Analysis of vertical forest structure in Siheung with vegetation indices derived from LiDAR data

Abstract

Remote sensing techniques have been developed to identify vegetation structure which affects the diversity and richness of wildlife as habitats. This study aims to develop the vegetation structure index which accurately illustrates the vertical structure of forests utilizing data derived from airborne LiDAR (Light Detection And Ranging), one of the emerging tool for surveying forest in remote sensing area with high accuracy and ability to identify the height of objects in the ground. The vertical structure of forest in the site Siheing is identified with utilizing non-ground returns of LiDAR data compared to the field survey data. The data treatment is classified in six layers, and percent cover of vegetation is estimated by density of LiDAR echoes. This showed distinct vertical structure between deciduous, coniferous and mixed forests in cover graph by layer. These metrics of cover by layer were statistically verified by correlation analysis with layer data from field measurement and utilized to build vegetation structure indices. This study finds that a vegetation index on vertical structure with high accuracy is vertical evenness index, which identifies the evenness of vertical forest structure. This index shows the highest accuracy verified by referenced field data and differentiates three forest types: deciduous, coniferous and mixed forest. Due to the lack of high accuracy of GPS measurement in field survey and impeded detection of understory vegetation by LiDAR, the overall accuracy of LiDAR data was not as high as prediction from other studies. However, the limitation could be overcome by minimizing error on GPS and constructing models considering the Korean forest specific factors such as reflectance of native vegetation and topographic characteristics and this will be the base for the future study on the vertical structure of wildlife habitat in forest using discrete pulse airborne LiDAR.