DETERMINATION OF THE GEOPHYSICAL MODEL FUNCTION OF THE ERS-1 SCATTEROMETER BY THE USE OF NEURAL NETWORKS

We present a geophysical model function (GMF) for the ERS-1 scatterome ter computed by the use of neural networks. The neural networks GMF (N N GMF) is calibrated with ERS-1 scatterometer sigma 0 collocated with European Center for Medium-Range Weather Forecasts (ECMWF) analyzed wi nd vectors. Four different NN GMFs have been computed: one for each an tenna and an average NN GMF. These NN GMFs do not present any signific ant differences which means that the three antenna are quasi-identical . The NN GMFs exhibit a biharmonic dependence on the wind azimuth with a small upwind-downwind modulation as found on previous GMFs. In orde r to check the validity of the NN GMF systematic comparisons with the European Space Agency (ESA) C band model (CMOD4) GMF (version 2 of Mar ch 25, 1993) and the Institut Francais de Recherche pour 1'Exploitatio n de la Mer (IFREMER) CMOD2 I3 GMF are done. It is found that the NN G MFs are highly accurate and relevant functions to model the ERS-1 scat terometer sigma 0.