A new AI-powered model claims up to 1.2m buildings in England are at risk of flooding but outside existing protection measures.
The model, said to be the first of its kind, has been launched by Ordnance Survey and AI data cloud company Snowflake. The two worked together to create the model, which used the former’s data to carry out the analysis.
In addition to identifying properties are risk of flooding, it also suggests that many of them are located in some of the country’s most deprived areas, raising concerns about their ability to recover from flood events.
The Intelligent Flood Readiness Model combines detailed mapping data with government datasets and current Flood Risk Management Plans (FRMPs) to assess both physical flood exposure and wider social vulnerability.
Its findings indicate that as many as 68% of the 1.2m buildings identified could be highly vulnerable to the after-effects of flooding, due to a combination of elevated risk and socioeconomic disadvantage.
Tim Chilton, a managing geospatial consultant at Ordnance Survey, said the model could help local authorities better prioritise flood defences and resource allocation. He added that it offers “data-driven, actionable insights – without the burden of analysing endless spreadsheets”.
Researchers found that the majority of at-risk buildings pre-date modern planning rules. Around 84% were constructed before 2001, when flood risk considerations became more firmly embedded in planning policy.
Older housing stock features prominently, with 15% of properties built before 1919 and a further 23% dating from between 1919 and 1959.
The model also highlights the complexity of England’s built environment, where older homes often sit alongside newer developments. This mix, researchers say, complicates flood planning and underlines the need for more detailed and up-to-date data.
By integrating six datasets – including building characteristics, deprivation indices and flood risk information from the Environment Agency – the model produces what developers describe as a “structural intelligence” layer. This allows policymakers to assess not only where flooding may occur, but how severely it could affect different communities.
The findings are said to challenge some assumptions about flood risk. The analysis suggests that 85% of vulnerable, undefended buildings are more likely to be affected by surface water flooding than by rivers or the sea. As a result, high-density urban housing may face greater overall risk than properties in traditionally recognised flood zones such as coastal or riverside areas.
Regionally, areas along the east coast of Yorkshire and the Humber were identified as having the highest concentration of vulnerable, undefended properties. Across England, nearly two-fifths of neighbourhoods were found to contain at least one such building, indicating that risk is widely dispersed rather than confined to specific locations.
Residential properties account for the majority of those affected, making up around 64% of the total. Commercial buildings represent approximately 15%, industrial sites 10%, and infrastructure about 5% – assets that can be critical to recovery efforts but may not be fully covered by existing plans.
Fawad Qureshi, global field chief technology officer at Snowflake, said: “Data is at the heart of making informed decisions. As this project shows, it’s rare that one body holds all the relevant data or that this data is in the same format.”
Flood Risk Management Plans are typically produced every six years and often rely on broad geographic assessments. Researchers say this can overlook significant variations within smaller areas, where flood impacts may differ street by street.
The model’s developers argue that AI-driven tools could support a shift from static planning towards more dynamic approaches, allowing authorities to test scenarios and identify emerging risks as data evolves. They also recommend greater focus on surface water infrastructure, building-specific vulnerabilities and the role of social deprivation in recovery outcomes.
While the model does not provide definitive predictions, its creators say it is intended to guide policymakers towards more targeted and evidence-based decision-making in the face of increasing flood risk.
Image credit: AC Rider/Shutterstock
Read next: AtkinsRéalis and Nvidia team up for nuclear-powered AI
Are you a building professional? Sign up for a FREE MEMBERSHIP to upload news stories, post job vacancies, and connect with colleagues on our secure social feed.
