Comparison of hybrid gain data assimilation update scenarios in an operational global forecast system

Hybrid gain data assimilation offers a flexible and practical alternative to traditional hybrid covariance data assimilation by combining gain matrices from ensemble-based Kalman filter (EnKF) and variational methods. This study is the first to compare two update scenarios of hybrid gain data assimilation, namely the sequential and parallel update scenarios, within an operational global weather prediction system, the Taiwan Global Forecast System (TGFS). By correcting the EnKF analysis with 3DVAR adjustments, the sequential update scenario with an optimal hybridization weight demonstrates advantages in the Northern Hemisphere and tropics during the boreal summer, but performs worse in the Southern Hemisphere compared to the parallel update scenario. In contrast, the parallel update scenario produces hybrid analysis by averaging the EnKF and 3DVAR analyses, exhibiting greater robustness to weight variations than the sequential one. In addition, the introduction of non-affine hybridization weights, which fix the EnKF weight at one, is found to be beneficial for the parallel update scenario even with a limited ensemble size. Overall, this study demonstrates the promising performance and flexible experimentation with the hybrid gain data assimilation algorithm in an operational system, and it further contributes practical insights into update scenario selection and hybridization weight configuration.