State estimation improvement in strongly coupled data assimilation with a two-scale Lorenz model

Coupled data assimilation (CDA) is a powerful strategy for integrating observations with coupled numerical models. This strategy holds great potential for enhancing weather and climate reanalysis and prediction. How to address cross-scale interactions in CDA is an important issue. In particular, the cross-scale interactions in the strongly coupled data assimilation (SCDA) framework pose substantial challenges. In this study, increasing the state estimation accuracy using an ensemble adjustment Kalman filter based on the two-scale Lorenz’96 (tsL96) model is investigated. Using the SCDA framework, we adopt cross-component localization factors and several covariance inflation schemes to address the filter divergence problem. The results show that ensembles of an appropriate size can achieve good assimilation results, the optimal localization parameters are scale-dependent for the model variables, and the adaptive inflation scheme outperforms the static fixed and relaxation-to-prior spread schemes. Although these experiments were carried out using an ideal framework, this study provides a valuable reference for improving estimation accuracy with the SCDA framework in operational simulation and prediction models.