A large dataset of labelled single tree point clouds, QSMs and tree graphs

Area of Science:

• Forestry
• Ecology
• Remote Sensing

Background:

• High-resolution individual tree data is crucial for forest monitoring and ecological studies.
• Existing datasets often lack detailed 3D structural information.
• Advancements in laser scanning technologies enable new data acquisition methods.

Purpose of the Study:

• To introduce the BioDiv-3DTrees dataset, a comprehensive collection of 3D tree data.
• To provide a valuable resource for forest inventory, biomass estimation, and ecological research.
• To facilitate algorithm development and data fusion in forest remote sensing.

Main Methods:

• Collected 4,952 individual tree point clouds using Terrestrial Laser Scanning (TLS) and Unmanned Aerial Vehicle Laser Scanning (ULS).
• Generated 3,386 Quantitative Structure Models (QSMs) and graph representations for 14 broadleafed species.
• Integrated laser scanning data with existing open-access forest inventory data (species, DBH, height).

Main Results:

• The BioDiv-3DTrees dataset contains detailed 3D point clouds and QSMs for 19 tree species.
• QSMs were validated against point clouds for accuracy in tree height, DBH, and crown projection area.
• The dataset links 3D structural data with traditional forest inventory attributes.

Conclusions:

• BioDiv-3DTrees offers a reliable and scalable resource for the forest science and remote sensing communities.
• The dataset supports diverse applications, including biomass estimation and tree structure analysis.
• This data facilitates the integration of detailed 3D information with conventional forest inventory methods.