Intel Supercomputer Tracks Oil Spill
**Guest post by Intel's Phillip Davis
As the catastrophic April 20 oil spill in the Gulf of Mexico grows worse by the day, there's one question everyone in the region is asking: Where is all that oil going to go?
Now, thanks to a complex ocean current model simulated on a supercomputer housed in Intel's New Mexico facility, residents of the Gulf Coast have a better idea--though the new picture is not a pretty one.
The new model suggests that the oil closest to the surface will be carried by the gulf's powerful Loop Current, which will carry it hundreds of miles, eventually hitting Florida. There it will rapidly curl around the bottom half of the state and dramatically pick up speed when it hits the Atlantic Ocean's Gulf Stream current, extending the spill's impact thousands of miles up the east coast of the United States.
"I've had a lot of people ask me, "Will the oil spill reach Florida?'" said Synte Peacock, a scientist at the National Center for Atmospheric Research (NCAR) in Boulder, Colorado, which created the model, known as the Parallel Ocean Program.
"Actually," she said, "our best knowledge says the scope of this environmental disaster will likely reach far beyond Florida, with impacts that have yet to be understood."
The idea to model the oil's flow came in the first days after the spill, she said. "A number of us were discussing why there were no longer-term scenarios about the impact of the spill," she recalled. "Then we realized we had a perfect model to do just that--the Parallel Ocean Program. It was a completely theoretical experiment we had done, sort of to see if we could measure the 'weather' in the ocean--its various flows and currents."
To model the oil spill, researchers reconfigured the model to have a definite starting point--the latitude and longitude of the "Deepwater Horizon" spill in the Gulf--and then let the model run to see what patterns emerged.
"We basically dropped a 'virtual dye' in the water, and then watched to see where it would go," she explained.
'Encanto' - one of the world's fastest supercomputers
But that simple description belies the tremendous amount of computer power needed to simulate the oil spill's problematic movements.
Intel's Xeon-powered Encanto supercomputer is housed in Intel's Rio Rancho campus. And that's where Intel and the state of New Mexico entered the equation.
Back in 2007, the state worked with Intel to create what was then the third-fastest supercomputer in the world (it's now ranked No. 32). Known as "Encanto," it's a water-cooled behemoth filling 28 tall cabinets and housing 3,500 quad-core Intel® Xeon® processors.
Simulations took thousands of computer hours
The first half-dozen simulations alone took Encanto more than 250,000 hours of computer time, using 1,000 cores in a massively parallel process. In other words, if all 1,000 processor cores had been running throughout all six simulations, the process would have taken about 10 days to complete.
"Each simulation started with a different 'eddy field,' or underlying current, and then we let it go forward into the future," Peacock said.
The simulation showed a big mass of oil deep underwater that was moving very slowly. But up near the surface - from about 65 feet in depth and higher - ocean flows were swifter, showing the oil spiraling up near the Louisiana and Florida coastlines.
But once the oil curls around Florida and hits the Gulf Stream, things really speed up--the current could possibly carry the oil up to 100 miles a day, or 3,000 miles in a month, according to the simulation.
Peacock stressed, however, that the simulations are not a forecast, but rather reflect a range of possible trajectories. A true forecast would have to take into account the oil's density and buoyancy, among other things. Scientists are working on that now.
As the work continues at NCAR and elsewhere to predict the path and impact of what's being called the nation's worst environmental disaster, Intel processors will continue to help experts deal with some of society's biggest challenges.
@bryanrhoads for Phillip Davis
Comments