A physically-based gmf includes two components. First, an ocean wave spectral model that relates the observed distribution of waves at the sea surface to a particular atmospheric forcing. Second, a model that predicts the radar backscatter that would arise from a particular configuration of the ocean surface. Because of model uncertainties and computational constraints, physically-based gmfs have had limited application, and so empirical gmfs have been relied upon for wind retrieval from scatterometer observations. Empirical gmfs, while quite useful operationally, do not include an explicit dependence on environmental parameters other than wind velocity. As such, these gmfs may be of limited use for studying the physical processes at the ocean surface that give rise to the observed radar signal. Air and sea temperature, boundary layer stratification, viscosity, sea state, directional spreading of wave energy relative to the wind direction, and other parameters describing the air-sea interface impact the radar return from the ocean surface but are not included in the empirical gmfs currently in use.
Predictions of wind speed and direction obtained using two different physically-based gmfs are analyzed. These results are compared with empirically-based wind vector estimates, as well as with wind vector measurements from collocated buoys. These comparisons are obtained under a range of atmospheric and oceanic conditions, so that the impact of a variety of environmental conditions on radar backscatter from the air-sea interface, and on wind-vector prediction may be evaluated.