Fine-scale Inventory of Forest Biomass with Ground-based LiDAR

Abstract

Biomass measurement provides a baseline for ecosystem valuation required by modern forest management. The advent of ground-based LiDAR technology, renowned for 3D sampling resolution, has been altering the routines of biomass inventory. The thesis develops a set of innovative approaches in support of fine-scale biomass inventory, including automatic extraction of stem statistics, robust delineation of plot biomass components, accurate classification of individual tree species, and repeatable scanning of plot trees using a lightweight scanning system. Main achievements in terms of accuracy are a relative root mean square error of 11% for stem volume extraction, a mean classification accuracy of 0.72 for plot wood components, and a classification accuracy of 92% among seven tree species. The results indicate the technical feasibility of biomass delineation and monitoring from plot-level and multi-species point cloud datasets, whereas point occlusion and lack of fine-scale validation dataset are current challenges for biomass 3D analysis from ground.