Analysis of UAV lidar information loss and its influence on the estimation accuracy of structural and functional traits in a meadow steppe

Accurate quantification of grassland structural and functional traits is the foundation for grassland management and restoration. Light detection and ranging (lidar), especially the unmanned aerial vehicle (UAV) lidar, has been recognized as an accurate and effective technique for local to regional-scale vegetation structural and functional traits estimation. However, in grassland ecosystems, it is more likely to be influenced by UAV lidar information loss caused by dense vegetation canopies. In this study, we investigated how UAV lidar information loss may occur and how it may influence the estimation accuracy of grassland structural and functional traits by comparing it with terrestrial laser scanning (TLS) and field measurements in a meadow steppe of northern China. Five structural traits (i.e., mean vegetation height, maximum vegetation height, standard deviation of vegetation height, canopy cover, and canopy volume) and one functional trait (i.e., aboveground biomass) were estimated from the UAV lidar data and TLS data for evaluation. The results showed that TLS-derived structural and functional traits had a much higher accuracy than UAV lidar-derived traits. By comparing with TLS data, we found that UAV lidar data had a much more prevailing information loss at canopy tops than at canopy bottoms. The average height loss of UAV lidar at canopy tops reached over 0.30 m, and the average relative height loss reached over 49%, comparing to a value of 0.03 m and 6% at canopy bottoms. Maximum vegetation height, standard deviation, and the distance from the UAV lidar system to the ground were the three most influential factors on UAV lidar information loss at canopy tops, indicating the commonly seen sharp canopy tops of grasslands were prone to be missed by the UAV lidar system. UAV lidar information loss at canopy tops had a much stronger influence on the estimation accuracy of grassland structural and functional traits than that at canopy bottoms. With the decrease of information loss at canopy tops, UAV lidar can be used to extract grassland structural and functional traits with a comparable accuracy to TLS. Among the five grassland traits, aboveground biomass was the least influenced by UAV lidar information loss. This study is a very first evaluation on the UAV lidar information loss in grassland ecosystems and its influence on grassland structural and functional trait estimation, which can provide guidance for UAV lidar data collection and processing in future grassland applications.