Artificial intelligence (AI) and machine learning (ML) could play a valuable role in improving flood prediction in Scotland, but new research shows that human judgment should be at the heart of the line.
As climate change increases the frequency and severity of floods, the DelugeAi project investigated the possibility that AI and ML technologies could increase forecast accuracy, strengthen early warning systems, and better inform public responses to flood risk across Scotland.
Commissioned by the Environmental Protection Agency of Scotland (SEPA), funded by Waters' expertise centre (crew), and led by Strathslide University, the project explored how AI and ML support flood forecasting in Scotland.
SEPA worked closely with the researchers as a national authorities for flood prediction. Project researchers recommend a step-by-step approach to adopting AI tools. This starts with applications that offer fast and visible benefits such as early warning enhancements.
Informational decisions
“We are pleased to announce that Christopher White, PhD, from the Faculty of Civil Environmental Engineering at Strathclyde, said: “Flood forecasting is complex and requires quick and informed decisions. Our findings show that AI can improve the way flood warnings are provided, and how risks are communicated, especially in areas with limited data.
“Such systems should always be chosen based on the specific problem intended to solve, the type of data available, the resources needed, how easy it is to understand and trust.
AI is not a silver bullet, but it offers real possibilities to support Scotland's flood resilience. With careful implementation, these tools can enhance existing systems, reduce response times and improve public safety in the face of increased flood risk.
“We are pleased to announce that we are a part of the world,” said Michael Cranston, a key expert flood forecast and warning at SEPA. “At SEPA, addressing climate emergency means being open to innovation based on trust and transparency.
“We have made great strides in the way we collect and interpret data. We can listen to our phones, from paper maps, whiteboards and a series of mallscode tones, even radar, satellites, and real-time modeling.
“We will carefully consider the findings of our research as we continue to evolve our services. AI will not replace professional judgment, but will help support our shared goals of processing more data, improving accuracy and ultimately protecting people's assets and the environment.”
Provide value
The project brought together hydrology and AI experts who conducted a global literature review and held an international workshop to assess how new technologies can add value to SEPA's existing flood warning development framework.
This study identified seven key areas within the flood forecasting process where AI and ML can provide value, from real-time data monitoring and improving weather input to decision support and alerting. While most operational systems still rely on traditional models, we found that blending approaches using AI can provide practical and immediate benefits.
Experts consulted during the project agreed that human judgment should remain central to forecasting, and trust, transparency and data quality are important for the responsible use of AI.
The researchers suggested that SEPA could consider adopting simpler AI tools over the next 1-2 years and enhance early warning and decision support systems before implementing more sophisticated applications such as localized flood monitoring and model calibration over a longer period of time.
The project also highlights the importance of increasing the training and forecasting workforce for safe and effective use of AI tools, ensuring that human expertise continues to guide all decisions.
Delgeai's team included Christopher White, Douglas Bertram, Robert Atkinson, Muhammad Usman, Camilla Nierazinzka and Victoria Marty Barkley.
