.jpg)
Plumes of smoke from fires worsened by the extreme temperatures in Moscow, Russia, in 2010. Some areas recorded pollution levels ten times the normal levels for the capital.
Image Credit: European Space Agency
(CC BY-SA 3.0 IGO)
Scientific Frontline: Extended "At a Glance" Summary: AI-Boosted Rare Event Sampling (AI+RES)
The Core Concept: A hybrid forecasting method that combines artificial intelligence with traditional physics-based climate models to efficiently and accurately predict the probability of extreme, once-in-a-millennium weather events, such as short-duration heat waves.
Key Distinction/Mechanism: Traditional physics models require massive computational resources to simulate rare extremes, while standard AI models often fail on these "gray swans" due to lack of training data. AI+RES overcomes this by using AI to intelligently score and guide a statistical technique called rare event sampling (RES). The AI identifies the atmospheric conditions most likely to cause rapid extremes, allowing the traditional climate model to focus its simulations only on those high-probability scenarios, rather than running tens of thousands of random variations.
Major Frameworks/Components:
- Physics-Based Climate/Weather Models: Traditional systems that compute scenarios based on physical conditions like atmospheric pressure and temperature.
- Rare Event Sampling (RES): A statistical method that speeds up simulations by scoring conditions to focus the model on promising scenarios; traditionally struggles with short-duration events.
- Artificial Intelligence (AI): Used to enhance the RES scoring mechanism by predicting which specific, short-term conditions will rapidly develop into extreme weather.
Branch of Science: Atmospheric Science, Climatology, Computational Science, and Data Science, Artificial Intelligence.
Future Application: The framework is designed to be scaled using existing state-of-the-art numerical prediction models and trained AI models. It could forecast other severe weather events like tropical cyclones and extreme precipitation, providing highly specific, regional probabilities under various climate change scenarios to inform local and federal adaptation planning. It could also generate rare-event datasets to further train and improve AI models.
Why It Matters: In testing, AI+RES achieved the accuracy of 50,000 traditional simulations using only one-hundredth of the computational power. This enables much faster, less resource-intensive forecasting of deadly extreme weather, giving policymakers the localized, reliable data necessary for life-saving mitigation efforts as severe weather events become more frequent.
.jpg) |
| A satellite recorded heat signatures from land during a record-breaking European heat wave on June 23, 2026. The deepest color on the chart represents temperatures at 55 degrees Celsius or higher (130 degrees Fahrenheit). Image Credit: Copernicus Sentinel/European Space Agency (CC BY-SA 3.0 IGO) |
Although day-to-day weather predictions have improved, forecasting events that might happen once in 1,000 years—like the deadliest heat waves—remains a challenge.
Traditional supercomputer-based models can forecast these events, but they require substantial time and energy. Meanwhile, newer forecasting models based on artificial intelligence are effective for day-to-day forecasts but often fail to predict outlier events not represented in their training data.
“AI weather and climate models are one of the great achievements of AI in science, but they are not magical—they fail on gray swans, the rarest and most extreme events,” said Pedram Hassanzadeh, an associate professor of geophysical sciences at the University of Chicago. “Detailed physics-based models can capture extremes, but they require prohibitively large amounts of time and energy.”
An international team of researchers in the United States and France, co-led by members of Hassanzadeh’s Climate Extremes Theory and Data Group, has developed a new hybrid method, published in Physical Review Letters, to solve this problem.
Their solution marries the efficiency of AI tools with the reliability of traditional models to predict the odds of rare events quickly and accurately while using far fewer resources.
“The power of this method,” Hassanzadeh said, “is that it combines the strengths of both AI and traditional physics and is particularly effective for extreme events, which are the hardest to simulate and have the greatest societal impact.”
The Statistics of Rare Events
Heat waves are one of the deadliest forms of extreme weather. In 2003, a heat wave led to roughly 70,000 deaths across Europe, and Russia suffered 56,000 deaths in 2010. This past June, nearly half of the United States—roughly 180 million people—experienced dangerous temperatures.
These waves are becoming increasingly frequent and severe, but the nature of outlier events makes them difficult to study and challenging to predict.
Forecasting has long relied on physics-based climate and weather models. These models help researchers determine how different conditions, such as atmospheric pressure, might affect temperature and other variables over time. They compute many potential scenarios, which researchers use to determine what is most likely to occur.
The challenge is this: if researchers want to know the odds that Chicago will reach 90°F in July—which is not uncommon—they do not need to run many simulations before one lands on that temperature. However, determining the odds of the temperature reaching 105°F requires many more attempts to observe that extreme. Running this volume of simulations demands substantial time and computational power.
Researchers can mitigate this issue by using a statistical technique called rare event sampling (RES). This approach accelerates the process by scoring conditions so that the climate model can focus only on the most promising scenarios and ignore the rest. However, RES is less effective for short-duration events, such as weeklong heat waves, than for an entire season that is overall unusually hot.
To address this, the team developed a new method named AI+RES, which enhances the scoring mechanism by incorporating AI’s ability to predict which conditions are most likely to lead to shorter, rapidly developing extremes.
“After doing this iteratively, you eventually get to a bunch of simulations that do indeed capture whatever rare extreme event that you’re interested in,” said Alexander Wikner, a Schmidt AI in Science Postdoctoral Fellow in Hassanzadeh’s group and co-first author of the study. “The more you get, the better you can estimate the probability of that event, and that ultimately gives you a lot more certainty.”
To test the method, the team ran 50,000 simulations using a traditional climate model to predict heat waves over areas of France and the US Midwest. Their new AI+RES method yielded nearly identical results using one-hundredth as many simulations.
Because this work was a proof of concept, they used a model that did not account for climate change, which introduces another level of complexity. The researchers hope to test their method on models running under different climate change scenarios to determine how results might shift as the planet warms.
Wikner noted that the method could also help generate rare-event datasets to train more advanced AI models, which would accelerate their method even further.
Scaling Up
According to the scientists, the hybrid method could be applied to other severe weather events, including tropical cyclones and extreme precipitation.
“What I like about this framework,” said Hassanzadeh, “is that it’s ready to be scaled. AI models trained on real weather observations are already built and in use, so the next step is to connect a state-of-the-art numerical weather or climate prediction model to one of them through this algorithm.”
This capability would give decision-makers access to the accurate information they need, such as the frequency of strong storms and heat waves in the current and future climate at regional scales—specifically over Texas, Florida, or California.
“This is exactly the kind of information federal, state, and local governments need as a first step for climate adaptation and mitigation planning,” added Hassanzadeh. “It’s very exciting that we could be scaling this up into real-world models that directly provide such information to the public and to policymakers.”
Funding: The Eric and Wendy Schmidt AI in Science Fellowship, France-Chicago Center FACCTs Award, US National Science Foundation, the Institute for Climate and Sustainable Growth at UChicago, RTE France, the National Agency for Research and Technology (ANRT), and the Institut des Mathématiques pour la Planète Terre (IMPT).
Published in journal: Physical Review Letters
Title: AI-Boosted Rare Event Sampling to Characterize Extreme Weather
Authors: Amaury Lancelin, Alexander Wikner, Laurent Dubus, Clément Le Priol, Dorian S. Abbot, Freddy Bouchet, Pedram Hassanzadeh, and Jonathan Weare
Source/Credit: University of Chicago | Maureen Searcy
Edited by: Scientific Frontline
Reference Number: as081726_02