Abstract
The aim of this study is to examine the performance of a numerical ellipsoid modeling methodology to estimate tree structural characteristics in mixed forest using airborne Light Detection And Ranging (LiDAR) data along with airborne Digital Matrix Camera (DMC) image. In three-dimensional numerical analysis using points cloud of LiDAR data, ellipsoid model has the potential to simultaneously estimate tree top position, diameter and shape of individual tree crown. A Japanese cedar plantation with randomly mixed pine trees was chosen in this study as this type of forest, which is typical of Japanese cedar plantation in Japan. We developed a methodology consisting of both tree species classification and estimation of characteristics of tree structure with the followings steps: (1) classification of area of cedar and pine trees in the mixed plantation by using ortho-DMC image, (2) estimation of number of trees and estimation of tree top location in horizontal plane by standard ellipsoid model for each species, derived from Crown Height Model (CHM) and based on random selections of points clouds on each of the classified areas, (3) estimation of tree top height and realistic shape of individual tree by using a truncated cone shape model and LiDAR points cloud in respective classified areas. The study area is a cedar plantation forest in Northern Japan. LiDAR measurements with a density of 14.65 pulses/m^2 and DMC imagery with a spatial resolution of 10cm are used in this study For validation, ground truth data of tree species, geographic tree position and tree height were measured at the study site. The developed methodology could correctly identify a total of 73 out of 89 cedar trees in the areas classified as cedar, and 12 out of 29 pine trees in areas classified as pine. Validation of estimated tree height resulted in coefficient of determination (R^2) of 0.72 and 0.78 for pine and cedar respectively. This study indicates that fitting the ellipsoid model and the truncated cone shape model to LiDAR points cloud is able to simultaneously estimate tree top position, crown shape and diameter of individual tree crown.
