I. Introduction
The two-scale model (TSM), also called composite model (CM), [1], [2] has been extensively used for computation of scattering from rough surfaces, and, in particular, from sea surfaces, see [3]–[5]. According to TSM, the rough surface is modeled as the superposition of a large-scale (or low-frequency) roughness, which includes surface spectral components with wavenumbers lower than a properly chosen cutoff wavenumber, and a small-scale (or high-frequency) roughness, which encompasses surface spectral components with wavenumbers higher than the cutoff wavenumber. The latter must be sufficiently smaller than the electromagnetic wavenumber, and sufficiently larger than times the inverse of the sensor resolution, so that a certain degree of arbitrariness is implied in its choice. Scattering from large-scale roughness is evaluated by using the geometrical optics (GO) approximation; it is dominant at near-specular directions (in backscattering, at small incidence angles) and the elements of the polarimetric covariance matrix, including the normalized radar cross sections (NRCS) at different polarizations, turn out to be dependent on the large-scale surface slope probability density function (pdf). Conversely, scattering from the small-scale roughness is computed by evaluating the scattering from a randomly tilted rough facet via the small perturbation method (SPM), and then averaging the obtained NRCS over the facet random slopes, statistically distributed according to the large-scale surface slope pdf. Backscattering from small-scale roughness is dominant at intermediate incidence angles and the elements of the polarimetric covariance matrix, including the NRCS, mainly depend on the small-scale roughness power spectral density (PSD). It must be noted that no depolarization effect can be obtained by the TSM if the facet random tilt is not accounted for, so that averaging over the facet random slopes is a necessary step for a polarimetric scattering analysis via the TSM.