An Investigation of the Spatiotemporal Variability of the Performance of Operational Global Ensemble Forecast Systems

This paper investigates the spatiotemporal variability of the performance of operational global ensemble forecast systems based on the THORPEX Interactive Grand Global Ensemble (TIGGE) data set. Special attention is paid to the relationship between the aforementioned variability and the dynamical processes of the atmosphere. The geographical distribution of statistics of ensemble performance is computed for a two-month period of the NH cold season. Particular cases in which the ensemble grossly underestimates the local forecast errors are analyzed in detail. While the spatiotemporal variability of the forecast errors and the ensemble statistics is dominated by synoptic scale variability, the quality of the prediction of the slowly varying large scale component of the flow plays an increasingly more important role in determining the quality of the ensemble predictions with increasing forecast time.