NOAA’s “next-generation” weather model (GFS-FV3) replaces the old Global Forecast System (GFS)

[Jesse Risley]

The GFS-FV3 was rolled out today.

From this morning's press release that was posted a short while ago to the FB page: "NOAA’s flagship weather model — the Global Forecast System (GFS) — is undergoing a significant upgrade today to include a new dynamical core called the Finite-Volume Cubed-Sphere (FV3). This upgrade will drive global numerical weather prediction into the future with improved forecasts of severe weather, winter storms, and tropical cyclone intensity and track. "

NOAA upgrades the U.S. global weather forecast model
AND
JACKSONVILLE, Fla - The Nation's premiere forecast model will go live on July 12, with an upgrade intended to improve weather forecasts around the world.

 

[JeremyS]

I'll be curious to see how this model performs. I believe this was supposed to replace the GFS several months ago, but they had to hold that off due to some issues. Presumably, they're happier/got the "bugs" worked out of it to officially replace the GFS now.

 

[rdale]

It has a MASSIVE winter cold bug. That was partially fixed but not completely - although given the horrible spring/summer so far in the Great Lakes it might need to be soon

It has been running in parallel for several months - sites like PV, TT, etc. have been displaying the data.

 

[Dan Robinson]

Forgive me if I missed this somewhere, but are all model sites now automatically displaying the FV3 data in place of the GFS, or does every site (COD, UCAR, etc) need to manually make the switch? How will we know which one we're looking at?

 

[Jesse Risley]

Forgive me if I missed this somewhere, but are all model sites now automatically displaying the FV3 data in place of the GFS, or does every site (COD, UCAR, etc) need to manually make the switch? How will we know which one we're looking at?

COD has a note on the model homepage that now that it is no longer running parallel, that tab is going away and the data is now incorporated under the regular GFS forecast model tab. Pivotalweather has a note, again at the top of the GFS forecast model tab, that the new data is fully operational into the model tab.

 

[rdale]

Dan - any model that was displaying the "old" GFS will automatically be displaying the new in its place. They would have to change their datasource to keep the old GFS going. Doing nothing always gets you the new model.

 

[GPhillips]

For those interested in also looking at the "old" GFS, pivotalweather calls it GFS Legacy in its model drop-down menu.

 

[Robert Edmonds]

Sounds like the last core was spectral, and this is moving away (if I am reading various sources correctly) from that formulation? Curious to know why, does it scale better? Spectral methods have the befit of less damping in waves, but also require a transform to that space.

 

[Jeff Duda]

For those interested in also looking at the "old" GFS, pivotalweather calls it GFS Legacy in its model drop-down menu.

But the "old GFS" is only going to run for another few weeks before EMC shuts that off.

 

[GPhillips]

But the "old GFS" is only going to run for another few weeks before EMC shuts that off.

Bummer, but not unexpected. They need to free up more computer resources I imagine.

 

[Bill Barlow]

Hi George, hope you are doing well. What is the grid spacing for the new version of the GFS? I have seen the newer GFS and it seems quite "noisy" looking at the 500Mb heights and vorticity forecasts. For all the years I have used the older version to forecast convection, hopefully this new physics will improve the model and produce less QPF bullseyes.

I'm glad to be promoted to an "enthusiast" from a "lurker".

 

[Jeff Duda]

Both the FV3-GFS and the "old" GFS were run at ~13 km grid spacing. You may be referring to other sources that re-grid the raw output onto coarser grids such as 40 or 80 km. I'm sure someone is still doing that with the updated GFS.