AI increases accuracy of National Water Model flood predictions
AI designed to weed out errors bolsters forecasts and could reduce damage costs.
Experts
Vinh Ngoc Tran
Assistant Research Scientist at the University of Michigan
Valeriy Ivanov
The accuracy of flood predictions from the National Water Model improves by four to six times when combined with AI modeling, according to new research led by University of Michigan researchers.
A study published in AGU Advances shows how the combination of the two approaches creates better results. The National Oceanic and Atmospheric Administration’s National Water Model is a hydrologic modeling framework that simulates “observed and forecast stream flow over the entire continental U.S.”
Researchers paired the water model with their deep learning AI model, called Errorcastnet, designed to identify errors. It was trained for the task by examining old floods paired with their National Water Model forecasts.
Resilient Infrastructure
Michigan Engineering is on the vanguard of the next infrastructure revolution
“So especially for floods, the performance of the pure AI model is quite poor,” Vinh Ngoc Tran, an assistant research scientist at U-M and co-author of the research paper, told AGU recently. “The advantage of the AI model is that they are very simple. You only need to use the data to train the model and provide the forecast, but the most important thing we need to be concerned about is ensuring prediction accuracy for flood events that can cause significant damage.”
AI forecasting programs, like Google’s, don’t take into account several physical factors, such as vegetation and elevation. That can lead to underpredicting flooding in some scenarios.
“You can’t throw away physics—by definition, you can’t,” Valeriy Ivanov, a U-M professor of civil and environmental engineering, told AGU. “You have to understand that systems are different. The landscapes are different. You have to account for dominant physical processes in your predictive model.”
More accurate predictions, researchers believe, could lead to better preparation and reduced damage in the aftermath of flooding events.