TLS-grapevine2024: A terrestrial laser scanner point cloud dataset of grapevines at different phenological stages

Abstract

Grapevines (Vitis vinifera L.) undergo structural and physiological changes throughout the growing season, progressing through distinct phenological stages that require regular monitoring. This dataset consists of high-resolution point cloud data acquired with a stationary terrestrial laser scanner (TLS) to document grapevine development from early leaf development to dormancy. Georeferenced point clouds were generated from 15 TLS scans along two vineyard rows at nine phenological stages. The dataset also includes multispectral and RGB photogrammetric point clouds and orthorectified raster products from an unmanned aerial vehicle survey conducted before harvest. Ground-truth measurements leaf area index, grape production, and pruning wood biomass were collected for each monitored grapevine. As a result, the dataset provides multi-temporal TLS observations that support grapevine structural analysis and development, phenological monitoring, and can be used for the development of AI-based models for precision viticulture.

Specifications Table

Subject	Earth & Environmental Sciences; Computer Sciences
Specific subject area	Terrestrial Laser Scanning (TLS), UAV remote sensing, multispectral imaging, LiDAR, geospatial modelling, artificial intelligence, machine learning for grapevine phenological and structural analysis.
Type of data	Point clouds (.las), raster products (.tif), images (.jpg, .tif), shapefiles (.shp), tabular data (.csv), and metadata files (.json, .md)
Data collection	TLS data acquired with Leica BLK360 G1 using Cyclone FIELD 360; data registration and filtering with Cyclone REGISTER 360 PLUS and CloudCompare. Complementary UAV data collected with DJI P4 Multispectral and Mavic 3T Agronomic measurements include leaf area index (LI-COR 2200C), grape production, and pruning wood biomass. Acquisition dates: April 2024–January 2025. TLS and UAV data at BBCH 89 were synchronized during the same campaign. Weather conditions: clear sky.
Data source location	Institution: University of Trás-os-Montes e Alto Douro City/Town/Region: Arroios, Vila Real, Norte Country: Portugal Coordinates: 41°17′28.83″N 7°43′17.90″W, Altitude: 435 m
Data accessibility	Repository name: Zenodo Data identification number: 10.5281/zenodo.16751663 Direct URL to data: https://doi.org/10.5281/zenodo.16751663
Related research article	None

1. Value of the Data

• •This dataset covers nine phenological stages of grapevine growth from April 2024 to January 2025, providing multi-temporal terrestrial laser scanner (TLS) observations for structural and phenological analysis.
• •It includes TLS point clouds collected at multiple stages and multispectral and RGB data acquired before harvest using unmanned aerial vehicles (UAVs), enabling the assessment of canopy geometry and spectral characteristics and allowing the comparison between sensing platforms.
• •Ground-truth measurements of leaf area index (LAI), grape production, and pruning wood biomass were obtained through standardized field protocols and linked to TLS and UAV data for validation.
• •The dataset can be used to calibrate and validate models for estimating grapevine biomass, canopy volume, and LAI, contributing to non-destructive monitoring techniques and improving the accuracy of proximal sensing methods.
• •Data are organized by BBCH phenological codes and provided in standard formats (.las, .tiff, .shp), ensuring accessibility for studies on grapevine structure, remote sensing, vegetation indices, geometric parameter extraction, sensor fusion, and AI-based modelling.
• •The dataset is openly available to promote reproducibility and collaborative research in deriving canopy metrics from TLS point clouds, correlating them with grape production, and comparing 3D canopy models from LiDAR and photogrammetry, as well as integrating spectral indices with structural data.

2. Background

Advances in precision agriculture have increased the use of three-dimensional data, particularly in precision viticulture [1]. Terrestrial laser scanning (TLS) systems enables non-destructive digitization of plant structures into point clouds [[2], [3]], supporting geometric and biophysical analyses such as canopy volume, light interception, transpiration dynamics, and plant vigour [4,5]. These factors influence crop management and yield [6,[7], [8]]. Compared with conventional measurements, TLS provides high-resolution spatial data, improving vineyard monitoring and decision-making [9,3] and complementing aerial observations for vineyard mapping and data calibration [10,11]. Colleting TLS data at multiple phenological stages allows the monitoring of structural changes throughout the growing season.

The presented dataset covers a full growing cycle (from April 2024 to January 2025), providing multi-temporal point cloud in nine phenological stages of TLS point clouds, integrates UAV multispectral and RGB data, and provides ground-truth validated measurements of variables such as leaf area index (LAI), grape production, and pruning biomass. This combination supports research on grapevine structural monitoring, AI-based modelling, and sensor fusion. Several datasets have been proposed for viticulture-related tasks (Table 1). However, most rely on UAV data [[12], [13], [14], [15]], which often have lower spatial resolution and have temporal coverage limited to a single or small time frame [12,13,15,16].

Table 1.

Overview of grapevine datasets providing point cloud data for vineyard monitoring.

Dataset	Sensors Used	Coverage	Ground-truth variables and/or provided data	Temporal scope
VineLiDAR [12]	UAV LiDAR	Two vineyards	Point clouds at different heights; can be used as ground truth to validate satellite-derived models	Sep 2021, Jul and Sep 2022
EscaYard [13]	UAV multispectral; Smartphone images	Two vineyards	Phytosanitary status per plant, location, number of bunches, geotagged images, UAV orthomosaics and point clouds	Jul 2022
FriûlBot [16]	Autonomous mobile robot (LiDAR, multispectral, IMU, GNSS, …)	Vineyard rows	GNSS waypoints, IMU and odometry data, LiDAR point clouds, multispectral images, vegetation indices	Jul 22–25 2024
Faucher et al. [14]	UAV RGB and multispectral	Two vineyards	Biomass, LAI, coordinates; RGB point clouds, elevation models, orthomosaics, vegetation indices	Jun 2021–Jul 2022; Jun 2023–Mar 2025
Vélez et al. [15]	UAV multispectral	Vineyard subsection	Coordinates of Botrytis-infected clusters, grapevine trunks and ground control points, multispectral images	Sep 2021

3. Data Description

The TLS-Grapevine2024 dataset provides multi-temporal 3D observations of grapevine canopy structure for precision viticulture and computational modelling. It includes processed TLS point clouds, raw UAV multispectral and RGB imagery, processed raster products and point clouds, agronomic measurements, and spatial reference files. The dataset is organized into 11 main folders: seven folders for principal growth stages, one main GroundTruth and one Vectors (Fig. 1). The total dataset size is approximately 10.7 GB.

Fig. 1.

Schematic representation of the data folder structure.

Each growth stage folder includes subfolders for each BBCH code [17] where data were acquired. Inside each BBCH folder, three other subfolders are provided: GroundTruth, TLS_PointClouds and UAV. Ground truth data are stored in CSV format. TLS point clouds are provided in LAS format, named according to its BBCH phenological code and acquisition date. The GroundTruth folder includes a CSV file with agronomic attributes for each monitored grapevine including: ID, variety, grape production (total weight, number of clusters, average bunch weight), pruning biomass, Ravaz index [18], and the geographic coordinates of each grapevine trunk (EPSG:4326 and EPSG: 3763 ETRS89 / Portugal TM06). LAI is provided is provided in CSV in the BBCH folder. Therefore, ground truth refers to physically measured variables which differ from vegetation indices derived from UAV multispectral data A summary of these variables is presented in Table 2. Missing data are indicated by the presence of a file named “Data_Not_Available.txt”. Each BBCH folder also contains a README.md file describing the folder contents, acquisition details, and data organization, and a metadata.json file providing machine-readable information on file names, formats, point counts, and quality parameters. The root folder includes a README.md summarizing the dataset structure and a master metadata.json file with global dataset information, including version, size, coordinate system, and acquisition platforms.

Table 2.

Variables included in the ground truth files.

Variable	Description	Unit
ID	Identifier for each grapevine	N/A
Grapevine variety	Cultivar of the grapevine	N/A
No. of clusters	Total number of harvested grape clusters per grapevine	N/A
Grape production	Total grape production per grapevine	kg
Weight per bunch	Average bunch weight per grapevine	kg
Pruned wood	Pruned wood weight per plant	kg
Ravaz index	Ratio between production and pruning weight	N/A
LAI	Leaf Area Index, leaf surface per ground surface area	m2/m2
Latitude	Geographic coordinate (EPSG:4326) – latitude	degrees
Longitude	Geographic coordinate (EPSG:4326) – longitude	degrees
X	Geographic coordinate (EPSG: 3763) – easting	metres
Y	Geographic coordinate (EPSG: 3763) – northing	metres

The UAV folder contains processed raster products (.tif) for blue, green, red, red edge and near-infrared bands, vegetation indices, RGB orthophoto mosaic, digital surface model (DSM) and digital terrain model (DTM). Dense photogrammetric point clouds (.las) generated from multispectral and RGB imagery are also included. The Vectors folder contains shapefiles (.shp) for the grapevine positions, TLS scanner locations, georeferencing targets, and the spatial extent of the TLS and UAV surveyed areas. Table 3 provides an overview of the data set components.

Table 3.

Dataset folders and components. All data use EPSG:3763 and were acquired using BLK360 G1 (TLS), DJI P4 Multispectral, and DJI Mavic 3T (UAVs).

Folder	File Types	Total Number of Files	Approx. Size	Purpose
TLS	.las	9	8.8 GB	3D canopy structure per phenological stage
UAV	.las, .tif, .jpg	TIF rasters: 13; LAS: 6; Raw images: 201 (TIF), 86 (JPG)	2.1 GB	Multispectral and RGB aerial data
GroundTruth	.csv	2	<1 MB	Agronomic metrics per grapevine
Vectors	.shp	5	<1 MB	Spatial reference and boundaries

4. Experimental Design, Materials and Methods

This section describes the experimental design and workflow for collecting multi-temporal TLS point clouds, UAV data, agronomic measurements, and ground-truth validation. The dataset covers nine phenological stages from the 2024 growing season and the dormant period between 2024 and 2025. The workflow included vineyard setup, sensor deployment (TLS, UAV, LAI), data acquisition at nine phenological stages, point cloud processing, ground-truth data acquisition, and dataset compilation and repository organization.

4.1. Experimental design and site setup

The study was conducted in a 0.10 hectare vineyard located in Arroios, Vila Real, Portugal (41°17′28.83″N, 7°43′17.90″W; altitude 435 m). Grapevines were planted in north–south rows with 2.1 m spacing between rows and approximately 1.1 m between plants. A 117 m² area comprising two rows was selected for TLS data acquisition, these rows were selected due to the health conditions of the grapevines and representation of multiple varieties. These rows contained 52 grapevines of the Touriga Nacional, Viosinho, and Cardinal varieties. A subset of 38 grapevines (29 Touriga Nacional and 9 Viosinho) was analyzed, as these varieties are representative of the Douro Demarcated Region [19]. The vineyard was managed by the farmer without additional interventions. The data collection campaigns were performed at the same time of day, around solar noon, in clear sky conditions. Table 4 summarizes the equipment and software used for data acquisition and processing.

Table 4.

. Summary of the equipment and software used.

Component	Model/Software	Version	Purpose
Terrestrial laser scanner	Leica BLK360 G1	—	Point cloud acquisition
RTK tablet	CHCNAV LT700 RTK	—	Acquisition of coordinates
UAV	DJI P4 Multispectral	—	Multispectral imagery acquisition
UAV	DJI Mavic 3T	—	RGB imagery acquisition
LAI Sensor	LI-COR LAI-2200C	—	LAI data acquisition
Software	Cyclone REGISTER 360	2024.0.1	TLS point cloud registration
Software	CloudCompare	v2.13.2	Point cloud filtering
Software	Pix4Dmapper	4.10.0	UAV photogrammetry
Software	FV2200	v2.1.1	LAI data processing

4.2. Temporal framework and phenological stages

Nine field campaigns were conducted between April 2024 and January 2025, covering different phenological stages defined by BBCH codes [17] (Fig. 3). Phenological stages (Table 5) were classified based on the dominant vegetative phase observed in the majority of grapevines. In addition to TLS data acquisition, LAI measurements and UAV multispectral and RGB imagery were collected at BBCH 89 (27 September 2024). Following this campaign, grape harvest was carried out (29 September 2024), with the number and weight of clusters recorded for each vine. Furthermore, pruning wood was collected and quantified at the end of December 2024 to compute the Ravaz index [18].

Fig. 3.

Illustration of the dominant phenological stages monitored during each data acquisition campaign.

Table 5.

Phenological stages of TLS data collection based on BBCH codes and growth stage.

Growth stage	BBCH Code	Date	TLS scans	UAV flight
Leaf development	13	Apr. 9, 2024	15	—
Inflorescence emerge	55	Apr. 19, 2024	15	—
Flowering	60	May 26, 2024	15	—
Fruit development	71, 75, 79	Jun. 7, 15, Jul. 11, 2024	15 each	—
Ripening	81, 89	Aug. 2, Sept. 27, 2024	15 each	Yes (BBCH 89)
Dormancy	00	Jan. 29, 2025	15	—

In the dataset, each phenological stage is organized as a within its principal growth stage directory including subfolders for TLS, UAV, and ground truth data, along with README.md and metadata.json files describing acquisition details, instrument metadata, and data quality.

4.3. TLS acquisition by phenological stage

The TLS data were collected using a BLK360 G1 scanner (Leica Geosystems AG, St. Gallen, Switzerland), a lightweight tripod-mounted device capable of generating high-density point clouds with RGB data assigned to each point (Fig. 4a). This scanner uses a high-speed time-of-flight system with Waveform Digitizing (WFD), offering 360⁰ horizontal and 300⁰ vertical field of view. It captures up to 360,000 points per second at 830 nm and can achieve a 3D point accuracy of 6 mm at 10 m and 8 mm at 20 m, with an acquisition range of 0.6 m to 60 m. RGB imagery is captured using an integrated 15-megapixel three-camera system producing 150-megapixel spherical images. The TLS was controlled via Leica Cyclone FIELD 360 (version 5.2.1) on an iPad Air 2 (Apple Inc., Los Altos, CA, USA), enabling the remote operation and real-time visualisation of the scan data and assess data quality during in the field.

Fig. 4.

Terrestrial laser scanner (TLS) and field setup: (a) scanner positioned adjacent to the permanent reference marker; (b) spherical georeferencing target; and (c) TLS data acquisition within the vineyard plot.

Data acquisition was performed between 13:00 and 15:00 using a medium density resolution and High Dynamic Range (HDR) imaging. Each scan required approximately six minutes, including equipment setup and stabilisation, resulting in approximately 1 hour and 30 min per campaign.

At each scanning point (locations in Fig. 2), a permanent reference marker (concrete block with a central mark) was placed, enabling the repositioning of the TLS through all scanning campaigns (Fig 4a). For spatial georeferencing purposes, spherical polystyrene targets with 0.15 m diameter (Fig. 4b) were placed at two heights (1.30 m and 1.73 m above ground level), spaced 6.3 m apart in a zig-zag layout, ensuring that a minimum of three targets was within the scanner’s field of view at each scanning position, introducing both planimetric and altimetric variability. The TLS was positioned at a maximum distance of 3.16 m from the plants, with an effective scanning range of up to 10 m. The spherical targets and TLS acquisition points were georeferenced to for accurate point cloud registration (point locations in Fig. 2). To prevent occlusion, branches obstructing the targets were removed prior to data acquisition.

Fig. 2.

Geographical location of the vineyard site used for data acquisition and phenological monitoring.

In each digitized phenological stage 15 scans were performed with the TLS positioned along the inter-row space (Fig. 4c) to capture both sides of the canopy structure in two rows. The scanner was moved five times per row at intervals of 6 m in three rows (Fig. 2).

4.4. TLS point cloud pre-processing

Point cloud registration, alignment, and overlay was performed in Leica Cyclone REGISTER 360 (version 2024.0.1) by integrating multiple scans performed at each acquisition date. After registration, individual point clouds corresponding to each phenological stage were grouped into bundles. The data were then cropped to the boundaries of the scanned vineyard rows and exported in LAS format. Noise filtering was applied using the Statistical Outlier Removal (SOR) method in CloudCompare (v2.13.2). The SOR filter parameters were configured to consider the twenty-five nearest neighbours (k = 25) and a standard deviation threshold of 3.0. This process removed outlier points, reducing noise while preserving the grapevine canopy structure. An overview of the point cloud data along the different phenological stages is presented in Fig. 5.

Fig. 5.

TLS point clouds generated for each phenological stage (April 9, 2024 to January 29, 2025) with their Biologische Bundesanstalt, Bundessortenamt und Chemische Industrie (BBCH) codes.

4.5. Leaf area index measurement

The LAI measurements were performed at BBCH 89, this stage represents full canopy development prior to harvest, making it relevant for structural validation and modelling. Measurements were obtained using the LAI-2200C Plant Canopy Analyzer (LI-COR, Nebraska, USA) to serve as physically validated ground-truth indicator of canopy density.

During measurements, the sensor was positioned parallel to the ground, and five readings were recorded for each grapevine: one above the canopy (reference) and four below at the ends and midpoints of horizontal transects. A 90° view cap was used to maintain sensor orientation towards the vineyard row.

Corrections for overestimations caused by sensor orientation or light variation, the built-in “clip” function was applied. This adjustment standardizes any below-canopy readings with transmittance higher than their corresponding above-canopy values by setting them to a maximum value of 1. Data were corrected using the FV2200 software (version 2.1.1), which applies a scattering correction and normalize data for solar angle, geographic location, and atmospheric conditions. These corrections ensure accuracy and reproducibility by following the manufacturer’s protocol for isolated row measurements.

For each grapevine, depth (X), height (Z), and length (Y) perpendicular to the row were recorded and were used as input into the “Change Canopy Model” function, which generated a virtual canopy profile. The software then identified which zenith angles intersected the plant canopy, and only those were used in the final LAI calculation. Each LAI record was stored in the corresponding BBCH stage folder and includes the grapevine ID and geographical coordinates and .

4.6. UAV data acquisition and processing

UAV flights were conducted at BBCH 89 using the P4 Multispectral and the Mavic 3T (DJI, Shenzhen, China) for multispectral and RGB data acquisition, respectively. The P4 Multispectral is equipped with six 1/2.9-inch CMOS sensors (2.08 MP resolution), including five monochrome sensors capturing blue (450 nm ± 16 nm), green (560 nm ± 16 nm), red (650 nm ± 16 nm), red edge (730 nm ± 16 nm), and near-infrared (840 nm ± 26 nm). The Mavic 3T has an integrated camera with a 1/2-inch CMOS sensor (12 MP resolution). Both UAVs are equipped with RTK modules and the sensors are stabilized by a 3-axis gimbal.

The flights were performed at a height of 40 m above ground level. The multispectral survey used 80 % longitudinal and 70 % lateral imagery overlap, while the RGB survey used 90 % longitudinal and 70 % lateral overlap. Radiometric calibration was applied to multispectral data using the onboard sunlight sensor and reflectance panel images acquired prior to the flight.

All imagery was processed in Pix4Dmapper (version 4.10.0, Pix4D SA, Lausanne, Switzerland) to generate photogrammetric point clouds (for RGB and for each multispectral band) and orthorectified raster products (digital surface model, digital terrain model, RGB orthophoto mosaic, reflectance of each spectral band, and vegetation indices). Data alignment was ensured through the RTK positioning and by using the spherical targets (Fig. 3b) as ground control points. RGB and multispectral datasets were processed using consistent workflows and referenced to the same coordinate system (ETRS89 / Portugal TM06, EPSG: 3763) to ensure interoperability.

Limitations

The dataset has limitations that should be considered when assessing its scope. Although it covers the main grapevine phenological stages, data were not collected for senescence (BBCH 91–99) and some phenological stages during flowering (BBCH 61–69), which restricts a complete analysis of the growth cycle. Placeholder folders containing a file named “Data_Not_Available.txt” are included in the repository for unrecorded stages and missing data to maintain structural consistency of the dataset.

Only one campaign of LAI measurements was performed (BBCH 89), limiting the temporal comparisons with TLS data in different phenological stages. Similarly, UAV data were collected in a single campaign (BBCH 89), reducing the potential for multi-temporal UAV-based analysis. However, the available UAV-based data include orthorectified raster products and photogrammetric point clouds in five spectral bands and RGB, enabling direct comparisons with TLS point clouds for geometric validation and vegetation index computation.

Another limitation is the absence of phytosanitary records. Grapevines may present issues that were not monitored, which can be cause misinterpretation of reflectance patterns, as canopy management and treatments were performed by the winegrower without plant health status inspection. Furthermore, the dataset does not include multi-year observations, which would be important for studying interannual variability in grape and biomass production and evaluating consistency in growth patterns.

Ethics Statement

The authors have read and follow the ethical requirements for publication in Data in Brief and confirming that the current work does not involve human subjects, animal experiments, or any data collected from social media platforms.

CRediT Author Statement

Leilson Ferreira: Conceptualization, Methodology, Software, Validation, Formal analysis, Investigation, Data curation, Writing - original draft, Visualization. Pedro Marques: Formal analysis, Investigation, Visualization, Data Curation, Writing - review & editing. Fernando Portela: Conceptualization, Methodology, Formal analysis, Visualization, Writing - review & editing. Joaquim J. Sousa: Resources, Writing - review & editing, Supervision, Funding acquisition. Raul Morais: Resources, Project administration, Funding acquisition. Emanuel Peres: Resources, Project administration, Funding acquisition. Luís Pádua: Conceptualization, Methodology, Validation, Investigation, Writing - review & editing, Supervision.

Data Availability

• ZenodoGrapevine Point Cloud Dataset Acquired via Terrestrial Laser Scanning at Multiple Phenological Stages (Original data)