CONDUCTIVITY TEMPERATURE DEPTH DATA ASSIMILATION INTO A 3-DIMENSIONALQUASI-GEOSTROPHIC OPEN-OCEAN MODEL

A variational assimilation technique is applied to processing the cond uctivity-temperature-depth (CTD) observations obtained in the Gulf Str eam extension in May 1990. The data are assimilated into a quasigeostr ophic open ocean model based on the evolution equations for buoyancy a nd vertical mean relative vorticity. The solution is controlled via th e initial and boundary conditions for model equations and the unknown forcing in the vorticity equation. The best fit for CTD data is compar ed with the results of optimal interpolation and the ordinary model fo recast. The assimilation solution displays obvious advantages of the a djoint assimilation technique in comparison with these traditional met hods of data processing. We have also performed an error analysis of t he resultant solution via estimation of the Hessian matrix elements as sociated with the minimized cost function. The results show the import ant role of regularization terms, which smooth the solution, making it less dependent on the first guess in the space of control variables a nd filtering out the data components which are not described by the mo del. The analysis has also revealed a number of problems connected wit h slow convergence of the assimilation procedure owing to non-efficien t descent in the subspace of boundary conditions.