Wind field models and model order selection for wind estimation

Abstract

Traditional scatterometer wind estimation inverts the model function relationship between the wind and backscatter at each resolution element, yielding a set of ambiguities due to the many-to-one mapping of the model function. Field-wise wind estimation dramatically reduces the number of ambiguities by estimating the wind for many resolution elements, simultaneously, using a wind field model that constrains the spatial variability of the wind. In this paper several wind field models are presented for use in field-wise wind estimation. Model accuracy, as a function of the number of model parameters, is reported for each model. This accuracy is evaluated using NSCAT JPL nudged L2.0 data. In order to reduce the computational load, automated classification schemes are developed to select the optimal number of model parameters necessary for a given wind field. Classification is performed through hypothesis testing on raw NSCAT data and point-wise estimates.