Development of a statistical method for eliminating improbable wind aliases in scatterometer wind retrieval

Wind velocities over the ocean can be estimated using measurements from spaceborne scatterometers by inverting the geophysical model function (GMF) which relates normalized backscatter to wind velocity. Current estimation procedures employ maximum-likelihood techniques. Unfortunately, there are several local maxima of the maximum-likelihood function. As a result, several (2-6) wind estimates are returned as possible solutions at each wind vector cell. An ambiguity-removal step is required to determine a wind field. In this paper, we develop a statistical test to distinguish among the maxima of a maximum likelihood equation, and apply it to wind estimation. An upper bound is derived on the probability of error if a lower likelihood wind estimate is discarded. This bound is used to eliminate improbable wind solutions. Using this procedure we show that for most ERS-1 wind vector cells the number of wind estimates can be reduced to two. This reduces the complexity of the ambiguity-removal step while at the same time increasing the confidence in the entire retrieved wind field