The model is a reduced-gravity, primitive equation, upper-ocean GCM with a variable-depth oceanic mixed layer. Wind stress and heat flux, calculated from the NCEP wind speed and cloud cover (ISCCP), are used to force the model. The assimilation scheme is an approximation to the extended Kalman filter in which the error covariances and corrections to the forecasts are only calculated in a reduced-dimension subspace spanned by a small number of empirical orthogonal functions (EOFs). To make the assimilation of the entire 7.5 year topex period feasible, the asymptotically stationary error covariances are used.
Initial experiments using simulated SSH data demonstrated the ability of the assimilation method to successfully constrain the circulation and sub-surface thermal structure. Results from assimilating actual altimetry data resulted in a 25% reduction in the rms misfit with observed SSH relative to a pure model integration. Evaluation of the impact on the sub-surface fields is more difficult due to a lack of independent measurements, but changes to the thermocline structure appear reasonable.