Digital terrain, surface, and canopy height model generation with dual-baseline low-frequency InSAR over forest areas

Interferometric synthetic aperture radar (InSAR) technology can be used to produce a high spatial resolution digital elevation model (DEM) on a global scale. However, above-ground vegetation strongly biases the DEM accuracy in forested areas by shifting the scattering phase center below the top of the canopy and above the underlying topography. Such a bias is related to the radar wavelength, forest structure, dielectric property, radar incidence angle, and terrain slope. In this paper, based on the strong penetration characteristics of the low-frequency (P-band) SAR signal, the time–frequency (TF) analysis is introduced to increase the InSAR observation space, where the decomposed sublook images can also be utilized to perform interferometry. By analyzing and modeling the sublook interferograms, a comprehensive dual-baseline framework for correcting the InSAR phase shifting is proposed to generate accurate digital terrain, surface, and canopy height models (DTM, DSM, and CHM) over forest areas. The proposed method is validated using P-band InSAR datasets acquired above the boreal coniferous forest in Remningstorp, southern Sweden, and the rainforest in Lopé, Gabon, Africa. For the boreal forest, the root-mean-square errors (RMSEs) in terms of the DTM, CHM, and DSM range from 2 to 4 m, while for the tropical rainforest with complicated topography, the RMSEs of the three elevation models range from 5 to 8 m compared with light detection and ranging (LiDAR) references. The high consistency between the InSAR-derived DTM/DSM and references demonstrates the effectiveness and stability of the proposed method and represents an improvement of 49–70% compared to the raw InSAR DEM.