The Random Volume over Ground (RVoG) model has been extensively applied to polarimetric synthetic aperture radar interferometry (Pol-InSAR) data for forest height inversion. The model assumes forest as a homogeneous volume of randomly oriented particles characterized by a constant extinction but does not take into account the forest vertical heterogeneity, to which interferometric coherence is sensitive. In order to integrate vertical heterogeneity in forest models, two complementary models, which take into consideration the forest natural structure, are investigated through analysis of volume interferometric coherence. The first model assumes a vertically varying extinction in the volume layer, and the second model considers predominant contributions localized in a finite height interval, modeled as a Gaussian-distributed backscatter. The two forest models are compared with constant extinction RVoG in the coherence and interferometric phase aspects. Finally, the contribution of these new models for forest height inversion using the Pol-InSAR technique is discussed in the context of a two-layer ground + canopy medium.