Tuesday, 24 January 2012: 9:30 AM
New metrics for verifying ensemble spread
Room 238 (New Orleans Convention Center )
The "Rank Histogram" and the "Matching of RMSE of ensemble mean forecast vs. ensemble spread (STD)" are two most common verification methods widely used nowadays to measure ensemble spread quality. However, these two tools are over-simplified and even misleading in many cases. Therefore, new verification methods are explored to measure ensemble spread quality in this study. The new method is based on spread-skill relation using NCEP Short-Range Ensemble Forecast (SREF) dataset. The proposed method can measure the degree of ensemble dispersion in one aspect, and to measure the usefulness of ensemble spread in another aspect. The usefulness of ensemble spread refers to its application to both predicting forecast error and possibly data assimilation.