Earth terrains are modeled by a two-layer configuration to investigate the polarimetric scattering properties of the remotely sensed media. The scattering layer is an anisotropic random medium characterized by a three-dimensional correlation function with lateral and vertical correlation lengths and variances. Based on the wave theory under Born approximations, this model is applied to derive the fully polarimetric backscattering coefficients of the Mueller and covariance matrices. A single scattering process is considered and all the multiple reflections at the boundaries are taken into account. For an anisotropic random medium with optic axis tilted off the vertical axis, the corresponding Mueller and covariance matrices do not contain any zero elements. To account for the azimuthal randomness in the growth direction of leaves in tree and grass fields, an averaging scheme is applied to obtain the backscattering coefficients. In this case, the Mueller matrix contains eight zero elements and the covariance matrix has four zero elements, and the cross polarization term σvhdoes not vanish. Theoretical predictions are matched with experimental data for sea ice and vegetation fields.