Predicting the unpredictable: How AI is revolutionizing flood management amid climate change
Two experts from SAS explain how AI is transforming flood management.
Published on December 16, 2024
Everything new is wildly interesting! That's the motto of our DATA+ expert, Elcke Vels. She writes stories about AI and how it affects our society, has a series on cyber security, and interviews Dutch innovation maestros. In her “What if...” column, she also explores intriguing scenarios that deviate from the status quo.
Climate change is causing more frequent and severe floods worldwide, with devastating impacts in countries like Spain, Switzerland, and the United States. Accurate flood prediction and timely warnings can save lives, and AI is becoming a key tool in this fight. Experts Katy Salamati and Remi Verdiesen from SAS, a leader in AI and advanced analytics, explain how AI is transforming flood management and helping communities prepare for future disasters.
In October, streets in Spain were overwhelmed by floodwaters, with cars floating away and rescue workers using ropes to save residents from a muddy current. The flooding in the Valencia region claimed more than 200 lives. This tragedy is part of a wider pattern of extreme weather events, with the effects of climate change becoming more apparent. Similar events in Europe, such as in Germany and Belgium in 2021, underscore the growing threat these increasingly frequent natural disasters pose.
The critical role of AI in modern flood management
In the past, predicting where and when floods would occur and understanding their potential impact was nearly impossible. We were largely dependent on reports from residents and limited data. However, we are gaining control today thanks to advanced AI systems, simulations, and IoT solutions. These technologies allow for more accurate predictions, real-time monitoring, and improved responses, enhancing our ability to manage and mitigate the effects of floods, explains Katy Salamati. She holds a PhD in civil engineering and is an adjunct assistant professor at North Carolina State University. Salamati also leads the development of IoT solutions for government and infrastructure at SAS, including the SAS Flood solution. The company develops AI systems and IoT applications to tackle environmental and operational challenges.
“There are three main phases in which AI can come into play,” she explains. “Before natural disasters happen, AI helps us better predict and prepare for them; during the events, it enables more effective responses through improved coordination, resource optimization, and faster reactions; and after the disasters, AI allows us to plan more effectively, learn from past incidents, and prepare for future ones, balancing both short-term responses and long-term strategies."
AI-based simulations, in particular, are highly valuable. Verdiesen, a research intern at SAS, specializing in extreme rainfall patterns and their implications: “Extreme weather events are rare, making them difficult to study through traditional observation. There is simply not a lot of historical data available.” However, he explains, AI and machine learning can fill this gap by simulating these rare scenarios, allowing us to explore and predict outcomes that might not have been observed in the past. “Using advanced modeling techniques, AI can generate and simulate these scarce moments, providing valuable insights into how such events might unfold under different conditions. This capability is crucial for improving preparedness and response strategies.”
Recent advancements in AI for flood prediction
However, although more is needed, more and more data is available. Thanks to that AI is advancing rapidly. Salamati: “Access to both real-time sensor data and historical information is crucial for training AI models and enhancing their accuracy. As more data is generated and stored—especially through IoT sensors—AI and machine learning algorithms become more adept at predicting future events.” However,” she adds as a critical note, “historical data is invaluable, predicting the future is also shaped by unforeseen external factors.”
North Carolina’s success story: AI in action
Across the globe, it is becoming increasingly clear how AI can transform flood prediction and management systems. A standout example is the Town of Cary, North Carolina, which has transformed its approach to flood mitigation. North Carolina experiences frequent flooding, primarily due to hurricanes and tropical storms. Cary now utilizes SAS’s advanced analytics and Microsoft’s IoT capabilities to consolidate data from diverse sources, enabling proactive and predictive responses to flooding events.
Traditionally reactive, Cary’s flood management relied on manual processes and citizen reports. “Now, integrated sensors and cloud-based analytics provide real-time insights into river levels, rainfall, and stormwater dynamics. This unified system eliminates data silos, ensuring timely decision-making and swift action to safeguard citizens and property”, Salamati Says.
How the Netherlands leverages AI for flood prevention
In the Netherlands, Rijkswaterstaat, the governmental body responsible for managing the country’s water systems and infrastructure, partners with SAS to ensure the safety and functionality of critical infrastructure. Using SAS Viya, Rijkswaterstaat leverages data from thousands of sensors deployed nationwide. These sensors monitor water levels, structural integrity, and other key indicators, enabling advanced anomaly detection.
Although the platform's focus is not exclusively on flood prediction, it plays a vital role in safeguarding the Netherlands' extensive network of dikes, bridges, and waterways. By detecting unusual patterns or deviations in sensor data, Rijkswaterstaat can act swiftly to address potential risks before they escalate into emergencies.
Overcoming challenges in AI-Driven flood management
In many ways, AI comes to the rescue in flood management. However, key challenges remain. Verdiesen: “Current research highlights the importance of incorporating fundamental physical laws into predictive models rather than relying solely on historical data. This is an approach we should prioritize more. Traditional modeling often overlooks the evolving dynamics of climate change, such as the increased capacity of warmer air to hold moisture in convective clouds, leading to more intense rainfall. Integrating these physical principles into models allows for more robust "how bad can it get" scenarios, enabling the development of climate stress-testing frameworks.”
Finally, while AI and machine learning can truly enhance predictions, a real challenge still lies in coordinating efforts and ensuring efficient communication. Recent events in Spain illustrate this: although the storm was predicted, failures in sharing information with local governments led to inadequate preparation.
"Addressing these systemic issues is essential for saving lives and minimizing the impact of disasters," Salamati concludes. "By improving our approaches now, we can better prepare for future floods, potentially saving lives and preventing significant damage."
Floods in Europe: innovations against climate change
“The magnitude of these events is unprecedented,” says Niko Wanders, lecturer in hydrology at Utrecht University.