Evaluating RayCloudTools to estimate single-tree volume

Wild, B., Özkan, T., Ali, M., Pöppl, F., Milenković, M., Hofhansl, F. ORCID: https://orcid.org/0000-0003-0073-0946, Pfeifer, N., Lau, A., & Hollaus, M. (2026). Evaluating RayCloudTools to estimate single-tree volume. Forestry: An International Journal of Forest Research 99 (1) cpaf08. 10.1093/forestry/cpaf087.

Preview

Text cpaf087.pdf - Published Version Available under License Creative Commons Attribution. Download (4MB) | Preview

Abstract

Above-Ground Forest Biomass (AGB) is vital for understanding the carbon cycle, for carbon accounting, and for climate projections. Single-tree AGB measurements or precise estimates are crucial for calibrating and validating remote sensing based AGB mapping (e.g. in the area-based approaches), but remain costly and challenging to acquire. The recently introduced open-source RayCloudTools (RCT) software includes an efficient QSM (Quantitative Structure Model) solution, RCT-QSM that uses Dijkstra’s algorithm to segment and volumetrically reconstruct trees, providing tree volume, which further requires density to obtain mass. The accuracy and practicability of RCT-QSM, however, have remained largely unassessed. This study provides a comprehensive evaluation of RCT-QSM, by comparing its volume estimates against: (i) three publicly available datasets of temporally coinciding TLS (Terrestrial Laser Scanning) scans and destructive measurements, (ii) four existing QSM methods (AdTree, TreeQSM, AdQSM, and SimpleForest), and (iii) allometric model outputs from two experimental plots in Austria, where point clouds were obtained with terrestrial and unmanned aerial vehicle (UAV)-based laser scanning. The comparison with destructively acquired single-tree data (n = 124) from three publicly available datasets shows an overall high correspondence between RCT-QSM derived volumes and destructively harvested volumes (CCC = 0.95) with a moderate negative bias (−7.3%) and an NRMSE of 5%. RCT-QSM outperforms other existing QSM solutions, such as AdTree, AdQSM, SimpleForest, and TreeQSM. TreeQSM metrics, however, show only small differences compared to RCT-QSM. An extensive point density sensitivity analysis featuring 1860 systematically downsampled point clouds from the same dataset demonstrates RCT-QSM’s high robustness to variations in point density. Accuracy and completeness of the results remain stable for point densities as low as one point per 10x10x10 cm voxel. Regarding the large-scale applicability, RCT-QSM provides reliable results for two experimental plots in Austria, which were scanned with TLS and UAV-LS, respectively. RCT-QSM efficiently derives single-tree volume, aligning well with allometric models, demonstrating its applicability across various data acquisition settings and forest conditions.

Actions (login required)

View Item