Robust retrieval of forest canopy structural attributes using multi‐platform airborne LiDAR

Abstract

AbstractLiDAR data acquired from airplanes and helicopters – known as airborne laser scanning (ALS) – are widely regarded as the gold standard for characterizing the 3D structure of forests at scale. But in the last decade, advances in unoccupied aerial vehicle (UAV) technologies have led to a rapid rise in the use of UAV laser scanning (ULS) for mapping forest structure. As both ALS and ULS data become increasingly available, they are being used to derive an ever‐growing number of metrics designed to measure different facets of canopy structure. However, which metrics can be robustly retrieved from both ALS and ULS platforms remains unclear. To address this question, we acquired coincident, high‐density ALS and ULS scans covering 115 plots (4‐ha in size) in an open‐canopy temperate ecosystem in Western Australia. Using this unique dataset, we quantified 32 canopy structural metrics related to canopy height, openness and heterogeneity, including metrics calculated directly from the point clouds and ones measured from derived canopy height models (CHM). Overall, we found that ALS and ULS‐derived metrics were strongly correlated (r2 = 0.90). However, this high degree of correlation masked considerable systematic differences between platforms. Specifically, point cloud metrics were less strongly (r2 = 0.87) correlated and had higher bias (10.7%) compared to CHM‐derived ones (r2 = 0.98; bias = 2.5%). Similarly, metrics of canopy openness and heterogeneity were less strongly correlated (r2 = 0.84 and 0.87) and exhibited greater bias (14.4 and 7.9%) than ones relating to canopy height (r2 = 0.96; bias = 3.8%). Our results indicate that only a small subset of the 32 metrics we tested were directly comparable between ALS and ULS platforms. Consequently, future efforts to combine laser scanning data across platforms and instruments should think carefully about which metrics are most appropriate, especially when working with point cloud data.