Numerical forecasting experiments on Typhoon Herb (1996)

A numerical prediction experiment is conducted on Super Typhoon Herb, 1996, that made landfall in northern Taiwan and caused severe wind and flood damage. A version of the Naval Research Laboratory's (NRL) Coupled Ocean/Atmospheric Prediction System (COAMPS) with a triple-nested grid, (81, 27, and 9-km horizontal resolutions) and 30 levels in the vertical is used in this experiment. Starting from archived global analysis fields, COAMPS generates equally good track predictions on all three grids. The average location errors of three grids were 58 km at 24 h, and 77 kin at 48 h. The prediction of wind fields, especially at the critical time just prior to Herb's landfall on both the medium mesh and the fine mesh agrees very well with subjective analyses by local experts. Quantitative precipitation prediction on the fine mesh is very good both in amount and distribution. Precipitation predictions on the medium and coarse meshes, however, are less skillful. The result from a supplementary numerical experiment with the same three-grid configuration of 9-27-81 km resolution, but with the medium resolution terrain (pertinent for 27 km) on the 9-km grid suggests that high-resolution dynamics, and the terrain afforded on the 9-km mesh, are both important ingredients for quantitative precipitation prediction. Therefore, for accurate quantitative precipitation prediction, it is important to place the high-resolution mesh over the complex terrain, so that the terrain effect on the structure of the storm can be represented on the finest possible resolution. This study demonstrates that a well-behaved, general-purpose, mesoscale numerical model can be a useful tool for tropical cyclone prediction.