Inversion study of the rough sea surface wind speed and sea surface salinity based on the support vector machine

In this paper, the support vector machine(SVM) regression techniques are applied to the inversion of sea state parameters (e.g. salinity and wind speed of the sea surface). The two scale model (TSM) is used to simulate backscattering coefficients of the rough sea surface with different radar parameters. After the sensitivity analysis, the L band (1.4 GHz) and the C band (6.8 GHz) are selected with appropriate angles as radar parameters. Then a variety of schemes of inversion are designed, in which single-frequency dual-polarization double angle, dual-frequency dual-polarization double angle and the ratio between the VV and HH polarization backscattering coefficients are chosen respectively as the samples information. After appropriate training, the SVM forecasting model is applied to inverse the salinity and wind speed of the sea surface. As shown by the results, at the C band, the inversion of the sea surface wind speed bears the highest accuracy, whereas at the L band, the inversion of the sea surface salinity demonstrates the highest accuracy when the ratio between backscattering coefficients is chosen as the samples information. The anti-noise performance of the SVM model is also examined, and the results show that the SVM model performs favorably in the sea state parameter inversion problem.