A multi-platform LiDAR dataset for standardized forest inventory measurement at long term ecological monitoring sites

A research consortium has released a comprehensive, multi-platform LiDAR dataset explicitly designed to serve as a benchmark for AI and robotics applications in forest ecology. The dataset, curated from a long-term ICOS monitoring site, integrates three complementary scanning technologies: UAV-borne laser scanning (ULS) for canopy coverage, terrestrial laser scanning (TLS) for detailed stem mapping, and backpack mobile laser scanning (MLS) using real-time SLAM for efficient sub-canopy data acquisition. The core control plot features a highly consistent registration with TLS point clouds containing approximately 333 million points, all complemented by the other modalities. The team employed marker-free, SLAM-aware protocols to streamline field work, and the final products are available in standard LAZ and E57 formats with UTM coordinates for full reproducibility.

This dataset is a significant resource for the AI and environmental science communities, providing a standardized testbed for critical tasks. Researchers can use it to benchmark and develop algorithms for 3D point cloud registration, evaluate the efficiency of different scanning platforms, and train models for automated tree segmentation and quantitative structure modeling. Crucially, by situating the data at an established ICOS site, it creates a direct link between high-resolution 3D structure and decades of concurrent ecological and carbon flux measurements. This enables more accurate allometric biomass estimation and paves the way for creating 'digital twins' of forest ecosystems, allowing for repeated, AI-powered inventories that track ecological change over time.