A Field-wise Wind Retrieval Algorithm for the Nasa Scatterometer

A FIELD-WISE WIND RETRIEVAL ALGORITHM FOR THE NASA SCATTEROMETER Charles G. Brown Electrical and Computer Engineering Master of Science An alternative to traditional wind retrieval techniques that does not rely on numerical weather prediction winds is developed and applied in this thesis. Traditional satellite scatterometer wind retrieval algorithms consist of point-wise wind estimation and ambiguity removal. Point-wise estimation yields multiple estimates, or ambiguities, for the average near-surface ocean wind in each scatterometer resolution element. The presence of ambiguities requires subsequent ambiguity removal to select unique wind estimates. Even though some point-wise ambiguity removal techniques perform well, they are still subject to error. Further, the most successful ones, such as the \nudging" method employed by the Jet Propulsion Laboratory (JPL), incorporate outside information. One alternative to the traditional methods is eld-wise wind retrieval. Like point-wise retrieval, the eld-wise method consists of estimation and ambiguity removal stages. However, eld-wise estimation and ambiguity removal employ a wind eld model that exploits the spatial correlation in wind elds to process large numbers of resolution elements simultaneously. A eld-wise retrieval algorithm that uses only scatterometer measurements is designed for the NASA Scatterometer (NSCAT) and is applied to data from that instrument. Test results indicate that the eld-wise method compares favorably with the standard NSCAT data product obtained from the Jet Propulsion Laboratory (JPL); however, the eld-wise method has the advantage of not relying on outside information. It is further concluded that the eld-wise method can be used to repair severe ambiguity removal failures in the output of the median lter, a traditional pointwise ambiguity removal scheme. An additional application of the eld-wise method is validation of the JPL product. The eld-wise method and a related model-based technique are used to assess the reliability of the JPL product, and both methods suggest over 96% accuracy, which is consistent with other recent validation results.