The Predictability of the Wind-Driven Ocean Circulation Investigated Through Data Assimilation

The major problem of ocean data assimilation is arguably the high dimension of the state vector, hence innovative methods must be used to obtain efficient, and affordable, data assimilation packages which can be used with fully realistic ocean general circulation models (OGCM’s). One such approach is to use an approximation to the extended Kalman filter in which the error covariances and corrections to the forecasts are calculated in a reduced dimension sub-space spanned by a small number of empirical orthogonal functions (EOFs) of the time-evolution of the unconstrained OGCM. The strategy used was to first test the method in an idealized, quasi-geostrophic model of the wind-driven circulation in a double gyre ocean as proof-of-concept ( publication 3), and then apply the methodology to a fully realistic OGCM of the Tropical Atlantic Ocean ( publication 4). In the first idealized application, the filter was implemented using both temporally evolving and asymptotically stationary error covariances, and also compared with the more traditional approach in which the error covariance is prescribed a simple functional form. The results showed not only that the reduced-rank filter outperforms the assimilation using the prescribed error covariance; but also that the performance of the filter with stationary error covariance is extremely similar to the much more computationally expensive filter with flow-dependent covariance. In the assimilation with the Atlantic OGCM, first synthetic SSHA simulated by the model are assimilated in a twin-experiment approach; second