Terrestrial laser scanning point clouds and RayExtract reconstructions of fire-affected Eucalyptus delegatensis and E. dalrympleana, Tumbarumba, NSW

================================================================================TLS POINT CLOUDS AND RAYEXTRACT RECONSTRUCTIONS60 fire-affected eucalypts, Tumbarumba, NSW, Australia================================================================================ This dataset supports the Methods in Ecology and Evolution article: 'Soft wood-leaf classification preserves crown height and improves tree volume estimates from terrestrial laser scanning under dense epicormic foliage' The paper introduces two enhancements to the RayExtract volume-reconstruction framework: a geometric ray-path correction (Point Direction Enhancement, PDE)and a per-point weighted radius-estimation scheme using return intensity andRandLA-Net wood-class confidence. This archive provides the input point clouds and the resulting tree reconstructions for all seven experimental configurations plus the manual reference. STUDY SITE AND ACQUISITION--------------------------------------------------------------------------------- Site: Tumbarumba Wet Eucalypt SuperSite, Bago State Forest, near Batlow, NSW, Australia (35.66 S, 148.15 E; ~1200 m a.s.l.). - Trees: 60 total - 30 fire-killed, fully defoliated Eucalyptus delegatensis (non-epicormic) and 30 surviving, densely foliated epicormic E. dalrympleana. - Scanner: RIEGL VZ-2000i full-waveform TLS (1550 nm), 121 scan positions on a 10 m grid, 0.03 deg angular step. Scanned 4-6 April 2022. - Source: Source acquisition archived by TERN, doi:10.25901/sd66-pq95. ARCHIVE CONTENTS--------------------------------------------------------------------------------Points.zip Points/ Manual/ manually wood/leaf-labelled cloud (label in alpha) - classification reference Manual_wood/ manual wood-only - VOLUMETRIC REFERENCE for all configs Intensity/ full cloud, calibrated return intensity in alpha RandLANet/ full cloud, RandLA-Net binary wood/leaf labels RandLANet_wood/ RandLA-Net wood-only points RandLANet_confidence/ full cloud, RandLA-Net wood-class probability in alpha (60 trees per folder, *.ply) Meshes.zip RayExtract watertight branch meshes, 60 trees per folder, *_trees_mesh.ply Meshes/ Reference/ Default/ PDE/ Intensity/ Intensity_PDE/ RandLANet_Wood/ RandLANet_Confidence/ RandLANet_Confidence_PDE/ Treefiles.zip (same 8 folders as Meshes) per-segment tree models, *_trees_info.txt Treefiles/ File naming: <treeID>[_epi]_<cloud>_trees_mesh.ply and ..._trees_info.txt. The "_epi" token marks epicormic E. dalrympleana; its absence marks non-epicormic E. delegatensis. FILE FORMATS--------------------------------------------------------------------------------- .ply point clouds (Points/) and triangle meshes (Meshes/), binary PLY. Readable with CloudCompare, Open3D, PDAL, MeshLab. - *_trees_info.txt treeinfo --branch_data output: per-segment x, y, z, radius plus branch attributes (DBH, height, volume, taper, branch order). Wood volume is integrated from these segment models. SOFTWARE--------------------------------------------------------------------------------- RayExtract / RayCloudTools: https://github.com/csiro-robotics/raycloudtools Both the weighted (--alpha_weighting, --segment_alpha_weighting) and ray-adjusted (--use_rays) fitting methods are optional flags in RayExtract. - Method reference: Devereux et al. (2026), "RayExtract: A fast, scalable method for tree volume reconstruction from terrestrial laser scanning", Remote Sensing of Environment 334:115162, doi:10.1016/j.rse.2025.115162. - Analysis code reproducing the figures, tables, and statistics is available in the article's code repository. LICENSE--------------------------------------------------------------------------------Creative Commons Attribution 4.0 International (CC-BY-4.0). Please cite the article and this dataset (Zenodo DOI) if you use these data.