Performance of the GFDL hurricane model during HFIP High-Resolution Hurricane Test Project

The HFIP High-Resolution Hurricane Test was conducted by the Developmental Testbed Center (DTC) with the primary goal of assessing the impacts of using higher horizontal resolution in hurricane numerical forecasting. As part of this project, a high-resolution version of the operational GFDL hurricane prediction system was run for 62 pre-selected cases. The model outputs were provided to the DTC Evaluation Team for generating objective verification statistics focused on quantifying the impact of resolution on hurricane prediction.

The GFDL hurricane model domain consists of a triply nested grid configuration, in which two inner grids are moveable and two-way interactive. The stationary outermost grid spans 75x75º with 1/2º resolution. The middle grid spans 11x11º with resolution 1/6º. The innermost grid spans 5x5º. We use 1/12º resolution for the innermost grid in the low resolution run and 1/18º resolution in the high resolution run. Both configurations were run with 42 vertical levels and a model top at approximately 0.27 hPa.

As in the operational GFDL system, the GFS global analysis and the storm message provided by NHC were used to generate initial conditions for the atmospheric model. An axisymmetric version of the prediction model was used to create an axisymmetric vortex based on the initial storm structure that was estimated from the data in the storm message. The initial conditions were calculated by adding back the model simulated vortex to the environmental fields that were determined from the GFS analysis. Six-hourly GFS forecasts output on 1/2 deg grid were used for lateral boundary conditions. The Princeton Ocean Model was run with a 1/6 deg horizontal grid spacing and 23 vertical sigma levels. The ocean model was initialized by a diagnostic and prognostic spinup of the ocean circulations using available climatological ocean data in combination with real-time SST and sea surface height data. During the ocean spinup, realistic representations of the structure and positions of the Loop Currents, Gulf Stream and warm- and cold-core eddies were incorporated.

The overall results of these test runs will be presented and the analysis of a few cases will be discussed in greater detail.