GFS vs. FV3 GFS

[Mitchel Volk]

I have been comparing the two models and my current assessment is the GFS has been a lot more consistent than the FV3 Model.  The FV3 seems to be too strong on many runs for east coast storms.   Just look at the upcoming storm for the 19-20th.  At one point NYC was going to get37" of snow and the latest 6Z run has mainly rain.  I love to see what other people think about this.

[WhiteoutWX]

I’ve also felt over the last few months the FV3 has generally been less consistent and more jumpy run-to-run. Does anyone have verification scores or comparisons for the two models? I’d like to see some data to back up my somewhat subjective observations.

[donsutherland1]

37 minutes ago, WhiteoutWX said:

I’ve also felt over the last few months the FV3 has generally been less consistent and more jumpy run-to-run. Does anyone have verification scores or comparisons for the two models? I’d like to see some data to back up my somewhat subjective observations.

The verification scores can be found at: https://www.emc.ncep.noaa.gov/gmb/STATS_vsdb/

The FV3 does not seem to be materially better than the GFS, so far.

[Mitchel Volk]

I saw that the verification score is for the whole world.  I believe for our neck of the woods in winter the models is not ready yet.  I hope they can improve this before it goes operational.

[BlizzardWx]

It would be ideal if the FV3 were better than the GFS at this stage, but at least the FV3 should have a higher ceiling than the GFS given its new capabilities. We will see!

For my neck of the woods the FV3 produces out of control lake effect when none occurs. Its kind of a pain. 

[Cheeznado]

Now that the shutdown is over, any word on when the FV3 will replace the GFS?

[WidreMann]

My amateur experience is that the FV3 is worse than the GFS. They said it has a cold bias, and that's very clear in comparing the charts. It needs to bake more.

[high risk]

19 hours ago, Cheeznado said:

Now that the shutdown is over, any word on when the FV3 will replace the GFS?

             mid-March at the earliest.   The final 30 day stability test has not yet been run.

[Stebo]

21 hours ago, WidreMann said:

My amateur experience is that the FV3 is worse than the GFS. They said it has a cold bias, and that's very clear in comparing the charts. It needs to bake more.

We have seen it struggle significantly locally on weak systems, clippers are a big part of our weather in winter, and it is failing hard even on a 3 day forecast compared to all other model data. Also cold sector precipitation it tends to be very light, I know the Euro also has the same issue but not nearly to this magnitude.

[RU848789]

It would be extraordinarily disappointing if, after all that time and money invested in its development, the FV3 isn't materially better than the old GFS.  The "world" 500 mbar scores, as per the attached link, were slightly better than the GFS, putting the FV3 in competition with the UK for 2nd best, a decent jump behind the ECMWF, still.  But as discussed above, more than a few of us on these boards want to know how it does with northeast winter storms.  Hasn't seemed too good so far.  And tonight's 18Z 1-2 foot blizzard isn't inspiring confidence, as it's kind of alone in that forecast and has way overdone snow several times earlier this season.  

https://arstechnica.com/science/2018/10/noaa-is-about-to-make-some-big-changes-to-its-global-weather-model/

[AfewUniversesBelowNormal]

FV3 has been discontinued? 

[isohume]

As far as another guidance tool goes, the FV3 can blow me. Well, right after the GFS does.  

[bdgwx]

I had read somewhere that the primary goal of the GFS upgrade this round was obviously the transition to the new dynamical core (GSM to FV3). It wasn't necessarily to make substantial improvements to the skill of the forecasts. That may come later. I agree about wanting skill scores posted for the CONUS. It's great ranking models in terms of their skill over the entire NH, but at the end of the day most of us really only care about the United States or even our own backyards.

[Jim Marusak]

so, given what is being seen so far, what tweaks would everyone like to make to the FV3 to help it along to get better in North America and beyond?

[wdrag]

My understanding from NOAA contacts, is that FV3 core implements very soon, with no signifiant overall improvement, but that significant improvement follows a year or so after implementation via updated physics and resolution. Yes, right now, it is my opinion in the ne USA that the GFS is better than the coming initial FV3 version, when looking at precipitation events, only.  Have not checked sfc temps nor any fields aloft. 

 

 

[donsutherland1]

Implementation of the FV3 has been delayed due to two primary issues:

-The snow depth and the water equivalent of snow depth at the surface have unrealistically large values when precipitation occurs in environments with low-level temperature profiles close to freezing. Techniques that use either of these variables for deriving snowfall will exhibit excessive snowfall values.

-The model forecasts exhibit a cold bias in the lower atmosphere that became more prominent after late September 2018.

https://www.weather.gov/media/notification/scn19-12gfsv15.pdf

[Great Snow 1717]

31 minutes ago, donsutherland1 said:

Implementation of the FV3 has been delayed due to two primary issues:

-The snow depth and the water equivalent of snow depth at the surface have unrealistically large values when precipitation occurs in environments with low-level temperature profiles close to freezing. Techniques that use either of these variables for deriving snowfall will exhibit excessive snowfall values.

-The model forecasts exhibit a cold bias in the lower atmosphere that became more prominent after late September 2018.

https://www.weather.gov/media/notification/scn19-12gfsv15.pdf

As things stand now, the FV3 needs major improvements.  

[TellicoWx]

Delayed indefinitely