A Geosat altimeter wind speed algorithm is derived by cross-calibrating Geosat and Seasat altimeter estimates of the normalized radar cross section &sgr;0 and modifying an existing Seasat altimeter wind speed model function to obtain a model function appropriate for Geosat observations. It is argued that the &sgr;0 distribution measured by an altimeter is relatively stable over a sufficiently large geographical region and a long enough time period. Systematic difference between &sgr;0 estimates from two altimeters can therefore be identified based on comparisons of their &sgr;0 histograms. Any such systematic differences can then be corrected using independent &sgr;0 estimates. When this method is applied to the Geosat and Seasat altimeters, a systematic difference between the two &sgr;0 histograms is shown to be consistent with differences between Seasat altimeter and nadir seasat scatterometer estimates of &sgr;0 deduced independently by a previous study. This supports the conclusions that (1) the &sgr;0 distribution is stable, and (2) the Seasat altimeter estimates of &sgr;0 were miscalibrated. After modifying the existing Seasat altimeter wind speed algorithm to account for this apparent &sgr;0 error, the resulting Geosat estimates of wind speed agree with high-quality buoy observations to within an rms difference of less than 2 m/s. ¿American Geophysical Union 1991