Determination of the geophysical model function of NSCAT and its corresponding variance by the use of neural networks

We have computed two geophysical model functions tone for the vertical and one for the horizontal polarization) for the NASA scatterometer (NSCAT) by using neural networks. These neural network geophysical model functions (NN GMFs) were estimated with NSCAT scatterometer sigma(o) measurements colloca ted with European Centre for Medium-Range Weather Forecasts analyzed wind v ectors during the period January 15 to April 15, 1997. We performed a stude nt t test showing that the NNGMFs estimate the NSCAT sigma(o) with a confid ence level of 95%. Analysis of the results shows that the mean NSCAT signal depends on the incidence angle and the wind speed and presents the classic al biharmonic modulation with respect to the wind azimuth. NSCAT sigma(o) i ncreases with respect to the wind speed and presents a well-marked change a t around 7 m s(-1). The upwind-downwind amplitude is higher for the horizon tal polarization signal than for vertical polarization, indicating that the use of horizontal polarization can give additional information for wind re trieval. Comparison of the sigma(o) computed by the NNGMFs against the NSCA T-measured sigma(o) show a quite low rms, except at low wind speeds. We als o computed two specific neural networks for estimating the variance associa ted to these GMFs, The variances are analyzed with respect to geophysical p arameters. This led us to compute the geophysical signal-to-noise ratio, i. e., K-p. The K-p values are quite high at low wind speed and decrease at hi gh wind speed. At constant wind speed the highest K-p are at crosswind dire ctions, showing that the crosswind values are the most difficult to estimat e. These neural networks can be expressed as analytical functions, and FORT RAN subroutines can be provided.