A Local Analysis of the Performance of the Ensemble Forecast Systems Included in TIGGE

We apply local diagnostics that measure the efficiency of the forecast ensemble in representing the forecast uncertainty to the ensembles of the major operational numerical weather prediction centers. For this analysis, we take advantage of the THORPEX Interactive Grand Global Ensemble (TIGGE) data set and the local diagnostic techniques developed by our research group. We find both important differences and similarities between the performance of the different ensembles. The most important differences are in the efficiency of the different ensemble systems in representing the space of the uncertainty and the magnitude of the uncertainty for the first 2-3 forecast days. The most important similarity is that while the different systems all tend to provide a good representation of the forecast uncertainty beyond the 2-3-day forecast time, they also tend to underestimate the magnitude of the uncertainty until the very late (about 8-10-day) forecast times. Finally, we discuss the implications of our results for future research and development in ensemble forecasting.