As Hurricane Earl brewed in the Atlantic, RENCI scientists kept watch. High-powered computer models allowed UNC researchers to construct forecasts and simulations (pictured) of its likely trajectory and strength.
Tracking hurricanes takes supercomputer brainpower and human expertise. Etherton (L), atmospheric research scientist, and Casey Averill, systems specialist, talk next to Blue Ridge, the Dell Nehalem supercomputer cluster at RENCI.
Networking cables connect RENCI to other external high-performance research networks.
Brian Etherton, atmospheric research scientist, with Hurricane Earl displayed on the 12-foot x 8-foot rear-projection display wall at the UNC Renaissance Computing Institute (RENCI) in Chapel Hill.
UNC’s hurricane trackers aim to save lives, prevent damage
Igor is an unpredictable, potentially destructive force of nature. It’s Brian Etherton’s job to try to figure out where Igor’s going and what he might do.
That’s because Igor is – literally – a hurricane, one of about a dozen such storms so far this 2010 hurricane season. Etherton, a senior atmospheric scientist at UNC’s Renaissance Computing Institute (RENCI) in Chapel Hill, works with high-powered computer models that try to predict the storms’ likely trajectory and intensity.
Predicting the threat of a storm is no easy task. But the UNC team seems to be staying several steps ahead of Mother Nature. When they compare their forecasts to other models and to actual results, their main program, known as HURNC, consistently produces relatively accurate predictions, especially for storms bearing down on the south eastern seaboard.
For instance, RENCI’s September 16 forecast shown below accurately predicts Igor’s path over Bermuda.
“We’ve designed the system and optimized it to predict storms that are going to hit North Carolina in four days,” Etherton says. “We’ve really been focused on that problem … that’s one of the things we want to do the most.”
The recent case of Hurricane Earl proved a good example of the RENCI model’s accuracy. Four days before the storm brushed the N.C. coast, the program had Earl heading toward the Outer Banks when other models had it somewhere between Bermuda and the coast. In the end, the eye of the storm did not make landfall, but Etherton says the model’s early predictions helped focus the attention of local National Weather Service forecasters and emergency management officials.
From the time a storm first forms out in the Atlantic, RENCI’s Dell/Intel supercomputer Blue Ridge, capable of 8 trillion calculations per second, gets to work. Twice a day, it churns out forecasts, plotting the storm’s likely course and intensity on a grid, with each point in the grid representing a 3-km-by-3-km box. That high resolution – much finer than the 9-km and 27-km resolution level of two other competing models – allows the program to better predict how multiple variables will play out.
For Etherton, the goal is clear: making good forecasts as early as possible can help save lives and prevent damage.
“It takes time to evacuate people,” he says. “On the Outer Banks, those islands with the roads and the causeways, you have to have your plan ready to go a couple of days ahead of time. So that’s really our focus, to make sure we’ve got information at a time that lets the emergency management community have a heads-up.”
The HURNC system was initially developed by Gary Lackmann, professor of marine, earth and atmospheric sciences at N.C. State University. Etherton and the UNC team continue to collaborate with Lackmann and graduate student Briana Gordon to further improve the program.