How does UAV-LiDAR attributes affect the measurement of individual mangroves?

Keywords: mangrove, LiDAR, individual tree, point density, discrete returns
Session Type: Virtual Guided Poster Abstract
Day: Friday
Session Start / End Time: 2/25/2022 02:00 PM (Eastern Time (US & Canada)) - 2/25/2022 03:20 PM (Eastern Time (US & Canada))
Room: Virtual 7

Abstract

Mangrove forest is among the most diverse and productive ecosystems, but its observation is extremely difficult due to its swampy location. Success has been achieved in remotely monitoring mangrove trees at individual-tree scale using unmanned-aerial-vehicle- (UAV-) based light detection and ranging (LiDAR). However, the impact of LiDAR on the accuracy is unknown. Hence, this study aims to investigate the influence of LiDAR point density and returns on individual mangrove tree delineation and measurement. In the experiment, based on an UAV-LiDAR dataset with 80.24 pt/m2 point density and four returns, three lower point density datasets (60.19 pt/m2, 40.23 pt/m2, and 20.28 pt/m2) and a single-return dataset were generated. All five LiDAR datasets were employed to conduct individual tree delineation and measurement. Treetops were detected using the variable window filtering method, crown boundaries delineated using the marker-controlled watershed segmentation algorithm, and tree height measured as the maximum height inside each crown segment. Results showed that (1) Individual trees were delineated with 47.83% to 50.93% detection accuracy, with the highest accuracy reached at 60.19 pt/m2 point density. Tree height was underestimated by all datasets, while the accuracy decreased as point density decreased. (2) The single-return dataset had lower point density than the original dataset. The accuracy of the single-return dataset follows the overall trend of accuracy change due to point density, indicating that the impact of returns is less than point density. Nevertheless, the generalizability of these findings needs to be tested in other study sites and using other LiDAR datasets.