【目的】叶绿素是植物生理状态的重要指示因子,研究基于实测叶绿素荧光数据和Fluspect-B模型模拟数据,探讨叶片叶绿素含量的估算方法。【方法】以黑龙江尚志市帽儿山林场11个典型树种不同冠层高度位置叶片为对象测定其叶绿素荧光光谱与叶绿素含量,结合通过机理模型模拟的不同树种叶绿素荧光光谱与叶绿素含量关系数据库,分别构建基于实测数据的统计模型(多元线性回归模型、人工神经网络模型、随机森林模型)、基于模拟数据的混合模型以及基于实测与模拟数据驱动的混合模型,估算叶片叶绿素含量,并分析不同树种、不同冠层高度处叶片叶绿素含量分布特征。【结果】随机森林模型的模拟效果最佳,叶片叶绿素含量估算精度决定系数(R²)达到0.830 5,均方根误差(RMSE)为7.109 8 μg/cm²;基于实测与模拟数据驱动的混合模型精度优于统计模型,R²和RMSE分别为0.913 3、6.374 9 μg/cm²。阔叶树种叶绿素含量拟合效果普遍优于针叶树种,不同冠层位置处上层叶片数据集拟合效果优于中层,下层的最差。【结论】基于实测与模拟混合数据的混合模型优于基于实测数据的统计模型,混合模型整体拟合效果较好,能较准确估测叶绿素含量。基于叶绿素荧光光谱数据,运用混合模型方法反演森林植被叶绿素含量具有可行性,可为林分叶绿素含量估算及森林生态系统碳汇估测研究提供数据基础。

【Objective】Chlorophyll is a crucial indicator of plant physiological status. This study explores methods for estimating leaf chlorophyll content using measured chlorophyll fluorescence and Fluspect-B model simulation data.【Method】The study analyzes the measured data of chlorophyll fluorescence spectrum and chlorophyll content of leaves at various canopy heights of 11 typical tree species in Maoershan Forest Farm. Additionally, it utilizes a database that simulates the relationship between chlorophyll fluorescence spectrum and chlorophyll content across different tree species. Statistical models are developed using multiple linear regression, artificial neural networks, and random forest modeling techniques based on the measured data. A hybrid model that integrates simulated data and a hybrid model combining measured and simulated data are employed to estimate leaf chlorophyll content. Additionally, the distribution characteristics of leaf chlorophyll content across different tree species and canopy heights are analyzed.【Result】Among the statistical models, random forest exhibits the highest effectiveness, achieving an estimation accuracy of coefficient of determination (R²) was 0.830 5 and root mean square error (RMSE) was 7.109 8 for leaf chlorophyll content (μg/cm²). The hybrid model incorporating both measured and simulated data demonstrates superior accuracy compared to the statistical models, yielding R² of 0.913 3 and RMSE of 6.374 9 μg/cm², respectively. The fitting accuracy for chlorophyll content of broad-leaved trees generally surpasses that of coniferous trees, particularly for upper leaf datasets at different canopy positions, which show better fitting effects than middle and lower layers.【Conclusion】The mixed model utilizing both measured and simulated data outperforms the purely statistical model based only on measured data. The mixed model exhibits good fitting accuracy, enabling precise estimation of chlorophyll content. The method based on chlorophyll fluorescence spectrum data proves viable for estimating forest vegetation chlorophyll content, laying a foundational dataset for large-scale chlorophyll content estimation and forest ecosystem carbon sink research.