Estimation of sea surface spectrum under non-stationary conditions

One of the goals of ESA's SMOS mission is the measurement of global sea surface salinity maps using MIRAS, a synthetic aperture interferometric L-band radiometer with full-polarimetric capabilities. To do so, sea state effects on the brightness temperature must be compensated, but coincident measurements of wind speed and/or significant wave height are uncommon. The objective of this work is to estimate the sea surface spectrum at a given time and location from the previous temporal and spatial evolution of the wind conditions. To do so, a neural network has been trained to estimate the spectrum parameters (height variance , spectral deviation /spl sigma//sub x,y/ and spectral peak of swell k/sub xm,ym/, wind velocity V/sub 10/') from the measured spectra. Different measured spectra have then been used to validate the network design. Finally, the error in the L-band brightness temperature computed from the actual and the estimated spectra is studied.