\n
![](\"https:\/\/gizmodo.com\/app\/uploads\/2026\/04\/Earth-from-space-1280x853.jpg\")
<\/p>\n
\n<\/p>\n
On November 12, 1970, the Bhola cyclone slammed into the coast of what was then East Pakistan. The storm brought maximum sustained wind speeds of 130 miles per hour (205 kilometers per hour) and a 35-foot (10.5-meter) storm surge, killing an estimated 300,000 to 500,000 people.<\/p>\n
Today, the Bhola cyclone remains the deadliest tropical storm on record. But if it had struck a decade later, it might not have been so devastating. Weather forecasting changed dramatically in the 1970s as meteorologists adopted physics-based computer models that improved storm prediction. With the rise of AI, forecasting is evolving again\u2014but this time, experts worry the new models may be less reliable when it comes to predicting unprecedented weather events.<\/p>\n
Researchers are calling this the \u201cgray swan\u201d problem. Gray swan weather extremes are physically plausible but so rare that they are poorly represented in training datasets. The trouble is, climate change is leading to more first-of-their-kind weather extremes. Think: the 2021 Pacific Northwest heatwave. This event was so severe that it would have been virtually impossible without climate change.<\/p>\n
Physical forecast models can simulate gray swan events like the Pacific Northwest heatwave, though they are labeled extremely rare. They can do that because they are built on the laws of physics. AI models are trained on past weather data, wherein gray swans are practically nonexistent.<\/p>\n
\u201cThey fail on gray swans,\u201d Pedram Hassanzadeh, an associate professor of geophysical sciences at the University of Chicago, told Gizmodo. He and his colleagues published a study last April that removed all Category 3 through 5 hurricanes from an AI model\u2019s training dataset, then tested it on Category 5 storms. The results showed that AI models cannot accurately forecast previously unseen events, as this would require extrapolation.<\/p>\n
\u201cThe concern isn\u2019t occasional misses. It\u2019s that AI models can miss silently, producing confident forecasts of unremarkable weather while a record-breaking event is unfolding,\u201d Rose Yu, an associate professor of computer science and engineering at the University of California San Diego, told Gizmodo in an email.<\/p>\n
\u201cOther risks matter too,\u201d she said. \u201cAI models can violate conservation laws in subtle ways that don\u2019t show up in standard metrics. When they bust a forecast, diagnosing why is harder. They depend on stable observing systems, which is a real concern given current pressure on satellite programs. And institutionally, if we consolidate around AI too quickly and let physics-based infrastructure atrophy, we lose the redundancy that currently catches AI\u2019s failures.\u201d<\/p>\n
Despite these pitfalls, meteorologists are rapidly adopting AI forecast models, and it\u2019s actually easy to understand why. They\u2019re faster, cheaper, and require far less computational infrastructure than physical models. When it comes to predicting typical weather patterns and events (not gray swans), their accuracy is comparable and improving rapidly.<\/p>\n
\u201cThe typical rate of progress for most state-of-the-art physical models has been something like a day more accurate per decade, which doesn\u2019t sound like a lot, but that\u2019s consequential,\u201d Andrew Charlton-Perez, a professor of meteorology and head of the School of Mathematical, Physical, and Computational Sciences at the University of Reading, told Gizmodo.<\/p>\n
\u201cThe rate of accuracy growth for machine learning models has vastly exceeded that,\u201d he said. \u201cThey are now competitive, and two-three years ago, they were not even in the same ballpark.\u201d<\/p>\n
During the 2025 Atlantic hurricane season, for example, Google DeepMind\u2019s model outperformed nearly every physical model on storm track and intensity. In fact, since 2023, leading AI models such as GraphCast, Pangu-Weather, and the ECMWF\u2019s AIFS have matched or outperformed the best physical models on medium-range forecasting metrics, according to Yu.<\/p>\n
AI models are proving especially valuable in parts of the world that lack traditional forecasting resources\u2014regions that are often on the frontlines of climate change. Hassanzadeh co-directed an initiative that provided 38 million farmers across India with AI-based monsoon forecasts, giving them up to four weeks\u2019 advance notice of the rainy season\u2019s onset.<\/p>\n
\u201c\u200b\u200bA lot of countries were left behind in that first revolution of weather forecasting, because [traditional] weather forecasting requires a supercomputer, hundreds of millions of dollars, various fields, workforce, and experts,\u201d Hassanzadeh explained. AI models, by comparison, are far more accessible to lower-income countries.<\/p>\n
Still, rapidly adopting these models without addressing the risks would be dangerous, especially in parts of the world highly vulnerable to the impacts of climate change. Shruti Nath, a postdoctoral research associate at the University of Oxford, recently co-authored an editorial calling for more rigorous testing of AI forecast models before public agencies widely adopt them.<\/p>\n
\u201cThere is still a lot of work to be done in understanding the limits of these models, alongside where they could supplement physical models and why,\u201d she told Gizmodo in an email.<\/p>\n
Nath\u2019s editorial outlines a framework for testing AI forecast models that would deliberately withhold a designated set of \u201ciconic\u201d extreme events (like the Pacific Northwest heat wave, for example) from the training dataset. These events would be reserved solely for testing in order to assess the models\u2019 ability to extrapolate unprecedented weather extremes, or gray swans.<\/p>\n
Actually implementing this AI Retraining Without Iconic Events (AIRWIE) protocol \u201cwould require the meteorological community to agree on which high-impact events constitute a rigorous benchmark,\u201d the editorial states. This would be a great undertaking, but Nath believes most researchers agree that there is an urgent need for this kind of testing.<\/p>\n
\u201cWe need to be a bit more organized, however, in ensuring that proper protocols can be followed and that robust safeguards are put in place and maintained by the community,\u201d Nath said. \u201cThis is difficult when things are in such a hype phase and no one wants to miss out on the bandwagon.\u201d<\/p>\n
Other researchers, like Hassanzadeh, are developing ways to teach AI forecast models to predict gray swans. He and his colleagues are investigating whether combining AI systems with \u201crelevant sampling\u201d methods\u2014which allow them to generate samples of gray swan events\u2014can improve the models\u2019 ability to extrapolate unprecedented extremes.<\/p>\n
Efforts to understand and address the limitations of AI forecasting will be critical, because there\u2019s no turning back now. AI is already reshaping the way we predict the weather, and as the climate becomes increasingly volatile, meteorologists will need every tool in their arsenal to be sharp and reliable. Despite their current limitations, there is much to gain from continuing to push these systems forward and figuring out how to best integrate them with physical forecasting.<\/p>\n
\u201cThe research agenda is about making AI models physically consistent, well-calibrated, and robust to distribution shift,\u201d Yu said. \u201cAbandoning this approach because of the gray swan problem means giving up the biggest improvement in forecasting in a generation.\u201d<\/p>\n<\/p><\/div>\n