Terrestrial laser scanning forest dataset of a 5 ha Pine stand in Brandenburg, Germany

Description

Table of contents

This dataset contains the following components:

1.       Point clouds (.laz)

Showing the 3D forest structure of the stand

The stand was divided into 25 tiles.

The tiles overlap by a 5 m buffer to avoid edge effects during data processing. In the shapefiles folder, we provide shapefiles of the unbuffered core areas, which can be used to clip point clouds if they are to be remerged into larger entities, or to clip results from single-tile processing before merging.

2.       Shapefiles (.shp) for

·         Scan positions
File: scanpositions_motzen.shp; EPSG: 25833; point

·         Covered area (convex hull of scan positions buffered by 15 m)
File: covered_area_motzen.shp; EPSG: 25833; polygon

·         Area of interest (fence surrounding the target stand)
File: AOI_fence_motzen.shp; EPSG: 25833; polygon

·         Buffered tiles (the area covered by the single point clouds)
File: tiles_buffered_by_5m_motzen.shp; EPSG: 25833; polygon

·         Tile core areas (tiles without buffers, to be used for clipping results of spatial processing of the point cloud data)
File: tiles_core_areas_motzen.shp; EPSG: 25833; polygon

·         Tree positions with DBH and height as extracted from 3DFin
File: treepositions_from3DFin_motzen.shp; EPSG: 25833; point

3.       Raster files (.tif) at a resolution of 10 cm showing the

·         Digital terrain model
File: dtm_res10cm_motzen.tif

·         Canopy height model
File: chm_res10cm_motzen.tif

For the whole covered area.

Raw scan data and height-normalized point clouds are not directly available due to size restrictions, but can be made available upon reasonable request.

Methods

Point cloud

The terrestrial laser scanning (TLS) point cloud was acquired with a RIEGL VZ-400i (RIEGL, Horn, Austria) using a gridded scanning pattern with 10 m distance between single scan positions. In total, 483 single scans were taken. Scans were recorded with 0.04° angular resolution and a pulse repetition frequency of 1200 kHz, resulting in an approximate point spacing of 7 mm at a distance of 10 m and a maximum measurement range of 250 m. Field sampling was conducted during the leaf-off season of winter 2024/25 (December 2024 to February 2025). To keep conditions comparable throughout the data collection, no scanning was done during wind, rain, or snow. Even though the period of data collection was relatively long, we expect no changes in the vegetation state, as all data collection tasks were completed in a period where the vegetation lies dormant.

→ Exact Scan positions can be found in this file: 2_shapefiles/scanpositions_motzen.shp

All scans were co-registered in RiSCAN PRO using outdoor non-urban processing settings. Point clouds were visually inspected for misaligned single scans. Single scans with a poor automatic alignment were removed and repeatedly aligned until a visually good alignment was achieved. Points with reflectance values below -23.8 dB were removed as noise. Additionally, scatter was filtered using statistical outlier removal (functions classify noise and sor from the package lidR in R, Roussel et al. 2020, Core Team 2024, Roussel and Auty 2026). For the outlier classification, the 15 neighboring points for each point were taken into account. Points farther removed than the average distance plus 7 times the standard deviation are removed as noise. The final point cloud was subsampled to 1 cm resolution and cropped to the plot extent. Georeferencing of the final cloud was improved using 12 ground control points measured with an Emlid REACH RS 4 Pro (Emlid, Budapest, Hungary) and SAPOS RTK correction data.

→ Point clouds are stored in folders with the prefix: 1_point_clouds

Digital terrain model

Ground points were classified using the cloth-simulation filter from Zhang 2016 using the classify_ground function from the lidR package in R (Roussel et al. 2020, Core Team 2024, Roussel and Auty 2026), with a class threshold of 0.05 m, a cloth resolution of 0.5 m and a rigidness of 3. Ground points were rasterized to obtain a digital terrain model using a triangular irregular network.

→ The 10 cm resolution DTM can be found in this file: 3_raster_files/dtm_res10cm_motzen.tif

Canopy height model

The previously described digital terrain model was used to height-normalize the point cloud. From the height-normalized cloud, a canopy height model was calculated using the rasterize_canopy function from the lidR package in R (Roussel et al. 2020, Core Team 2024, Roussel and Auty 2026).

→ The 10 cm CHM can be found in this file: 3_raster_files/chm_res10cm_motzen.tif

Tree positions, height, and diameter

Tree positions and measurements were extracted from the cloud using 3DFin (Laino et al. 2024). Missing diameter at breast height (DBH) values for detected trees were filled using the expanded 3DFin output table by calculating the median of fitted circles at heights of 0.5 to 2.3 m. Circles with a size of 1.2 m or larger were excluded from those calculations, as this is an unrealistically large diameter for this stand. Tree positions with fewer than 3 fitted circles were rated as unreliable and falsely detected and deleted.

Trees with a DBH < 7 cm were removed from the final output, as 7 cm is the DBH threshold for considering trees as mature in the German national forest inventory.

→ Tree positions can be found in this file: 2_shapefiles/treepositions_from3DFin_motzen.shp

References

During data processing, we used the following software and software packages:

D. Laino, C. Cabo, C. Prendes, R. Janvier, C. Ordonez, T. Nikonovas, S. Doerr, C. Santin, 3DFin: a software for automated 3D forest inventories from terrestrial point clouds, Forestry (Lond) 97 (2024) 479–496. https://doi.org/10.1093/forestry/cpae020.

R Core Team, R: A Language and Environment for Statistical Computing, Vienna, Austria, 2024. https://www.R-project.org/.

J.-R. Roussel, D. Auty, Nicholas C. Coops, Piotr Tompalski, Tristan R.H. Goodbody, A. Sánchez Meador, Jean-François Bourdon, F. de Boissieu, A. Achim, lidR: An R package for analysis of Airborne Laser Scanning (ALS) data, Remote Sensing of Environment 251 (2020) 112061. https://doi.org/10.1016/j.rse.2020.112061.

J.-R. Roussel, D. Auty, Airborne LiDAR Data Manipulation and Visualization for Forestry Applications, 2026. https://cran.r-project.org/package=lidR.

W. Zhang, J. Qi, P. Wan, H. Wang, D. Xie, X. Wang, G. Yan, An Easy-to-Use Airborne LiDAR Data Filtering Method Based on Cloth Simulation, Remote Sensing 8 (2016) 501. https://doi.org/10.3390/rs8060501.

Notes

Other

CRediT Authorship statement

Miriam Herrmann: Data Curation, Data collection; Marius Derenthal: Project Administration, Data collection; Florian Plewnia, Ute Eggert, Johann Meindl, Anika Sieber, Marion Stellmes: Data collection; Fabian Fassnacht: Resources, Supervision.