A Comparison of Tree Height Estimates Using three Remote Sensing Methods - Stereo Photogrammetry, Airborne Laser Scanning and Terrestrial Laser Scanning - in Tropical Forest

By Jianqi Ding (China)

Abstract
Tree height measurement in the tropical forest is an issue with high uncertainty due to the complexity of the forest structure and dense canopy. Various remote sensing techniques are introduced for the tree height detection. Terrestrial laser scanning (TLS), airborne laser scanning (ALS) are two LiDAR based remote sensing techniques acquiring high spatial-resolution point cloud. Photogrammetry based point cloud is recently attached more importance in the tree height detection with the development of unmanned aerial vehicle (UAV). This study is aimed at gaining insight of the capability of point clouds of TLS, ALS and photogrammetric point cloud (PPC) to simulate the canopy height and detect the individual tree height. On the canopy level, the canopy height model (CHM) is retrieved from the difference between digital surface model (DSM) and digital terrain model (DTM), which are both converted from point cloud. On the individual tree level, tree height is sought for based on the canopy height model (CHM), and Inverted watershed analysis (IWA) is adopted to delineate the tree crown, thus the maximum value within each crown is considered as the individual tree height. Subsequently, to investigate the difference between the traditional and remote sensing method, individual tree information collected in field survey is linked to TLS, based on the diameter at breast height (DBH) and their position in the plot. And the tree height and the above ground biomass (AGB) are extracted and estimated for comparison, using field survey, TLS, ALS and photogrammetry methods.

TLS and ALS give a similar canopy height, and PPC produces a smoother CHM and notable lower height. The delineation of the discrete trees is effective, but the overlapping trees are often merge as one. TLS and ALS provide comparable detected tree number and height, and photogrammetry is less efficient in crown delineation and underestimate tree height, compared to TLS and ALS. Field survey shows an overestimation of height, and the TLS, ALS and photogrammetry methods estimate less ABG with respect to field survey, by 7.65%, 6.29% and 4.54% respectively, in the tropical forest.

Keywords: tree height; terrestrial laser scanning; airborne laser scanning; photogrammetry; canopy height model; inverted watershed analysis; above ground biomass