On the Use of Atmospheric Ensembles to Generate Physical-biogeochemical Ocean Model Uncertainties

The representation of ocean model errors is required in ensemble-based data assimilation methods. In this study, we assess ocean model uncertainties in a high-resolution Bay of Biscay configuration, using atmospheric ensembles and stochastic ocean simulations. The atmospheric ensemble consists of 50 members produced by the European Centre for Medium-Range Weather Forecasts ensemble prediction system (ECMWF-EPS). The stochastic methods focus on uncertainties associated with erroneous atmospheric forcing, improper ocean model parameterizations, and ecosystem state uncertainties. Ensemble simulations are performed over repeated periods to mirror operational forecasting cycles and estimate the performance of the model within a given forecast lead time. The stochastic model is the main method introducing ocean model uncertainties, whilst the ECMWF-EPS system appears to have a moderate impact on augmenting the ensemble spread. Results are assessed in several ways, including event-oriented probabilistic scores. The study is based on two Copernicus Marine projects aim at contributing to ensemble-based ocean data assimilation systems.