The UK’s land – urban and rural environments – supports the basic requirements for millions of lives; from the food we eat and the ecosystems we rely on for clean air and fresh water, through to the spaces we use to live, work and play.
But we also know that land is a finite resource, with increasing demands for new housing and infrastructure alongside the need to consider sustainability and ensure our communities can adapt to the changing climate.
Decisions made at national and local level about land use can impact on economic growth, health and the environment but these decisions are rarely straightforward, as debates around everything from the location of solar farms to roads and rail lines and housing developments show.
So how do we make effective and evidence-based decisions which fairly distribute opportunities for prosperity and meet the demands of a growing population, while protecting the environment?
Applying the latest data and technology to land use planning
The Geospatial Commission, part of the UK government’s Department for Science, Innovation and Technology, is responsible for setting the UK’s geospatial strategy and coordinating public sector geospatial activity. As part of their Land Use Programme, they have partnered with the Turing to further enhance the UK’s spatial data capabilities and land use decision-making through artificial intelligence.
If you are a decision maker trying to ensure that housing is close to public transport and employment, that green energy infrastructure is placed in optimal locations, that rivers are kept clean and that nature is protected, data and technology are playing an increasing role.
Recent advances in spatial data science and technologies like machine learning and A I have improved our ability to process and analyse data quickly and efficiently.
These powerful technologies can also help us see into the future, bringing together large amounts of information from different sources (collected at ground level and through satellite or aerial imagery) to model scenarios depending on the choices we make, and providing visualisations that give more information on proposed changes to local stakeholders and communities.
AI and machine learning can also help us to deal with complexity, reflecting the many interactions between decisions on land use such as uses of land that don’t sit well alongside each other (like industrial activity and residential properties), versus fruitful opportunities to use the same land for more than one purpose (such as rooftop solar energy).
First steps
The first phase of the Turing-Geospatial Commission partnership, completed in June 2023, involved collaboration with Newcastle City Council to develop a prototype scenario modelling tool called DemoLand.
The Council’s planners were able to feed local policy input, targets and priorities into the development of the tool, which then used machine learning and AI to explore a range of interventions that could achieve the required objectives.
A visualisation tool was built to help users to explore the trade-offs between scenarios, evaluating the impact of decisions against policy priority indicators like air pollution, access to jobs and green space and house prices.
Extending the project into geospatial AI
Now in its second phase, the partnership is developing the tool to expand and enhance the role that geospatial AI plays in its functionality.
Firstly, through applying satellite data. Images from satellites are revolutionising the collection of spatial data, allowing the analysis of vast areas of land in detail, monitoring land changes over time to track the impact of policies. This phase is exploring the potential to integrate satellite sources through vision foundation models to improve the underlying models that power the tool.
Secondly, technology based on large language models will be incorporated into the tool to make interacting and exploring its outputs more understandable. This will be crucial in adding value to non-technical users such as those without a background in data science or analytics.
The teams working on the project are also committed to learning broader lessons and developing new insights on the application of geospatial AI to land use, drawing on experiences from the DemoLand project and also engaging with the wider Turing network of partners interested in disciplines such as urban analytics.
The project will also make suggestions on how geospatial AI could be applied to land use decisions more broadly. These recommendations will be incorporated into the Geospatial Commission’s Land Use Programme.
Next steps for geospatial AI
In the long term, we would like to see the tool and these approaches rolled out more widely across the country, and to do this we plan to seek further local authorities who would like to collaborate as beta testers.
In time we also plan to expand the areas of impact the tool can model, such as further sustainability and climate change scenarios as well as further integrating AI capabilities, with an ultimate aim of seeing data and AI becoming an integral part of decisions about the way we use land across the nation.
Top image: Zhou
