To better anticipate the next Sandy-size hurricane, the U.S. National Oceanic and Atmospheric Administration (NOAA) is upgrading the supercomputers it uses for predicting the weather.
By October, the agency will have 10 times as much computing power to devote to predicting the weather as it does today, thanks to a $44.5 million upgrade of two supercomputer systems now being carried out by IBM and subcontractor Cray.
Part of the cost of the upgrade has been funded by a $25 million allocation from the Disaster Relief Appropriations Act of 2013, passed by Congress after Sandy hit the East Coast in October 2012. The severity of the hurricane surprised many meteorologists, and NOAA’s National Weather Service was criticized for not predicting the storm with as much accuracy as did a European weather agency.
When the upgrade is finished, the two NOAA supercomputers will each be able to execute more than 5 petaflops (quadrillion floating point operations per second). The new gear will allow NOAA to update its weather forecasting software to achieve greater accuracy, the agency said.
Today, the agency can predict weather patterns up to 10 days in advance within an area, or what it calls a resolution, of 27 kilometers (16.77 miles), meaning each prediction covers a 27 square kilometer area. With the new equipment in place, it can sharpen the resolution to 13 kilometers (a little over 8.07 miles).
In general, the smaller the area of resolution, the more accurate the forecast will be, a NOAA spokesman said.
Weather prediction is a computationally intensive task, one that involves the analysis of how many separate variables could interact.
“The number of computations go up very, very quickly as you increase the resolution,” said Per Nyberg, senior director of worldwide business development at Cray. Simply doubling the resolution will require eight times the computation. “You get into exponential increases.
“Any weather service could use any computational power you give it,” Nyberg said. Cray, which specializes in supercomputers that can be easily scaled out, counts weather services among its most ardent customers.
On the software side, NOAA’s National Weather Service has been refining its forecasting models as well. The agency’s Hurricane Weather Research and Forecasting model accurately charted the approach of Hurricane Arthur, which menaced North Carolina last year. It has also pressed into service the High-Resolution Rapid Refresh model, which now delivers 15-hour forecasts each hour.
NOAA, which is part of the United States Department of Commerce, is also in the process of updating its satellite system for watching weather patterns from space. The new satellites are due to be operational in 2017.
Although the National Weather Service offers a range of weather forecasts and alerts directly to the public free of charge, its material is most widely distributed through reuse by commercial services, ranging from television stations to online weather sites.