17:15 〜 18:30
[MGI29-P03] Local Ensemble Transform Kalman Filter Experiments with Hybrid Background Error Covariance
キーワード:データ同化、アンサンブルデータ同化、DFS、ハイブリッドデータ同化、rank deficiency、LETKF
Recent operational numerical weather prediction (NWP) systems have achieved significant forecast improvements by using hybrid background error covariance (HBEC) that linearly combines climatological and ensemble-based error covariance. Currently, the HBEC has been used mainly in variational data assimilation systems to use the flow-dependent error covariance in addition to static error covariance. This study explores using the HBEC within the ensemble Kalman filter (EnKF). EnKF approximates the error covariance matrix by sample estimates using the ensemble perturbations. EnKF potentially results in suboptimal analyses if the ensemble size is smaller than the number of assimilated observations. Kretchmer et al. (2015) proposed adding a collection of climatological perturbations to the forecast ensemble mean to boost the rank of the background error covariance. This study followed the Kretchmer et al. (2015)’s approach and implemented the HBEC coupled with the local ensemble transform Kalman filter (LETKF). In addition, this study investigates relationship between the rank deficiency and the degrees of freedom for signal (DFS). With the 40-variable Lorenz-96 model, we found that HBEC outperformed the standard LETKF especially when the ensemble size is smaller than the number of assimilated observations. With the HBEC, the optimal localization radius and DFS increased. These suggest that the HBEC mitigate the rank deficiency and assimilate more observations effectively than the standard LETKF. We also test the HBEC with an intermediate atmospheric model known as the SPEEDY (Simplified Parameterizations, Primitive Equation Dynamics). This presentation will include the most recent progress up to the time of the conference.