Boundary conditions for studying branch-scale tree growth strategies using tree quantitative structure model time series

Abstract

Advances in Light Detection and Ranging (LiDAR) technology, along with point cloud modeling techniques like Quantitative Structure Models (QSM), have improved the accuracy of non-destructive above-ground tree biomass estimation. However, branch-level tree growth analysis using QSM time series remains underexplored due to challenges in data quality and methodology. This study investigates the boundary conditions in terms of data and species to facilitate robust QSM generation, which could enable branch-level growth analysis using multi-temporal LiDAR data and QSM. A multi-scan terrestrial laser scanning dataset and a dataset from a tower-based system were used to assess the impact of data acquisition setup and data quality on QSM reconstruction for birch (Betula pendula), pines (Pinus sylvestris), and spruces (Picea abies). The results show that reliable QSMs for detecting branch-level growth require a minimum spatial resolution of approximately 500 pts./m3 with a uniform point density, a maximum uniform 3D point distance of 2 cm, and gaps smaller than around 20 cm. Although smaller spherical noise clusters can be removed using denoising techniques, larger and denser noise clusters (e.g., > 9500 points within 1 m radius) are more likely to be misidentified as additional branches. Foliage removal methods risk modeling accuracy by inadvertently removing woody points. Regarding species, birches were more accurately modeled than pines and spruces. While QSMs are reproducible for single time points, comparing branches over time is challenging due to inconsistencies in modeled branching order and scanner positioning. Nonetheless, tree-level QSM metrics remain statistically consistent, revealing diverse growth strategies within and across species.