Arboreus: A Realistic Boreal Forest Simulator with Topographic Terrain and Satellite-Informed Parametric Tree Distributions

F. Labra Caso, A. Saradagi, A. Koval, G. NikolakopoulosarXiv 2026 · arXiv

TL;DR

Arboreus is an open-source simulation framework that generates "data-informed plausible" Boreal forest environments by fusing satellite imagery, topographic maps, and species-specific parametric tree models. It enables researchers to statistically validate robot autonomy and perception algorithms in complex, large-scale synthetic worlds that maintain high fidelity to real-world Nordic forest sites.

What Problem Does This Paper Address?

Testing autonomous robots in under-canopy forest environments is a logistical nightmare. Field trials are expensive, time-consuming, and fraught with risks like adverse weather, safety concerns for pilots, and potential harm to the ecosystem. While simulation offers a safer alternative, existing tools often fall short: they either rely on simplistic "cylinder" trees, require labor-intensive manual design, or lack the statistical variability needed for rigorous validation.

There is a critical need for simulated environments that aren't just random, but are "data-informed"—meaning they reflect the actual tree densities, species distributions, and terrain profiles of specific real-world locations where a robot might eventually be deployed.

Methods

The authors propose a multi-stage procedural pipeline to bridge the gap between abstract simulation and real-world complexity:

  1. Species Characterization: The system uses parametric models for dominant Boreal species (Birch, Spruce, and Pine) based on actual Nordic growth studies. Parameters include trunk height, diameter at breast height (DBH), and canopy geometry.
  2. Satellite-Informed Distribution: Tree placement is guided by the Copernicus satellite dataset. By analyzing Tree Cover Density (TCD) and Dominant Leaf Type (DLT) at 10m resolution, Arboreus uses a Binomial distribution to determine how many and what kind of trees to place in a given patch.
  3. Topographic Terrain Generation: Instead of flat ground, the simulator pulls elevation data from the Swedish cadastral agency (Lantmäteriet). It uses Poisson Surface Reconstruction to create smooth, realistic terrains from sparse elevation samples.
  4. Tree Reconstruction: Point clouds are generated for each tree, which are then converted into skeletons and meshes using acyclic graphs and allometric rules for canopy interpolation.
Overview of the Arboreus forest generation pipeline, showing the transition from satellite data to 3D meshes.

The framework is designed to be engine-agnostic, with demonstrated compatibility for GazeboSim, Unreal Engine (via AirSim), and NVIDIA Isaac Sim.

Datasets

The authors introduced a new open-source dataset to accompany the tool:

  • 25 Boreal Forest Worlds: Each world covers $1000m^2$ and corresponds to one of three real forest sites in northern Sweden: Mjölkuddsberget (hilly, mixed forest), Jämtön (flat, pine-dominant), and Storforsen (riverside, birch/spruce mix).
  • Input Data: The framework leverages the Copernicus High Resolution Layer for vegetation and Lantmäteriet for topography.

Key Insights and Ideas

  • "Plausible" vs. "Exact": The authors emphasize creating "plausible" worlds rather than exact digital twins. The goal is to match the statistics of a site (density, species mix) while allowing for stochastic variation in the fine details, which is better for testing the robustness of autonomy algorithms.
  • Controllable Realism: Users can inject field observations to tune the "age" of the forest (scaling tree dimensions) or the Level of Detail (LOD) of the branching structures. This allows a trade-off between visual/geometric fidelity and computational performance.
  • Multi-Modal Synthetic Data: The paper demonstrates that Arboreus can generate synthetic data for both aerial (ALS) and terrestrial (TLS) perspectives. Fusing these allows for more accurate forest inventory tasks, like estimating "stem sweep" (the curvature of a trunk), which is often hidden from aerial-only scans.

Key Results

  • Density Accuracy: Validated against manual ground truth annotations, the satellite-informed distribution achieved a density error of only 0.1345 (roughly equivalent to a 2-tree difference per $100m^2$).
  • Terrain Fidelity: The interpolated terrain achieved an average Root Mean Square Error (RMSE) of 0.1685 meters compared to real-world elevation API values.
  • Cross-Platform Utility: The authors successfully demonstrated a variety of robotic tasks:
  • SLAM: Lidar-based localization and mapping in GazeboSim.
  • Reinforcement Learning: Validating a control policy for the Boston Dynamics Spot quadruped in Isaac Sim.
  • Forestry Mapping: Using synthetic data for tree instance segmentation and volume estimation.

Limitations and Future Work

While Arboreus provides a robust geometric and statistical framework, it does not currently model complex ecological processes like tree growth, decay, or seasonal foliage changes (though these can be delegated to the physics engine's rendering system). The authors also note that very high-quality tree reconstructions can lead to large file sizes that may challenge some simulation hardware.

Future work aims to quantify the "sim-to-real" gap more precisely by comparing robot performance in paired simulated and real-world environments.

Summary updated

Abstract

This work presents Arboreus, a realistic Boreal forest simulator, developed to support statistical validation of robotic autonomy algorithms in Boreal forest environments. Arboreus presents an automated methodology for synthesizing large-scale, data-informed, plausible forest worlds from parametric trees, height-variant terrain, and satellite-informed tree density distributions. The parametric generation of dominant Boreal tree species such as Birch, Spruce, and Pine is guided by trunk, canopy, and foliage parameters, extracted from growth studies in the Nordic areas. Tree distribution in an area of interest is guided by the Copernicus satellite dataset, which provides tree cover density and dominant species type at up to 10m raster resolution. The forest floor generation is guided by topographic data from the Swedish cadastral agency to produce realistic data-informed terrains. Given coordinates and extents of forested areas of interest, Arboreus synthesizes data-informed plausible forests, while offering flexibility to inject user observations of real forests (detail, age, dimensionality and tree count), which is critical for realistic synthetic data generation and for the testing of feature-sensitive autonomy algorithms. The applicability of Arboreus for benchmarking and validation of robot autonomy for different robot mobilities is demonstrated in Gazebosim, Unreal Engine, and Isaac Sim physics engines. The applicability of Arboreus as a source of synthetic forest data and its potential to support downstream forest resource mapping research (volume estimation, tree instance segmentation and stem attribute estimation) is also demonstrated. Arboreus has been open-sourced along with a dataset comprising 25 Boreal forest worlds of size 1000m2 corresponding to three real forest sites.
BibTeX
@misc{caso2026arboreus,
  author = {F. Labra Caso and A. Saradagi and A. Koval and G. Nikolakopoulos},
  title = {{Arboreus: A Realistic Boreal Forest Simulator with Topographic Terrain and Satellite-Informed Parametric Tree Distributions}},
  year = {2026},
  eprint = {2611.22908},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2611.22908},
}
Figure