A Comparison of Methodological Approaches for Measuring Tree Height and Crown Projection in Middle Taiga Forest Using UAV LIDAR Data

Boreal forests are the most important reservoirs of terrestrial carbon. Their woody stores up to 30–35% of the total forest carbon stock, thereby making its inventory a task of national importance. In recent decades, the global practice has shifted towards scaling forest inventory data using unmanned aerial vehicles (UAVs), which are an effective and inexpensive means of collecting information at the level of individual ecosystems. The variety of survey and processing methods requires finding optimal approaches to wood stock inventory using UAVs. This article compares two methods (Matlab (ML) and TerraScan (TS)) for estimating the number, height, and projective cover of tree crowns by processing point clouds of different densities (100 and 2800 points/m2; ML100, ML2800, TS100, and TS2800). A LiDAR survey was carried out across three 50 × 50 m sample plots. SP 1: forest crossed by a stream; SP 2: forest on a flat surface; and SP 3: forest bordering an oligotrophic bog. The number of trees identified using ML100/ML2800 for SPs 1, 2, and 3 was 55/66, 67/87, and 174/220. The corresponding figures for TS100 and TS2800 were 66/95, 85/156, and 82/166, respectively. Average tree height reaches 10–24 m, with little difference between the two processing methods at each site. The total projective crown cover for SPs 1, 2, and 3 was 2569, 2494, and 2059 m2 in ML; 2091, 2424, and 1506 m2 in TS. Remote assessment enabled a reasonably complete estimation of the number of trees exclusively in the first layer: the ML100 method was sufficient for this task. However, for a more comprehensive inventory that includes lower canopy layers, the TS2800 method is preferable provided an appropriate minimum tree height value is selected.