Using sUAS-Derived Point Cloud to Supplement LiDAR Returns for Improved Canopy Height Model on Earthen Dams

Abstract

Abstract Of the nearly 2,400 regulated dams in South Carolina, USA, 97% are earthen dams. Many of these dams are covered in dense vegetation that impede inspections and increase slope instability. Because of their relatively spatially coarse collections, LiDAR datasets are often insufficient on small earthen dams for adequately characterizing both the ground surface and vegetative cover. This study tests the feasibility of small Unmanned Aerial Systems (sUAS) derived “image returns” in extracting a canopy height model (CHM) on a private dam in Columbia, South Carolina. Following traditional methods for creating a CHM from 3D point clouds, both LiDAR- and sUAS image-derived datasets were used to create a composite CHM. A validation process used a set of 30 in situ measured tree heights compared to their modeled heights from the sUAS only, LiDAR only, and composite CHMs. In terms of the root mean squared error (RMSE), the LiDAR only model (5.60 m) outperformed the sUAS only (8.17) and LiDAR + sUAS (7.23 m) composite models. Findings from this study suggest that sUAS derived point clouds have potential, but the structure from motion (SfM) technology and supporting ground control need to be improved before operational application on canopy height modeling for dense vegetation over earthen dams.