全极化SAR数据能够有效刻画森林冠层的垂直结构特征，在森林地上生物量（AGB）反演中具有重要应用价值，而合理的极化分解方法是提取关键极化特征并构建高精度定量反演模型的重要支撑。然而，在复杂森林冠层条件下，基于Barakat极化度的无模型分解方法在应对去极化效应、噪声干扰及观测几何变化时易产生偏差，导致极化特征的稳定性与泛化能力受限。为此，本研究在无模型分解框架中引入有效极化度（Effective Degree of Polarization, EDoP），提出一种融合EDoP的改进型无模型极化分解方法。以湖南省三个典型森林区域的L波段全极化SAOCOM影像为数据源，分别采用模型分解、传统无模型分解以及融合EDoP的无模型分解三种策略提取极化特征，并结合前向特征选择与四种机器学习回归模型开展森林AGB定量反演。结果表明，融合EDoP的方法显著改善了散射目标的功率分配：体散射（VOL）能量占比由约40%降至30%，二次散射（DBL）分量提升约20%，有效抑制了由噪声和去极化引起的能量偏差，增强了各散射分量与森林AGB之间的相关性，尤其在冠层结构异质性较高的林分中表现更为突出。相较于传统无模型分解，融合EDoP所提取的极化特征与AGB的相关性显著增强；在森林AGB定量反演性能方面，本文提出的方法使R2提升0.15~0.30，rRMSE降低4%~7%，且生成的森林AGB空间分布图更加合理，低估与高估现象均得到有效缓解，延缓了森林AGB遥感反演的饱和效应。本研究进一步验证了融合EDoP的无模型分解方法在复杂林分中的稳健性与可迁移性，证实其具备缓解森林AGB遥感反演饱和效应、提升复杂森林区域AGB反演精度的潜力，对实现大范围森林生物量与碳储量的精细化监测具有重要应用价值。

Objective: Full-polarimetric synthetic aperture radar (SAR) data are highly effective in characterizing the vertical structure of forest canopies and play a crucial role in forest above-ground biomass (AGB) estimation. A proper polarimetric decomposition method provides important support for the effective extraction of key polarization features and the construction of high-accuracy quantitative inversion models. However, under complex forest canopy conditions, the traditional model-free decomposition method based on the three-dimensional Barakat degree of polarization (DOP) is prone to errors caused by depolarization effects, noise, and variations in observation geometry. These limitations reduce the stability and generalization ability of polarimetric features in AGB retrieval. Method: To address these issues, this study introduces the Effective Degree of Polarization (EDoP) into the model-free decomposition framework and proposes an improved model-free polarimetric decomposition method that integrates EDoP. L-band full-polarization SAOCOM images from three representative forest regions in Hunan Province, China, were used as data sources. Three decomposition strategies—model-based decomposition, traditional model-free decomposition, and the proposed EDoP-integrated model-free decomposition—were applied to extract polarization features. Subsequently, forward feature selection was combined with four machine learning regression models, including multiple linear regression (MLR), k-nearest neighbor (KNN), support vector regression (SVR), and random forest (RF), to perform quantitative AGB inversion. Result: The results demonstrate that the EDoP-integrated decomposition method significantly improves the power allocation among scattering mechanisms. The proportion of volume scattering (VOL) decreased from approximately 40% to 30%, while the double-bounce scattering (DBL) component increased by about 20%, effectively mitigating energy estimation bias caused by noise and depolarization effects. This adjustment enhanced the physical consistency between scattering components and forest AGB, particularly in heterogeneous canopy structures. Compared with the traditional model-free approach, the proposed method achieved stronger correlations between extracted polarimetric features and AGB. In quantitative inversion performance, the proposed method increased the coefficient of determination (R2) by 0.15–0.30 and reduces the relative root mean square error (rRMSE) by 4%–7%. Moreover, the spatial distribution maps of predicted AGB were more consistent with actual forest patterns, effectively reducing both underestimation and overestimation, and alleviating the saturation effect commonly observed in high-biomass regions. Conclusion: This study verifies the robustness and transferability of the EDoP-integrated model-free decomposition method in complex forest environments. The proposed approach effectively suppresses decomposition bias and enhances the accuracy and stability of AGB retrieval. By mitigating the AGB saturation effect and improving model generalization in heterogeneous forest stands, this method provides a promising pathway for large-scale, high-precision monitoring of forest biomass and carbon storage using polarimetric SAR data.