Retrieval of forest stem volume from synthetic aperture (SAR) backscatter and interferometric SAR (InSAR) coherence is generally performed using a model-based approach, where in situ measurements are necessary to estimate the unknown model parameters. Problems arise when in situ data are either not available or of low quality or the observables present spatial variations. In this work we present three approaches for automatic modeling and inversion of forest backscatter and coherence to retrieve forest stem volume. The three approaches exploit statistical distributions of the observables to obtain estimates for the unknowns in the model. Results shows remarkable agreement with those obtained by means of traditional modeling approaches based on in situ data.