Evaluation of Different Model-Error Schemes in Mesoscale Ensemble Forecasts (Invited)

The performance of several different model-error schemes and selected combinations is verified for probabilistic forecasts with the WRF-ARW mesoscale ensemble system over the Contiguous United States. Including a model-error representation leads to more spread and small, but significant increases in forecast skill. In the free atmosphere, a stochastic kinetic-energy backscatter scheme performs best, while multiple-physics schemes tend to be superior near the surface. Combing multiple stochastic and deterministic parameterizations results in the biggest improvement throughout. To investigate if the model-error schemes are able to represent structural uncertainty or if the improved skill is solely the result of an increase in ensemble spread, two additional computations were performed: First, the Brier score is decomposed into reliability, resolution and uncertainty, which have different sensitivities to spread. Secondly, all forecasts are calibrated to have the same variance as the observations, which results in similar ensemble spreads. In the raw and re-calibrated ensemble systems, the decomposition of the Brier score improves both, the resolution and reliability component, indicating that the benefits of including a model-error scheme goes beyond increasing the ensemble spread. The improvements are quantified for biased and de-biased forecast. We find that the relative performance of the different model-error schemes remains similar in the raw and postprocessed ensemble experiments.