Determination of the surface geostrophic velocity field from satellite altimetry

[1] Presently available marine geoid models are not accurate enough to extract the mean surface circulation directly from satellite altimetry. A novel method for estimating the mean velocity field of major ocean current systems is derived from the free surface boundary condition. With a given quasi-geostrophic balance for the horizontal surface flow, a scaling analysis of this boundary condition indicates that although the vertical velocity w is mostly balanced by the local change of the free surface, partial derivativeeta'/partial derivativet, useful information on the mean current ((u) over bar,(v) over bar) is contained in a small ageostrophic departure (partial derivativeeta'/partial derivativet-w). Our method consists in the development of a simple algebraic equation with two unknowns ((w) over bar,(v) over bar) and an adjustable parameter e alpha associated with partial derivativeeta'/partial derivativet, assuming that the latter is proportional to (partial derivativeeta'/partial derivativet-w). Most interestingly, partial derivativeeta'/partial derivativet and all other coefficients of the equation can be determined from altimetry. The altimeter data used is combined TOPEX/Poseidon-ERS gridded data, and the solution is obtained by least squares, minimizing the contribution from the time-variable part of the parameter alpha a and prescribing the zonal direction of the mean current. The method, which is found to be particularly useful for quasi-zonal high-energy current systems, has been validated against direct observations in the Gulf Stream and Southern Ocean. Comparisons with direct observations and Monte Carlo experiments suggest an overall solution error of about 10 cm s(-1). Once calibrated against regional velocity statistics, this method will be able to determine from altimetry the mean or instantaneous surface velocity field down to the frontal scale, with a realism that has been inaccessible because of the geoid constraint.