Multi-model superensemble forecasts for Weather and seasonal climate

In this paper we illustrate the performance of a multi-model superensemble that shows superior forecast skills compared to all participating models. The model comparisons include global weather, seasonal climate and hurricane track and intensity forecasts. The emphasis of this presentation will be over the tropics. The superensemble concept is first illustrated for a low order spectral model from which multi-models and a "nature run" were constructed. We divide 200 time units into a training period (70 time units) and a forecast period (130 time units). The multi-model forecasts and the observed fields (the nature run) during the training period are subjected to a simple linear multiple regression to derive the statistical weights for the member models. The superensemble forecasts, generated for the next 130 forecast units, outperform all the member models of the superensemble. This procedure was deployed for the multi-model forecasts of global weather, multiseasonal climate and hurricane track and intensity forecasts. For each forecast type we clearly demonstrate a superior performance of the multi-model statistical superensemble compared to the performance of the multi-models. Seasonal and multi-seasonal forecasts demonstrate a major success of this approach for the atmospheric general circulation models (AGCMS) where the sea surface temperatures and the sea ice are prescribed. In many instances, a major improvement in skill of the superensemble over the best models is noted. A number of other issues related to the errors of climatology, optimum number of models required for superensemble, errors of straightforward ensemble averages and the bias of models are also addressed in this analysis.