FOREST PARAMETER ESTIMATION FROM AIRBORNE LIDAR DATA IN RUGGED MOUNTAINOUS AREAS

Abstract

During the past decade, the procedure for quantification of forest parameters using LiDAR data has been rapidly improved. Among various forest parameters, biomass is the paramount in understanding the potentials productivity of forests. Various methods have been developed to estimate biomass at both plot and individual tree levels. In order to quantify biomass at the individual tree level, tree crown delineation must be conducted, which is sometimes challenging especially for multi-layer dense forests in rugged mountainous areas. In this study, Light Detection and Ranging (LiDAR) data were used to delineate tree crowns and estimate biomass in a mountainous forest. Firstly, a novel algorithm was proposed to identify individual tree crowns using the concept of live crown ratios based solely on LiDAR data. Then, above ground biomass (AGB) was estimated using machine learning approaches based on tree crowns delineated in the previous step. LiDAR-derived metrics related to forest parameters such as tree height and crown areas as well as topographic characteristics extracted based on the delineated tree crowns were used to estimate AGB. Three machine learning models— random forest, Cubist, and support vector regression—were evaluated for AGB estimation and relative importance of input variables was examined.