Forest Inventory: SILFORE

The dataset integrates field-based dendrometric measurements, spatial data representing plot geometry, and raster-based canopy height models (CHM). Each sampling unit includes a central point, a circular plot with variable radius, a 30 m buffer, and a CHM clipped to the buffer extent. All spatial data are referenced in the ETRS89 / Portugal TM06 coordinate system (EPSG:3763).

The data structure includes a GeoPackage file (parcelas.gpkg) containing plot centres, plot polygons, buffer areas, and a CHM index table; a set of CHM raster files in GeoTIFF format stored in the “CHM” folder; a QGIS (v3.40.15) project file (DATASET.qgz) for visualisation; and a README file. The QGIS project includes pre-configured layers and symbology and uses relative paths to ensure portability. Each sampling plot is uniquely identified by the attribute PARCELA, which ensures the linkage between field data, spatial data, and raster files.

Each CHM raster file is named according to the corresponding PARCELA identifier and represents the canopy height model clipped to the 30 m buffer. The relationship between raster and vector data is described in a metadata table provided as a separate CSV file (chm_index.csv), including the plot identifier (PARCELA), site name, and raster file name. Additional information on LiDAR acquisition dates and field data collection is provided in a separate table (acquisition_dates.csv), which also includes mission identifiers and the time difference (in days) between LiDAR acquisition and field measurements (date_diff).

Field inventory data were aggregated at plot level, with each record corresponding to a sampling plot (PARCELA). These data include variables such as location, species, plot area, stand age, number of trees, diameter at breast height (DBH), mean and maximum tree height, dominant height, basal area, and volume. The complete field inventory table is provided as a CSV file (field_inventory.csv).

Field data were collected in circular plots ranging from 250 to 500 m², randomly distributed and georeferenced using RTK GNSS with centimetre-level accuracy. Plot size was adapted to stand density. DBH was measured for all trees, while sample trees were selected using the Draudt method and heights measured using a Haglöf Vertex 5 hypsometer.

Stand age was determined when necessary. Tree-level data were aggregated to plot level, and basal area and volume were computed using standard forestry procedures. Field and LiDAR data were acquired on different dates, with time differences ranging from 0 to 148 days, as explicitly reported in the acquisition_dates.csv table.

These differences are considered negligible due to low growth rates, although they should be taken into account in analyses sensitive to temporal variation. LiDAR data were collected using a DJI Matrice 300 RTK UAV equipped with a Zenmuse L1 sensor, flying at 70 m above ground level, with 80% overlap, a pulse repetition rate of 160 kHz, and an average density of approximately 350 points per square meter. Mission planning and execution were performed using DJI Pilot (v2.5.1.10).

The system recorded up to three returns (first, intermediate, and last) in repetitive scanning mode. Point cloud processing was performed using DJI Terra software (v4.0.10), with ground point classification adapted to local conditions. Digital terrain models (DTM) and digital surface models (DSM) were generated using Agisoft Metashape (v2.1.0), and canopy height models (CHM) were derived as the difference between DSM and DTM.

Additional processing was carried out using QGIS and SAGA GIS (v9.5.1). The resulting CHM rasters have an approximate spatial resolution of 0.042 m and high positional accuracy ensured by RTK GNSS. Forest stand volume was estimated using species-specific allometric equations based on diameter and, where applicable, tree height, following established models for Pinus pinaster, Quercus rotundifolia, and Quercus pyrenaica, as described in the referenced literature.

This dataset can be used for LiDAR calibration, validation of canopy height models, and forest productivity analysis. Users are advised to consider differences in stand structure, species composition, and acquisition dates when performing comparative analyses. The dataset follows FAIR data principles (Findable, Accessible, Interoperable, Reusable), providing a persistent identifier (DOI), an open license, a defined coordinate reference system, standardised measurement units, and detailed documentation of data processing methods and software used.

For further information, users may contact the dataset author at jpmc@ipb.pt.