Predicting forest stand variables from airborne LiDAR data using a tree detection method in Central European forests

Abstract

Abstract In this study, the individual tree detection approach (ITD) was used to estimate forest stand variables, such as mean height, mean diameter, and total volume. Specifically, we applied the multisource-based method implemented in reFLex software (National Forest Centre, Slovakia) which uses all the information contained in the original point cloud and a priori information. For the accuracy assessment, four reference forest stands with different types of species mixture and the area of 7.5 ha were selected and measured. Furthermore, independent measurements of 1 372 trees were made for the construction of allometric models. The author’s ITD-based method provided slightly more accurate estimations for stands with substantial or moderate dominance of coniferous trees. However, no statistically significant effect of species mix on the overall accuracy was confirmed (p < 0.05). The root mean square error did not exceed 1.9 m for mean height, 3.0 cm for mean diameter, and 12.88 m3 ha−1 for total volume.