Comparison of a single-model EPS with a multi-model ensemble consisting ofa few operational models

Since the introduction of operational ensemble forecasts in Numerical Weath er Prediction (NWP) more than 5 years ago, the dispute on how to best deter mine the initial perturbations has largely dominated the direction of resea rch in the field of ensemble prediction. While it is important to consider uncertainties in the initial condition, errors due to model physics or the model numerics and truncation provide another source of forecast errors and might also be considered in ensemble prediction. In this study, we compare the performance of 2 fundamentally different ensemble schemes. First, the ensemble prediction system (EPS) of the European Centre for Medium Range Fo recasts is taken as a representative of the single-model approach based on the perfect model assumption and thus taking only the uncertainty in the ob servations into account. Second, a virtual ensemble comprised of the operat ional forecasts of 4 NWP centers as a "gratis" candidate of the multi-model approach which, in addition, takes model errors into account. The comparis on is based on forecasts of 500 hPa fields over Europe for a summer and a w inter period in 1997 and on diagnostics ranging from various measures for t he performance of the ensemble means to the statistical consistency and dis crimination properties of the ensembles. The different sizes of both ensemb les poses the main difficulty for the interpretation of the results. If the ensemble size is not considered as a criterion for the evaluation, the res ults lead to controversial conclusions; but when penalizing for an overly l arge and inefficient ensemble the results are for the most part consistent, and one has to conclude that the multimodel ensemble performs better in mo st forecast aspects.