A detailed geophysical evaluation of the initial NASA scatterometer (NSCAT) wind data sets was performed in order to determine the error characteristics of these data and their applicability to ocean surface analysis and numerical prediction. The first component of this evaluation consisted of collocations of NSCAT data to ship and buoy wind reports, special sensor microwave imager wind observations, and National Centers for Environmental Prediction and Goddard Earth Observing System (GEOS) model wind analyses. This was followed by data assimilation experiments to determine the impact of NSCAT data on analysis and forecasting. The collocation comparisons showed the NSCAT wind velocity data to be of higher accuracy than operational ERS 2 wind data. The impact experiments showed that NSCAT has the ability to correct major errors in analyses over the oceans and also to improve numerical weather prediction. NSCAT data typically show the precise locations of both synoptic-scale and smaller-scale cyclones and fronts over the oceans. This often results in significant improvements to analyses. Forecast experiments using the GEOS model show approximately a 1-day extension of useful forecast skill in the southern hemisphere, in good agreement with the results of Observing System Simulation Experiments conducted prior to launch. ¿ 1999 American Geophysical Union