Strongly Coupled Data Assimilation Using Leading Averaged Coupled Covariance (LACC) for Real World Observation

Strongly coupled data assimilation (SCDA) is believed to be a promising method for initializing the coupled Earth system models by providing a more physically consistent and balanced initial condition. However, its implementation in the coupled model still faces many scientific and technical challenges, a significant one of which is the different time scales between model components. The leading averaged coupled covariance (LACC) approach takes advantage of the strong asymmetry feature of ocean-atmosphere correlation to enhance the signal-to-noise ratio when calculating the cross-component coupled covariance. Two important challenges of the implementation and evaluation of a SCDA system in real world are the absence of truth and the existence of model bias. Here we evaluate the performance of SCDA using the LACC approach to assimilate real world observation in a fully coupled general circulation model systematically and quantitatively. To provide a reasonable and convincing evaluation when not knowing the truth, four kinds of comparison schemes are conducted for the performance assessment using the weakly coupled one (WCDA) as the benchmark. Both the root-mean-square error (RMSE) and anomaly correlation coefficient (ACC) are utilized to analyze the results, with full-value RMSE divided into the climatology part and the anomaly part. The results show that compared with WCDA, though the improvement of anomaly RMSE is not consistently clear, the SCDA using LACC can reduce the climatology RMSE. And the ACC can also be significantly improved both in the ocean and in the atmosphere. With appropriately selected averaging length, the SCDA using LACC can also outperform the simultaneous coupled covariance approach.