VARIATIONAL DATA ASSIMILATION IN A REDUCED GRAVITY MODEL OF THE NORTHWESTERN INDIAN-OCEAN - EXPERIMENT WITH KELVIN WAVES

Variational method of assimilating data in a reduced gravity model of the northwestern Indian Ocean is discussed. A cost function representi ng the misfit between the model and the data is minimised with the mod el equations acting as constraints. The initial conditions of the mode l are used as control parameters and the best fit initial conditions a re determined as a result of minimisation carried out using a variable -storage quasi-Newton method. Experiment has been done with simulated data generated from a control sun of the model describing propagation of Kelvin waves. Assimilation of upper layer thickness data alone has been found to be sufficient for a good recovery of the initial state. The method has been found to have good predictive potential and is par ticularly suitable for assimilating data from satellite altimeters.