A new insight into localisation methods in ensemble Kalman filter systems

Pavel Sakov and Laurent Bertino conducted a comparative study of two non-adaptive localisation methods popular in operational oceanographic and atmospheric data assimilation systems: covariance localisation (CL) and local analysis (LA). Both methods are applied mainly on empirical basis, and the relation between them remains largely unknown.

The authors conclude that while CL and LA are essentially different methods, in practice they normally produce essentially similar results. Apart from this main conclusion, the study contains a number of interesting theoretical advances. The two methods are compared after being formulated in a common framework using representation via tapering of the ensemble anomalies and left multiplied ensemble square root schemes.

The research has been conducted in the course of eVITA-EnKF project by Research Council of Norway.


Add comment

Login to post comments