Tree branch structure is a crucial determinant of timber quality. Metrics such as branch angle and diameter, traditionally measured through labour-intensive manual measurements or assessed subjectively, are required to evaluate knot sizes, which determine log grades or branching characteristics of a given genotype. Over the past 25 years, remote sensing tools have enabled significant advances in the field of branch characterisation, addressing the limitations of field-based methods. Despite these advances, the application of these technologies to characterise small branches and juvenile trees has proven challenging, largely due to the technical difficulties associated with occlusion and the slender target that small branches present. The recent development of proximal sensing tools, such as personal laser scanners (PLS) that enable sub-canopy scanning of forest environments from a mobile platform, has demonstrated potential for addressing occlusion issues. However, their effectiveness for characterising branch structure is little known.

To address these gaps in the research, a comprehensive literature review was conducted to establish the current state-of-the-art for branch characterisation, including technologies, analytical methods, and applications. A study was then conducted, divided into two research components, which aimed to characterise juvenile Pinus radiata branch structure using the most promising technologies and analytical methods identified from the literature. The first component assessed PLS as a tool for branch characterisation, which demonstrated the greatest potential for balancing accuracy and data capture efficiency. The second component assessed two additional scanners with higher accuracy and reduced beam divergence to investigate the impact of scanner quality on branch characterisation. Owing to the lack of specific analytical tools available for PLS, quantitative structural modelling (QSM), a method originally developed for terrestrial laser scanner (TLS) data, was employed for the characterisation of the point clouds.

Data was captured in a 3-ha breeding trial of nine-year-old Pinus radiata in Kaingaroa Forest, New Zealand, from which five trees were selected for crown mapping to ground-truth the laser scanner measurements. Branch height, azimuth, diameter and angle were measured for a total of 440 branches from the five trees. Proximal sensing data were captured by a statically mounted Faro Focus TLS, and second- and third-generation PLS units, the Hovermap PLS and the Hovermap ST-X (STX). Data were processed through custom pipelines to extract individual trees and prepare them for analysis within TreeQSM. The resulting QSM models were then analysed to assess branch detection rates and the accuracy of branch diameter and angle measurements.

The results demonstrated that the combination of QSM and PLS was able to characterise branch diameter with a moderate level of accuracy (R² = 0.82, RMSE = 20.2%). Branch detection rates, however, were low (detection accuracy = 0.07, F1 = 0.13). This was largely due to the low number of true positive detections (51) and the high number of false positives and false negatives (301 and 389 respectively). The use of more powerful and precise scanners had minimal impact on branch characterisation results, with similarly robust levels of branch diameter measurement for the TLS (R² = 0.92, RMSE = 21.7%) and STX (R² = 0.78, RMSE = 18.4%), but low levels of detection accuracy (TLS = 0.08, STX = 0.07) and F1 score (TLS = 0.15, STX = 0.13). Virtually no correlation was observed between field-measured branch angles and measurements from any of the laser scanners. A proof of concept looking at tree-level phenotypes demonstrated some promise, although a more robust study with a greater sample size is required to ascertain the utility of this approach.

Overall, the findings of this study indicate that none of the laser scanners evaluated were suitable for characterising the branch structure of young P. radiata trees using the QSM approach. Discussion of the results implied that current laser scanning technologies are likely not suitable for this method of characterisation, indicating more suitable analytical methods are required to gain meaningful results from the point clouds. This study represents a solid foundation for the operational use of branch characterisation in commercial forest phenotyping and offers valuable insights for future research in this field.