林龄结构信息能够有效反映区域森林群落不同生长阶段的固碳能力，对于评估森林生态系统的健康状况具有重要意义。本研究以中国温带典型优势树种落叶松林为研究对象，分别选择其芽萌动期、展叶期和落叶期时段的Sentinel-2影像，采用多元线性回归（MLR）、随机森林（RF）、支持向量机回归（SVR）、前馈反向传播神经网络（BP）以及多元自适应回归样条（MARS）等5种方法依次构建落叶松林龄反演模型。通过相关性分析首先确定最佳遥感反演物候期，并在此基础上根据相关性差异筛选出5个最优特征变量用于模型反演，分别为冠层含水量（CWC），归一化水体指数（NDWI），叶面积指数（LAI），光合有效辐射吸收率（FAPAR）和植被覆盖度（FVC）。研究结果表明，展叶期为落叶松林最佳遥感反演物候期。除植被衰减指数（PSRI）以及落叶期的NDVI、RVI外，落叶松林龄与各指标之间均呈负相关关系，其中与冠层含水量（CWC）的相关性最高，pearson相关系数达到-0.74（p<0.01）。此外，不同模型反演结果表明，随机森林模型（RF）为最佳落叶松林龄估测模型，其平均决定系数R²和平均均方根误差RMSE分别为0.89和2.91 a；多元线性回归模型（MLR）的林龄估测结果最差，其平均决定系数R²和平均均方根误差RMSE仅为0.57和5.69 a，非线性模型能更好的解释林龄与建模变量之间的关系。

The information of forest age structure can effectively reflect the carbon sequestration capacity of regional forest communities at different growth stages. This way is important for assessing the health status of forest ecosystems. In this study, the typical dominant tree species Larix gmelinii forest in temperate zone of China is selected as the object, and Sentinel-2 images of its bud germination period, elongating period of leaf, and defoliation period are selected. The retrieval model of Larix gmelinii stand age is constructed using Multiple Linear Regression (MLR), Random Forest (RF), support vector regression, feedforward back propagation neural network, and multiple adaptive regression spline. The optimal phenophase of remote sensing retrieval is first determined through correlation analysis. On this basis, five optimal characteristic variables, namely, Canopy Water Content (CWC), normalized difference water index, leaf area index, fraction of absorbed photosynthetically active radiatio, and fractional vegetation cover, are selected for model retrieval according to the difference in correlation. Results show that the elongating period of leaf is the optimal remote sensing retrieval phenophase. Except for the plant senescence reflectance index and NDVI and RVI in defoliation period, a negative correlation exists between the stand age of Larix gmelinii and each index, among which the correlation between the stand age and (CWC is the closest, and the correlation coefficient of Pearson reaches -0.74 (p<0.01). The results of different model retrievals indicate that RF model is the best model for estimating the age of Larix gmelinii, and its average coefficient of determination (R²) and mean Root Mean Square Error (RMSE) are 0.89 and 2.91 a, respectively. MLR is the worst for estimating Larix gmelinii forest age, and its average R² and RMSE are 0.57 and 5.69 a, respectively. Nonlinear models can better explain the relationship between stand age and modeling variables.