The aftermath of a tornado in in Aroma Park, Illinois this past March. Stacey Wescott/Chicago Tribune via Getty
This story was originally published by the Guardian and is reproduced here as part of the Climate Desk collaboration.
As the United States prepares for hurricane season and a summer of record-breaking heat, experts fear the Trump administration’s cuts to climate and weather data programming could make the federal government’s weather forecasts less reliable when they are needed most.
The National Oceanic and Atmospheric Administration (NOAA) late last year launched a suite of artificial intelligence-powered global weather forecast models which it said would improve “speed, efficiency, and accuracy.” In March, an agency official said those models are being trained with centuries of weather data.
Artificial intelligence is a valuable tool for weather prediction, but only when it is well-trained with ample data, said Monica Medina, who served as NOAA’s principal deputy undersecretary of commerce for oceans and atmosphere from 2009 to 2012.
Under Trump, climate and weather data collection has declined, said Medina. This year, the Trump administration proposed a modest budget increase for the National Weather Service, but a 40 percent cut to NOAA overall.
“We absolutely need AI to help us crunch the data faster and to make sense of more and more data that we can collect,” said Medina, who under Joe Biden also served as assistant secretary of state for oceans. “But right now, what we’re doing is cutting back the data collection…we’re going in the wrong direction.”
Crucially, when it comes to predicting extreme weather events, new models still “underperform.”
In an emailed comment, Erica Grow Cei, a National Weather Service spokesperson said: “Despite the misinformation circulating about missing weather and climate data, there is, in fact, a wealth of weather data collected each day, from satellites in space, to a network of weather balloons, to buoys in the ocean, and land-based sensors.”
But widespread reports show staffing cuts have forced NOAA’s National Weather Service to scale back satellites and balloon launches, key parts of the country’s data collection system. And shrunk climate programs threaten ocean buoy networks and other observation systems, experts say. Research into effects of the climate crisis on Earth’s systems is also being slashed, along with funding for researchers who analyze data and identify new sources.
“Weather times time equals climate,” said Craig McLean, NOAA’s former acting chief scientist and head of NOAA Research. “Cutting climate research impacts the skill of our weather forecast, and it arrests our advancement of weather forecasts.”
Those impediments are coming as the US is preparing for more extreme weather. A “super El Niño” is expected to spike temperatures, smash heat records nationwide and may boost hurricane activity in some regions.
Noaa will issue its outlook for the 2026 Atlantic hurricane season on Thursday. For decades, scientists used traditional physics-based models to predict future weather conditions, using complex mathematical equations to simulate the dynamics at work in the atmosphere. New AI-based models instead identify patterns in decades of historical data to forecast weather outcomes.
That new technology uses less computing power than traditional models—which must run thousands of mathematical equations to work—and has been found to outperform traditional models for some aspects of weather forecasting. But it also seems to have major shortcomings, experts have found.
Crucially, when it comes to predicting extreme weather events, new models still “underperform,” according to an April study published in Science Advances. Because their forecasts are based on past weather events, the authors found, they seem to have trouble simulating the record-breaking weather events that are becoming increasingly common amid the climate crisis, instead tending to predict weather more similar to historical events.
It’s kind of a snowball effect. You need accurate data for inputs for our forecast models, but we’re running on less data currently with this current administration.”
Traditional physics-based models don’t have this problem, because they assess and predict the weather outcomes that certain physical conditions yield.
“They don’t really care if there’s a different situation than we’ve seen before, because they can understand based on a rules-based [analysis] what will happen tomorrow,” said Sebastian Engelke, a professor at the University of Geneva who co-authored the study.
Chris Gloninger, a forensic meteorologist who in 2023 received death threats after speaking about the climate crisis on television, likened the problems with AI-powered models to the ways other kinds of infrastructure struggles to manage a world experiencing global warming.
“You have infrastructure systems in this country that are built on having a steady or static climate, and we know that that’s not the case as extremes are increasing,” he said.
Like stormwater systems that were not designed to keep up with climate-fueled heavy rainfall events or roads that were not designed to withstand climate-fueled extreme heat, “the AI weather models were trained on a climate that no longer exists,” Gloninger said.
This problem already has real world implications, said Gloninger, noting that conventional models outperformed AI-based ones when forecasting a historic February 2026 blizzard in the north-eastern United States.
If the government scales up its reliance on AI-powered models while reducing the amount of data that powers them, that problem could compromise federal forecasts, said Gloninger. “It’s kind of a snowball effect,” he said. “You need accurate data for inputs for our forecast models, but we’re running on less data currently with this current administration.”
Long before Trump re-entered office, the National Weather Service had faced decades of understaffing. Recent cuts have exacerbated the problem, Gloninger said.
NOAA has not wholesale switched to AI forecasting. Instead, it says it is employing more artificial intelligence in its ensemble models, which blend multiple techniques to produce a range of probable outcomes. Cei said NOAA’s new AI-powered model suite is “an addition to our stable of weather models, not a replacement,” adding that it was “built on data” from the agency’s flagship physics-based Global Forecast System model.
But Gloninger said he is still concerned that rolling any AI technology into federal models could raise problems, particularly amid cuts to weather data collection and climate research.
“There could still very much be issues when you have a component of artificial intelligence that isn’t really trained when it comes to extreme weather and climate,” he said.
Neil Jacobs, current Noaa administrator, is “probably one of the preeminent modeling scientists”, said John Sokich, a former director of congressional affairs for the National Weather Service. “I don’t believe he would rush implement something that has not been tested,” said Sokich.
But though Jacobs is “committed to advancing weather forecasting,” Jacobs is also “a Trump appointee who must back the Trump budget or leave his job,” said McLean. The administrator defended Trump’s NOAA cuts at a House environment subcommittee hearing in April, McLean noted.
“I don’t think Dr Jacobs would be in a rush to be replacing capacity with AI that’s not ready yet,” he said. “But at the same time, the man has demonstrated his willingness to be obedient to the president who appointed him [and who is] destroying the National Oceanic and Atmospheric Administration.”
Weather forecasts serve “indispensible” practical functions, powering early disaster warnings, enabling safe aviation and shipping, and helping officials optimize sectors of the economy from energy production to agriculture, said Medina. Less accurate forecasting could pose dangers to Americans, she said.
“Weather forecasts are vital to our economy, to our health, and to public safety,” she said.
