Introduction#
Meta has unveiled Canopy Height Maps v2, a new global forest mapping system developed in collaboration with the World Resources Institute. This open-source model utilizes advanced DINOv3 technology to provide precise measurements of tree canopy height, which is essential for monitoring forest health and tracking restoration efforts.
Improved Accuracy#
The updated model shows significant improvements in accuracy compared to its predecessor, Canopy Height Maps v1, released in 2024. The accuracy measurement, known as R², has risen from 0.53 to 0.86, indicating a stronger correlation between the model's predictions and actual measurements. DINOv3 replaces the earlier DINOv2 backbone and has been trained on a vast dataset of satellite images, enhancing the clarity of canopy maps and minimizing bias in measuring tall trees.
Enhanced Training Dataset#
To improve the model further, the training dataset for Canopy Height Maps v2 has been expanded to include a wider variety of lidar examples from different geographical areas. Lidar, which stands for Light Detection and Ranging, is a technology that uses laser light to measure distances and create detailed 3D maps. Meta has also developed automated tools to align satellite images with lidar data, ensuring more accurate canopy height estimates.
Applications and Future Use#
The original model, Canopy Height Maps v1, has been utilized by organizations such as Forest Research in the UK for national forest inventory and climate tracking. The European Commission’s Joint Research Centre has applied it in their Global Forest Cover map for 2020 and plans to use the new version for future mapping and the 3 Billion Tree Initiative, which aims to plant 3 billion trees in the EU by 2030. In the U.S., the maps are being integrated into city planning tools for the Cities for Smart Surfaces initiative, which includes participation from mayors of ten cities, such as Atlanta and Boston.