A Diagnostic Study on the Statistical Predictability of Tropical Cyclone Motion

Abstract

Statistical tropical cyclone prediction systems typically fall into one of three categories: 1) those using meteorological predictors derived from observed synoptic data; 2) those using purely empirical predictors such as climatology, present motion, past motion, analogs, etc.; and 3) those using combinations of both synoptic and empirical predictors. The variance-reducing potential of each of these prediction systems on given acts of dependent data is examined in detail. In general, it is found that empirical prediction systems are always superior in the shorter range forecast periods and even for extended forecast periods before storm recurvature. During and after storm recurvature, however, the synoptic-type predictors provide a better means of reducing the variance of tropical cyclone motion. It is shown that statistical tropical cyclone forecasting systems should make judicious use of both synoptic and empirical predictors.