Comparison of the performance of the GFDL hurricane model, using the AVN, UKMET and NOGAPS analysis

Since 1995, the Geophysical Fluid Dynamics Laboratory (GFDL) hurricane model has been run at the National Centers for Environmental Prediction (NCEP) to provide track and intensity forecasts for storms in the Atlantic, Eastern Pacific and Central Pacific basins. This implementation of the GFDL model (to be referred to as GFDA) obtains initial conditions and boundary conditions from the NCEP aviation model (AVN). Since 1996, a version of the GFDL hurricane model has also been run operationally by the U.S. Navy's Fleet Numerical Meteorology and Oceanography Center (FNMOC) for prediction of tropical cyclones in other ocean basins throughout the world. This version of the model is identical to the version running at NCEP, except it uses fields from the Navy Operational Global Atmospheric Prediction System (NOGAPS) in place of the AVN fields. This past summer, the Navy version of the GFDL model (referred to as GFDN) has been used for backup support for NCEP in the Atlantic and Eastern Pacific. Although run for a limited number of cases where comparisons could be made between GFDA and GFDN, it was found that forecasts produced by GFDA and GFDN were sometimes quite different, even though the underlying forecast model was identical. On average, the GFDL forecasts run from the NOGAPS fields (GFDN) exhibited superior performance, suggesting sensitivity of the model to the supporting global fields. Results showing comparisons of the storm tracks and track errors of GFDA and GFDN will be presented for cases of Hurricanes Dennis, Floyd and Gert.

Since 1996, the U.K Meteorological Office (UKMET) global model has also been providing operational forecast guidance for storms in the Atlantic basin. On average, these forecasts have been highly skillful. An effort is underway to rerun the GFDL forecast system for Hurricanes Dennis and Floyd using the UKMET global fields. Detailed comparisons will be made in the performance of the GFDL system run from the UKMET and AVN fields. Attention will be paid to differences in the storm track and average skill for the two sets of forecasts. Evaluation will be made of the differences in the environmental flow surrounding the storm represented by both global models, which likely contributed to the differences in the model performance. These results will also be presented for various cases.