Data Assimilation and Predictability Studies for the Coupled Ocean-atmosphere System

A S OCEANIC DATA SETS increase dramatically in quality and quantity in the near future, and both oceanic and atmospheric models improve apace, the predictability of the coupled ocean-atmosphere system will become more important on the theoretical level and more critical on the practical level. Predictability of the atmosphere with prescribed sea-surface temperatures (SST) has been evaluated; numerous studies indicate that two initially very similar atmospheric states will lead to time evolutions that on the average diverge and become uncorrelated over an interval on the order of 2 weeks. There is also a growing literature on the predictability of the upper ocean with prescribed atmospheric wind stress and heat fluxes. But the variability, and hence predictability, of the coupled system is quite different from the sum, product, or any other simple function of its parts (Ghil et al., 1991a). The long-term goal of our work at UCLA is to provide a description, understanding, and prediction of the coupled ocean-atmosphere system as complete and reliable as that which now exists for the atmosphere alone. Our approach is to develop methods for data assimilation from sequential estimation and control theory and for predictability studies from dynamical systems and statistical turbulence theory: these methods are then tested on a variety of models, ranging from simple models amenable to analytical treatment to coupled ocean-atmosphere general circulation models (GCMs).