Learn More

In 2023, the UK government launched the Environmental Improvement Plan, with goals including: every English resident should have access to green space within a 15-minute walk and be able to see at least three trees from their home; and increase England's tree canopy cover by 2% by 2050. Achieving these targets requires frequent and accurate tree canopy monitoring.

Forest Research, the research agency of the UK Forestry Commission, is responsible for monitoring trees, woodlands, and forests. In April 2025, the agency released a new 'Trees Outside Woodland' map, making data publicly available for free for the first time. The map was funded by the UK government's Natural Capital and Ecosystem Assessment program, which aims to complete a baseline assessment of England's natural assets by 2029.

Previously, Forest Research combined surveys and LiDAR data to assess tree canopy cover, but LiDAR data acquisition was costly and unsustainable. Monitoring isolated trees, tree groups, and woodlands smaller than 0.5 hectares was particularly difficult, yet such woodlands account for about 30% of England's tree canopy.

Meta, in collaboration with the World Resources Institute, used the DINOv2 model trained on 18 million satellite images to generate an open-source global tree canopy height map with 1-meter resolution, capable of detecting individual trees worldwide. After the map's release in April 2024, several governments expressed interest in using the model to improve afforestation efforts.

Freddie Hunter, Remote Sensing Lead at Forest Research, stated that the high-resolution canopy height model based on DINOv2 is 'the most powerful open-source AI model released in recent years' and is transformative for national-level single-tree detection and monitoring. The team is now applying the canopy height map to national aerial imagery, aiming to generate more timely estimates of tree canopy cover and height, including isolated trees, tree groups, and small woodlands, as well as estimates of timber volume loss due to felling.

Applying the model to aerial imagery could produce synthetic canopy height models whose quality (at least in single-tree identification) may consistently surpass the Environment Agency's National LiDAR Survey. This would enable Forest Research to estimate tree density and canopy area within a three-year rolling timeframe.

Using Meta and WRI's open-source model may allow Forest Research to avoid building and training its own canopy height prediction model, thereby reducing reliance on LiDAR and survey data. This could significantly save data procurement costs, improve update frequency and data quality, and ultimately enhance monitoring of government tree canopy targets, directly influencing environmental policy.

Forest Research also plans to use the structural information from the canopy height map to enhance national tree species distribution mapping, particularly for canopy estimation in complex environments such as towns and cities.

Meta recently released DINOv3, aiming to further advance visual intelligence. Meta hopes that by improving the accuracy of canopy height maps, global governments can choose to use open-source models (including DINOv3) to monitor afforestation investments.