Forecasting Tropical Cyclone Motion Using Empirical Orthogonal Function Representations of the Environmental Wind Fields

Abstract

Tropical wind fields from the U.S. Navy Global Band Analyses (GBA) are studied to depict the synoptic flow surrounding tropical cyclones. The composite fields of the zonal and meridional wind components on a grid centered on the tropical cyclone indicate physically realistic flow patterns. Scalar empirical orthogonal function (EOF) analysis is used to represent the zonal and meridional GBA wind component fields. The representation of these components in terms of the first 35 out of a total of 527 EOF coefficients accounts for at least 80% of the total variance and eliminates much of the noise from the fields. This truncation requires only 7% of the storage needed for the original gridpoints. The eigenvectors can be interpreted as representing different synoptic flow patterns. Statistical regression equations are derived to predict the future zonal and meridional translation of the tropical cyclone. The EOF coefficients are used as predictors to represent the synoptic information for the scheme. The track forecast errors are slightly smaller than those from the Joint Typhoon Warning Center. Further reduction in forecast errors results from stratification of the cases. Stratification by prior 12-h motion results in 72-h weighted mean forecast errors of only 481 km for the dependent sample. Stratification by synoptic situation based on the EOFs is also tested. From discriminant analysis, the second zonal eigenvector at 700 mb best relates to the 72-h zonal storm motion while the second meridional eigenvector at 250 mb has the best correlation with the meridional motion. Weighted mean forecast errors for regression equations derived within the synoptic subgroups are 484 km at 72 h. Thus, stratification by synoptic situations in terms of wind-based EOFs is as effective as stratification by past storm motion in improving track forecasts.