The NASA scatterometer (NSCAT) estimates the wind speed and direction of near-surface ocean wind. Several possible wind vectors (termed ambiguities) are estimated for each resolution element known as a wind vector cell (WVC). Typically, the speeds of the possible wind vectors are nearly the same, but the directions are very different. The correct wind must be distinguished in a step called ambiguity removal. Unfortunately, ambiguity removal algorithms are subject to error. In an attempt to evaluate the accuracy of the Jet Propulsion Laboratory NSCAT product, we use a new model-based quality assurance algorithm that uses only NSCAT data. The algorithm segments the swath into overlapping 12¿12 WVC regions and classifies each region according to estimated quality. The 9-month NSCAT mission data set is analyzed. In 82% of the regions the ambiguity removal is over 99% effective, with the ambiguity errors correctable using a model-based correction technique. In 5% of the regions, areas of significant ambiguity error are found. For remaining regions, all of which have root-mean-square (rms) wind speeds less than 4 m s-1, there is too much uncertainty in the wind field model or too much noise in the measurements to uniquely evaluate ambiguity selection with sufficient confidence. We thus conservatively conclude that for the set of regions with rms wind speed greater than 4 m s-1, NSCAT ambiguity removal is at least 95% effective. ¿ 1999 American Geophysical Union