Verfahren zur Berücksichtigung von Modellfehlern in der numerischen Wettervorhersage

Various model-error methods are being introduced and their merits and disadvantages discussed. The performance of the different model-error representations is documented for a particular meso-scale ensemble system: for this ensemble system stochastic perturbations and the multi-physics approach resulted in clearly better probabilistic forecasts, while the multi-parameter method showed little impact.

Best results were obtained by combining all model-error schemes suggesting that no single representation alone can currently describe the complex nature of model error.