Cloud and AI are ‘fundamentally changing’ ability to forecast weather, NOAA chief says

LAS VEGAS — National Oceanic and Atmospheric Administration head Rick Spinrad said this week that even as hurricanes, wildfires and other weather events worsen, the use of innovative tech like cloud computing and AI is improving the agency’s ability to predict and forecast them.

“This is fundamentally changing our ability to do probably the most important thing in our business, and that's predict, project and forecast,” said Spinrad, speaking at AWS’ re:Invent 2024 conference in Las Vegas.

Spinrad said NOAA, which houses the National Weather Service and owns or operates 18 weather satellites, archives about 230 terabytes of environmental data per month from various observational platforms, ranging from the sea, to ships and even space.

That data is then hosted by commercial cloud service providers and made available to the public and nascent climate industry through NOAA’s Open Data Dissemination program, or NODD. AWS, Spinrad said, now hosts 32 petabytes of environmental information, or about “half of [NOAA’s] holdings.”

“Our ability to collect these data, put them on NODD, store them in a system that allows AI applications, is really critical,” Spinrad said. “And it's not just for our use at NOAA. We are supporting a burgeoning climate industrial effort, a commercial weather enterprise. We are now collecting enough environmental information — enough environmental intelligence — to make a fundamental difference.”

According to Spinrad, the difference includes improved access to critical weather data for incident meteorologists in the field. Using another tool from Amazon, AppStream 2.0, incident meteorologists dispatched to the scenes of floods, droughts, storms and wildfires can “have the same kind of access to our advanced weather information and interactive processing system that they would back at the home office.”

Spinrad said forecasters have been able to make a “seven year advancement in forecast capability in a matter of months using AI, using the training data [NOAA] now has access to.” Such advancements would not be possible without the ability to assimilate new data types, he added, noting that autonomous systems, satellites and surface-based robots all provide data that can go into forecasts.

“It is the cloud, the AWS cloud, that’s both allowing us to store and distribute real-time high-resolution data,” Spinrad said. “The ability then to take those data, use them as training data to improve our forecast, is one of the strongest areas for AI applications, or what we call ‘AI for numerical weather prediction.’”

Forecast advancements are especially critical after another costly U.S. hurricane season. Hurricanes Helene and Milton could both exceed $50 billion disasters, as Hurricanes Katrina and Harvey did. Through December, the U.S. dealt with 24 weather and climate disasters in 2024 that cost more than $1 billion in damage, on par with 28 such disasters in 2023.

In 1980, there were only three such events and their total cost was less than either Hurricanes Helene or Milton.

“The environmental world as we know it — climate, weather, oceans, ice — is not your parents’ world. The hurricanes are different. The tornados, the drought, the flood, the coastal inundation, they are different,” Spinrad said. “We're facing increasingly powerful, fundamentally different kinds of weather and climate events.”

Spinrad said NOAA’s future plans include incorporating AI into more NOAA applications, as well as tailored forecasts for other industries, like agriculture and the energy sectors, and new commercial sectors, like construction, insurance and retail.

“At the end of the day, what we are going to do is improve lives, livelihoods and lifestyles,” he said.