目前国内外都正积极的进行基于简缩极化(Compact Polarimetry, CP)SAR的多方面研究,但鲜有研究将其应用在森林地上生物量(Above Ground Biomass, AGB)方面。为探究CP SAR数据在森林AGB反演中的可行性,以宜良小哨林场为研究区,提取水平线性CP Stokes1模式、垂直线性CP Stokes2模式、π/4线性模式(π/4 Transmit and Orthogonal Linear Receive)及CTLR(Circular Transmit and Dual Orthogonal Linear Receive)模式的4种CP SAR数据,并基于波的二分性原理,分别提取了各种模式的若干SAR参数,利用基于快速迭代特征选择的k最近邻(KNN-FIFS)算法开展了研究。研究结果表明：基于CTLR模式的森林AGB反演结果最优,R_²=0.52,RMSE=13.02 t/hm_²；此外,联合4组CP SAR数据的森林AGB反演结果精度有明显提升：R_²=0.58,RMSE=12.16 t/hm_²。KNN-FIFS适合于采用CP SAR参数进行森林AGB反演,其反演结果与采用全极化SAR数据进行反演的差别并不明显。提取的CP SAR参数中,线极化度m_l ,倾斜角45度或-45度时的线极化分量功率值g₂等特征在森林AGB反演中表现出较高的适用性,说明其能更好的表征森林信息。

[Objective] Accurate estimation of forest above ground biomass (AGB) is of great significance to the research of carbon cycle and carbon neutrality. The purpose of this study is to evaluate the feasibility of using compact polarimetric SAR data in forest AGB estimation. [Method] This study takes Xiaoshao Forest Farm in Yiliang county as the research area, and extracts the horizontal linear compact polarimetric Stokes1 mode, the vertical linear compact polarimetric Stokes2 mode, the π/4 linear (π/4 Transmit and Orthogonal Linear Receive) mode and CTLR (Circular Transmit and Dual Orthogonal Linear Receive) mode of 4 compact polarimetric SAR data, and based on the principle of wave dichotomy, several SAR parameters of various modes were extracted respectively, and then the k Nearest Neighbor with Fast Iterative Feature Selection(KNN-FIFS)method was applied to estimate the AGB of the study area. Finally, the accuracy of the KNN-FIFS inversion results was verified using the LOO(Leave-One-Out) method. [Result] The forest AGB estimation result of Stokes1 mode, with R_² = 0.28, RMSE = 16.36 t/hm_²; the forest AGB estimation result of Stokes2 mode, with R_² = 0.35, RMSE = 14.96 t/hm_²; the forest AGB estimation result of π/4 mode, with R_² = 0.34, RMSE =15.21 t/hm_²; the forest AGB inversion result of Stokes characteristic parameters in CTLR mode, with R_² = 0.52, RMSE = 13.02 t/hm_². The forest AGB inversion result combining 4 sets of compact polarimetric data, with R_² = 0.58, RMSE = 12.16 t/hm_². [Conclusion] The biomass estimation result with CTLR mode compact polarimetric data performs best; when the four groups of compact polarimetric parameters are combined for the forest AGB estimation, the estimation result is better than the result using each set of parameters alone; KNN -FIFS is suitable for forest AGB estimation using compact polarimetric SAR parameters, and the difference between the polarimetric results and the polarimetric using full-polarizion SAR data is not obvious. Among the extracted compact polarimetric parameters, the characteristics of the degree of linear polarization (m_l), and the power of the linear polarization component at a tilt angle of 45 degrees or 135 degrees (g₂) show high applicability in the forest AGB estimation, It shows that it can better characterize the forest information.