Marc Bocquet, CEREA joint laboratory Ecole des Ponts ParisTech and EdF R&D, University Paris-Est, Champs-sur-Marne, France
Nonlinear four-dimensional ensemble variational data assimilation for chaotic geophysical models
Abstract: This talk will offer a theoretical overview of a selection of data assimilation methods for nonlinear chaotic models, with the exception of particle filters. The focus will be on nonlinear four-dimensional ensemble variational methods, from which the iterative ensemble Kalman smoother (IEnKS) is a sleek archetype. Why the IEnKS should theoretically be more accurate than the ensemble Kalman filter (EnKF), 4D-Var or current implementations of 4DEnVar, will also be explained. Important directly related issues will be discussed, such as the generation of updated perturbations, the role of model adjoint (or not) in these schemes, four-dimensional localisation, the theoretical importance of quasi-static variational implementations, as well as the convergence and the geometry of the ensemble with respect to the unstable subspace of the dynamics.