A review of the theoretical potential, & limitations of the Ensemble Kalman Filter:
a comparison with 4D-Var


UK Met Office

This review was undertaken to address the question "Should the Met Office have an EnKF R&D project?".

After a brief review of published EnKF methods, various properties of the EnKF are examined in simple idealised examples. Strengths and weaknesses compared to 4D-Var are discussed. A method for using ensemble generated covariances in a variational analysis scheme, injcluding localisation, is presented.

The conclusion is that the Met Office should continue to have a main "control" high-resolution forecast, with a 4D-Var based assimilation, but with background error covariances enhanced using errors of the day patterns from available ensembles. Methods of generating ensembles of pertubations are being investigated.