The Trump Administration is cutting personnel and funding for federal agencies such as the National Weather Service and the Federal Emergency Management Agency as hurricane season 2025 begins.
But wait! Artificial intelligence is rolling out lightning-fast data-crunching powers that can fill in for some of the missing meteorologists and emergency managers.
But wait! Those AI platforms rely on precise data to perform their miracles. And those data largely come from scientists at the National Weather Service and its parent, NOAA, the National Oceanic and Atmospheric Administration, whose data collection capabilities are being curtailed.
So where does that leave the American people, particularly Floridians, when it comes to forecasting and recovering from severe weather?
That is decidedly unclear.
Republican Virginia Gov. Glenn Youngkin captured the angst Tuesday, May, 20, at the first meeting of President Donald Trump’s panel to remake the agency that Homeland Security Secretary Kristi Noem says will soon be known as something other than FEMA. As in North Carolina, emergency managers in Virginia were caught off guard by the far-reaching inland impacts of Hurricane Helene last September, despite good forecasting, and now it’s hurricane season again.
“We’re going to be changing the tires on this car while this car is barreling 100 miles an hour … down the freeway,” Youngkin told his fellow panelists. “We’re going to be massively transforming a response system — while that response system HAS to be effectively responding. And we will pray every day that this is going to be a zero hurricane season, and that we will not have to deal with any massive natural disasters.”
Zero hurricanes?
That is not in NOAA’s forecast.
Going dark
As hurricane season starts — it runs June 1 through Nov. 30 — several NOAA-administered databases were retired unexpectedly. One, while not a forecasting tool, tracks “Billion Dollar Weather and Climate Disasters,” showing how climate changes have fueled an upward trajectory of physical damage from hurricanes, tornadoes, wildfires, winter storms, and other severe weather with losses exceeding $1 billion.
The data cover 1980 through 2024, detailing 403 high-dollar events that collectively inflicted nearly $3 trillion in losses, most of it the past five years. The data-curious will not find a 2025 update. Instead, the site explains NOAA’s new instructions:
“In alignment with evolving priorities, statutory mandates, and staffing changes, NOAA’s National Centers for Environmental Information (NCEI) will no longer be updating the Billion Dollar Weather and Climate Disasters product. Additional details and the opportunity to submit comments are available at the NESDIS Notice of Changes website. All past reports, spanning 1980-2024, and their underlying data remain authoritative, archived, and available via the Billion-Dollar Disasters dataset landing page.”
Another source of data that has dwindled this year partly due to helium shortages followed by waves of layoffs and early retirements at NOAA, is the National Weather Service’s fleet of humble weather balloons. Carrying sensors called radiosondes, the balloons traditionally were released by staff into the atmosphere (click for a demonstration) twice daily from 92 stations across the country.
They measured humidity, pressure, temperature, wind profile and other data that feed evening and morning weather forecasts in cities across the country. Balloon launches at a dozen stations were halted or reduced in April, as explained in this Weather Service advisory:
“Until further notice, the National Weather Service (NWS) may temporarily reduce or suspend scheduled radiosonde launches at selected NWS upper air sites due to staffing limitations or operational priorities. Office(s) will continue to conduct special observations as resources allow and in response to emerging weather events,” says an April 17 advisory, issued just days before a line of tornadoes and hail storms pelted Oklahoma, Texas, Nebraska, Colorado and Wisconsin.
Twenty-six databases and data sources going dark for various reasons are listed at Notice of Changes | NESDIS | National Environmental Satellite, Data, and Information Service.
‘You’ve got to prioritize’
Still, universities and the private sector carry on developing data-fueled AI tools they hope will speed up analysis and application of big data, wherever they can get it. Beta applications range from analyzing decades of hurricane data to learn when forecasts and outcomes did or did not match, to artificially voicing streams of written storm advisories for swift broadcast in multiple languages.
W. Craig Fugate, FEMA director under President Obama and former director of emergency management in Florida under Gov. Jeb Bush, sounds confident that AI tools can help FEMA and NOAA weather the cuts, as long as data collections by reliable sources such as NEXRAD (radar), dropsondes (sensors dropped from planes), and measurements gathered by Hurricane Hunter aircraft are preserved.
Flying into a storm
- Click to see a Hurricane Hunter flying through Hurricane Beryl last year.
- Click to see a night flight through Hurricane Lee amid lightning in 2023.
“Budget cuts means you’ve got to prioritize. They’re obviously going to prioritize the severe weather. They’re obviously going to prioritize hurricanes,” Fugate said.
A NOAA Corps pilot flies a WP-3D Orion N43RF Hurricane Hunter to measure atmospheric conditions during Hurricane Ida in late August 2021. NOAA’s aircraft fleet is stationed in Lakeland, Fla. Credit: Lt. Cmdr. Kevin Doremus, NOAA Corps He is less certain that the nation’s premier source of high-quality weather data — which feeds private sector platforms including The Weather Channel, local weather broadcasts, and commercial weather apps — will continue to deliver as before.
“Don’t know. I’ve seen the proposed budgets, but I don’t know,” Fugate said, referring to federal budget proposals being debated in Congress. “The private sector is heavily leveraged to use the government’s data. The question will be, are there impacts to the ability to run the models and the data collection?
“I get where people are saying, ‘We’re seeing all these cuts. We have all these retirements. It’s going to have to have an impact.’ And I’m like, probably, but is it going to affect the life safety products, or is it going to affect more routine operations? I don’t know,” he said. “I’ve been in government long enough, that we’ve cut big chunks of government and they still were able to do the functions.”
Promising technology
Perhaps to the rescue artificial intelligence, from generative AI to machine learning, is driving development of products such as Google DeepMind’s GenCast, which “learns” by reanalyzing historical data, comparing how a given storm was forecast and how it actually played out, then correcting for errors.
Ilan Price, lead author of the official report on GenCast, said the tool performed beautifully in predicting the track of Hurricane Milton last year from 8.5 days before landfall, though it is not yet able to forecast intensity. The tool is still being taught to forecast elements such as intensity, wind speed and cloud cover. GenCast is an open source product, available for non-commercial uses. It relies on weather data from NOAA and from the European Centre for Medium-Range Weather Forecasts.
Simpler examples of AI in development are found in the realms of forecast messaging and aggregating information via the so-called Internet of Things – a network of physical objects including private satellites, security cameras, and smartphones that share information online. (Here are samples from Hurricane Ian making landfall at Fort Myers Beach in 2022.)
Troves of such information can be collected and analyzed by AI, according to a 2023 analysis by Deloitte Insights. AI can inform emergency managers in real time what a storm is doing to a given location before, during and after – inundating roads, tearing off roofs, leaving deep floodwaters, or maybe none of those.
Masses of video and drone footage from multiple public and private sources can be collected and analyzed to provide rapid damage assessments. (Similarly, Deloitte says, multi-source data analyzed by AI can help track the spread of diseases, and AI tools helped detect and fill fast-breaking shortages in PPE protective gear during the COVID-19 pandemic).
Before a storm, images and messages posted to the internet can be gathered by AI-powered tools to help emergency managers assess in real time how many evacuees are on highways and where they are headed, as well as the number and location of shelters and hotel rooms still available.
Fugate predicts AI can eventually run traffic control and shepherd motorists in real time to the safest routes and shelters based on their actual positions in relation to the forecast track of an approaching storm.
At the University of Florida, a new tool called BEACON employs AI to prioritize, localize and translate written emergency bulletins from forecasters and emergency managers and transmit them continuously in artificially generated, human-sounding voices and in multiple languages and dialects.
Randy Wright, executive director of UF’s media and college operations, which helped develop the tool, describes BEACON as an “always on” alerting and advisory channel for radio — the most resilient of platforms for disseminating emergency information — and for mobile apps, which work well until power and internet connectivity fail.
BEACON broadcasts via app and on the Florida Public Radio Emergency Network (FPREN), a branch of NPR, with funding from the Florida Division of Emergency Management, and uses technology from Futuri Media, a company that specializes in AI technology for media and public safety.
(Disclosure: WGCU Public Media, Southwest Florida’s source for PBS and NPR and a member-supported service of Florida Gulf Coast University, utilizes information provided by FPREN including forecasts and reports during severe weather.)
“Recently, an area of strong thunderstorms moved through the Gainesville region and BEACON was automatically taking a variety of official National Weather Service special weather statements … and airing those updates [in human-sounding voices] as the situation evolved. It was an extraordinary early test case scenario where the system performed wonderfully,” Wright said by email in answer to questions from the Florida Trident.
Where future danger will strike
Another test was during Hurricane Milton last October. A sample of how BEACON performed can be heard here.
Wright said BEACON relies entirely on official “trusted and reliable” government sources.
Flood Mapp, a tool being developed in Australia, is using AI to predict how water will behave in inland flood events such as torrential rainfall in urban areas. Fugate, who is an adviser in developing both BEACON and Flood Mapp, said the latter’s value is not only in issuing alerts but in analyzing masses of available data to learn where future danger will strike, under what circumstances, and how severe the damage will be. It can help at-risk communities build smarter and become more resilient.
“It’s a real-time tool, but also can I use the tool to see how vulnerable my community is, and what can I do differently?” Fugate said. He feels strongly that emergency managers must communicate better to the public about the danger of storm surge and inland flooding, which he says are often more dangerous in a hurricane than wind. Likewise, AI tools are being developed to detect wildfires and predict their behavior, forecast rip currents, and forecast where climate change will bring future flooding to areas not accustomed to it.
At its best, Fugate predicts, AI-driven tools may soon help emergency managers analyze all available data, including data on human behavior, to make the toughest of calls, when the best forecasting in the world can’t prescribe exactly when, where and how to evacuate their citizenry.
“Where AI is really going to be a powerful tool is helping in decision support,” he said. “It’s easy when [a hurricane] is coming right at you for three days. It’s harder when it’s one of those storms where there’s less certainty or just a small deviation can mean a big difference in impact.”
‘A serious worry’
For the last word, computer scientist Amy McGovern warns that all tools powered by artificial intelligence or human intelligence are only as good as the data they use. She is director of the National Science Foundation AI Institute for Research on Trustworthy AI in Weather, Climate, and Coastal Oceanography, based at the University of Oklahoma.
McGovern told The National Weather Desk “Off The Radar” podcast in February that AI will be a boon to weather forecasters, climate scientists and emergency managers — but only if the integrity of weather data is ensured.
“It’s fun stuff. I’m very excited about the future of AI for weather,” said McGovern, an adviser to forecasting AI startup BrightBand, founded by a former Google X executive. And yet, she added, “I have a concern that there might be people who are putting out models that, they’re selling them, overselling them, before they’re ready — because of the AI hype.” She wants the most vital weather data to come from and be verified by non-commercial sources.
“It’s a concern for me that somebody’s going to ‘predict’ the weather and people are going to die because it wasn’t verified we. It’s a serious worry for me.”Amy McGovern, computer scientist and director of the National Science Foundation AI Institute for Research on Trustworthy AI in Weather, Climate, and Coastal Oceanography
“NOAA doesn’t have the competing interests that private industry does. Academia doesn’t have the competing interests that private industry does. …We’re not selling the product, whereas private industry needs to sell the product. But maybe we can do some of that exploration together so that it can actually be a great risk-management synergy.”
Otherwise, she warns, commercial products built on bad data in the current rush to privatize government functions can lead to terrible results.
“It’s a concern for me that somebody’s going to ‘predict’ the weather and people are going to die because it wasn’t verified well,” she said. “It’s a serious worry for me.”
Laura Cassels is a veteran Florida journalist and former Capitol Bureau chief who specializes in science, the environment, and the economy. The Florida Trident is an investigative news outlet focusing on government accountability and transparency across Florida. The Trident was created and first published in 2022 by the Florida Center for Government Accountability, a non-profit organization that facilitates local investigative reporting across the state.