The data describe vegetation outlines and tree tops above 1m in height as polylines and points. Data have been processed from a digital terrain model (DTM) and digital surface model (DSM), converted from raw LiDAR data. The LiDAR dataset was acquired for Cornwall and Devon (all the land west of Exmouth) during the months of July and August 2013. The data were created as part of the Tellus South West project. Full details about this dataset can be found at https://doi.org/10.5285/78dba959-989b-43d4-b4da-efd2506e0c8e

Terrestrial Laser Scanning (TLS) offers complementary and non-destructive methods for capturing three-dimensional information on forest structure. ITS is an essential process for modern forestry and ecological studies as it enables the precise isolation and characterization of individual trees from surrounding forest, facilitating critical analyses such as biomass estimation and advanced forest structure assessment. This study explores the applicability of two established Individual Tree Segmentation (ITS) algorithms, dalponte2016 and watershed, to TLS data for the assessment of forest structure. Conducted within the David C. Lam Asian Garden of University of British Columbia, the research employed a Trimble TX08 scanner to collect high-density point clouds, aiming to improve urban forest management and sustainability practices. Despite the benefits of TLS’s high spatial resolution and rapid data collection, challenges such as over-segmentation, under-segmentation and occlusion by forest canopy impeded accurate tree stem and understory detection. Both algorithms, traditionally used in Airborne Laser Scanning (ALS), demonstrated limitations when applied to TLS data, highlighting the need for algorithm refinement and the incorporation of machine learning techniques. This study contributes to the literature by critically assessing the performance of ITS algorithms in TLS data processing, offering insights for future advancements in precise forestry inventory. The absence of ground truth data and inaccessibility of ground control points (GCPs) underscored the necessity for a more comprehensive approach with regards to data collection and the validation of tree segmentation methods.

lidaRtRee is an R package that provides functions for forest analysis using airborne laser scanning (LiDAR remote sensing) data: tree detection (method 1 in Eysn et al., 2015) and segmentation; forest parameters estimation and mapping with the area-based approach. It includes complementary steps for forest mapping: co-registration of field plots with LiDAR data (Monnet and Mermin, 2014); extraction of both physical (gaps, edges, trees) and statistical features from LiDAR data useful for e.g. habitat suitability modeling (Glad et al., 2020) and forest maturity mapping (Fuhr et al., 2022); model calibration with ground reference; maps export. It is available on CRAN. Tutorials are available in the documentation. lidaRtRee est un package R pour l'analyse de la structure des forêts à partir de données acquises par scanner laser (LiDAR) aéroporté : détection d'arbres (méthode 1 dans Eysn et al., 2015) et segmentation ; estimation de variables forestière et cartographie par approche surfacique. Il propose des fonctions additionnelles telles que : géoréférencement des données de terrain avec les données LiDAR (Monnet and Mermin, 2014); extraction de statistiques et d'objets (trouées, lisières, arbres) utilisables par exemple pour la modélisation d'habitat (Glad et al., 2020) et la cartographie de la maturité des forêts (Fuhr et al., 2022); calibration de modèles avec des données de terrrain ; production de cartes. Le package est disponible sur CRAN. Des tutoriels sont disponibles dans la documentation.

This dataset contains UAV photogrammetric products, UAV LiDAR point clouds, hand-made tree crown segmentation polygons, and georeferenced tree height and stem diameter measurements in carbon sequestration plantations in Quebec, Canada. The products included in this dataset are RGB orthomosaics, LiDAR and photogrammetry point clouds, digital surface models and processing reports. The tree measurements include species, height, stem diameter and location.

Supplementary Material

This material regards the paper entitled "A Density-Based Algorithm for the Detection of Individual Trees from LiDAR Data".

The Readme.txt file explains all the contents of the data package, which consists of the data supporting the paper and the MATLAB script for the Individual Tree Detection and Measurement (ITDM).

Please cite the related article if using the data or the script.

Latella, M., Sola, F., & Camporeale, C. (2021). A Density-Based Algorithm for the Detection of Individual Trees from LiDAR Data. Remote Sensing, 13(2), 322.

This data set provides high-resolution (1-m) tree canopy cover for states in the Northeast USA. State-level canopy cover data are currently available for Pennsylvania (data for nominal year 2008), Delaware (2014), and Maryland (2013). The data were derived with a rules-based expert system which facilitated integration of leaf-on LiDAR and imagery data into a single classification workflow, exploiting the spectral, height, and spatial information contained in the datasets. Additional states will be added as data processing is completed.

Our objective is to improve the measurements of canopy structure and biomass of a forest stand and detect their annual changes via a ground-based laser scanning technology, also known as terrestrial lidar (TLS). A TLS instrument utilizes lasers to scan an environment, measure 3D locations of objects encountered by lasers and detect intensities of laser lights scattered by those objects back to the TLS instrument. TLS have shown abilities and is being further explored to retrieve stem diameter, stem count density, stand height, leaf area index, foliage profile, foliage area volume density, aboveground biomass and other useful forest structural parameters rapidly and accurately. Three TLS instruments used in this project include: (1) the Echidna (R) Validation Instrument (EVI), built by CSIRO Australia; (2) Dual-Wavelength Echidna® Lidar (DWEL), built by Boston University, University of Massachusetts, Lowell, University of Massachusetts, Boston and CSIRO Australia; (3) Compact Biomass Lidar (CBL), built by University of Massachusetts, Boston. To validate the forest structural parameters retrieved using these TLS instruments, we set up a one-ha (100 m by 100 m) forest site and collected tree inventory data including: tree location, tree species, DBH, tree height and crown dimension since 2007 with a two-year gap of 2008 and 2009. Lidar data are available from the ORNL DAAC (http://dx.doi.org/10.3334/ORNLDAAC/1045).

Laser scanning point clouds of forest stands were acquired in southwest Germany in 2019 and 2020 from different platforms: an aircraft, an uncrewed aerial vehicle (UAV) and a ground-based tripod. The UAV-borne and airborne laser scanning campaigns cover twelve forest plots of approximately 1 ha. The plots are located in mixed central European forests close to Bretten and Karlsruhe, in the federal state of Baden-Württemberg, Germany. Terrestrial laser scanning was performed in selected locations within the twelve forest plots. Airborne and terrestrial laser scanning point clouds were acquired under leaf-on conditions, UAV-borne laser scans were acquired both under leaf-on and later under leaf-off conditions. In addition to the laser scanning campaigns, forest inventory tree properties (species, height, diameter at breast height, crown base height, crown diameter) were measured in-situ during summer 2019 in six of the twelve 1-ha plots. Single tree point clouds were extracted from the different laser scanning datasets and matched to the field measurements. For each tree entry, point clouds, tree species, position, and field-measured and point cloud-derived tree metrics are provided. For 249 trees, point clouds from all three platforms are available. The tree models form the basis of a single tree database covering a range of species typical for central European forests which is currently being established in the framework of the SYSSIFOSS project.

This dataset contains data underlying the publication: "Estimation of above‐ground biomass of large tropical trees with terrestrial LiDAR" and consists of the following data folders:

1_ReferenceMeasurementData (Destructive sampling tree AGB data): Destructive sampling measurement data of 29 large tropical trees from 3 sites (Indonesia, Peru and Guyana) for estimating tree wood volume and tree AGB.

2_AllometricEqInventoryData (Forest inventory data): Forest inventory data of 29 individual large tropical trees from 3 sites (Indonesia, Peru and Guyana) for estimating tree AGB with allometric equations.

3_QsmCylinderData (Quantitative Structure Models): Quantitative Structural Models (QSM) cylinder model outputs (3D tree architecture models), generated from the individual TLS point cloud data of 29 large tropical trees from 3 sites (Indonesia, Peru and Guyana).

4_LidarTreePoinCloudData (TLS tree point cloud data): TLS point cloud data for 29 large tropical trees in 3 study sites: Indonesia (peat swamp forest in Central Kalimantan, Borneo), Peru (lowland tropical moist forest in Madre de Dios) and Guyana (lowland tropical moist forest in Cayuni-Mazaruni).

This is an airborne LiDAR data, which is part of Saihanba National Forest Park, including white birch and larch.

Tropical rainforest canopies, where stems and crowns reside, are hotspots of biological diversity, mediate the global biochemical processes and are the interface between organic nature and the atmosphere. Ecosystem functions such as growth, competition and mortality, depend on the spatial arrangement of tree crowns that varies significantly across forest types and disturbance gradients. The exact nature and function of tropical tree canopies are not well known. Field inventories often focus on measuring the horizontal component of the canopy (tree diameter and species identification) and rarely measure tree height or crown size. Remote sensing airborne lidar (light detection and ranging) is considered a complementary method to circumvent the difficulty of field techniques in measuring forest canopies, with recognized ability to estimate canopy height and vegetation vertical profiles. A recent lidar-based method (adaptive mean shift, AMS3D) shown improved capabilities to deal with the complexity of tropical forest and to provide spatial explicit measurements of crown morphology and packing by modeling individual trees.

Here, we report tree-level forest inventory (field and LiDAR) over nine 1 ha forest plots in the La Selva Biological Station, Costa Rica. The field inventory includes stem location (tree height > 10m), tree diameter, tree height and wood density. Crown radii measurements for a tree subsample (n=285) were used to derive local tree allometry. The lidar forest inventory includes the point clouds measurements over the 1 ha plots with an additional attribute to cluster the individual points into tree crowns. The lidar tree crowns were modeled to estimate tree location, tree height, crown depth, crown area, crown volume and crown surface. Additionally, we provide tree-level LiDAR forest inventory for 45360 plots (1 ha) distributed over old-growth, selectively-logged, secondary and swamp forests. The field inventories have been compared and show good agreement in terms of stem number density, size-frequency distributions (stem diameter, tree height), basal area and aboveground biomass.

Methods The methods used to derive the field and lidar forest inventory are described in

Clark DB, Ferraz A, Clark DA, Kellner JR, Letcher SG, Saatchi S (2019) Diversity, distribution and dynamics of large trees across an old-growth lowland tropical rain forest landscape. PLoS ONE 14(11): e0224896. https://doi.org/10.1371/journal.pone.0224896

and

Ferraz A, Saatchi S, Longo M, Clark DB (2020) Tropical tree size-frequency distributions from airborne lidar. Ecological Applications

respectively.

We present a comprehensive UAV-based dataset combining LiDAR and multispectral imagery over five certified reclaimed wellsites (each ~100 m × 100 m) near Grande Prairie, Alberta (55° 10′ 15″ N, 118° 47′ 46″ W). Data were acquired on three dates—August 2023, October 2023, and May 2024—using a DJI Zenmuse L1 LiDAR sensor and a MicaSense RedEdge-P multispectral camera. The LAZ files include georeferenced, RGB-colored point clouds from the L1 sensor, with per-point reflectance in five spectral bands—blue, green, red, red-edge, and near-infrared—interpolated from the RedEdge-P imagery. Please note that multispectral reflectance is unavailable for Wellsite 442 in May 2024. Each point is manually annotated as vegetation or ground, and we also provide CSV files containing the manually corrected 3D coordinates of individual tree tops. This dataset supports the development and evaluation of algorithms for individual-tree detection, vegetation filtering, and long-term monitoring of revegetation in post-reclamation landscapes.

Tree Canopy polygon data was derived using Maryland LiDAR Charles County - DEM Meters, and boundary corrections/updates based on 2017 six inch orthoimageryLiDAR Description: Survey National Geospatial Program Base LiDAR Specification, Version 1. The data was developed based on a horizontal projection/datum of UTM Zone 18 North, NAD83, meters and vertical datum of NAVD1988 (GEOID12A), meters. LiDAR data was delivered in RAW flight line swath format, processed to create Classified LAS 1.2 Files formatted to 2283 individual 1500m x 1500m tiles, and corresponding Intensity Images and Bare Earth DEMs tiled to the same 1500m x 1500m schema, and Breaklines in ESRI Shapefile format. The data was then converted to a horizontal projection/datum of NAD83 Maryland State Plane Coordinate System, Feet. LiDAR was delivered in Classified LAS 1.2 Files formatted to 1927 individual 4000' x 6000' tiles, and corresponding Intensity Images and Bare Earth DEMs tiled to the same 4000' x 6000' schema, and Breaklines in ESRI Shapefile format. Ground Conditions: LiDAR was collected in Winter 2014, while no snow was on the ground and rivers were at or below normal levels. In order to post process the LiDAR data to meet task order specifications, Quantum Spatial established a total of 59 QA control points and 95 Land Cover control points that were used to calibrate the LiDAR to known ground locations established throughout the SANDY_Restoration_VA_MD_DC_QL2 project area.

This dataset comprises co-aligned hyperspectral and LiDAR data collected of European beech (Fagus sylvatica) forest within core protected areas of the UNESCO Rhӧn Biosphere Reserve, Germany. Data was collected using the Headwall Hyperspec Nano sensor flown from a unmanned aerial vehicle (UAV) in September 2020. The dataset comprises image and LiDAR data of four sites, each approximately 8ha in size. The study forests were subject to the extreme drought event that impacted central Europe in 2018/2019 and this project sought to collect data to enable individual tree and stand level assessment of the response (canopy damage and defoliation) of European beech trees to extreme drought events. The hyperspectral images available in this dataset have approx. 5cm pixel size with an associated LiDAR dataset and are suitable for identifying individual trees and the degree of canopy damage (defoliation, discolouration, and mortality) sustained by individuals/stands within the forest. The work was supported by the Natural Environment Research Council (Grant NE/V00929X/1). Full details about this dataset can be found at https://doi.org/10.5285/23d6a61c-c1cf-4c1b-a65c-f3fe42fc0e76

IntroductionThe unmanned aerial vehicle -based light detection and ranging (UAV-LiDAR) can quickly acquire the three-dimensional information of large areas of vegetation, and has been widely used in tree species classification.MethodsUAV-LiDAR point clouds of Populus alba, Populus simonii, Pinus sylvestris, and Pinus tabuliformis from 12 sample plots, 2,622 tree in total, were obtained in North China, training and testing sets were constructed through data pre-processing, individual tree segmentation, feature extraction, Non-uniform Grid and Farther Point Sampling (NGFPS), and then four tree species were classified efficiently by two machine learning algorithms and two deep learning algorithms.ResultsResults showed that PointMLP achieved the best accuracy for identification of the tree species (overall accuracy = 96.94%), followed by RF (overall accuracy = 95.62%), SVM (overall accuracy = 94.89%) and PointNet++(overall accuracy = 85.65%). In addition, the most suitable number of point cloud sampling of single tree is between 1,024 and 2048 when using the NGFPS method in the two deep learning models. Furthermore, feature value of elev_percentile_99th has an important influence on tree species classification and tree species with similar crown structures may lead to a higher misidentification rate.DiscussionThe study underscores the efficiency of PointMLP as a robust and streamlined solution, which offers a novel technological support for tree species classification in forestry resource management.

This dataset represents tree crowns derived from LiDAR data. Tree crowns are defined as circles that fitted to the approximated radius of a tree's branches and leaves. The tree crowns were derived using LiDAR data. The operation was constrained to those areas of tree canopy, using the tree canopy dataset developed separately for this project, which employed automated techniques coupled with manual editing to extract tree canopy from imagery and LiDAR. Mapping of tree crowns was performed using an automated feature extraction technique that incorporated segmentation and morphology routines. The automated routine first created objects from the tree canopy using an inverse watershed segmentation algorithm applied to the LiDAR nDSM (normalized digital surface model) datasets. These objects were then refined using the spatial properties of the objects. Centroids were computed by finding the geometric center of the tree object. Attributes include the tree height and radius. The height was calculated using the 98th quantile of the LiDAR nDSM height to reduce outlier values. The radius was then calculated from the tree centroid using the formula. This radius was used to derive the tree crowns.For more information, please see the 2021 Tree Canopy Assessment.

Title

Terrestrial lidar data collected from four large tropical rainforest trees in Floresta Nacional de Caxiuanã

Authors

A. Burt M. Boni Vicari A. C. L. da Costa I. Coughlin P. Meir L. Rowland M. Disney

Contact

a.burt@ucl.ac.uk

License

These data are distributed under the terms of the Creative Commons Attribution 4.0 International Public License (CC BY 4.0) - see the LICENSE file for details.

Overview

Terrestrial lidar data were acquired from four large tropical rainforest trees prior to harvest (diameter range: 0.6-1.2m, height range: 30-46m) in a natural closed forest stand in Floresta Nacional de Caxiuanã, Pará, Brazil (approx. coordinates in the WGS-84 datum: -1.798, -51.435 degrees), during August/October 2018. This dataset includes: i) raw lidar data, ii) tree-level point clouds, and iii) quantitative structural models. A complete description of the four trees, these data, and the companion destructive harvest data can be found in our paper entitled: ‘New insights into large tropical tree mass and structure from direct harvest and terrestrial lidar’.

Acquisition

Neighbouring vegetation surrounding each tree was removed before data collection. Lidar data were acquired using a RIEGL VZ-400 terrestrial laser scanner. A minimum of 16 scans (upright and tilt) were collected from 8 scan positions around each tree. The angular step between sequentially fired pulses was 0.04 degrees, and the distance between scanner and tree varied. This arrangement provided a 45 degree sampling arc around each tree, and a complete sample of the scene from each position. The laser pulse has a wavelength of 1550nm, a beam divergence of 0.35mrad, and the diameter of the footprint at emission is 7mm. The instrument was in ‘High Speed Mode’ (pulse repetition rate: 300kHZ), ‘Near Range Activation’ was off (minimum measurement range: 1.5m), and waveforms were not stored.

Processing

i) Individual scans were registered onto a common coordinate system using RIEGL RiSCAN PRO (v2.7.0, http://riegl.com). ii) Tree-level point clouds were extracted from the larger-area point cloud using treeseg (v0.2.0, https://github.com/apburt/treeseg). iii) Points were classified as returns from wood or leaf material using TLSeparation (v1.2.1.5, https://github.com/TLSeparation). iv) Points from buttresses were manually removed using CloudCompare (v2.10.3, https://cloudcompare.org). v) Quantitative structural models were constructed using TreeQSM (v2.3.2, https://github.com/InverseTampere/TreeQSM) via optqsm (v0.1.0, https://github.com/apburt/optqsm).

File and directory naming convention

The four trees are identified: CAX-H_T1, CAX-H_T2, CAX-H_T3 and CAX-H_T4. The various files and directories are described as follows:

./CAXH-H/ ├───CAX-H_T1/ (Directory: tree-level directories) ├───CAX-H_T2/ ├───CAX-H_T3/ ├───CAX-H_T4/ │ ├───2018-10-06.001.riproject/ │ │ ├───ScanPos001/ (Directory: individual scan directories containing raw lidar data and other auxiliary files; odd: upright, even: tilt) │ │ ├───ScanPos.../ │ │ ├───ScanPos020/ │ │ │ ├───181006_194253.rxp (File: measurement data stream) │ │ │ ├───181006_194253.mon.rxp (File: monitoring data stream) │ │ ├───matrix/ (Directory: contains the registration matrices) │ │ │ ├───001.dat │ │ │ ├───....dat │ │ │ ├───020.dat (File: 3x4 matrix used to rotate and translate scan 20 into the coordinate system of scan 1) │ │ ├───clouds/ (Directory: contains tree-level point clouds) │ │ │ ├───CAXH_T4.txt (File: point cloud of CAX-H_T4 as extracted by treeseg) │ │ │ ├───CAXH_T4nb.txt (File: CAXH_T4.txt with buttress points manually removed using CloudCompare) │ │ │ ├───CAXH_T4w.txt (File: CAXH_T4.txt with leafy returns removed using TLSeparation) │ │ │ ├───CAXH_T4l.txt (File: CAXH_T4.txt with woody returns removed using TLSeparation) │ │ │ ├───CAXH_T4wnb.txt (File: CAXH_T4.txt with buttress points manually removed using CloudCompare, and leafy returns removed using TLSeparation) │ │ ├───models/ (Directory: contains quantitative structural models constructed from the tree-level point clouds) │ │ │ ├───CAXH_T4.mat (File: quantitative structural model of CAXH_T4.txt) │ │ │ ├───CAXH_T4nb.mat │ │ │ ├───CAXH_T4w.mat │ │ │ ├───CAXH_T4wnb.mat │ │ │ ├───CAXH_T4.models.dat (File: reports the volume (m3) and standard deviation (m3) of the QSMs) │ │ │ ├───intermediate/ (Directory: contains intermediate QSMs generated by optqsm) │ │ │ │ ├───CAXH_T4/ │ │ │ │ ├───CAXH_T4nb/ │ │ │ │ ├───CAXH_T4w/ │ │ │ │ ├───CAXH_T4wnb/ │ │ │ │ │ ├───CAXH_T4wnb-1.mat │ │ │ │ │ ├───CAXH_T4wnb-....mat │ │ │ │ │ ├───CAXH_T4wnb-10.mat

General description

Data about individual tree segments detected from the Finnish national 5 pt/m^2 airborne laser scanning data. The dataset contains 281304 individual trees from a test site of 3km x 3km located in Padasjoki, Finland. The ALS data has been collected in 2019.

Data description

The individual tree segment dataset has been published as a CSV file. The tree height, DBH, stem volume and above-ground biomass have been predicted using Random Forest Machine Learning models. The rest of the features have been calculated directly from the ALS point cloud. The CSV file contains the following columns:

• LX and LY: tree location (location of max height in the canopy height model) in the ETRS-TM35FIN coordinate reference system (EPSG:3067)
• Hmax: tree height (based on highest returns)
• Gele: ground elevation at tree location (interpolated from ground points)
• lowBranch: height of lowest branch (limited accuracy)
• crownV: crown volume as 3D convex hull
• species: predicted species: 1 = pine, 2 = spruce, 3 = deciduous tree
• H: predicted tree height
• DBH: predicted diameter at breast height
• Volume: predicted stem volume
• Biomass: predicted above-ground biomass (dry mass)

Citation

Any scientific publication using the data should cite the following paper:

Hyyppä, M., Turppa, T., Hyyti, H., Yu, X., Handolin, H., Kukko, A., Hyyppä, J., & Virtanen, J. -P. (2024). Concepts Towards Nation-Wide Individual Tree Data and Virtual Forests. ISPRS International Journal of Geo-Information, 13(12), 424. https://doi.org/10.3390/ijgi13120424

This study provides urban forest tree inventories for the University Endowment Lands, Pacific Spirit Regional Park, Musqueam First Nation, and Southlands of British Columbia for 2015 and 2021. It describes urban forest structure and composition based on LiDAR-derived urban forest attributes including tree count, maximum height (zmax), crown area, and species type (coniferous/broadleaf). Hotspot analyses revealed spatially biased distributions of hot and cold spots, indicating heterogeneous urban forest structure and composition across the study area. Hot spots characterized by taller coniferous trees and higher mean crown areas were found across coastal, park & forest, or residential zones. Cold spots characterized by shorter trees and smaller mean crown areas were found in zones with more infrastructure. Overall, forest attributes differed between species type and inventory-year. Results suggest that anthropogenic drivers, such as urban development and land-use change, interact variably with tree attributes to influence the structure, composition, and distribution pattern of urban forests. The study highlights the applicability of using LiDAR-derived urban forest inventories to prioritize conservation and management strategies at the local level. By understanding the complex interactions between urban forest attributes and change drivers, communities can optimize green space utilization, sustainability, and equality within urban forest ecosystems.

Data from the paper:

Dalagnol, R. et al. Large-scale variations in the dynamics of Amazon forest canopy gaps from airborne lidar data and opportunities for tree mortality estimates. Sci Rep 11, 1388 (2021). https://doi.org/10.1038/s41598-020-80809-w

Link: https://www.nature.com/articles/s41598-020-80809-w

This repository contains:

1) Data frame with data from static and dynamic gaps used in Figure 2 (Dalagnol_2020_Data_Multitemporal_gaps.csv). Each row is the aggregated measurement at 5-km resolution. The site component referes to the five site studied with multitemporal data. Site order from 1 to 5 is DUC, TAP, FN1, BON and TAL.

2) Data frame with data from static gaps and environmental factors used in Table 1, Figure 3, 4, 5 (Dalagnol_2020_Data_Singledate_gaps_Modeling.csv). Each row is the aggregated measurement of one site observed by airborne lidar data.

3) Raster file at 5-km resolution with dynamic gap fraction estimates presented in Figure 5 (dynamic_gap_fraction_amazon.tif).

If you need anything else, please contact the corresponding author: Ricardo Dalagnol (ricds@hotmail.com).