Pol-InSAR sensitivity to hemi-boreal forest structure at L- and P-bands

•Ground phase bias is minimized at P-band in the HH-HV dual-pol space.•L-band is less sensitive to forest structure and then lead to accurate and unbiased forest height.•Spruce and pine trees can be distinguished using GBm at P-band. A dual-frequency approach is proposed to discriminate hemi-boreal forest species using L- and P-band data. First, it is shown that L-band is less sensitive to the forest structure than P-band. Second, the underlying ground topography is estimated using polarimetric sub-space decomposition at P-band to minimize any potential bias. Then, the Gaussian backscatter model is employed to extract the forest vertical backscatter at P-band. The inverted profiles differ significantly over the various forest stands and in the polarimetric channels, indicating important P-band sensitivity to the forest vertical structure. According to these measurements, it is shown that tree species, as pines and spruces, were presenting different signatures related to the vertical distribution of their structural elements. On the contrary, sensitivity to the forest vertical heterogeneity is low at L-band, allowing to accurately invert the forest height from an empirical structure function. A general workflow taking advantage of both frequencies is finally proposed to discriminate the forest species.