A tornado passes across south Oklahoma City, on May 20, 2013.
The atmosphere never gets a moment’s privacy. It can barely stir enough to move a leaf without some piece of high-tech equipment—often many, many pieces—knowing about it. The U.S. alone has up to 30 satellites at any one moment that devote at least part of their time to monitoring global and national weather patterns; 122 Doppler radar systems scattered across the country looking up from the ground; and a web of computers that just got a massive upgrade—increasing their data-crunching capacity 30-fold—to process the information that all that other hardware gathers.
It’s no surprise then that the National Weather Service was able to see a broad pattern of tornadoes coming days before they actually struck with such tragic consequences in Moore, Oklahoma on Monday. It’s no surprise either that the 16 min. warning between final alert and tornado touchdown, though brief-seeming, was twice what was possible in the past. The unpleasant surprise is that this entire weather-forecasting infrastructure is much wobblier than it seems, and without a lot of care—and a fair bit more money—the whole thing could start to come undone.
(MORE: Prelude to Disaster: Inside the Oklahoma Weather Center)
Running a 21st century meteorological system on the cheap is barely possible, but by federal budget standards, predicting the weather is a real bargain. Last year’s total outlay for the National Oceanic and Atmospheric Administration, which does much of the heavy lifting when it comes to forecasting, was just shy of $5 billion—or less than 1% of annual Defense Department expenditures and barely one-third of what even chronically budget-crunched NASA gets. The recent computer upgrade, which affected two main processing centers in Reston, Va., and Orlando, Fla., cost just $25 million, little more than milk money in a nearly $4 trillion federal budget. But while those modest expenditures may be enough, they leave very little room for error.
Last October, in the run-up to Hurricane Sandy, a critical forecasting satellite known as GOES-East went briefly offline, effectively blinding NOAA’s surveillance of exactly the portion of the coastal U.S. that the storm was approaching. A back-up satellite parked in orbit nearby was moved into position and filled in the gap until repairs could be made, but had that gone down too, the surveillance system would have been crippled. As it was, American forecasters did not anticipate the sharp left turn Sandy made as it approached the New York and New Jersey regions—the very turn that led to 200 deaths and $50 billion in damage—and had to rely on European satellites and computers to make that call for them. That scared a lot of people.
“Gaps are opening in both our operational and research satellites,” says J. Marshall Shepherd, director of the atmospheric sciences program at the University of Georgia. “If one of them goes out, it’s not like simply replacing a burned-out light bulb. These things take years to build and launch.”
(WATCH: Bryan Walsh: Why Tornadoes are So Hard to Predict)
GOES-East, as its geosynchronous label suggests, orbits at a high enough altitude that the speed of its orbit matches the rotation of the Earth, meaning that it can simply hover in the sky above the eastern half of the country. A counterpart, GOES-West, monitors the other half. Up to five other U.S. satellites move in polar orbits, circling the Earth longitudinally instead of latitudinally, allowing them to collect global weather data—the large-scale systems that are the engine of our continental system, which in turn drives the in-your-state, on-your-block systems.
It’s the satellite fleet that’s in the most immediate need of capital investment—far more so than the computer and radar systems are. NOAA sought—and got—a $117 million increase for satellites in the upcoming federal budget, though the exact dollar amount could change as details are hammered out in Congress. That money will have to be made to go a long way. It’s not just GOES-East that’s having problems. GOES-West and the back-up satellite are getting old too, and all could begin failing irreparably by 2015. The polar satellites are not expected to last past 2016. Replacement satellites for all of these are not projected to go up until the very years the existing ones begin flickering, which means that any slippage in the design, construction and launch schedules could mean a wide gap in coverage.
“All of these satellites are in the twilight of their existence, especially the polar-orbiting ones,” says Shepherd. “The GOES are all living on borrowed time.”
(PHOTOS: Tornado Flattens Oklahoma City Suburb, Kills Dozens)
The $25 million that so effectively boosted NOAA’s computing power came from a supplemental appropriations bill that was voted on and passed in the wake of Hurricane Sandy, and it was a welcome infusion of much-needed cash. But it was a one-time infusion, and even a system that’s 30 times more powerful than what it was is still woefully short of where it should be. “What we really need is to be 100 times better than what we were,” says atmospheric sciences professor Cliff Mass of the University of Washington. “This gets us roughly to parity with the power of the European system, but we have to do much better.”
There’s more than just bragging rights at work there. The European computers conduct only global modeling and leave the smaller-scale stuff—the equivalent of state and local forecasting—to the individual countries. The U.S. system has to do it all—covering our entire 3.7 million sq. mi. land mass (6 million sq. km), while the Europeans get to parcel out their 3.9 million sq. mi. (6.3 million sq. km). “So until we began upgrading,” says Mass, “Europe had ten times the computing power we did and was being called on to do a lot less with it.”
More computer muscle will make it possible for the U.S. to conduct what are known as ensemble forecasts. Instead of running a weather model once or perhaps twice or three times in a row to check it for accuracy, you conduct it 50 times, simultaneously, on multiple systems, producing a much higher resolution forecast. “This gives you a resolution down to about 2 km (1.25 mi.), which is what you need,” says Mass. “The weather service simply has not had that kind of capability.”
The bright spot in all of this—the one area that wins only raves—is the Doppler radar web. It was radar that may have made the greatest difference in Oklahoma—adding critical minutes to the warning time—and radar that will continue to look out for people on the ground as the tornado season plays out and the hurricane season begins. A lot of the system’s accuracy is due to the relatively recent implementation of what’s known as dual-polarization radar—pulses that are sent out both horizontally and vertically (compared to just horizontally in the past) and can better distinguish, say, rain from snow from hail, or from dust or even insect swarms for that matter, a not-insignificant ability as a 17-year locust infestation begins. “We have a magnificent radar system,” Mass says simply. “We have capabilities we just didn’t have at all 30 years ago.”
That is a very good thing, especially in the wake of the very, very bad things that unfolded in Oklahoma—but our ability has to get better still. Spending tens of millions of dollars now to avoid tens of billions of dollars in damages later is a wise investment by any measure. Avoiding the loss of homes, livelihoods and lives is an investment dividend that can not even be calculated.