Coupled Data Assimilation for Enso Prediction∗

An outstanding problem with present ENSO forecast systems is that most of them are initialized in an uncoupled manner, that is, no feedbacks are allowed between the ocean and the atmosphere during data assimilation and model initialization. Such an approach may produce realistic initial states, but not necessarily the optimal conditions for skillful forecasts, because model-data mismatch can cause serious initialization shock. A more reasonable approach is to initialize forecast systems using a coupled approach, which assimilates data, both oceanic and atmospheric, into the coupled models that are used for forecast. Here we briefly review the progress in this important research area. In particular, based on the evolution history of an intermediate couple model, and some preliminary results from a state-of-the-art forecast system, we demonstrate the impact and necessity of coupled data assimilation, and we suggest it may hold a key for further improvement of the predictive skill of present ENSO models.