Treegraph: tree architecture from terrestrial laser scanning point clouds

Wanxin Yang,Phil Wilkes,Matheus B. Vicari,Kate Hand,Kim Calders,Mathias Disney

DOI: https://doi.org/10.1002/rse2.399

IF: 5.7874

2024-06-05

Remote Sensing in Ecology and Conservation

Abstract:We present Treegraph, an open‐source Python software explicitly designed to retrieve tree architectural traits from the whole tree scale down to individual branches and internodes from TLS point clouds. We validate the method against manual measurements and demonstrate its application to 603 trees across three forest types for testing metabolic scaling theories. This tool has the potential for advancing large‐scale ecological studies on tree architecture. Accurate quantification of tree architecture is critical to interpreting the growth, health and functioning of trees and forests. Terrestrial laser scanning (TLS) offers millimetre‐level point cloud data, but current approaches to 3D tree reconstruction from TLS point clouds primarily focus on retrieving total volume at tree scale for aboveground biomass (AGB) estimation. Few methods have been designed specifically to provide tree architectural properties, including branch‐level morphology and topology, rather than AGB; derived topological traits have tended to be a compromise, and of secondary importance to volume. We present Treegraph, a new approach explicitly designed to retrieve the architectural traits of trees at multiple scales, from the whole tree scale down to individual branches and internodes, using TLS data with limited assumptions about tree form. It provides morphological traits such as branch length and diameter, alongside topological traits including parent–daughter connections of branches and internodes, furcation (branching) number and branch order. We compare Treegraph‐derived morphological and topological traits with manual measurements of branches from eight destructively harvested trees, yielding RMSE values of 0.60 m (5.96%) for branch length, 2.99 cm (33.45%) for branch diameter, 0.46 (19.38%) for furcation number and 0.08 m (33.16%) for internode length, respectively. In a broader application to 603 trees from tropical, temperate and urban forests, we demonstrate that the derived morphological and topological traits support testing of structure‐related metabolic scaling theories. Testing branches over 10 cm in diameter across 18 657 branching nodes shows that Treegraph‐derived branch‐level scaling exponents deviate from WBE predictions, exhibiting area‐preserving behaviour while displaying asymmetry in length and diameter of daughter branches. Available as open‐source Python software, Treegraph provides fine‐level branching network information, promoting improved insights into tree structure and function. This data‐driven approach reduces the need for empirical heuristic parameters, which has the potential for advancing large‐scale ecological studies on tree architecture.

ecology,remote sensing