UAV-enabled evaluation of forestry plantations: A comprehensive assessment of laser scanning and photogrammetric approaches

The use of unmanned aerial vehicles (UAVs), particularly with high-density point clouds obtained through UAV laser scanning (ULS) and UAV structure from motion (UAV-SfM) techniques, offer cost-effective alternatives for forest inventory. However, the literature lacks comprehensive assessments of their limitations across diverse ranges of age classes and site conditions. This study addressed this gap by evaluating the estimation accuracy of crucial tree attributes, diameter at breast height (DBH) and tree height, in a range of age classes within forest plantations. In addition, this study thoroughly evaluated the performance of ULS and UAV-SfM in diverse site conditions using point clouds obtained from Pinus radiata D. Don plantations, a widely planted commercial timber species worldwide. To achieve this, UAV and field data were gathered from twelve sites, including multitemporal data for four sites. By employing an automated data processing pipeline, individual trees were segmented and structural metrics extracted from tree segments to estimate DBH and tree height at an individual tree level. Results indicated that UAV-SfM and ULS performed comparably in estimating DBH over the entire dataset, with R 2 values of 0.67 and 0.74 and RMSE values of 2.05 cm (11%) and 2.13 cm (11%) respectively. However, ULS generally outperformed UAV-SfM at the site level, achieving higher R 2 values (0.46-0.90 vs 0.21-0.85) and RMSE values (0.33–7.24 cm at 7–24% vs. 0.35–6.15 cm at 8–17%). ULS also consistently outperformed UAV-SfM in tree height measurements across sites, with an average per site RMSE of 0.68 m (5.4%) compared with 1.21 m (11.59%), demonstrating its robustness in diverse conditions. Site-specific factors such as stand maturity and logging debris affected measurement reliability in both datasets, with accuracy improving for younger sites, sites with a more open canopy and more favourable site conditions (less logging debris and weed cover). The study also indicated a moderate relationship between ground sampling distance (GSD) of the imagery and UAV-SfM accuracy. The findings highlight the significance of considering site-specific variables when choosing UAV technologies for conducting forest inventory, ensuring informed decisions in UAV-based forest inventory practices. Consequently, the insights gained from this research hold significant importance for practical forestry applications.