Boreal forest height inversion using E-SAR PolInSAR data based coherence optimization methods and three-stage algorithm

As an important forest parameter, forest height can be used to estimate forest above-ground biomass through specific allometric equation. Recent works have shown polarimetric synthetic aperture radar (SAR) interferometry (PolInSAR) is a powerful technique for forest height monitoring.. Three-stage algorithm is a geometrical and most widely useful approach for forest height inversion based on the RVoG model using PolInSAR data. However, the inversion accuracy of this algorithm is affected by the estimation accuracy of underlying topography phase and volume interferometic coherence. One of important reasons is the unreasonable choices of polarization channels which can not separate scattering centers effectively. In practice, typical selections involve Pauli polarization channels (HH+VV, HH-VV, HV) and linear channels (HH, VV), which can not guarantee the reliability of inversion result. In this paper, we propose a relatively robust inversion algorithm to estimate forest height. Firstly, this algorithm utilizes two different coherence optimization methods (C&P optimization and Phase diversity optimization) to generate five optimum polarizations which are characterized by effective separation of phase centers. Then, it combines these optimum polarizations and three-stage algorithm to estimate forest height. The performance of the inversion algorithm is demonstrated using full polarimetric single baseline interferometic data acquired by E-SAR airborne system at P-band over Krycklan boreal forest in northern Sweden. In-situ ground truth heights in stand-level have been used to validate the inversion result. The preliminary result indicates that the proposed inversion algorithm can estimate height with relatively high accuracy, which decreases by 60.4% for mean absolute bias and 54% for root-mean-square error respectively.