机载激光雷达及高光谱的森林乔木物种多样性遥感监测

利用机载LiDAR和高光谱数据并结合37个地面调查样本数据,基于结构差异与光谱变异理论,通过相关分析法分别筛选了3个最优林冠结构参数和6个最优光谱指数,在单木尺度上利用自适应C均值模糊聚类算法,在神农架国家自然保护区开展森林乔木物种多样性监测,实现了森林乔木物种多样性的区域成图。研究结果表明,(1)基于结合形态学冠层控制的分水岭算法可以获得较高精度的单木分割结果(R~2=0.88,RMSE=13.17,P0.001);(2)基于LiDAR数据提取的9个结构参数中,95%百分位高度、冠层盖度和植被穿透率为最优结构参数,与Shannon-Wiener指数的相关性达到R~2=0.39—0.42(P0.01);(3)基于机载高光谱数据筛选的16个常用的植被指数中,CRI、OSAVI、Narrow band NDVI、SR、Vogelmann index1、PRI与Shannon-Wiener指数的相关性最高(R~2=0.37—0.45,P0.01);(4)在研究区,利用以30 m×30 m为窗口的自适应模糊C均值聚类算法可预测的最大森林乔木物种数为20,物种丰富度的预测精度为R~2=0.69,RMSE=3.11,Shannon-Wiener指数的预测精度为R~2=0.70,RMSE=0.32。该研究在亚热带森林开展乔木物种多样性监测,是在区域尺度上进行物种多样性成图的重要实践,可有效补充森林生物多样性本底数据的调查手段,有助于实现生物多样性的长期动态监测及科学分析森林物种多样性的现状和变化趋势。

Forest species diversity mapping using airborne LiDAR and hyperspectral data

Forest species diversity, as a key component of biodiversity, plays an irreplaceable role in maintaining ecological balance, processes, and services. In recent years, forest tree species diversity is facing a serious threat with intensifying human activities and influence of climate change. The status and trends of forest tree species diversity must be dynamically monitored to develop effective forest biodiversity conservation approaches.In this study, an airborne light detection and ranging (LiDAR) (>4 points/m²) and hyperspectral (PHI-3 sensor with spatial resolution of 1 m) data combined with 37 field sample data are used to detect tree species variation in the structural and spectral properties in the Shennongjia Forest Nature Reserve of China. First, we use the morphological crown control method based on a watershed algorithm to isolate individual tree crowns by using LiDAR. We select optimal structural indices from nine commonly used structural indices derived using LiDAR based on the theory of structural and spectral variation hypothesis. Meanwhile, we select optimal vegetation indices from 16 VIs based on the hyperspectral data by conducting a correlation analysis with the field samples. Second, a self-adaptive fuzzy C-means clustering algorithm is applied to map the species diversity (i.e., richness and Shannon-Wiener index) in the study area for each 30 m×30 m window at the individual tree crown scale.Result indicates that the individual tree isolation by using the watershed algorithm can obtain a high accuracy (R²=0.88, RMSE=13.17, P<0.001). The 95th quintile height, canopy cover, and vegetation permeability are the optimal structural indices, and their correlation with the Shannon-Wiener index are from 0.39 to 0.42 (R²=0.39-0.42, P<0.01). The correlation among RI, OSAVI, narrow band NDVI, SR, Vogelmann index 1, PRI, and field inventory Shannon-Wiener index are relatively high based on the airborne hyperspectral data (R²=0.37-0.45, P<0.01). Finally, we use the selected three structural and six vegetation indices to predict the optimal clustering numbers (species richness) and the Shannon-Wiener index by using the self-adaptive fuzzy C-means clustering algorithm. The result shows that the maximum tree species that can be predicted is 20. The prediction accuracy of species richness is R²=0.69, RMSE=3.11, and the Shannon-Wiener index is R²=0.70, RMSE=0.32.This method shows the potential of LiDAR combined with hyperspectral data in mapping species diversity of a subtropical forest. Moreover, it could provide an effective method for analyzing the current situation and its changing trend of forest biodiversity at a regional scale.