Synthetic aperture radar tomography is a technique that can be used to retrieve the digital surface model of a studied area. It is based on the acquisition of N images of the same scene obtained at different viewing angles to retrieve backscattering profile. Many standard inversion techniques exist in the literature, but in this paper, we mainly focus on increasing the performances of Capon inversion. This algorithm is a non parametric estimation method, it is able to retrieve the volume height of forested area. Although, Capon inversion gives acceptable results, it is possible through using backscattering statistical models to increase the performances of Capon algorithm. In this paper, we present new hybrid algorithms, based on the combination of Capon inversion with statistical models of the reflectivity signal. We selected three models for the volume height, uniform distribution, exponential distribution, and Gaussian distribution, and for each probability distribution, we have computed analytically the corresponding hybrid inversion. To validate the proposed approach, we used the BioSAR 2008 dataset obtained in a boreal forest situated in northern Sweden. Results of the proposed algorithms are validated quantitatively by measuring the detection rate of the forest height according to the relative error for the whole area. Qualitatively, by analyzing images of DSM, relative error and generated histograms. It was proven that the combination between Capon inversion and a Gaussian model yields the best detection rate of 50.4% for a relative error of 0.3. The exponential model achieved 47.6%, the uniform model 42.6% and the standard Capon 38.6%. [ABSTRACT FROM AUTHOR]