An ultra-high-density dataset of UAV-LiDAR collected in Eucalyptus plantations in Guangxi, China

Saplings were planted along contour lines at an initial spacing of 2 × 3 m (equivalent to 1,665 individuals ha-1). The forests in each plot were all even-aged stands of the same strain. Within the stand, some trees occurred as single-stemmed sprouts, others as double-stemmed, and a few even as triple-stemmed individuals, with their proportions varying among plots.

Additionally, some trunks bore bark that had loosened and exfoliated but remained attached to the trunk. A brief description of each site is provided below.l Huangmian test site (HM) is located within the Panlong Division of the state-owned Huangmian Forest Farm and is an unbounded plot (109.785°E, 24.677°N). The trees in this plot underwent a qualitative thinning in the four year, leaving approximately 900 trees/ha; about 92.3% of the trees were single‑stemmed sprouts.

A few saplings, shrubs, and climbing vines were also present. Canopy closure, determined from the ULS point clouds as the proportion of vegetation points above 2 m relative to all point clouds, was 0.75. The topography is mountainous, with elevations ranging from 262 to 282 m and a mean slope of 25°.

The region lies in a transitional zone between the southern and central subtropical climatic belts, with an annual mean temperature of 20.4°C and a mean annual rainfall of approximately 1,483.8 mm.l Binyang test site (BY) is located in central Binyang County and is an unbounded plot (108.868°E, 23.096°N). Of the trees, 48.3% were double‑stemmed sprouts. The plot supported abundant shrubs and lush herbaceous vegetation.

Canopy closure was 0.69. The topography is mountainous, with elevations ranging from 190 to 230 m and a mean slope of 26°. The region lies south of the Tropic of Cancer and has a subtropical monsoon climate, with a mean annual temperature of approximately 20.8°C and an annual rainfall of about 1,550–1,650 mm.l Datang (DT), Gaofeng (GF), and Wutang (WT) test sites are geographically close, and all fall within the jurisdiction of the State-owned Gaofeng Forest Farm.

The DT and WT test sites are unbounded plots, with central coordinates of 108.558°E, 22.995°N and 108.565°E, 22.888°N, respectively. The GF test site is a rectangular plot of 900 m² (30 m × 30 m) at 108.383°E, 22.996°N. Shrubs, herbs, and lianas were sparse across all three plots. Canopy closure was 0.74, 0.83, and 0.72, respectively.

Most trees in the DT and GF plots were single‑stemmed sprouts, whereas 45.6% of trees in the WT plot were double‑stemmed sprouts. All three sites feature mountainous terrain. Their elevations range from 218 to 230 m, 327 to 344 m, and 182 to 210 m, with mean slopes of 31°, 30°, and 17°, respectively.

The region has a mean annual temperature of 21.6°C and a mean annual rainfall of 1,300 mm.l Nachen test site (NC) is located in Jiangnan District, Nanning City, and comprises two unbounded plots with central coordinates of 108.292°E, 22.462°N and 108.298°E, 22.466°N. Trees in both plots were almost exclusively single‑stemmed sprouts. Both plots had sparse shrubs but lush herbaceous vegetation. Canopy closure was 0.72 and 0.77, respectively.

The topography is hilly, with elevations ranging from 252 to 264 m and 265 to 276 m, and a mean slope of 17° for both plots. Climatic conditions are similar to those of the Gaofeng test site.l Shangsi test site (SS) is located in Shangsi County and is an unbounded plot (107.998°E, 22.216°N). Of the trees, 24.1% were double‑stemmed sprouts.

Shrubs were nearly absent, and herbaceous cover was very sparse. Canopy closure was 0.63. The site is mountainous, with elevations ranging from 304 to 318 m and a mean slope of 26°.

This area has a southern subtropical monsoon climate, with a mean annual temperature of 21.2°C and a mean annual rainfall of 1,217 mm.2. Field data Nine field plots were established across seven test sites. For plots with boundaries (i.e., the Gaofeng test site), a diameter tape was used to measure the DBH (at 1.3 m) of trees with a DBH greater than 5 cm within the boundaries.

For plots without boundaries, DBHs were measured continuously for 20–30 trees per row across 3–4 rows of trees. The coordinates of each sample tree stump were measured using the network real-time kinematic global navigation satellite system (RTK-GNSS). Due to errors in RTK-GNSS measurements, tree position adjustments were required.

To this end, in the ArcGIS software environment, the field-survey tree map (in ArcGIS shapefile format) was overlaid with the trunk point cloud file (height range: 0.2–2.4 m, .las format), and the positions of the sample trees were carefully moved manually to the centers of adjacent point cloud clusters, with the expected positioning error of the sample trees not exceeding 5 cm. After recalculating the planar coordinates, the final obtained sample tree positions were taken as the initial centers of the sample trees, which could be regarded as the centers of the cross-sections at a height of 1.3 m above the ground.

The field measurement outputs consist of eight DBH record files (.xlsx) and eight tree-location files (ArcGIS shapefile). The DBH filenames encode the test-site abbreviation and measurement date. Notably, the NC site contains two plots, and for this site, the sensor abbreviation is additionally included in the DBH filenames.l 1_HM_DBH_field_20251110.xlsx;l 2_BY_DBH_field_20250923.xlsx;l 3_DT_DBH_field_20250316.xlsx;l 4_GF_DBH_field_20240824.xlsx;l 6_WT_DBH_field_20251110.xlsx;l 7_NC_DBH_field_SA_20260601.xlsx;l 8_NC_DBH_field_SAmini_20251110.xlsx;l 9_SS_DBH_field_202601270.xlsx.

The tree-location filenames contain the study-area abbreviation. For the GF and NC sites, where two sensors were used for data collection, the sensor abbreviations are also included in the filenames.l 1_HM_tree_location.shp;l 2_BY_tree_location.shp;l 3_DT_tree_location.shp;l 4_GF_tree_location_SA.shp;l 5_GF_tree_location_L2.shp;l 6_WT_tree_location.shp;l 7_NC_tree_location_SA.shp;l 8_WT_tree_location_SAmini.shp;l 9_SS_tree_location.shp.

The projection system for all of the aforementioned tree distribution files is WGS_1984_World_Mercator.3. LiDAR data Three ULS systems were deployed to acquire nine datasets of laser point clouds across the seven study areas. These devices and their manufacturers are listed below: l The Fly2Map Done SF-1200 UAS integrated with an AcuteLas SA130 laser scanner (Abbreviated as SA, South Surveying & Mapping, Guangzhou, China), The URL for the equipment presentation is southinstrument.com/product/details/pro_tid/1/id/211.html and southinstrument.com/product/details/pro_tid/1/id/270.html.l The second system was a Fly2Map SF‑1200 UAV equipped with an AcuteLas SA130mini laser scanning system (SAmini, southsurvey.com/products_view/id/224.html).

The SA130mini sensor is a lightweight version of the SA130, with technical specifications similar to those of the SA130.l The DJI Matrice 300 RTK UAS equipped with a Zenmuse L2 sensor (Abbreviated as L2) (Shenzhen Dajiang Innovation Technology Co., Ltd., Shenzhen, China), The URL for the equipment presentation is enterprice.dji.com/cn//matrice-350-rtk?site=enterprise&from=nav and enterprice.dji.com/cn/zenmuse-l2?site=enterprise&from=nav.

In order to acquire ultra-high-density point cloud data and ensure the completeness and uniformity of point clouds around individual tree trunks, this study adopted a predefined dual-grid crossing flight planning scheme featuring dual 45° oblique intersections, with detailed workflows elaborated as follows: (1) Constructing the primary base grid (0° and 90° orthogonality). The UAV first entered the test site via the designated waypoint and executed a lawnmower-style zigzagging flight along a path parallel to the tree-planting rows at a fixed strip spacing of 6‒10 m to complete the initial full coverage of the operational zone.

After finishing the first flight traverse, the operator precisely rotated the flight heading by 90° and executed a second round of reciprocating coverage was executed with the identical spacing. This action form the first set of base grids composed of two mutually orthogonal flight trajectories. (2) Generating the secondary crossing grid (45° and 135° oblique intersection).

The entire flight-route matrix was rotated by 45°, centered on the geometric centroid of the pre-constructed base grid. The fixed-strip-spacing reciprocating flight logic was replicated the generation of the second set of orthogonal grid paths. These paths formed a 45° angular offset relative to the primary base grid trajectories.

Through the mutually staggered arrangement and cross-coverage of the two independent flight grids, this planning scheme achieved a comprehensive, blind-spot-free scanning of the target plot. After the full mission was completed, the UAV exited the test site along the pre-defined egress route, forming a fully closed operational loop. The flight route map for the HM test site is shown below; the flight route for other test sites are similar.l 1_HM_trajectory.shp Raw data were preprocessed using the software provided by the laser scanner manufacturer.

Subsequent point-cloud processing included denoising and classification into ground and non-ground points via the improved triangulated irregular network (TIN) densification (IPTD) filtering algorithm implemented in LiDAR360 (v 5.0) (Green Valley Inc., Beijing, China). The ground points were then used for elevation normalization. A total of nine datasets were derived from eight plots across seven study areas, with two datasets acquired at the GF site using the SA and L2 sensors, respectively.

Each dataset was named by combining the test-site abbreviation and sensor abbreviation—for instance, HM-SA denotes the SA sensor dataset at the Huangmian site, while NC-SAmini refers to the SAmini sensor dataset at the Nachen site. Owing to the large file sizes, only the trunk-layer point clouds (0–5 m height range) are provided in this release. The filenames encode the study area, sensor, and acquisition date, along with an indication of the original point-cloud density.

The nine point-cloud filenames are listed below:l 1_HM_SA_5m_20251110.las (Original point cloud density: 46,8882 pts/m2);l 2_BY_SA_5m_20251023.las (63,193 pts/m2);l 3_DT_SA_5m_20260415.las (46,604 pts/m2);l 4_GF_SA_5m_20240828.las (72,809 pts/m2);l 5_GF_L2_5m_20241104.las (65,927 pts/m2);l 6_WT_SA_5m_25050317.las (51,102 pts/m2);l 7_NC_SA_5m_20260528.las (64,775 pts/m2);l 8_NC_SAmini_5m_20260617.las (66,915 pts/m2);l 9_SS_SA_5m_20260310.las (32,592 pts/m2).

The projection system for all of the point cloud files is WGS_1984_World_Mercator.