Towards Bayesian estimator selection for QuikSCAT wind and rain estimation

The QuikSCAT scatterometer infers wind vectors over the ocean using measurements of the surface backscatter. During rain events the QuikSCAT observations are subject to rain contamination. Three separate estimators have been developed: wind-only, simultaneous wind and rain, and rain-only, which account for rain contamination in varying degrees. This paper introduces a Bayes estimator selection technique to adaptively choose a best estimator from among the three types of estimators at each measurement location. Bayes estimator selection is introduced from a general perspective after which it is applied specifically to QuikSCAT wind and rain estimation. Bayes estimator selection is demonstrated in a case study to illustrate improvements in wind and rain estimation which can be obtained.