TreeScanPL10k: A Central European tree species dataset of annotated terrestrial laser scanning point clouds

The TreeScanPL10K dataset provides a comprehensive collection of high-resolution terrestrial laser scanning (TLS) point clouds with detailed annotations for individual tree segmentation and species classification. The dataset comprises 10,417 individually segmented trees across 272 circular forest plots (15 m radius) collected from six forest districts across Poland, representing diverse Central European forest ecosystems with varying species composition, age structures, and topographical conditions.

The dataset includes 30 identified tree species, with species labels provided for 7,465 trees (71.7% of the dataset). Each point cloud contains the following attributes beyond standard XYZ coordinates and intensity:

• treeID: Integer values identifying individual tree instances (0 for non-tree points)

• treeSP: Species index according to taxonomic identification (0 for non-tree points and unknown species)

• completelyInside: Binary flag indicating whether the tree crown is completely within plot boundaries (1 = complete, 0 = cropped)

The dataset follows the annotation and point attribute conventions introduced in the ForInstance dataset (Puliti et al., https://zenodo.org/records/8287792), ensuring compatibility through the use of treeID and treeSP attributes.

Data are organized alphabetically with 10 plots per zip file for efficient download and processing. The dataset includes a species_id.csv file providing the mapping between species indices and taxonomic names (Latin and English). A comprehensive plot_summary.csv file accompanies the dataset, containing per-plot information including time of scanning, the plot location rounded up to nearest kilometer (EPSG 2180), TLS device used, tree counts, species distribution, and plot classification. Plots are categorized as Broadleaved, Coniferous, Single species, or Mixed when no single group dominates with more than 70%. The dataset includes an individual_tree_summary.csv file containing tree-level attributes for all segmented trees: plot filename, tree ID (indexed per plot), species code, Latin name, completeness flag (1 = full crown, 0 = cropped), point count, tree height, crown area (2D convex hull of top-down view), and projected area statistics (mean and standard deviation from side views at 0°, 45°, 90°, and 135°). Trees not completely within plot boundaries have NaN values for all area metrics.

Data were collected using FARO Focus 3D X130 and Trimble TX5 scanners with four scanning positions per plot. The annotation process combined automated tree detection with manual segmentation by trained experts. Rigorous quality control ensured accurate segmentation and species assignment through spatial alignment with field reference data.