The analysis of altimetry measurements along repeated satellite ground tracks, as in the Geosat Exact Repeat Mission, provides an excellent way of determining temporal variability in the sea surface height. This collinear analysis is usually carried out in two steps: first, by determining a mean sea level profile and then by removing a radial orbit error. We propose a new unified analysis method for the simultaneous estimation of orbit error, mean sea surface height, and residual sea surface heights (variability) with respect to each pass. All passes are treated in exactly the same way; the consequent rank or datum defect in the model is overcome by appending a minimal set of conditions, chosen either to give no overall distortion of the orbits or otherwise a best fit to a reference profile, such as the geoid. Gaps in the altimeter data are easily accommodated. Numerical results of the method are shown using data from the first year of the Geosat Exact Repeat Mission. Some consideration is given to the choice of radial orbit error model. The results indicate that for orbit error removal, a sinusoid model is the most appropriate over a one-revolution cycle. For passes less than one revolution in length, a component of the longer wavelength ocean signal is lost if using the sinusoid model. If passes longer than about two revolutions (some 200 mins) are analyzed, additional terms are needed to eliminate nongravitational perturbations. ¿ American Geophysical Union 1992