Supercomputer Makes Predicting Floods a Whole Lot Easier

NOAA has rolled out a computer system that can make hourly forecasts of flood risks for 2.7 million locations.

A man navigates a boat of rescued goats past a partially submerged car after flooding on Aug. 16, 2016, in Gonzales, Louisiana. (Photo: Brendan Smialowski/AFP/Getty Images)

Aug 18, 2016· 2 MIN READ
John R. Platt covers the environment, wildlife, and technology and for TakePart, Scientific American, Audubon, and other publications.

In the wake of this week’s devastating floods in Baton Rouge, Louisiana, the ability to predict torrents and droughts in the United States just got a major upgrade.

The National Oceanic and Atmospheric Administration this week rolled out the supercomputer-powered National Water Model, a system that can generate hourly forecasts of flood and drought risk for 2.7 million locations in the contiguous United States. NOAA’s previous system could only calculate the risk for 4,000 locations every few hours.

“We’re going to have river and stream forecast information in locations we never had it before,” said Thomas Graziano, director of NOAA’s new Office of Water Prediction at the National Weather Service. “We want to show where the water’s going to be, how deep it’s going to be, and when it’s going to be there, so that the emergency management community can pre-position people and resources to most effectively mitigate the impact of the floods.”

Both the new and old systems rely on data from more than 8,000 U.S. Geological Survey–managed gauges that measure water flow in rivers around the country. The new system adds the processing power of a Cray XC40 supercomputer along with 20 years of analysis by the USGS and the Environmental Protection Agency to understand how water flows through the country’s rivers and streams. “We call this the hydro-fabric for the entire country,” Graziano said.

RELATED: Texas’ Record Floods Are the New Normal

The new system will do more than predict floods. “It’s a water resource model,” Graziano said, that will inform on everything from floods to droughts, as well as agricultural resources, ecosystems management, reservoir supplies, hydropower generation, recreation, and river transportation.

In addition to hourly forecasts, the National Water Model will provide forecasts 10 and 30 days ahead. It could even go much further into the future. “Over time, we would look to expand the range of our forecast to better support the water supply folks who need to make decisions about what their reservoirs are going to look like on a six-month, nine-month, or yearlong time basis,” Graziano said.

Lehigh University professor David Casagrande, who has studied the impact of floods in the Midwest and other locations, called the enactment of NOAA’s new system “huge.”

“It’s nice that what was previously available in a part of the country is now available to everybody else, because it’s proven to make a huge difference,” Casagrande said.

He pointed to farmers on the floodplains of the Mississippi, who for years have received important advance warnings about the potential for floods. “They know as much as a week in advance when a major flood event could happen,” he said.

Would the upgraded National Water Model have made a difference in this week’s flooding in Baton Rouge if it had been activated a few weeks earlier? Graziano said the team looked into that and found the new system’s models would have done what he called “a reasonably good job” predicting the crests and timing of the flood two to three days ahead. “It certainly would have been guidance that would have augmented what our staff did,” he said. “It will be useful moving forward.”

The National Water Model won’t yet be able to anticipate coastal flooding from hurricanes or climate change–inspired sea-level rise, but Graziano said NOAA will begin work to combine freshwater and coastal estuary models in the next fiscal year. “That will be a big step forward,” he said.

Casagrande believes the system is important because major floods are clearly increasing.

“We’re at a really critical juncture,” Casagrande said, pointing out that the National Flood Insurance Program, which he calls “the only policy tool for managing flood risk in the country,” is more than $23 billion in the red after expenses from Hurricanes Katrina and Sandy. “The climate change bill is here, and we better start thinking about doing something.”

Other partners in the development of the model were the National Center for Atmospheric Research, the National Science Foundation, and the Consortium of Universities for the Advancement of Hydrological Sciences, which brought in experts from more than 120 academic institutions.