A new artificial intelligence model developed by an international research team can map forest canopy height with sub-meter precision using only standard RGB satellite imagery, potentially revolutionizing how forests are monitored for carbon sequestration and management. The model achieves near-lidar accuracy at a fraction of the cost, addressing a critical challenge in global carbon accounting and sustainable forestry.
The research, published in the Journal of Remote Sensing on October 20, 2025, introduces a framework that combines large vision foundation models with self-supervised learning to estimate tree heights from ordinary satellite images. According to the study available at https://spj.science.org/doi/10.34133/remotesensing.0880, this approach addresses the limitations of traditional monitoring methods that have hindered accurate, scalable forest assessment.
Forests and plantations play a vital role in carbon sequestration, yet accurately monitoring their growth has remained costly and labor-intensive. Traditional lidar systems provide accurate height data but are limited by high costs and technical complexity, while optical remote sensing often lacks the structural precision required for small-scale plantations. The new AI model bridges this gap by delivering detailed canopy height maps without expensive specialized equipment.
The research team from Beijing Forestry University, Manchester Metropolitan University, and Tsinghua University developed a canopy height estimation network composed of three modules: a feature extractor powered by the DINOv2 large vision foundation model, a self-supervised feature enhancement unit to retain fine spatial details, and a lightweight convolutional height estimator. When tested in Beijing's Fangshan District, the model achieved a mean absolute error of only 0.09 meters and an R² of 0.78 compared with airborne lidar measurements, outperforming traditional CNN and transformer-based methods.
"Our model demonstrates that large vision foundation models can fundamentally transform forestry monitoring," said Dr. Xin Zhang, corresponding author at Manchester Metropolitan University. "By combining global image pretraining with local self-supervised enhancement, we achieved lidar-level precision using ordinary RGB imagery. This approach drastically reduces costs and expands access to accurate forest data for carbon accounting and environmental management."
The model's significance extends beyond technical achievement to practical applications in climate change mitigation. It enabled over 90% accuracy in single-tree detection and strong correlations with measured above-ground biomass, making it suitable for carbon accounting under initiatives such as China's Certified Emission Reduction program. The ability to reconstruct annual growth trends from archived satellite imagery provides a scalable solution for long-term carbon sink monitoring and precision forestry management.
When applied to a geographically distinct forest in Saihanba, the network maintained robust accuracy, confirming its cross-regional adaptability. This generalization capability makes the model suitable for both regional and national-scale carbon accounting, offering a powerful tool for tracking forest growth, optimizing plantation management, and verifying carbon credits globally.
The AI-based mapping framework offers particular promise for global afforestation and reforestation monitoring programs as nations work toward net-zero goals. Future research will extend this method to natural and mixed forests, integrate automated species classification, and support real-time carbon monitoring platforms. As the world advances toward climate change mitigation targets, such intelligent, scalable mapping tools could play a central role in achieving sustainable forestry and accurate carbon accounting.
