The purpose of this work is twofold. The first involves a change in the vortex initialization within GFDN, and an evaluation of these changes based on a number of cases from the 1998/1999 tropical cyclone season. The second is to compare track forecasts produced from NOGAPS, GFDN and COAMPS with "best track" information for a number of cases in the 1999 tropical storm season.
The modification to the initial vortex formulation in GFDN involves a change in the way asymmetries are added into an initial vortex. In the new formulation, asymmetries are added from a previous forecast of GFDN for the time corresponding to the model start time in order to lessen the spin up time needed for the model to represent an observed storm. Asymmetries are added to the momentum, temperature, and moisture fields via a simple nudging algorithm. A comparison of the old and new vortex initialization schemes suggests significant improvements in short range track forecasts. For a total of 12 different storms there is a 10, 7, 6, 1 and -3% improvement in the mean track error for the 12, 24, 36, 48 and 72-hour forecasts using the new scheme.
FNMOC is currently producing twice daily forecasts from nine operational areas using COAMPS. Three of these areas (West Atlantic, Central America and Korea) cover regions in which tropical cyclones are frequently observed. Another area, which was created specifically to test the automated tracker on COAMPS forecasts, is located in the Western Pacific and is run twice daily, although not operationally. The tracker works with 27 km resolution COAMPS fields to produce tropical cyclone track information in a standard output format for use in the Automated Tropical Cyclone Forecast (ATCF) system. This system is used to compare the regional model tracks (COAMPS and GFDN) with the global model tracks (NOGAPS) and with CLIPER and best track information. Also, a comparison between different COAMPS tracks for the same storm is made for a number of cases. This is possible because there is significant overlap between the West Atlantic and Central America areas and between the Korea and Western Pacific area so that the same storm frequently appears in both areas. This last comparison is used to identify the influence of domain size and location on producing accurate track information in a regional model.