Mapping of Complex Vegetation Communities and Species Using UAV-LIDAR Metrics and High-Resolution Optical Data

Abstract

Mapping of sensitive species in environments such as upland swamps is important for management of anthropogenic impacts. However, the task is challenging due to the inherently complex distribution of species. Unmanned aerial vehicle (UAV) based remote sensing has the potential to map such complex environments. In this study, an integrated UAV and light detection and ranging (UAV-LiDAR) system was employed to map swamp and non-swamp vegetation in a complex upland swamp environment. The paper details the process of data acquisition, pre-processing, extraction of LiDAR metrics, dimensionality reduction, and different classification methods. Independent component analysis with support vector machine (ICA+SVM) produced the best classification results with a 69.9% overall accuracy (OA) and 0.62 kappa coefficient (k). The OA and k accuracy further improved to 73.6 % and 0.67 by adding high resolution optical (RGB) data along with the LiDAR data. The UAV-LiDAR technology offers an effective approach to distinguish between swamp and non-swamp vegetation communities.