England Environmental Improvement Plan and Green Space Targets
When Trees Became a National Priority
Imagine an ambitious goal: every resident of England should live no more than a 15-minute walk from a green space and be able to see at least three trees from their window. Sound like a pipe dream? It's actually a real government plan.
In 2023, the UK government launched the «Environmental Improvement Plan» – a massive program to restore the country's ecology. England's parks and green spaces contribute about £6.6 billion (approx. $8.4 billion) to the economy annually through their positive impact on public health and the climate. Yet, one in three residents still lacks access to quality natural areas, even though 80% of the population lives in cities.
To reach these targets, we first need to understand how many trees already exist and exactly where they are growing, and then track changes year over year. That's where things get complicated.
Challenges of Monitoring Tree Cover and LiDAR Costs
The Problem: Counting Trees Is Too Costly
Previously, Forest Research – the government agency responsible for monitoring the UK's forests and parks – used a combination of field surveys and LiDAR technology. Simply put, LiDAR is a system that scans terrain with lasers from aircraft or satellites to create detailed 3D maps.
The issue is that obtaining up-to-date LiDAR data is incredibly expensive. So much so, in fact, that regular updates are simply unfeasible within the budget. But the world doesn't stand still: trees grow, they are cut down, and new ones are planted.
It is particularly difficult to keep track of individual trees and small patches of woodland smaller than half a hectare. Meanwhile, about 30% of all tree cover in England falls into such areas. For a long time, a significant portion of green space remained without reliable oversight.
In April 2025, Forest Research published the new «Trees Outside Woodlands Map» – the first public tool showing vegetation outside of large forests. The map was developed as part of the «Natural Capital and Ecosystem Assessment» program, a government project aimed at creating a complete digital picture of the country's ecosystem by 2029.
To make this monitoring more accurate and affordable, the agency turned to a tool originally created for a completely different purpose: Meta's DINOv2 computer vision model.
Using DINOv2 Computer Vision for Satellite Tree Mapping
How AI Learned to See Trees from Space
DINOv2 is an open-source computer vision model developed by Meta for object recognition. In 2024, the company partnered with the World Resources Institute to train DINOv2 on 18 million satellite images to create a global map of tree canopy heights.
The map's resolution is one meter. This means the model can detect individual trees worldwide. Not just clusters, but every single tree. Best of all, the map is free and open to everyone.
The core of the approach is simple: the model analyzes satellite imagery and determines the canopy height at every point. This makes it possible to understand where trees are, how tall they are, and their density across a given area.
Following the map's release in April 2024, government agencies from various countries showed interest. The UK was among the first to decide to integrate DINOv2 into its national monitoring system.
Improving National Forest Monitoring with AI Integration
What This Changes for the Country
Freddie Hunter, Head of Remote Sensing at Forest Research, called the DINOv2-based solution «the most powerful open-source model released in recent years», noting that it is a total game-changer for tracking individual trees.
Currently, the Forest Research team is working on integrating height maps with national aerial photography. The goal is to obtain up-to-date data on canopy cover and height for individual trees and small woodlands nationwide. Additionally, AI will help more accurately estimate the volume of timber lost during logging.
If the technology meets expectations, Forest Research will be able to obtain data with precision rivaling the expensive LiDAR scans from the Environment Agency. When it comes to identifying standalone trees, the results may even be superior.
This will allow for information to be updated every three years, rather than waiting decades for new laser-scanning cycles. And, crucially, it can be done without massive spending on commercial satellite data.
There is another objective: Forest Research plans to use the structural information from the maps to more accurately identify tree species (types). This is especially important in complex urban environments, where vegetation is closely intertwined with buildings and road infrastructure.
Benefits of Open Source AI for Environmental Policy and Budgets
Why This Matters
Using Meta's model saves Forest Research from the need to develop and train their own neural networks from scratch. This reduces dependence on costly field research and external data providers.
Implementing AI to define canopy boundaries and planting density will lead to significant budget savings. Ultimately, this directly impacts how the government monitors environmental indicators, which in turn helps adjust state policy.
Meta has already introduced the next version – DINOv3 – which promises even more advanced visual recognition. The idea is for governments worldwide to use open algorithms to monitor investments in reforestation.
In the UK's case, it's not just about ecology; it's about the economy. Understanding where and how many trees are growing allows for urban planning where parks truly benefit public health and provide a return on investment.
Testing DINOv2 as a Global Standard for Ecosystem Tracking
What's Next
Forest Research is currently actively testing the model on national datasets. While it's too early to talk about a complete departure from traditional methods, the direction of travel is clear. If DINOv2 proves its effectiveness, it could become the gold standard for monitoring, not just in Britain, but worldwide.
The model's open nature plays a key role here: any organization can adapt it to its needs without licensing costs. This is a prime example of how artificial intelligence solves practical problems – not just by generating content, but by helping the state care for nature and save taxpayers' money.