Wave spectra retrieval from complex ENVISAT Wave Mode data using a parametric inversion scheme

Spaceborne synthetic aperture radar (SAR) is still the only instrument providing directional ocean wave information on a global and continuous basis. Different retrieval schemes for the derivation of two-dimensional ocean wave spectra Fk from SAR data have been developed (Krogstad, 1994; Hasselmann et al., 1996; Mastenbroek, 2000; Dowd, 2001; Johnson et al., 2003). The potential of the data for improving wave model forecasts has been demonstrated [Heimbach et. al, 2000]. More recent schemes are based on SAR cross spectra, which allow to resolve the directional wave propagation ambiguity present in conventional SAR image variance spectra (Engen, 1995). With the launch of the European satellite ENVISAT cross spectra have become available on anoperational basis. Actvities therefore exist at different weather centres to use the data for the assimilation of numerical ocean wave forecast models. The cross spectra method is based on a special processing technique, where two SAR images (looks), which are separated by about half a second (Engen, 1995) are generated. By calculating the cross spectrum Φk of the two looks information can be obtained about the wave propagation direction. This information had to be taken from wave models in earlier retrieval schemes which were based on symmetric SAR image variance spectra. A difficult problem in SAR wave retrieval schemes is the fact that SAR data contain information mainly about longer waves, whereas shorter waves in particular those propagating in the flight direction (azimuth) are strongly distorted or completely filtered out in many cases. In order to obtain a complete two-dimensional wave spectrum a retrieval scheme therefore has to blend SAR information and prior information in some consistent way. The Partition Rescaling and Shift Algorithm (PARSA) scheme described in this paper is able to deal with this problem and has several additional features compared to the scheme presented in Hasselmann (1996):