Assessing the influence of ground-based LiDAR devices, processing solutions, and forest stand characteristics on tree diameter and height estimation

Diameter at breast height (DBH) and tree height (TH) are essential parameters for accurately estimating tree volume, biomass, and other forest metrics. Conventional measurement methods often involve considerable time investment, which has motivated the development of approaches using 3D point cloud technology. These point clouds can be generated by terrestrial laser scanning (TLS), handheld mobile laser scanning (HMLS), or smartphone-based LiDAR. This study aims to determine which factor most significantly affects DBH and TH measurement accuracy. We evaluated multiple devices – specifically, TLS (FARO Focus S150), two HMLS devices (GeoSLAM ZEB Horizon and Stonex X120GO), and a smartphone (iPhone 13 PRO MAX), and processed their point clouds using two most perspective algorithms: FSCT and 3DFin. Data collection spanned forest plots differing in tree species composition and density: low- and high-density beech plots (Fagus sylvatica L.); high-density spruce plot (Picea abies [L.] H. Karst.); and high- and low-density mixed plots comprising Abies alba Mill., Acer pseudoplatanus L., Fagus sylvatica L., Fraxinus excelsior L., and Picea abies [L.] H. Karst. Across all devices and approaches used for DBH estimation, coefficient determination (R2) ranged from 0.692 to 0.998, RMSE from 1.24 to 6.73 cm, and relative RMSE from 4.30% to 42.7%, and bias from –1.84 to 3.50 cm. For TH estimation, R² ranged from 0.539 to 0.997, RMSE from 0.38 to 3.72 m, relative RMSE from 1.68% to 12.9%, and bias from –1.84 to 1.76 m. The results definitely showed that the accuracy of DBH estimation is significantly impacted by the device used and plot characteristics playing a secondary role, whereas TH accuracy depends largely on the selected algorithm. FSCT exhibited notably lower root mean square error (RMSE) for TH, with an average 1.6 m, compared to 2.4 m for the 3DFin. Across all test conditions, TLS consistently reached the smallest errors, while HMLS devices produced comparable results, and smartphone-based LiDAR demonstrated the highest variability.