Regionally bred modes and ensemble forecasting of tropical cyclone motion

The techniques of ensemble forecasting have been shown in some of the work by the author to possess potential skill in improving tropical cyclone track forecasts using a simple barotropic model. In particular, the application of regionally bred modes was shown to be much better than random perturbations added to the model grid. This study intends to (a) obtain skill information of these methodologies in a full-physics baroclinic environment, and (b) gain further understanding of the dynamical characteristics of different kinds of perturbations.

Some preliminary results using a modified version of the NCAR/PSU MM4 indicate that the regionally bred modes are indeed more well-organized than random perturbations throughout a period of 72 h examined. The skill of a verification case using bred modes as perturbations is also slightly better than the corresponding barotropic ensemble. However, the general results differ from those obtained in some other previous studies where useful perturbations were extracted from the EOF components of grew random errors. Therefore, diagnostic tools such as EOF expansion and energetics consideration will be used to resolve these apparent discrepancies. Results based on a larger verification statistics will be presented in the conference.