Extended Abstract (272K)10C.3
A Summary of Ugrades to the Operational GFDL Hurricane Model for 2003
Morris A. Bender, NOAA/GFDL, Princeton, NJ; and T. Marchok, I. Ginis, B. Thomas, R. E. Tuleya, and H. L. Pan
Since 1995 the GFDL Hurricane Prediction System has provided operational guidance for forecasters at TPC in both the Atlantic and Eastern Pacific ocean basins. In addition, a version of the GFDL model (GFDN) has been used by the Navy to provide operational guidance for storms in most of the other ocean basins as well. In 2002 a new high resolution, two nested version of the model was made operational for both versions of the GFDL model. The new implementation was quite successful with significantly improved track performance. However, many problem areas remained. To address these problems, a new version of the GFDL model was developed and extensively tested for operational implementation in 2003. In the new version, the current GFDL model cumulus parameterization and Mellor Yamada boundary layer formulation were replaced with a simplified version of the Arakawa-Schubert scheme (SAS) and a nonlocal diffusion scheme, respectively, both of which runs in NCEP's Global Forecast System (GFS) model. In addition, the vertical resolution of the GFDL model was increased from 18 to 42 sigma levels a new mass initialization was developed and an improved ocean initialization was added. Since parallel tests indicated significant track improvements, reduction in model bias, and some improved intensity prediction, the new system was made operational in 2003. A detailed summary of the various changes will be outlined. The new model showed significantly improved track performance compared to the previous GFDL system in both the Atlantic and East Pacific basins during its first year of operational implementation. For example, in the Atlnatic basin, the model exhibited skill relative to CLIPER of 48, 57, 55, 57 and 58% at forecast days 1, 2, 3, 4 and 5. This compared to skill relative to CLIPER of the GFS model (the next best model performer), of 41, 51, 54, 58 and 59%, respecitvely. In the East Pacific, the forecast skill averaged about 30% in the 1-4 day time period. Final verifications will be shown for both track and intensity, as well as case examples that highlight the model's excellent and poor performances during the season. Finally, a brief summary of the changes planned to be made operational for the 2004 hurricane season will also be outlined.
Session 10C, Tropical Cyclone Prediction and Predictability III: Operational Models
Wednesday, 5 May 2004, 10:15 AM-11:45 AM, Napoleon II Room
Previous paper Next paper
Recorded presentation