Retrieval of integral ocean wave parameters from SAR data using an empirical approach

Abstract

Spaceborne synthetic aperture radar (SAR) is still the only instrument providing continuous two-dimensional (2-D) ocean wave measurements on a global basis. From the European satellites ERS-1, ERS-2 and ENVISAT more than a decade of continouos global SAR measurements taken over the oceans is available. It is well known that the derivation of ocean wave parameters from SAR data is not straightforward and different methods have been proposed. It this study we present a new technique, which is based on an empirical approach. The method is called CWAVE1 and uses a set of 11 parameters derived from the SAR measurement as input to a quadratic model function. This set includes the radar cross section, the image variance and 9 additional parameters computed from the SAR spectrum using a set of orthogonal functions. The CWAVE1 function has a number of tuning variables which are fitted for the estimation of different integral ocean wave parameters like the significant wave height Hs, the wave height H12 of waves with periods longer than 12 s, the mean period and the spectral moment m4. The approach is very efficient from the computational point of view with the Fourier transformation of the SAR image as the only demanding processing step. The fit procedure is based on a stepwise regression method, which is able to identify the SAR parameters which are dominating the wave parameter estimates. Two different data sets of globally distributed ERS-2 wave mode image spectra and collocated ocean wave spectra computed with the numerical ocean wave model WAM are used to fit the model and to retrieve the ocean wave parameters. Each data set comprises 5000 colocation pairs. Scatterplots and global maps with the derived wave parameters are presented. It is shown that the rms of the retrieved significant waveheight with respect to the WAM model waveheight is about 0.5 m. The respective rms of H12 is about 0.4 m and the mean period shows an rms of 0.8 s with a high frequency cutoff period of 5 s. In a second step the model is tuned with simulated SAR data using the full nonlinear SAR ocean transform. This approach enables the adjustment of the model to different incidence angles, which is the basis for the application of the method to image mode data.