Ensemble Kalman filter

Summary

The ensemble Kalman filter is a method of data assimilation. It replaces the covariance matrix of the Kalman filter by the sample covariance of a finite ensemble of model states. Each member of the ensemble is run forward by the full, possibly nonlinear, model, and is then corrected with the observations by a Kalman-type update. It avoids storing and propagating a covariance matrix of the size of the state. That is impossible for weather and ocean models with millions of variables. It was introduced by Geir Evensen in 1994 and is now used in operational weather and ocean forecasting.