The ocean surface roughness affects the radar cross section measured by altimeters. The wind speed is responsible for this roughness and most of present algorithms use the radar cross section (RCS) to infer it. A few authors (Monaldo and Dobson, 1989; Glazman and Greysukh, 1993) emphasized the influence of the sea maturity on satellite measurements. They found a marginal improvement in wind speed retrieval by including significant wave height in their algorithm. In this paper several previously established algorithms relating altimeter radar cross section to ocean surface wind speed are first analyzed. The shapes of the RCS versus wind speed curves are shown to depend mainly on the minimization methods used to generate the model functions. An empirical wind speed algorithm is then derived from the two altimeters (ALT and SSALT) on board TOPEX/POSEIDON (T/P) satellite using a quality controlled data set in which North Atlantic operational surface wind and wave analyses are collocated with altimeter ALT measurements.

Unlike usual algorithms, this new function depends on both the radar cross section and the significant wave height. The improvement in the T/P wind speed estimate seems significant at the 99.9% level. The accuracy of the derived function is evaluated using an independent collocated SSALT and numerical weather prediction models data set. Here again the improvement is significant, but the 90% level because of the small amount of data avaiable. The T/P wind speed estimates are furthermore compared to collocated estimates from National Oceanic and Atmospheric Administration data buoys: the new algorithm retrieves wind speed from Geosat measurements with an accuracy compatible with usual algorithms. ¿American Geophysical Union 1994