Assessing lean and positional error of individual mature Douglas-fir (Pseudotsuga menziesii) trees using active and passive sensors

Abstract

There is a growing demand for point cloud data that can produce reliable single-tree measurements. The most common platforms for obtaining such data are unmanned aircraft systems with passive sensors (UAS), unmanned aircraft equipped with aerial lidar scanners (ALS), and mobile lidar scanners (MLS). Our objectives were to compare the capabilities of the UAS, ALS, and MLS to locate treetops and stems and to estimate tree lean. The platforms were used to produce overlapping point clouds of a mature Douglas-fir (Pseudotsuga menziesii (Mirb.) Franco) stand, from which 273 trees were manually identified. Control trees were used to test tree detection accuracy of four algorithms and the number of stems detectable using each platform. Tree lean was calculated in two ways: using the stem location near the canopy and using the treetop. The treetops were detected more accurately from ALS and UAS clouds than from MLS, but the MLS outperformed ALS and UAS in stem detection. The platform influenced treetop detection accuracy, whereas the algorithms did not. The height estimates from the ALS and MLS were correlated (R2 = 0.96), but the MLS height estimates were unreliable, especially as distance from the scanner increased. The lean estimates using the stem locations or treetop locations produced analogous distributions for all three platforms.