A probabilistic approach to scatterometer model function verification

Abstract

The scatterometer backscatter model function has a threshold wind speed, below which there is no backscatter return from the ocean surface. Owing to the variability associated with the wind at the scatterometer footprint resolution, the sharp cutoff in the model backscatter response is not observed in Ku band spaceborne scatterometer data. A merged data set is generated by temporal and spatial collocation of the National Center for Environmental Prediction (NCEP) 10‐m neutral winds with individual National Aeronautics and Space Administration Scatterometer (NSCAT) backscatter measurements from level 1.7. A model is set up to compare the observed and expected probability distributions of σ0 in an NCEP grid resolution cell (1.875°×1.904°), after incorporating the variabilities associated with the wind and the backscatter measurements. The differences in the model responses arrived at using the NSCAT 1 and the new threshold model function (TMF) are presented. The TMF model does well in fitting the observed σ0 distributions for all incidence angles, antenna azimuths, and both polarizations over the wind speed range considered (1–10 m s−1). A comparison of the modes of the model‐predicted and observed distributions revealed the model function cutoff behavior at the threshold wind speed. The model also predicts increasing percentage occurrences of negative σ0 with decreasing wind speeds and increasing incidence angles, in line with the observations. The TMF model is corrected for the superthreshold winds to yield a better fit to the observed probability distributions.