Ideas for Ensemble Kalman Filtering

Eugenia Kalnay


In this seminar we present clean comparisons between EnKF and 4D-Var made in Environment Canada, showing that their operational 4D-Var and EnKF systems had identical scores in the NH, but EnKF was ahead in the SH. 4D-Var with a background error covariance based the EnKF (one of several possible hybrid approaches) was also shown to be superior to 4D-Var with constant B.

We briefly describe the Local Ensemble Transform Kalman Filter (LETKF) as a representative prototype of Ensemble Kalman Filter, and give several examples of how advanced properties and applications that have been developed or explored within 4D-Var can be simply adapted to the LETKF without requiring an adjoint model. These include a no-cost smoother, "outer loop" and "running in place" algorithms that allow dealing with nonlinearities and accelerate spin-up, a comparison of several simple methods to handle model errors, a coarse resolution LETKF analysis by weight interpolation without degrading the analysis, adjoint forecast sensitivity to observations without an adjoint model, and how to estimate the optimal inflation and observation errors within the EnKF.

Although the Ensemble Kalman Filter is less mature than 4D-Var, its simplicity and its competitive performance with respect to 4D-Var suggest that it could become the method of choice. It is frequently stated that the best approach should be a hybrid that combines “the best characteristics” of both EnKF and 4D-Var (e.g. Lorenc 2003). Unfortunately this also brings the main disadvantage of 4D-Var to the hybrid system, namely the need to develop and maintain an adjoint model. This makes the hybrid approach attractive to operational centres that already have appropriate linear tangent and adjoint models, but not otherwise.