Statistical post-processing of the navy nested tropical cyclone model and the operational tropical cyclone model

Description

A statistical technique proposed by Elsberry and Frill (1980) for adjusting dynamical tropical cyclone motion forecasts is extended to the Two-Way Interactive Nested Tropical Cyclone Model (NTCM) and the operational One-Way Interactive Tropical Cyclone Model (TCMO). The technique utilizes linear regression equations to reduce systemic errors. Backward extrapolation positions are presented as a less expensive, but inferior, alternative to the backward integration positions required by the original technique. A scheme is developed for applying the technique in storm-motion coordinates as well as zonal-meridional coordinates. Tests with 186 NTCM cases indicate moderate improvement in forecast errors by the zonal-meridional regression technique, and slight improvement by the storm-coordinate scheme. In TCMO tests with 212 cases, the zonal-meridional regression equations reduced the forecast errors, but the storm-coordinate equations did not. The technique failed to improve forecast errors in independent tests with NTCM 1981 data, presumably due to differences in error biases, which indicates a need for a larger sample size. Alternatively backward integration positions may be necessary to achieve consistent improvements from this statistical technique. The technique was able to improve 60h-72h forecast errors in TCMO 1981 cases. (Author)

Subjects