UAV-LiDAR based tree species identification under leaf-on, leaf-off, and combined canopy conditions in a mixed temperate forest

Accurate identification of individual tree species is essential for assessing forest biodiversity and supporting sustainable ecosystem management. This study investigates the capability of UAV-LiDAR features to identify six tree species in a mixed temperate forest in Germany. Pre-acquired UAV-LiDAR data collected under leaf-on and leaf-off conditions were used to evaluate how structural and intensity-based features contribute to individual tree species identification. A total of 69 LiDAR-derived features describing structural and intensity characteristics at the individual tree level were extracted, and 24 important features were retained after assessing correlation. A Random Forest (RF) algorithm was then applied to identify the tree species and evaluate the importance of features. The results showed that intensity-based features, particularly the mean intensity of first-or-single returns and median intensity, were the most effective for species discrimination. Combining leaf-on and leaf-off conditions achieved the highest identification (overall accuracy = 80%), while leaf-on and leaf-off condition exhibited lower accuracies (75-76%). Coniferous species such as Douglas-fir and Norway spruce, together with the deciduous specie European beech, were consistently identified with high accuracy, whereas morphological similarity between European hornbeam and European beech led to misidentification among deciduous species. These findings demonstrate that UAV-LiDAR derived features exhibit strong potential in distinguishing individual tree species in mixed temperate forest. This study further advances LiDAR based tree species identification by demonstrating the capability of UAV-LiDAR to integrate fine-scale structural and intensity information for improved species identification across canopy conditions.