Low-cost UAV LiDAR reconstructs poplar branch architecture for high-throughput breeding
image: Study site details. (a) Location and coverage of the study site; (b) field photograph; (c) UAV LiDAR point clouds covering the test area.
Credit: Plant Phenomics
Researchers have developed a new way to reconstruct the branching architecture of poplar trees from low-cost unmanned aerial vehicle (UAV) light detection and ranging (LiDAR) point clouds, turning dense point clouds into detailed branch skeletons and measurable traits. The work focuses on first- and second-order branches, using material transport flux and object-level geometric features to identify branch tips, trace branch paths and estimate branch length and angle. Applied to a 5-hectare breeding plot, the method captured structural variation across 248 poplar genotypes and revealed links among branch traits. The approach is designed to support high-throughput phenotyping, helping breeders assess complex tree architecture at scale without destructive sampling. This offers a practical route toward faster, architecture-driven selection in tree breeding.
Poplar is a major forestry species in China, valued for timber, ecological restoration and shelterbelts. Branch architecture strongly affects growth, photosynthesis and water use, making it a key target for breeding. Yet traditional measurements rely on climbing or felling trees, which is labor-intensive and limits sample size. Ground-based light detection and ranging (LiDAR) can capture branches but is slow for large plots, while unmanned aerial vehicle (UAV) LiDAR offers higher efficiency but lower resolution. Existing branch reconstruction algorithms were mainly built for conifers, and their performance drops for broadleaf species with complex, irregular crowns. Low-cost UAV LiDAR further challenges reconstruction because point clouds are sparser and less uniform. Based on these challenges, in-depth research is needed on scalable branch reconstruction for broadleaf trees using low-cost UAV LiDAR.
The study, led by researchers at the Institute of Forest Resource Information Techniques, Chinese Academy of Forestry, and the Key Laboratory of Forestry Remote Sensing and Information System, National Forestry and Grassland Administration, was published (DOI: 10.1016/j.plaphe.2026.100180) in Plant Phenomics (2026, Volume 8, Article 100180). The team developed a branch reconstruction algorithm that combines material transport flux with object-level geometric features. Using DJI Zenmuse L1 UAV LiDAR data from a 5-hectare poplar breeding station in Hubei Province, China, they reconstructed first- and second-order branches and estimated branch length and angle for phenotypic analysis.
Data were collected in January 2024 at the Shishou National Poplar Breeding Station using a DJI Zenmuse L1, a low-cost UAV LiDAR system. Flights at 70 m and 8 m/s, with a 45° gimbal tilt, produced an average point density of 1,100 points/m². After noise removal, ground filtering, normalization and individual tree segmentation, the algorithm extracted skeleton points with a voxel-based near-centroid method. A connected graph and Dijkstra shortest paths then modeled material transport from the root to every skeleton point. Branch tips were identified where transport flux dropped to a local minimum. The highest endpoint defined the first-order branch, while second-order branches were traced using path length and curvature. Recursive smoothing reduced zigzag artifacts before parameter estimation. In 20 validation trees, first-order branch identification achieved an F1-score of 1.00; second-order branches averaged 0.69. Second-order branch length estimation had a root mean square error (RMSE) of 0.47 m, and branch angle RMSE was 7.06°. Across 248 genotypes, second-order branch length showed the highest variation, with a coefficient of variation (CV) of 29.68%. Correlation analysis found a moderate positive relationship between first-order branch length and average second-order branch length (r = 0.34), and K-means clustering separated the genotypes into six architectural groups.
The authors said the method addresses a critical gap for broadleaf trees. By combining biological transport principles with object-level geometry, they said, complex poplar branches can be reconstructed from lower-cost UAV LiDAR even when point clouds are sparse or uneven. They added that the approach moves branch phenotyping from individual trees toward entire breeding plots, giving breeders a scalable way to compare architecture across many genotypes. The authors said future work will test the method under leaf-on conditions and in more complex canopies.
The method could help poplar breeding programs screen large numbers of trees for branch architecture, a trait tied to light capture, space use and yield. By linking reconstructed branches to genotypes, breeders can identify superior lines and group them by branching strategy. The six phenotypic clusters found in the study illustrate how length and angle trade-offs create distinct canopy forms. The framework may also support forest management, carbon and biomass modeling, and gene mapping. However, the approach currently relies on leaf-off UAV LiDAR data; foliage occlusion, uneven point density and data gaps remain limitations. Further validation under leaf-on conditions and across species will be needed before broad operational use.
###
References
DOI
Original Source URL
https://doi.org/10.1016/j.plaphe.2026.100180
Funding information
This research was funded by the Biological Breeding-National Science and Technology Major Project (grant number 2023ZD0405605) and the National Natural Science Foundation of China (grant number 42301510).
About Plant Phenomics
Plant Phenomics publishes breakthrough research that integrates genomics, genetics, physiology, molecular biology, bioinformatics, statistics, mathematics and computer science to advance plant phenotyping—from the cellular level to entire populations. The journal tackles key scientific challenges in phenomics by showcasing cutting edge technologies for data acquisition, management, interpretation and modelling. Its scope spans practical applications across all plant sciences, aiming to translate novel phenotyping tools into real world solutions for breeding, crop management and fundamental plant biology. By fostering interdisciplinary innovation, Plant Phenomics serves as a hub for researchers developing high throughput, accurate and scalable methods to decipher complex plant traits, ultimately accelerating progress in agriculture and plant science.