Accurate tree disc volume estimation using TLS: validation and improvement via point cloud repair

Tree trunk volume is a key parameter in forest inventory. Traditional forest surveys typically rely on sample trees and trunk volume equations to estimate tree trunk volume; however, the collection of sample trees is destructive, and trunk volume equations often involve considerable estimation errors. As an emerging technology, terrestrial laser scanning (TLS) has been regarded as an efficient and high-precision alternative for tree trunk volume estimation. Nevertheless, the accuracy of TLS in tree-level trunk volume estimation still lacks systematic evaluation.

To this end, this study used TLS to scan disc samples cut from standard trees, and evaluated the reliability of TLS-based tree trunk volume estimation by comparing point cloud-derived disc volumes with those obtained using the water displacement method. Utilizing the Leica RTC360 scanner, 123 disc samples from four tree species (Altingia excelsa, Robinia pseudoacaci, Platycladus orientalis, and Quercus suber) were collected. A novel bottom surface filling algorithm based on point cloud projection was developed to mitigate data loss at disc bases, followed by Poisson surface reconstruction and trunk volume calculation via the Divergence Theorem.

The results demonstrated high accuracy (R² = 0.940, CCC = 0.9745, rRMSE = 14.92%), with a slight underestimation bias (-5.31 cm³). Species-specific analyses indicated significant differences in estimation accuracy (Kruskal-Wallis, H = 21.1606, p = 0.0001), with Platycladus orientalis exhibiting the highest accuracy (rRMSE = 4.37%) due to its smooth bark and uniform wood structure, while Quercus suber showed the largest errors (rRMSE = 7.10%) attributed to its rough, blocky bark.

Bark characteristics and wood structure were identified as key factors influencing TLS accuracy. The analysis revealed that smoother scanned surfaces-comprising both bark surfaces and cross-sections-resulted in higher estimation accuracy. These surface characteristics are closely linked to species-specific external texture and internal wood structure. This study elucidates the influence mechanisms of species-specific physical characteristics on the accuracy of TLS-based trunk volume estimation and proposes targeted strategies for optimizing scanning parameters and point cloud processing. The study provides a robust theoretical and technical foundation for high-precision, non-destructive tree trunk volume measurement in forestry applications.