Abstract
In this paper, a Lagrangian assimilation method is presented and implemented in a realistic OPA OGCM with the goal of providing an assessment of the assimilation of realistic Argo float position data. We focus on an application in the Mediterranean Sea, where in the framework of the MFSTEP project an array of Argo floats have been deployed with parking depth at 350 m and sampling interval of 5 days. In order to quantitatively test the method, the "twin experiment" approach is followed and synthetic trajectories are considered. The method is first tested using "perfect" data, i.e. without shear drift errors and with relatively high coverage. Results show that the assimilation is effective, correcting the velocity field at the parking depth, as well as the velocity profiles and the geostrophically adjusted mass field. We then consider the impact of realistic datasets, which are spatially sparse and characterized by shear drift errors. Such data provide a limited global correction of the model state, but they efficiently act on the location, intensity and shape of the described mesoscale structures of the intermediate circulation.