Predicting Floods with Computer Simulations
Originally Posted on: June 8th, 2017
In September of 2016, Cedar Rapids was expecting another massive flood.
The National Weather Service was predicting a 23-foot crest on Tuesday, Sept. 27 (later revised to 22 feet). The city shifted into overdrive. Residents and officials deployed more than 10 miles of sand and earthen barriers to successfully protect the city, including a quarter of a million sandbags.
Meanwhile, at the Iowa Flood Center (IFC), Ricardo Mantilla was running the IFC flood prediction model. Mantilla, who is a research engineer at IIHR and an assistant professor of civil and environmental engineering, says that four days in advance, the IFC model was predicting a crest of 22 feet in Cedar Rapids. If two more inches of rain fell over the weekend, the model predicted a crest of 26 feet.
“Our prediction, made on Friday, was right on the spot,” Mantilla says. He admits that the IFC model predicted the arrival of the crest in Cedar Rapids a day early, which tells him that the model needs a bit more refinement.
Still, Mantilla is encouraged. “We have more and more evidence that we are on a good path,” he says.
Welcome to Iowa
Mantilla, a native of Colombia, arrived at the University of Iowa as a postdoc just in time for the flood of 2008. In 2009, IFC Director Witold Krajewski was formulating a vision for the Iowa Flood Center to study flooding and create new tools to predict and communicate flood risk, including the Iowa Flood Information System (IFIS). Mantilla had focused on surface hydrology and water transport geographical information systems. He says it was a natural transition to direct his efforts toward developing a flood prediction model.
Expanding the model to include the entire state of Iowa was a major challenge, but the team was able to develop tools to see flows on the statewide scale. They call their model HLM-Async; HLM stands for Hillslope Link Model, and Async is the numerical equation solver that runs on high-performance computers.
“The scientific challenges were difficult, and they still are today,” Mantilla says. Nevertheless, he calls the flood of 2016 a small success story, because the systems worked, producing reasonably accurate predictions as the flood was happening.
A Forecast for Cedar Bluff
HLM-Async is a distributed model, designed to make predictions accurately and consistently, all the time, throughout the state. “That’s not an easy task,” Mantilla says. “I think the Iowa Flood Center is definitely at the forefront of that in the world.”
Mantilla says he continues to be fascinated by water and river networks. The math and physics he uses to understand water movement are the same math and physics needed to understand complex systems such as the economy or society in general, he says. “Everything that matters—love, food, water—I can always connect what I know to these aspects of life.”