Deriving comprehensive forest structure information from mobile laser scanning observations using automated point cloud classification

Abstract

The advent of mobile laser scanning has enabled time efficient and cost effective collection of forest structure information. To make use of this technology in calibrating or evaluating models of forest and landscape dynamics, there is a need to systematically and reproducibly automate the processing of LiDAR point clouds into quantities of forest structural components. Here we propose a method to classify vegetation structural components of an open-understorey eucalyptus forest, scanned with a 'Zebedee' mobile laser scanner. It detected 98% of the tree stems (N = 50) and 80% of the elevated understorey components (N = 15). Automatically derived DBH values agreed with manual field measurements with r2 = 0.72, RMSE = 3.8 cm, (N = 27), and total basal area agreed within 1.5%. Though this methodological study was restricted to one ecosystem, the results are promising for use in applications such as fuel load, habitat structure, and biomass estimations. Show more