When Hurricane Matthew tore through Hilton Head, South Carolina, in 2016, it smashed Palmetto Bay Marina with such force that the water rose above the pilings. Winds were nearly 90 miles per hour. More than a foot of rain crashed down from the sky. The scene that a local Sea Tow owner described at the time seemed almost apocalyptic: “The entire marina lifted off of its anchor and then floated away.”

But just 5 miles away at Shelter Cove Marina, the 42-foot Grand Banks Shangri-La was fine. The owners received a text message from the dockmaster saying “all is good,” along with a video of their boat resting at the still-standing dock. “Feel so bad for folks in Palmetto Bay,” they wrote on a cruisers’ forum. “Total carnage.”

The experience was yet another example of a lesson that boaters continue to learn year after year: Just a few miles can make an enormous difference when it comes to hurricane strikes. It’s a lesson that experts with the National Oceanic and Atmospheric Administration also fully understand, and that they are preparing to address with a pair of new supercomputers that got their initial test run in June.

The supercomputers—nicknamed “Dogwood” and “Cactus,” and ranked as the 49th and 50th fastest machines of their kind in the world—will allow for a number of things, including a new hurricane forecast model for 2023. It’s called the Hurricane Analysis and Forecast System, and it’s expected to let forecasters predict storm details at a scale of 2 kilometers (about 1.2 miles) versus the 4.5-kilometer (nearly 3-mile) resolution they can achieve today.

“All of this is set up for people who are trying to have a good time, and making sure they don’t run into crazy weather,” says Avichal Mehra, National Weather Service chief of dynamics and coupled modeling group. “The mission is all about saving life and property.”

The supercomputers are three times faster than the previous system that NOAA used. For those who want to be in the mathematical know, they operate at a speed of 12.1 petaflops, with each petaflop being the equivalent of a quadrillion floating-point operations per second. For everyone who studied literature or psychology instead of higher-level mathematics in college, suffice it to say that all the extra computing power will essentially allow forecasters to run bigger, faster models that should improve predictions about what storms are going to do next, and where they’re going to do it.

All the computing power means forecasters will be able to increase the number of “grid points” that go into the mix of numbers the machines are crunching, alongside additional data that comes from ocean buoy sensors, “hurricane hunter” aircraft and more. The supercomputers also will let forecasters reduce the amount of spacing between the grid points, and run the computations more frequently, among other things. The result is expected to be more granular detail in forecasts, giving boaters even more precise information before and during major storms.

“Now, we can distinguish between towns that are just 2 kilometers apart,” Mehra says. “We can see the changes in the atmosphere or the ocean in those two locations.” In addition, he says, because the inputs going into the supercomputers are getting better, and because the models will be run more frequently, forecasters also should be able to give boaters a greater length of time in the forecast before a storm actually hits.

“Right now we put out five days of forecasts,” Mehra says. “We are hoping that with these new supercomputers, we can go to seven days.”

He says “hoping” because there’s still a lot of work to be done before the Hurricane Analysis and Forecast System becomes fully operational next year. Right now, teams of data scientists, physicists and others are doing what’s called “initialization” to assimilate existing data into the new system. It’s a staggering amount of data coming from all kinds of sources, including U.S. Department of Energy and U.S. Navy. The teams working to bring the new hurricane-forecast system online need to run tests to make sure the inputs are working correctly at the supercomputers’ scale and speed, and then need to move personnel around to ensure the most efficient results going forward.

“It is quite resource-intensive,” Mehra says, adding that he and his colleagues are thrilled to do the work because it will let them process information for the public that they wanted to provide in the past, but couldn’t because of limited computing power.

Going forward, he says, boaters should feel confident that even though the mission is challenging, its goals are absolutely clear.

“Our forecasters have a really tough job,” Mehra says. “I have analyzed more than 100 hurricanes, and each one of them is unique. They have their own life cycles. We have some idea, say, that things will intensify, but how much it will intensify, and when it will start intensifying, and when it will start weakening, these are very challenging questions. The job is to warn boaters and everybody else about what is happening.”
—Kim Kavin

This article was originally published in the October 2022 issue.