Flow-dependent, inexpensive, high-resolution ensembles for Coupled Ensemble Prediction

Flow-dependent, inexpensive, high-resolution ensembles are generated based on the idea that the truth/verification can often be well approximated by a relatively small and smooth shift of the forecast in space and time. It is difficult to predict the space-time shifts associated with a particular forecast error, but it is possible to estimate their distribution. Here, we generate random fields using a stochastic model that allows the temporal and spatial correlations to be strictly controlled. The technique produces perturbations of sea surface temperature for the COAMPS atmospheric ensemble forecast and atmospheric forcing perturbations for the NCOM ocean ensemble forecast. The model sensitivities to the SST uncertainties and atmospheric forcing uncertainties are evaluated with observations.