Evaluation of the Performance of the GFDL FV3 Forecast Model on Tropical Cyclone Track and Intensity Prediction

The FV3 dynamical core, developed at GFDL, will replace the current GFS operational spectral model as the new dynamical
core that will power the National Weather Service’s (NWS) next generation global prediction system. The GFS is recognized as one
of the primary tools for Tropical Cyclone (TC) prediction and the agencies responsible for providing operational TC track and intensity guidance (e.g., National Hurricane Center, Joint Typhoon Weather Center), rely heavily on its TC forecasts. Thus, a key requirement to replace the GFS with the new generation global model is that the TC track and intensity skill be comparable or exceed the current GFS. As the new model is developed, rigorous evaluation of the FV3 performance on TCs is continuing at both GFDL and NCEP in all ocean basins, particularly the Atlantic.
This presentation will summarize a thorough evaluation of track and intensity skill using two versions of the FV3 dynamical core. The first version consists of a horizonal global resolution of 13 km model (global fvGFS)and has been run in near-real time for all of 2017.

In the second version, the 13km global model is nested with a
high resolution 3km domain which covers the entire Atlantic basin including the Gulf of Mexico. Forecasts using the high-resolution nest were run from early August through November, which covered all of the significant storms of the active 2017 Atlantic hurricane season.

In both versions of the model, the 6-class GFDL micro-physics replaced
the scheme operational in the GFS, which provided the physics required to produce a realistic inner-core structure particularly in the version with the high resolution.
Preliminary results indicated that both versions of FV3 provided very skillful track guidance, which exceeded the current NWS operational guidance (e.g., the GFS and HWRF). Although the global fvGFS exhibited a general negative bias, the overall intensity skill was still almost comparable to HWRF when evaluated over a large sample size that included both the Atlantic, East and West Pacific. The negative bias and intensity error was reduced significantly with inclusion of the 3km nest in the Atlantic. Intensity guidance was nearly comparable with operational HWRF for the intense 2017 Atlantic hurricanes in the 2-5 day lead times, despite a significant model spin-up that resulted from starting with the GFS initial condition and vortex.
In this presentation, comparison of performance will also be made between these 2 versions of the FV3, as well as all other
operational guidance (e.g., GFS, ECMWF, UKMET, HWRF, COAMPS-TC).