GFS Upgrade coming - November 2014

[snowking]

Thought I would set up a thread for all the model nerds out there (myself included) with a really significant upgrade coming to the GFS just in time for winter.

 

I won't go too in depth here in terms of each specific change as this is not my area of expertise. You can find a full run down here:

 

http://www.emc.ncep.noaa.gov/GFS/impl.php

 

The single most significant change will be the horizontal resolution upgrade - we will see things improved from 23km up to 13km (switching at the same time from a Eulerian to a Semi-Lagrangian grid). This should bring things more on a par with the likes of the ECMWF (currently at a resolution of 16km), though the ECMWF itself is due an upgrade later this decade I believe. In addition to this I believe there is also a change due to the data assimilation process, once more designed (after consultation with the ECMWF) to bring the GFS more on a par with it's rivals. In addition to this the high resolution stretch of the operational run will be extended from its current 8 day period (+192) out to 10 days (+240).

 

There is so much more to it than just this though. For the amateur forecasting community I cannot stress highly enough how significant this upgrade could be. If we bear in mind that up until last year many on here religiously followed the limited area 12km resolution NAE runs for just 48 hours ahead, which covered just the North-East Atlantic. Now take that idea and push it out globally, with many more parameters available completely free of charge. This really could be a game changing moment from that point of view.

 

There is no concrete date set as yet for this other than some time in November 2014, though I believe the upgraded data will be ready to run in automated forecasts from next month. It will be fascinating to see how this upgrade improves things a we head in to the most eagerly followed time of the year here and we should all bear in mind when analysing runs that we cannot make the same old GFS vs the european runs arguments as before as we are entering totally unchartered waters with this one.

 

There are also several stratospheric and o-zone improvements which I know will please many too.

 

I'll try and post as soon as I can find a concrete date for the completion of the upgrade. In the meantime if anyone with more knowledge of the physics of modelling cares to chip in then please feel free to

 

SK

 

[Cloud 10]

I and a few others had thought that the PRHW14 model which appeared on the model verification graphs was an early taster of the upgraded GFS,but it seems that it was some sort of experiment by one of the programmers,although it is probably related in some way.

 

 

http://www.emc.ncep.noaa.gov/gmb/wd20rt/vsdb/

 

 

Hopefully the new "bells and whistles" GFS can improve its verification stats against the 

all-conquering ECM.

[lorenzo]

It would be shame to lose those dartboard lows... and the random 06z and pub run madness.

 

This could be fun. Hopefully they switch it on just as an Easterly is on the cards..

 

A great sub plot for Winter..

[Paul]

Bear in mind that although the model will run at that resolution, the available output will be a bit different from that - currently the output is 0.5 degrees (around 55km), that'll be bought back to 0.25 degrees - which will match the current output of the ECM as well. 

[mulzy]

Question

By implementation, does that mean full replacement of the existing GFS or will the implementation involve some form of parallel running?

[Paul]

It's already running parallel and when the update happens it'll be a direct replacement.

[lorenzo]

Briefly asked one of the mets over the pond around this tonight, they are quite excited about the capabilities of the new GFS.

 

Looks like we have a Battle Royale for Winter, GFS improving layering on the strat for starters.

 

ECM and GFS also were days out with each other on MJO knock on effects, something that had to be compared to UKMO Control and JMA for a good read.

 

This is going to be good...

[mulzy]

Is there anywhere on the internet we can see this 'parallel' run?

[bluearmy]

implementation dec 9th. hopefully, we will be able to see it running in parallel before that. meteociel usually have parallel gfs upgrade when available. i understand that there is an upgrade on the gefs at the same time.

[Suburban Streamer]

Sylvain has just confirmed to me on twitter that the new gfs will be running in parallel from some point in the next two weeks

[Paul]

We'll also be running it

[Dancerwithwings]

We'll also be running it

:yahoo: :yahoo:

[pdjakow]

GFS parallel results are already available on 0.25, 0.5, 1.0, 2.5deg.

ftp://ftp.ncep.noaa.gov/pub/data/nccf/com/gfs/para/

 

Meteociel.fr results (1.0deg grid):

http://meteociel.fr/modeles/gfse_cartes.php?runpara=1&ech=6

[Steve C]

We'll also be running it

Any timescale? I prefer using the site here, as my never more than basic French skills are not much more than merci, bonjour!, or au revoir! now. 

[Paul]

Hot off the press, nicely timed for the new thread. We have the newly upgraded (0.25 degree resolution) GFS parallel runs available to view now, ahead of the official upgrade which is currently planned to be December 9th.

 

Free chart viewer

http://www.netweather.tv/index.cgi?action=gfs-update-2014;sess=

 

Netweather Extra chart viewer

http://www.netweather.tv/secure/cgi-bin/premium.pl?action=newgfs;sess=

[bluearmy]

I wonder when sylvain will begin to run parallel GEFS on meteociel ?

[Summer Sun]

I wonder when sylvain will begin to run parallel GEFS on meteociel ?

 

He's running this GFS parallel - if thats what you mean?

[Cloud 10]

He's running this GFS parallel - if thats what you mean?

 

BA means the gefs ensembles

[Paul]

BA means the gefs ensembles

 

There isn't a parellel of the GEFS running currently. 

[Cloud 10]

There isn't a parellel of the GEFS running currently. 

 

Maybe just an upgrade to the existing gefs?

 

• Unification of analysis ensemble with GEFS ensemble
• Ensemble resolution increase to T574L64

[bluearmy]

 

Maybe just an upgrade to the existing gefs?

 

• Unification of analysis ensemble with GEFS ensemble
• Ensemble resolution increase to T574L64

 

 

yes cloud but they must surely be running them in parallel before going live.

 

incidentally, weatherbell stating that new gfs now going live 17th dec.

[Paul]

They most likely will be available, but they're not there yet.

[radiohead]

 

Hi Paul, just looking at the UK Max Temperature charts, it seems when the figures get into double digits the temperature is displayed the wrong way round. 12 is displayed as 21, 13 as 31 etc.

[Paul]

Thanks, it's not that they're backwards, they're just too close together and overlap as the resolution is that much higher. We should get them sorted on Monday

[Cloud 10]

Upgraded GEFS taking over in September.

 

Meteociel - Cartes Ensemble GFS (ENS/GEFS)

 

 

These maps show the Z500 géopotentiel and pressure at sea level of each of the 21 scenarios proposed by all GEFS 1. The maps are updated from 6:40, 12:40, 6:40 p.m. ET 0:40. (3 hours later than parallel) This run GEFS Parallel (or GEFSP //) is a test version of the new GEFS that will replace the current GEFS late September 2015. It runs in "parallel" of the former GEFS (hence the name). It has a larger horizontal and vertical resolution and uses the latest GFS model (released in early 2015). Its resolution is 0.5 ° 1.0 ° instead of before. Moreover maturities range from 3 hours to 3 hours up to 192h. It is slightly more efficient than the current GEFS and often repproche more operational GFS. Warning: The control run is totally different operational run / deterministic, since the two runs are made ​​at different resolutions. Furthermore the update of GFS January 2015 has not yet been reflected in GEFS, the modeling can thus seem to be more different.