A Quailty Assurance Algorithm for SeaWinds

The scatterometer wind retrieval process produces several possible wind vector choices or ambiguities at each resolution cell. Ambiguity selection routines are generally ad hoc and often result in ambiguity selection errors. It is important to locate areas of ambiguity selection error to assess the quality of scatterometer wind data. A quality assurance algorithm is presented based on comparing ambiguity-selected winds from SeaWinds on QuikSCAT to a low order wind field model fit. Regions exceeding error thresholds are rated and flagged as possible ambiguity selection errors. Appropriate error thresholds and additional flagging criteria are set through an analysis of false alarms versus missed detections on a manually-inspected training data set. The algorithm correctly identifies 97% of the regions manually flagged as ambiguity selection errors in the training set with a false alarm rate of less than 2%. Applying the algorithm to the entire QuikSCAT data set, we conclude that the ambiguity selection is over 95% effective on regions of rms wind speed greater than 3.5 m/s. The algorithm validates that higher noise occurs at nadir and in low wind speed regions. Additionally, fewer estimated ambiguity selection errors occur at nadir and on the swath edges due to a larger ambiguity set in those regions. The percentage of ambiguity selection errors are found to be highly correlated with cyclonic storm and rain occurences.