Statistical Analysis of Ensemble Forecasts of Tropical Cyclone Tracks over the North Atlantic.

摘要

The skill of individual ensemble prediction systems (EPS) is evaluated in terms of the probability of a tropical cyclone (TC) track forecast being within an expected area. Anisotropic probability ellipses are defined from each EPS to contain 68% of the ensemble forecast members. Forecast reliability is based on whether the forecast verifying position is within the ellipse. A sharpness parameter is based on the size of the EPS probability ellipse relative to the main operational forecast probability product, the Goerss Predicted Consensus Error (GPCE). For the 2008-2011 Atlantic TC seasons, the ECMWF ellipses have the highest degree of reliability of the EPSs. Additionally, the ECMWF ellipse has a higher resolution than the GPCE operational product over all forecast intervals. The sizes and shapes of the EPS ellipses varied with TC track types, which suggests that information about the physics of the flow-dependent system is retained compared to isotropic probability circles that may not reflect variability associated with track type. It is concluded that the ECMWF ensemble contributes the most to a combined EPS-based product called the Grand Ensemble (GE), and further modification of the GE to reflect this has a potential for reducing the sizes of warning areas.