Improving the GFS Boundary Layer Parameterization for Global Tropical Cyclone Predictions

Global weather prediction models are getting more skillful in tropical cyclone (TC) forecast as the horizontal resolution increases. NOAA/GFDL’s new global atmospheric model fvGFS (FV3 dynamical core with GFS physics) demonstrates very good skill in TC track and intensity predictions at 13-km resolution. We further improve the TC intensity forecast skill of fvGFS by updating the boundary layer (BL) turbulence parameterization. From TC structure analyses in retrospective 10-day forecasts of fvGFS, we find that the GFS BL parameterization tends to over-estimate the BL turbulent diffusivity in TCs, resulting in a weaker supergradient jet, weaker radial inflow and deeper inflow layer compared to observations. The over diffusive issue is due to the inappropriate representation of the BL height parameter in the GFS BL scheme, which affects both the magnitude and vertical extent of the parameterized turbulent diffusivity. Different approaches are used to reduce the turbulent diffusivity in the TC BL. It is found that the simple reduction approach used in the HWRF model (Gopalakrishnan et al. 2013) generally downgrades the TC track and intensity skill of fvGFS. Alternatively, we modify the turbulent diffusivity by adjusting the height parameter in the GFS BL scheme. Preliminary results show some promising improvements of the vertical TC structure. Other impacts on TC track and intensity will also be discussed in this presentation.