Statistical adjustment of the HATRACK typhoon track forecasting model

Abstract

A statistical technique proposed by Elsberry and Frill (1980) for adjusting dynamical tropical cyclone motion forecasts is extended to the Hurricane and Typhoon Tracking (HATRACK) scheme of Renard, et al (1968). The objective is to eliminate systematic HATRACK errors through use of linear regression equations. Predictors derived from comparison of backward integration of the HATRACK model with known positions of the storm are included to relate past HATRACK performance with future performance. The predictors also include translation speeds along the -36 hr to +72 hr track plus position and radius of the storm at the initial time. Regression equations are derived from dependent samples of 60 and 61 storms for the 500, 700, and 850 steering levels. Tests of the equations with the dependent samples indicate improvement over typical forecast error averages. The regression scheme approaches the same error distribution with time regardless of steering level. In tests with an independent sample of 31 storms, the regression equations were able to reduce the systematic longitudinal/latitudinal errors at all levels. In these tests, the regression scheme reduced HATRACK bias more efficiently that the bias correction of the Modified Hurricane and Typhoon Tracking (MOHATT) model.