Distillation of key storm evolution characteristics from ensemble forecasts through path clustering

Ensemble forecasts provide important guidance on the range of possible atmospheric evolutions for a known initial state. In the case of tropical cyclones, summary forecast information commonly derived from ensemble forecasts includes the geographic track and intensity evolution. Additional information on the structural evolution of the storm (particularly if there is a possibility of extratropical transition [ET]) is also valuable.

Physically-based distillation of the information contained in these ensemble forecasts of the evolving tropical cyclone provide new insights into the sources of the ensemble spread and thus to the sources of forecast sensitivity. In this study we characterize the initially tropical cyclone evolution in terms of a five dimensional space comprising the geographic path (φ(t), λ(t)) and the path of the system through the cyclone phase space (B(t), VTL (t), VTU (t)). Path clustering within this framework provides objective partitions of the 51-member ECMWF ensemble. Results are demonstrated for ensemble forecasts of Typhoons Sinlaku and Jangmi, both well observed during the THORPEX TPARC international field program.