Ensemble Data Assimilation and Downscaling in the Bay of Biscay

Using an ensemble data assimilation strategy we revisit the problem of downscaling where an ensemble of lower resolution model runs is used to force a higher resolution model. The objective is to obtain the best estimate of the ocean state in a coastal domain given different sources of information: ocean state and uncertainties from the parent run, local observations, and response of the child model to local forcing uncertainties.

To understand and characterise the main model uncertainties we run ensembles of model experiments for a coastal region over a continental shelf in the Bay of Biscay. The ensembles are formed by perturbing the wind forcing and by applying open boundary conditions from the ensemble of lower resolution experiments. We then perform twin experiments, where we assimilate synthetic data using an ensemble Kalman filter.