A model for the ocean surface normalized radar cross section &sgr;0 is derived from 3 months of NASA scatterometer (NSCAT) observations (September 15 to December 18, 1996). The model expresses &sgr;0 as a function of wind speed, relative wind direction, incidence angle, and polarization. The dependence of &sgr;0 on wind speed is based on collocated special sensor microwave/imager (SSM/I) satellite wind retrievals and European Center for Medium-Range Weather Forecasts (ECMWF) model winds. We find that at low winds (<5 ms-1), the SSM/I winds are more reliable than ECWMF, probably owing to small location errors in the ECMWF wind features. The primary wind direction dependence of &sgr;0 (i.e., the second harmonic) is found from histograms of the &sgr;0 difference between the forward and aft antennas. The &sgr;0 versus wind speed relationship is adjusted for cross-swath incidence angle differences and is then incorporated into the NSCAT 1 model used to process the 10-month (September 15, 1996, to June 29, 1997) NSCAT data set. The resulting NSCAT 1 wind vectors are compared to ECMWF wind fields and buoys. The mean and standard deviation of the NSCAT minus ECMWF (buoy) wind speed difference are 0.05 and 1.78 ms-1 (-0.29 and 1.26 ms-1), respectively. The wind direction mean and standard deviation differences are 0.8¿ and 18.5¿ (7.9¿ and 15.7¿), respectively. The difference between the NSCAT and the ECMWF (buoy) direction exceeds 90¿ only 1.1% (1.2%) of the time. We have no explanation for why the buoy wind directions are biased 8¿ relative to both NSCAT and ECMWF. ¿ 1999 American Geophysical Union