Javier Amezcua, University of Reading
A weak-constraint 4D-Ensemble-Var
CoauthorsPeter Jan van Leeuwen, Michael Goodliff
Abstract: 4D-Ensemble-Var is a hybrid data assimilation method which avoids the need to compute tangent linear and adjoint operators for the evolution model and observation operators. This is done by taking advantage of 4D space-time covariances from an ensemble. Localisation in this case is not trivial, since localising a 4D covariance with a static localisation function can lead to loss of important information. We present a new 4D-En-Var formulation that uses the weak-constraint framework (model error). We demonstrate how this formulation can ameliorate the problem with localisation.