Development of a Biasing Scheme to Improve Initial Dynamical Model Forecasts of Tropical Cyclone Motion.

摘要

A technique of adjusting the initial steering currents to improve the short-term forecasts of tropical cyclone motion with a dynamical model was tested. Initial tests were performed with hand-analyzed data, although 41 objectively-analyzed cases were also examined to evaluate operational feasibility. Several schemes to track the storm objectively resulted in the use of the streamfunction minimum to define the storm centers. Two biasing approaches were developed. The first used statistical regression equations to predict the initial storm motion, while the second used an empirical relationship to define the initial steering current. Various areal distributions of the bias corrections were then applied to the initial wind fields. Results from these tests indicated improved forecast accuracy if the largest corrections were applied at some distance from the storm center. The biasing decreased average forecast errors for the first 30 hours by about 40% for the hand-analyzed cases and about 30% for the objectively-analyzed cases, with 25% and 15% decreases after 48 hours, respectively. Significant improvements were also observed in the standard deviation of track errors, with averaged decreases of 60% for hand-analyzed cases and 45% for objectively-analyzed cases. (Author)