Evaluation of Tropical Cyclone Structure Forecasts in a High-Resolution Version of the Multiscale GFDL fvGFS Model

A nested version of the cubed-sphere finite-volume dynamical core (FV3) with GFS physics (fvGFS) is capable of tropical cyclone (TC) prediction across multiple space and time scales, from subseasonal prediction to high-resolution structure and intensity forecasting. Here, a version of fvGFS with 2-km resolution covering most of the North Atlantic is evaluated for its ability to simulate TC track, intensity, and finescale structure. TC structure is evaluated through a comparison of forecasts with three-dimensional Doppler radar from P-3 flights by NOAA's Hurricane Research Division (HRD), and the structural metrics evaluated include the 2-km radius of maximum wind (RMW), slope of the RMW, depth of the TC vortex, and horizontal vortex decay rate. Seven TCs from the 2010-16 seasons are evaluated, including 10 separate model runs and 38 individual flights. The model had some success in producing rapid intensification (RI) forecasts for Earl, Edouard, and Matthew. The fvGFS model successfully predicts RMWs in the 25-50-km range but tends to have a small bias at very large radii and a large bias at very small radii. The wind peak also tends to be somewhat too sharp, and the vortex depth occasionally has a high bias, especially for storms that are observed to be shallow. Composite radial wind shows that the boundary layer tends to be too deep, although the outflow structure aloft is relatively consistent with observations. These results highlight the utility of the structural evaluation of TC forecasts and also show the promise of fvGFS for forecasting TCs.