A kilo-member ensemble is evaluated for track forecasting based on semioperational runs for 2001-2003 Atlantic and Eastern Pacific tropical cyclones. A multigrid barotropic model (MUDBAR) is perturbed across a multidimensional parameter phase space to simulate uncertainties in the environment and vortex structure. The results of this work can be viewed as a sensitivity study on factors affecting storm motion or more generally as an evaluation of ensemble methods.

Several products are developed and evaluated from the ensemble output. The spread-error relationship is examined to determine whether skillful forecasts of forecast skill are possible. Ensemble-based maps of cumulative strike probabilities are verified using several metrics. These results are compared to the track forecasts generated by NCEP's GFS ensemble forecasting system. Finally, the problem of calibrating the ensemble output is discussed.