Snowflake and Ordnance Survey Use AI to Map England’s Undefended Flood-Risk Buildings

Snowflake and Ordnance Survey have collaborated on an AI-powered flood readiness model that identifies around 1.2 million buildings in England that may be exposed to flood risk without being covered by existing flood protection measures. The Intelligent Flood Readiness Model combines geospatial data, government flood risk information and machine learning analysis of planning documents to help policymakers understand where physical flood exposure overlaps with social vulnerability.

The project comes at a time when flood resilience is becoming a more urgent climate adaptation priority for governments, infrastructure owners, insurers and local authorities. The Environment Agency’s latest national flood risk assessment estimates that around 6.3 million homes and businesses in England are in areas at risk of flooding from rivers, the sea or surface water.

How the Model Works

The model brings together six critical data streams into what Snowflake and Ordnance Survey describe as a shared “structural intelligence” layer. It uses Ordnance Survey building datasets, the Indices of Deprivation in England, Environment Agency flood data, defended and undefended flood risk areas, and AI-driven analysis of more than 3,000 pages of statutory Flood Risk Management Plan documents.

This approach allows the system to compare building-level characteristics, such as height, footprint and construction type, with broader indicators of social vulnerability. It can then identify places where flood risk may be particularly difficult to manage, either because buildings are physically exposed, communities have fewer recovery resources, or existing flood plans do not fully cover the area.

For local authorities, the practical value lies in moving beyond broad flood-zone mapping. Instead of treating large areas as having the same risk profile, the model can highlight clusters of vulnerable properties, including those that may cross administrative boundaries or fall between flood management planning zones. This could help councils, water companies and emergency planners prioritise investment in drainage, property-level resilience, warning systems and community preparedness.

Older Buildings and Deprived Communities Face Higher Exposure

According to the model’s findings, up to 68% of the identified undefended buildings are in deprived areas, where households and communities may have fewer financial and social resources to recover quickly after flooding. The analysis also suggests that 84% of these buildings were constructed before 2001, before flood risk was more systematically incorporated into planning permission requirements.

The age profile of the affected building stock is significant. The model found that 15% of at-risk premises date from before 1919, while 23% were built between 1919 and 1959. These buildings may have been constructed before current flood maps, planning rules, or climate risk assumptions existed. They may also require different adaptation measures from newer developments, especially where basements, older drainage systems, dense urban layouts or heritage constraints limit conventional flood defence options.

The findings also point to the importance of social equity in climate adaptation. Flood impacts are not only determined by water depth or proximity to rivers and coastlines. Recovery depends on insurance access, savings, housing condition, mobility, local infrastructure and the ability of residents to relocate temporarily or repair damage. This makes flood risk both an environmental and socioeconomic challenge.

Surface Water Flooding is a Major Concern

One of the most important findings is that 85% of vulnerable and undefended buildings identified by the model are at risk from surface water flooding, rather than river or coastal flooding. Surface water flooding typically occurs when intense rainfall overwhelms drainage systems, especially in paved urban areas where water cannot soak into the ground.

This has major implications for net-zero and climate resilience strategies. While decarbonisation remains central to limiting long-term climate impacts, adaptation measures are increasingly needed to manage risks that are already materialising. For cities and developers, this means investing in sustainable drainage systems, permeable surfaces, green roofs, rain gardens, urban wetlands and better maintenance of drainage infrastructure.

For asset owners and financial institutions, the results underline the growing importance of climate risk data in property valuation, lending, insurance and infrastructure planning. Buildings that appear safe under traditional river and coastal flood maps may still face material exposure from intense rainfall and overwhelmed drainage systems.

Implications for Policy, Insurance and Investment

The model’s use case extends beyond emergency response. It could inform capital allocation for flood defences, local adaptation plans, housing policy and insurance risk assessment. By identifying where buildings are both physically exposed and socially vulnerable, public bodies can target limited resources more effectively.

Insurance is another area where granular risk data is becoming increasingly important. Flood Re, the UK’s joint government and insurance industry scheme, was created to help make flood cover more affordable for households in high-risk areas. However, as climate risks rise, insurers, lenders and policymakers are under pressure to understand whether current approaches remain adequate for future conditions.

The UK government has said it is investing £10.5 billion in flood defences and related infrastructure. The Environment Agency has also reported improvements in flood protection for tens of thousands of homes over recent years. However, the Snowflake and Ordnance Survey analysis suggests that better data integration may be needed to ensure that adaptation funding reaches the places where risk, vulnerability and lack of protection overlap most sharply.

A Wider Role for AI in Climate Adaptation

The Intelligent Flood Readiness Model illustrates a broader trend in climate technology: using AI not only to measure emissions, but also to manage physical climate risks. For businesses, public authorities and infrastructure operators, the ability to combine asset data, hazard maps, planning documents and socioeconomic indicators could support more resilient decision-making.

However, the model should not be seen as a replacement for local expertise, engineering assessment or community engagement. AI can help identify patterns and gaps that would be difficult to detect manually, but flood resilience still depends on implementation. That includes planning reform, infrastructure upgrades, property-level adaptation, nature-based solutions and public communication.

For the net-zero transition, the message is clear. Climate strategy can no longer focus only on emissions reduction. Organisations also need to understand how changing weather patterns affect buildings, supply chains, public services and vulnerable communities. Data-driven tools such as Snowflake and Ordnance Survey’s model can help make those risks more visible, but the next step is turning insight into targeted investment and measurable resilience.

Source: sustainabilitymag.com