A huge performance upgrade to the supercomputers that predict American weather could enable companies to keep delivering their products even in the face of severe storms.
Prior to a storm, stores like Home Depot or Walmart want to get supplies of products people will want to the places that are going to be affected, and pause shipments when the weather hits.
Get what matters in tech, in your inbox every morning. Sign up for Source Code.
"Knowing where that would hit, they would be able to better quantify what to ship, where and when," Chris Caplice, executive director at the MIT Center for Transportation & Logistics, told Protocol.
But weather is difficult to predict: The sheer number of variables to control for in modeling the Earth makes it a daunting task. And the weather events that have the most profound effect on U.S. industry, like hurricanes, tend to have a large amount of uncertainty about them before they make landfall.
Now, the U.S. wants to take a real stab at finally — hopefully — giving us more-accurate weather reports with powerful new supercomputers.
The U.S. National Oceanic and Atmospheric Administration announced Feb. 20 that it would be upgrading the supercomputers it uses to model the Earth's weather patterns over the next two years. The new computers will initially be tasked with improving NOAA's medium-term weather forecasting abilities — for events from one day to 16 days out — as well as better modeling for severe regional weather, like hurricanes, Brian Gross, director of the Environmental Modeling Center at the National Weather Service, told Protocol.
For companies looking at the potential path of a hurricane and seeing it covers the entire state of Florida, that's not helpful, Caplice said. But if that path can be significantly narrowed with NOAA's new supercomputers, "retailers would love it," he said. That could have "tremendous value" for some companies, especially those that don't have a lot of slack in when they can deliver products, Caplice added.
NOAA's new machines from Cray will be in Virginia and Arizona, and each will have a capacity of 12 petaflops. For reference, the current fastest computer in the world clocks in at 148.6 petaflops, although the tenth-fastest machine can achieve roughly 18 petaflops — so NOAA's computers aren't that far off the pace.
The Cray supercomputers will triple the computing power at NOAA's disposal, the administration said. And they will be used to process data collected by the NWS from satellites, weather balloons, radar, buoys, and other sources above and on the Earth's surface.
"This is a pretty big upgrade for us," Gross said.
In the long term, the computers will help NOAA improve its forecasting models in three specific ways: increasing the model resolution, which means they can focus on smaller events like flash floods or thunderstorms; how complex the model is, and the different physical factors they can include in a single model; and the number of forecasts they can run at once to determine how confident NOAA can be about an upcoming weather incident.
NOAA's computational upgrades are actually to bring the U.S. "on par" with many of its peer agencies across the world, said David Michaud, NWS' director of the Office of Central Processing. He said that agencies like the U.K.'s Met Office and Europe's ECMWF have similar capabilities to what the U.S. will have, and that they all can work together. "Across the modeling community, there's a real collaborative feel, in terms of interacting and sharing best practices with science," Michaud said. "We're staying on par in terms of our contribution with the science in collaboration with the community."
Implementing massive supercomputers like these is not like setting up a new desktop computer. "Generally the transition between one contract to another takes about a two-year timeframe," Michaud said. But in about a year's time, if everything goes according to plan, NOAA developers will start transitioning code over to the new system, checking it's all configured correctly. And if that goes off without a hitch, Michaud said, the computers should be fully operational by February 2022.
Get in touch with us: Share information securely with Protocol via encrypted Signal or WhatsApp message, at 415-214-4715 or through our anonymous SecureDrop.
Given that weather information is so critical for businesses that rely on rapid supply chains, some companies will be eagerly awaiting the results.
"If you know, for example, the upper plains of the Mississippi is going to flood, then companies can take advantage of that, and maybe advanced-ship or reroute," Caplice said. "I can see in weather disasters it could be very helpful."