As the earthquake-rattled East Coast prepares for another battering in the form of Hurricane Irena, researchers from the University of Texas at Austin and IBM say they've developed new flood prediction technology that can simulate rivers 100 times faster than real life.
The partnership brings together IBM's analytics expertise with UT's experience with river systems, weather and sensor data. The technology could help provide several days' warning before a flood occurs.
Conventional flood prediction techniques in the U.S. focus only on the main stems of the largest rivers. With limited resources, it makes sense to start there -- but water systems are called "systems" for a reason, and the tributary networks are the main stage for the actual flooding that causes property damage or even death.
So scale is a problem. But what if you could simulate all those tributaries at once using supercomputers? (Suddenly, the challenge doesn't seem so daunting.) With IBM's Deep Thunder weather simulation model on hand, the researchers are working to predict the entire 230-mile length of the Guadalupe River and 9,000 miles of tributaries in Texas.
It's fast, too: the system can generate up to 100 hours of river behavior in one actual hour of time.
"Effective flood preparedness can be looked at as a large scale computing problem, with a huge number of relevant data and independencies," IBM researcher Frank Liu said in a statement.
The value is in the advance notice for municipalities and disaster response teams, who can make and execute emergency plans in the time between prediction and the real deal.
It's even more useful for regional planners, who can potentially unearth inter-system dependencies previous research didn't expose, scaling the idea of a "systems" approach even further and across state lines. It also offers more granular data to manage irrigation efforts as they pertain to water conservation.
Here's a look in a video:
The University of Texas researchers say they plan to link the river model directly to the NEXRAD radar precipitation system to better predict flood risk on a creek-by-creek basis.