In this paper, a technique is described for the tomographic characterization of forested areas through multiple synthetic aperture radar (SAR) observations, based on either single or multipolarimetric acquisitions. This technique is based on the idea of characterizing the Fourier spectrum of the multibaseline data as being constituted by two effective scattering centers displaced along the vertical direction, plus the associated decorrelation terms. As a result, SAR tomography will be formulated as the problem of detecting the number of scattering centers within the resolution cell, estimating the parameters that describe their spatial structure, and evaluating the associated backscattered powers. Parameter estimation is carried out through the covariance matching estimation technique, which provides an asymptotically optimal solution. The results of an experiment performed on a real P-band multibaseline fully polarimetric data set relative to the forested site of Remningstorp, Sweden, are reported.