From individual trees to virtual forests: a comprehensive 3D structural dataset of Central European tree species.

This dataset provides a comprehensive, multi-scale collection of forest structural data for Central European tree species. The data were mainly derived from extensive terrestrial laser scanning campaigns conducted between 2009 and 2025 across 47 forest sites in the Czech Republic, capturing 12 tree species in both leaf-on and leaf-off conditions. The repository is comprised of three components. First, it includes a library of 3618 individual segmented terrestrial laser scanning tree point clouds, accompanied by detailed metadata coupling field-measured traits with structural parameters derived from a quantitative structure model (QSM). Second, the dataset contains 15,718 virtual airborne laser scanning scenes (50 × 50 m plots) simulated using the HELIOS++ radiative transfer model. These synthetic plots were procedurally assembled using the Forest Factory growth model and provide multi-resolution point clouds (8 points/m² and 16 points/m² densities) alongside precise spatial and structural metadata for every tree. Third, the repository features 273 fully reconstructed three-dimensional (3D) tree representations, combining a QSM for woody architecture with biologically realistic foliage distribution based on empirical traits. This dataset serves as a robust ground-truth resource for training, testing, and validating machine learning algorithms aimed at individual tree segmentation, species classification, and structural parameter extraction from airborne laser scanning data. Furthermore, the ready-to-use 3D representations and virtual scenes facilitate advanced 3D radiative transfer modelling and the development of a virtual forest.