Forest species diversity mapping using airborne LiDAR and hyperspectral data in a subtropical forest in China

• Published in: Remote sensing of environment 2018-08, Vol.213, p.104-114
• Main Authors: Zhao, Yujin, Zeng, Yuan, Zheng, Zhaoju, Dong, Wenxue, Zhao, Dan, Wu, Bingfang, Zhao, Qianjun
• Format: Article
• Language: English
• Subjects: Adaptive algorithms; Airborne sensing; Biochemistry; Biodiversity; Clustering; Forest biodiversity; Forest conservation; Forest management; Forest resources; Forests; Image processing; Imaging spectroscopy; Leaf biochemical components; LiDAR; Mathematical analysis; Plant diversity; Properties (attributes); Resource management; Scientific imaging; Shannon-Wiener; Spatial discrimination; Spatial resolution; Species diversity; Species richness; Spectroscopy; Trees; Tropical forests; Watersheds; Wildlife conservation
• ISSN: 0034-4257; 1879-0704
• DOI: 10.1016/j.rse.2018.05.014

Monitoring biodiversity is essential for the conservation and management of forest resources. A method called “spectranomics” that maps the diversity of forest species based on species-driven leaf optical traits using imaging spectroscopy has been developed for tropical forests in earlier studies. In this study we applied the “spectranomics” method in combination with airborne hyperspectral (PHI-3 sensor with 1 m spatial resolution) and LiDAR (>4 points/m2) data to first identify interspecies variations in biochemical and structural properties of trees and then estimate the tree species diversity within the Shennongjia Forest Nature Reserve in China. Firstly, we used the watershed algorithm based on morphological crown control to isolate individual tree crowns (ITCs) from the LiDAR data. For each ITC, we then calculated seven vegetation indices (VIs) representing key biochemical properties from the hyperspectral data and additionally derived the LiDAR-based tree height which was identified to support the discrimination of the tree species in a preceding analysis. Finally we utilized the combination of the seven selected VIs and tree height as input to a self-adaptive Fuzzy C-Means (FCM) clustering algorithm. The FCM algorithm was applied to fixed subsets of 30 m × 30 m and it was assumed that the number of clusters identified within a subset represents the number of occurring species. The species richness and Shannon-Wiener diversity index calculated from the clustering outputs correlated well with the field reference data (R2 = 0.83, RMSE = 0.25). The results show that forest species diversity can be directly predicted using the suggested clustering method based on crown-by-crown variations in biochemical and structural properties in the examined subtropical forest without the need to distinguish the individual tree species. •Isolated individual tree crowns and extracted tree heights•Determined 7 optimal biochemical VIs for predicting forest diversity•Clustering species richness using crown-by-crown biochemical and structural traits•Directly derived species diversity without species discrimination