A Synoptic-Dynamic Analysis of Cases of Major Underestimation of the Forecast Uncertainty

Ensemble forecasts are prepared to account for the effects of uncertainties in the initial conditions and the formulation of the model on the forecasts. The THORPEX Interactive Grand Global Ensemble (TIGGE) database allows researchers to obtain ensemble forecast data from the leading operational global forecast centers of the world. Recent results of our research group suggest that, on average, the ensembles included in the TIGGE data set underestimate the forecast uncertainty. Our investigation shows that the overall problem is due to a few episodes of large underestimation of the forecast uncertainty. We selected two such episodes for the 96-hour lead time UKMO ensemble forecasts by using both NCEP and ECMWF analyses for the verification of the forecasts. The two verification times are 0000 UTC January 6, 2012 and 1200 UTC February 10, 2012. In the first case, the ensemble underestimates the errors associated with the prediction of an upper tropospheric wave packet over the North Pacific, while in the second case it underestimates the uncertainty in the prediction of a cyclogenesis over the north Atlantic region.