The American GFS Model (Global Forecast System) has been replaced or upgraded with a new “digital dynamical core” called the VF3 (Finite-Volume Cubed-Sphere) the first core upgrade in almost 40 years.

In theory this will result in more accurate output from the model. But as in many human endeavors you may gain something on the one hand but lose something on the other hand. As someone who uses these tools on a regular basis I’ve witnessed allegedly “upgraded” models perform worse than the old version (over the past 5 years other changes have been made to the GFS and to the renowned ECMWF model). These models often did NOT provide forecasters with better guidance.

The other major problem with the constant “fiddling” with models is that a key component of the value humans can add is adjusting the models for their biases and performance characteristics (strengths and weaknesses) with local knowledge to account for those.

But that takes many years of experience with an unchanging model. The constant changes make the forecasters job all the more difficult.

NOAA claims its a big improvement. We shall see. They have shown some impressive precipitation forecast improvements, but of course they showed us no failures so who knows.

We used to call the GFS model the “Good for Sh**” model. Now the ECMWF has gone down hill with numerous changes and now who knows how the GFS will do in the real world, years from now we’ll have an idea.

Meanwhile I suspect they will have to keep “tweaking” (not twerking) it as they discover problems as it is run daily. And on-going tweaks will prevent a forecaster from trusting its reliability until they are largely done with new “upgrades”.

For now consider my reaction to this major upgrade as “Underwhelmed”.

For more follow me on Twitter @MellishMeterWSB.