Application of the model space vertical localization in a global LETKF at JMA
The covariance localization is applied to the ensemble-based data assimilation system to reduce the impact of a sampling error and a rank deficiency of background error covariance due to the limited ensemble size. Ensemble Kalman Filters (EnKF) including the Local Ensemble Transform Kalman Filter (LETKF) utilizes the localization in the observation space. However, it is known that the observation space localization could have a substantial detrimental impact on its accuracy especially when non-local observations such as satellite radiance observations are assimilated. Recently, the model space vertical localization scheme using modulated ensemble perturbations that can be applied to the LETKF framework with a relatively reasonable computational cost has been proposed. In this presentation, the application of the model space vertical localization to a global LETKF at JMA will be explained and its comparison with the observation space localization will be presented.
Mr. Yoichiro Ota Japan Meteorological Agency