By combining skeleton graph optimization with fractal self-similarity, the approach reduces errors common in existing models, such as incorrectly connected branches and gaps caused by incomplete data scans. Tests on 29 trees across tropical forest sites showed the method achieved near-perfect accuracy, with a concordance correlation coefficient of 0.994, outperforming widely used models TreeQSM and AdQSM.
Trees play an essential role in regulating ecosystems, maintaining biodiversity, and mitigating climate change. Precise 3D models allow scientists to calculate fundamental parameters such as diameter at breast height (DBH), above-ground biomass, and wood volume, which are vital for estimating carbon stocks and analyzing forest structure. LiDAR (Light Detection and Ranging) technology has become a key tool for capturing point cloud data to build such models, but traditional methods still face challenges. Current approaches often produce fragmented or erroneous branch connections, struggle with incomplete canopy data, and are sensitive to outliers. These limitations hinder reliable application in large-scale forest monitoring and biodiversity studies. To overcome these barriers, researchers sought a method that could adapt to data quality issues while maintaining high fidelity in structural reconstruction.
A study (DOI: 10.1016/j.plaphe.2025.100060) published in Plant Phenomics on 1 June 2025 by Zhenyang Hui’s team, East China University of Technology, offers valuable tools for forestry, carbon monitoring, and ecological conservation.
The proposed 3D individual tree modeling method was rigorously evaluated using LiDAR datasets collected from three tropical forest sites in Peru, Indonesia, and Guyana, which represented diverse ecological types including moist terra firme, peat swamp, and lowland forests. A total of 29 trees were scanned with a Riegl VZ-400 terrestrial laser scanner operating at 1550 nm, providing fine resolution and accurate point clouds that were co-registered with 1 cm precision. To establish reliable reference values, all trees were destructively harvested, and volumes of stems, buttresses, and large branches were calculated using standard forestry formulas. This provided benchmark data against which the modeling performance of the new method, termed SfQSM, could be tested. Results showed that SfQSM delivered highly accurate volume estimates, with most deviations from harvested volumes falling between −1 m³ and 1 m³ across all sites. Notably, smaller-diameter trees from Indonesia produced lower deviations, while larger Peruvian trees showed greater discrepancies, indicating that tree size influences modeling precision. Quantitative evaluation revealed that SfQSM achieved a mean deviation of 0.162 m³, a root mean square error of 1.023 m³, and relative errors as low as 0.01% and 0.09%, significantly outperforming two benchmark methods, TreeQSM and AdQSM. TreeQSM’s errors were more than twice as high, and AdQSM’s deviations exceeded SfQSM’s by over thirtyfold. Moreover, the concordance correlation coefficient of SfQSM reached 0.994, confirming superior alignment with true tree volumes. Visual comparisons further highlighted its robustness: TreeQSM often produced fragmented trunks, and AdQSM generated overfitted or non-existent branches, while SfQSM consistently produced continuous, realistic models that conformed to point clouds. Collectively, these findings demonstrate that the proposed method not only surpasses existing approaches in statistical accuracy but also delivers high-fidelity structural reconstructions, making it a reliable tool for ecological applications.
By producing highly accurate individual tree models, the method provides a reliable foundation for biodiversity assessments, species classification, and habitat analysis. It also enables precise estimation of tree volume and biomass, which is essential for calculating carbon stocks and tracking forest contributions to the global carbon cycle. In addition, accurate 3D reconstructions enhance virtual ecological landscapes and can inform sustainable forestry practices, reforestation programs, and climate change mitigation strategies.
###
References
DOI
10.1016/j.plaphe.2025.100060
Original URL
https://doi.org/10.1016/j.plaphe.2025.100060
Funding information
This work was supported by the Funding of National Key Laboratory of Uranium Resources Exploration-Mining and Nuclear Remote Sensing (2024QZ-TD-26), Outstanding Young Talents Funding of Jiangxi Province (20232ACB213017), Double Thousand Plan of Jiangxi Province (DHSQT42023002), National Natural Science Foundation of China (NSF) (42161060, 41801325) and the Natural Science Foundation of Jiangxi Province (20242BAB25176, 20192BAB217010) for their financial support.
About Plant Phenomics
Plant Phenomics is dedicated to publishing novel research that will advance all aspects of plant phenotyping from the cell to the plant population levels using innovative combinations of sensor systems and data analytics. Plant Phenomics aims also to connect phenomics to other science domains, such as genomics, genetics, physiology, molecular biology, bioinformatics, statistics, mathematics, and computer sciences. Plant Phenomics should thus contribute to advance plant sciences and agriculture/forestry/horticulture by addressing key scientific challenges in the area of plant phenomics.