In this study, real Geosat altimetric sea level observations for the 1-year period extending from January to December 1987 were assimilated into a realistic wind-driven numerical synoptic ocean model of the California Current. The objective was to evaluate the effectiveness of using a realistic synoptic ocean model to interpolate (dynamically) real altimetric sea level observations from the Geosat ERM onto a regular grid. First, the model mean sea level was shown to simulate qualitatively the observed mean sea level found in in situ CalCOFI hydrographic data. Second, the model was shown to simulate approximately the space-time statistics of mesoscale eddy variability contained within the altimetric sea level observations. Third, prior to assimilation the altimetric sea level observations were referenced using this model mean. Then the referenced altimetric sea level observations were assimilated into the model, with the resulting sea level residuals compared with estimates from in situ (expendable bathythermograph) observations. This comparison yielded nearly exact agreement at low frequency (i.e., the semiannual cycle), but less agreement on month-to-month time scales of variability (yet still significant), probably owing to the unfiltered nature of the in situ estimates. Subsequent comparisons of the dynamically interpolated altimetric sea level residuals with those obtained from the statistical interpolation yielded marked improvement in the latter over the former; i.e., dynamical interpolation allowed observed mesoscale variability to be conducted from the Geosat ERM repeat tracks into the regions between tracks with little change in magnitude and structure, not possible with the statistical interpolation.

The latter tended to produce extrema on grid points that coincided with the track lines. As such, dynamical interpolation demonstrated marked improvement in resolution, time-space continuity, intensity, and gradient structure of mesoscale eddy activity over that possible with statistical interpolation. Model-data assimilation also allowed for the determination of the vertical structure of the mesoscale eddy activity. Moreover, it yielded a forecasting capability that had a statistically significant improvement over persistence in specifying mesoscale eddy activity 2--3 weeks into the future. ¿ American Geophysical Union 1990