Tuesday, May 15, 2018

FV3GFS Moving Forward

It is well known that the US is lagging other forecasting centers, such as the European Center for Medium Range Weather Forecasting and the UK Meteorological Office, in global model forecast skill. 

In testing now is an entirely new dynamical core for the US Global Forecast System, developed by the Geophysical Fluid Dynamics Laboratory (GFDL) and known as the Finite Volume Cubed-Sphere Dynamical Core, or FV3 for short.  See https://www.gfdl.noaa.gov/fv3/ for gory details. 

Source: https://www.gfdl.noaa.gov/fv3/
Experimental "FV3GFS" runs are now being produced by NCEP and I've started to download them to weather.utah.edu.  Look for FV3-13km (experimental) in the left hand nav bar.  For now, I'm only downloading wind and precipitation forecasts.  These should be at near-native resolution and hopefully my post-processing of the data is calculating the 3-hour precipitation correctly.  If you see anything odd, let me know.

In addition to a new dynamical core, the FV3GFS is also using the GFDL cloud microphysics parameterization to simulate cloud process.  This should be an upgrade compared to the old GFS cloud microphysics scheme, but I'm not sure if it has been tested much in winter storms or complex terrain. 

From a research standpoint, I'm not that interested in model performance during the warm season (i.e., right now and during the summer).  I am, however, hoping to get my paws on some of the runs from this past cool season so we can see how it does for winter storms over the west.

Whether or not this proves to be the game changer that NCEP hopes it will be remains to be seen.  I suspect we will see the FV3GFS become the operational GFS at some point later this calendar year or in early 2019. 


  1. Nice work.

    Looks like you are just doing maps, hard to tell if anything is awry from these. I guess if all you see is precip for the entire forecast period, then you know something is wrong. Casual look suggests its forecast of the next week is what we are expecting.

    A meteogram would be an easier way to compare.

    You doing snow at this point? Doesn't much matter for the next 5 months, eh

    Do you put this on the LCC guidance page at some point?

  2. I compared frame by frame with the GFS 13km and it looks fine. Both track fairly well with each other with similar precip and wind fields.