More and more micro-wave data are available from spatial and airborne measurements over sea surface. An accurate backscattering model which is capable of taking the multi-scale aspect of the sea surface into account, is required to model correctly the data as well as a precise sea spectrum. Several unified backscattering models have been developed in recent years under Gaussian statistics. However, these models are not able to give a correct modelization of the backscattered signal when different data sets are studied together. One of the objectives of this study is to improve the modelization of the backscattered signal to get better agreement with the data.The first step of this study is to include non Gaussian statistics into backscattering model as it is well known they have a significant impact on the normalized radar cross section (NRCS). Then, a non Gaussian version of the Weighted Curvature Approximation was developed taking the kurtosis of slopes into account. This work was based only upon vertical polarization.It is then shown that the corrections allow a better agreement with the data but they are not sufficient to get a good estimation of the NRCS for all incidences and electromagnetic frequencies. This induces the hypothesis of a modification of the short wave sea spectrum.Then, a new parametrisation of the omnidirectional sea spectrum is suggested to get a better agreement with the multiband data sets and is based on the spectrum developed by Elfouhaily et al. The new omnidirectional short wave sea spectrum is quite alike the Elfouhaily’s spectrum with some noticeable differences for the decimetric scales.
