The Super Model forecasting method applied to the imperfect model scenario
Recently a new forecast method that goes beyond the traditional ensemble forecasting was proposed. While traditional ensemble forecasts try to determine the uncertainty of the dynamics by initializing the model with many different states, the so called super-model (SUMO) ensemble method tries to reduce the uncertainty of the forecast by coupling different models and let them exchange information during run-time. As we have seen so far SUMO was evaluated in what we would call the perfect model scenario (PMS), that is the models and the system have structurally the same equations. Therefore the only mismatch between the model and the system tested so far is that the ensemble of models and the system have different coefficients or control parameters. Here we report on the much more realistic imperfect model scenario (IPMS). We assume that in the modeling process there occurred an error and the flow equations governing the models and the systems are therefore different. It is far from obvious that techniques developed for the PMS work in the IPMS as well.