Multilevel Extraction of Vegetation Type Based on Airborne LiDAR Data

Abstract

Precise determination of vegetation type is important in remote sensing of the ecological environment. Many studies have explored ecosystem structure on explicit spatial scales using specific remote sensing data, but few studies have considered vegetation information extraction at various landscape levels using LiDAR-derived raster layers. This study determined vegetation information based on LiDAR-derived canopy height model (CHM) and LiDAR-derived forest canopy coverage (FCC) from the typical natural secondary forest of Maoershan Experimental Forest Farm (Northeast China). Geographic object-based image analysis was adopted for all experiments. Optimal classification characteristics and thresholds were determined and classification rule sets established for vegetation type extraction at four levels: vegetation-not-vegetation, vegetation type, forest type, and canopy—canopy gap. We compared and analyzed the capability of LiDAR-derived raster layers extraction to describe vegetation features at these four levels. An area-based assessment method was used for accuracy verification. The research showed that physical information such as vegetation height and canopy density provided by LiDAR point cloud is effective for extracting vegetation characteristics and categories. Moreover, results showed that high-resolution LiDAR-derived raster layers could provide more detailed vegetation information. This study represents a reference for data selection and mapping strategies for hierarchical and multiscale vegetation type extraction.