Mark Buehner, Environment and Climate Change Canada
An Ensemble Kalman Filter for Numerical Weather Prediction based on Variational Data Assimilation: VarEnKF
CoauthorsRon McTaggart-Cowan, Sylvain Heilliette
Abstract: Several NWP centers currently employ a variational data assimilation approach for their deterministic forecasts and a separate EnKF system that is used for both initializing ensemble forecasts and for providing ensemble covariances for the deterministic system (e.g. ECCC, NCEP, Met Office). This presentation describes a new approach for performing the data assimilation step within a perturbed-observation EnKF by adapting the variational algorithm used for deterministic data assimilation. The analysis increment is computed with a variational assimilation approach separately for the ensemble mean and for all of the ensemble perturbations. Several practical benefits may be realized by using the same data assimilation approach for both deterministic and ensemble prediction in terms of code development and the testing of system changes. In addition, the use of a variational approach may allow the assimilation of a larger volume of observations for updating the ensemble mean than with current EnKF algorithms. Preliminary results with the Canadian global 256-member system show that a particular configuration of VarEnKF can provide significantly improved ensemble forecasts than the current EnKF with only a modest increase in computational cost. Moreover, since the analysis update for each ensemble perturbation is computed independently, this part of the VarEnKF approach would scale perfectly up to a very large number of processors.