attachicon.gifgraphe6_0000_306_141___Londres.gif

 

P around 7-8c higher than the mean where as the OP and Control are close to the mean. So in all probability P is outside the SD.

 

The verification of means -v- op is between 30-50% better from around D10 so if you are going to forecast from that period over the long run you will be more successful using the means. I use them with clusterings, 8-10 and 10-14 day anomalies and try to get a pattern and trend. They are not perfect but better than using the ops for sure.

 

That's perfectly fine, but it doesn't fit with posting mean charts at a long range and assigning any meaning to them. The range of outputs is key then, not the mean, the mean chart on its own is misleading. Posting the number and nature of the clusters is much more illuminating.

Share this post


Link to post
Share on other sites

wild model swings causing chaos today? id still follow johns lead and view these charts to get a pretty good idea of what the future upper pattern might be.

http://www.cpc.noaa.gov/products/predictions/610day/500mb.php

 

these are far more consistent then the current ops which are clearly prone to wild swings, which make them pretty useless at trying to pick out longer term trends imho.
 

Share this post


Link to post
Share on other sites

I post charts that far out to highlight that even the variation from the 0z and 06z op will make little difference to the long wave pattern as we go forward. I was saying that the collapse of the Pacific Ridge is more of a driver for the medium term than the variation on the track of the weekend's storm. Of course if the GFS P goes in a different direction then all bets are off, but at the moment this is a low risk, it has three ensemble members supporting it on the 06z (16%).

IDO you can not make a forecast based on mean charts that far out!!! Naturally there will be several clusters and outcomes on the table and the thing to remember with the gfs p is the resolution it runs at at 192-240 hours. It can pick up a more likely trend within that timeframe which via the butterfly affect leads to a different outcome down the line. Other ensemble members run at a lower resolution cant. By all means use a mean chart if it is consistent with the ops and other models as a trend but with so much uncertainty even at 120 hours its outrageous to claim with any cetainty the forecast for 11 - 14 days based on a blended mean!!

Share this post


Link to post
Share on other sites

The ops are the best for up to D6 IMO. But always worth comparing to the mean at that time as well. You can see the mean of 2m temps for this weekend and compare it to the GFS P:

 

attachicon.gifgraphe6_0000_306_141___Londres.gif

 

P around 7-8c higher than the mean where as the OP and Control are close to the mean. So in all probability P is outside the SD.

 

The verification of means -v- op is between 30-50% better from around D10 so if you are going to forecast from that period over the long run you will be more successful using the means. I use them with clusterings, 8-10 and 10-14 day anomalies and try to get a pattern and trend. They are not perfect but better than using the ops for sure.

To use the phrase I made on someone's status was that given the differences at 4/5 days out then most of the solutions no matter how obscene are probably fair game to be honest. One or more of the models will be miles off at a time frame where you would expect them to be close to the mark.

To complicate matters further, the metoffice 6-15 day output is completely different to yesterdays (mild westerly pattern with dry weather at times has now become a rather potent spell of northerlies which make or may not break down at the end of the forecast). I don't think anyone has a clue, I certainly don't  :pardon:

Share this post


Link to post
Share on other sites

That's perfectly fine, but it doesn't fit with posting mean charts at a long range and assigning any meaning to them. The range of outputs is key then, not the mean, the mean chart on its own is misleading. Posting the number and nature of the clusters is much more illuminating.

 

You make a pretty good point.  If the spread of solutions within the ensemble suite is so deviant it makes the mean almost pointless in such a chaotic situation.

 

T+120 850mb wind ensemble mean and OP

 

post-5488-0-23530800-1419252376_thumb.gi

 

30m/s 850mb wind signal remains aligned over much of England and Wales with previous ECM runs favouring SE/E/NE England for most pronounced wind signal tied in with 27/28 Dec low.  SD between 4.0 and 6.0 still indicates quite a spread in synoptic evolutions but interestingly the mean spread 25m/s signal with an SD of 1.8 overlaps with the 30/ms OP across N England and NE England down toward N Norfolk.  The mean track of the low due S of Iceland, N of N Scotland toward Scandinavia.  GFS06Z tracking further S as a weaker/delayed feature with explosive cyclogenisis occurring N Sea/Norway.

 

So much uncertainty still but a consistent signal with N, NE England and East Anglia for wind signal.

Share this post


Link to post
Share on other sites

The GFS parallel "official" release has been put back another week. Hardly surprising!

I have been unable to see any of the runs in full since Saturday's 18z: http://www.netweather.tv/index.cgi?action=gfs-update-2014;sess=7bc3d6eaac28dcf2b4ea715771f2e707

Share this post


Link to post
Share on other sites

Reasons to be cheerful..well it's christmas, and the ecm 00z and other models look more wintry so far today and the latest met office update covering next week is a peach, what a flip from the bland update yesterday. Here is a juicy part for y'all..

 

It will become colder with snow showers developing across the north, these may spread southwards at times and it will be frosty with freezing fog patches possible.

 

This is the perfect tonic, gin and tonic even, for those who were pulling their hair out in frustration wondering when winter was going to start. Let's hope this is the defining moment. :w00t:  :smiliz19:  :yahoo:

Share this post


Link to post
Share on other sites

Right in a probably vain effort to reduce the fog to mist I've been looking at the NCEP state forecasts to see what they have to say:

 

Firstly New York:

 

I STARTED TO TREND TOWARD SUPERBLEND HEADING INTO THE FINAL
WEEKEND OF 2014 AS THERE`S MORE DISAGREEMENT ON THE LARGE-SCALE
PATTERN. GENERAL TREND WILL BE FOR NEAR/JUST SLIGHTLY ABOVE NORMAL
TEMPERATURES.

 

New Hampshire:

 

SUNDAY...

CONFIDENCE IN THE FORECAST GOES DOWN CONSIDERABLY OVER THIS TIME.
DEVELOPING UPPER LEVEL RIDGING INTO ALASKA WILL SEND THE POLAR JET
SOUTHWARD INTO THE CENTRAL U.S.
 

 

Indiana:

 

REMAINDER OF THE EXTENDED QUITE A BIT MORE CONVOLUTED AS MODELS
EXHIBITING DIFFERENCES IN HANDLING OF SUBSEQUENT WAVES ALOFT
TRAVERSING THE LOWER 48.

 

NOT
ANTICIPATING ANY BETTER MODEL CONSENSUS FOR AT LEAST A FEW MORE
DAYS...ESPECIALLY CONSIDERING THE CHAOTIC MODEL PERFORMANCE
RECENTLY. DEFINITELY SOMETHING TO MONITOR GOING FORWARD.

 

At least NCEP agree with us re model performance!

 

I think the suggestion of the polar jet going south into the central USA is a positive for cold and allied with the ridge into Alaska this suggests some amplitude remaining in the flow upstream, the problem in the USA is the lack of clarity with these shortwaves because they do impact on the movement of the PV.

 

At this point if some deeper cold is to arrive I think it will be more likely from the east or ne rather than north or nw.
 

Edited by nick sussex

Share this post


Link to post
Share on other sites

IDO you can not make a forecast based on mean charts that far out!!! Naturally there will be several clusters and outcomes on the table and the thing to remember with the gfs p is the resolution it runs at at 192-240 hours. It can pick up a more likely trend within that timeframe which via the butterfly affect leads to a different outcome down the line. Other ensemble members run at a lower resolution cant. By all means use a mean chart if it is consistent with the ops and other models as a trend but with so much uncertainty even at 120 hours its outrageous to claim with any cetainty the forecast for 11 - 14 days based on a blended mean!!

 

Please can you email UKMO, ECMWF, CFS, JMA, EUROSIP and tell the they are wasting their time with LRF. I am sure they will be happy to save a few quid by scrapping their efforts. they will probably pay you a finders fee. The models go out to FI and I think it is fair to draw conclusions based on your own interpretations. I remember the non-Easterly of last year and many a D10 mean chart was posted and not a lot of flack was thrown around then!

 

The reason they run the ensembles is to counteract the butterfly effect and the mean is the measure of that. 

 

To use the phrase I made on someone's status was that given the differences at 4/5 days out then most of the solutions no matter how obscene are probably fair game to be honest. One or more of the models will be miles off at a time frame where you would expect them to be close to the mark.

To complicate matters further, the metoffice 6-15 day output is completely different to yesterdays (mild westerly pattern with dry weather at times has now become a rather potent spell of northerlies which make or may not break down at the end of the forecast). I don't think anyone has a clue, I certainly don't  :pardon:

 

That 6-15 day forecast ties in with the ECM backtrack to join the GFS op; its not really backing the GFS P. It goes on to say"It may turn more unsettled with time, with wetter and windier conditions returning from the west, and temperatures becoming nearer normal." Pretty much ties in with the GFS mean :D 

 

 

 

You make a pretty good point.  If the spread of solutions within the ensemble suite is so deviant it makes the mean almost pointless in such a chaotic situation.

 

 

 

 

Agree; mean at D5 is of nominal value for surface conditions but I don't believe anyone has suggested that!

Share this post


Link to post
Share on other sites

The GFS parallel "official" release has been put back another week. Hardly surprising!

Avoiding comment on the 06z (I did warn you not to bother after the 00z's!), the para Gfs is in danger of becoming a parody rather than a parallel as it suffers the same issues as the ecm op in over amplifying an incorrect call early on in its run due to it's very high resolution. The Gfs para likely to be worse because of the ncep data initialisation not being as good as ecmwf (for some strange reason).

Share this post


Link to post
Share on other sites

Avoiding comment on the 06z (I did warn you not to bother after the 00z's!), the para Gfs is in danger of becoming a parody rather than a parallel as it suffers the same issues as the ecm op in over amplifying an incorrect call early on in its run due to it's very high resolution. The Gfs para likely to be worse because of the ncep data initialisation not being as good as ecmwf (for some strange reason).

Isn't this really disappointing in that the higher resolution might end up causing more problems than it solves?

 

In an effort to improve the model in a sense they've made it worse!

Share this post


Link to post
Share on other sites

Avoiding comment on the 06z (I did warn you not to bother after the 00z's!), the para Gfs is in danger of becoming a parody rather than a parallel as it suffers the same issues as the ecm op in over amplifying an incorrect call early on in its run due to it's very high resolution. The Gfs para likely to be worse because of the ncep data initialisation not being as good as ecmwf (for some strange reason).

Good point ba.Would you consider it has a lesser reliability factor than the gfs op then?Sm seems to rate it but im not sure why tbh.thanks

Share this post


Link to post
Share on other sites

Please can you email UKMO, ECMWF, CFS, JMA, EUROSIP and tell the they are wasting their time with LRF. I am sure they will be happy to save a few quid by scrapping their efforts. they will probably pay you a finders fee. The models go out to FI and I think it is fair to draw conclusions based on your own interpretations. I remember the non-Easterly of last year and many a D10 mean chart was posted and not a lot of flack was thrown around then!

 

 

Can you provide any evidence that they create LRFs based on isolated mean charts? No interpretation can be drawn from them, only that, the further they get from zero hours, the more worthless they are.

 

See below.

 

gens-22-1-300.png

Share this post


Link to post
Share on other sites
Guest
This topic is now closed to further replies.
Sign in to follow this  

×
×
  • Create New...