ASSIMILATION OF SEA-SURFACE HEIGHT DATA INTO AN ISOPYCNIC OCEAN MODEL

Two different methods for assimilating sea surface height data into an isopycnic primitive equation model were developed and tested for the idealized case of an analytical Gaussian warm core (anticyclonic) eddy ; then implemented using remotely sensed data from the Brazil-Malvinas confluence region. The first method makes a geostrophic assumption ab out the flow to relate sea surface height field gradients to the model velocities. The second method nudges the model sea surface height its elf toward the observed values using a linear vertical influence funct ion in the upper layers. The relationship between the surface height o bservations and the layer interface displacements is derived from obse rvations of eddies in different regions of the Atlantic Ocean. Both as similation methods were successful in transferring the dynamical influ ence of the sea surface height measurements deep into the water column , but a combination of both gave the best results. The application of both methods reproduced the detailed mesoscale features of the real oc eanic circulation when assimilating Geosat sea surface height measurem ents from the Brazil-Malvinas confluence region into an isopycnic ''bo x'' ocean model. The velocity fields show deep anticyclonic (cyclonic) circulations of 66 cm s(-1) (52 cm s(-1)), which are present in obser vations of eddies and meanders. When compared to the no assimilation r un, the assimilated sea surface height held exhibits a richer spectrum , with energy in all spectral bands strongly correlated with the obser ved values. The resulting band of energy higher than 150 cm(2) between 300 and 600 km is in agreement with previous studies of this region. The rms error between the model sea surface height and the Geosat data was reduced from similar to 13 cm in the no assimilation run to 2 cm rms after the assimilation. These experiments demonstrate the flexibil ity and effectiveness of the Newtonian relaxation (nudging) method for assimilating real data into a complex, multilayer primitive equation ocean model.