Consistency of Aerial LiDAR-Derived Forest Metrics Across Multiple Acquisitions

Abstract

As remote sensing technology advances, there is growing interest in using LiDAR to compare structural attributes between forest ecosystems and to monitor forests following active management. However, these emerging applications cannot be confidently employed without quantification of the consistency of repeat LiDAR acquisitions. To address this problem, I explore the impact of different sensors, flight directions, and flight altitudes on LiDAR-derived metric and canopy surface model consistency in the Dinkey Creek Watershed. In addition, I analyze whether structural changes in the forest are detected by LiDAR metrics. I conduct metric comparisons at 15-, 30-, and 46-meter resolutions and canopy surface model comparisons at 1-, 2-, and 3-meter resolutions to determine if processing pixel size affects metric consistency. The results show that collecting LiDAR using different sensors, flight directions, and flight altitudes does not reduce the stability of the metrics. Elevation percentile metrics, canopy proportion metrics, structure classes, and canopy rumple metrics maintain > 0.90 r2 values; and elevation mean, elevation standard deviation, and total cover > 2 meters maintain r2 values above 0.95 between acquisitions collected two years apart. Additionally, the r2 values of the canopy proportion metrics are lower in areas harvested between acquisitions, indicating these metrics reflect structural changes in the forest. Metric consistencies increase with pixel size, from 15-46 meters. Canopy surface model consistency varies, but maintains r2 values above 0.90 for pixel sizes ranging 1-3 meters.