Image
How AI Is Changing Climate Science
Artificial intelligence is beginning to transform climate science, not just by improving forecasts, but by helping researchers understand the physical forces shaping the planet’s future.
A new study led by Antonios Mamalakis of the University of Virginia School of Data Science and Department of Environmental Sciences demonstrates how advanced AI systems can uncover the climate patterns driving winter precipitation across the United States while also revealing whether the models are learning meaningful science or merely identifying statistical shortcuts.
Published in "Artificial Intelligence for the Earth Systems," the research combines deep learning and explainable artificial intelligence, or XAI, to analyze one of climate science’s persistent challenges: predicting seasonal precipitation months in advance.
The findings could eventually help communities better prepare for droughts, floods, wildfire conditions, and even water shortages, particularly across the southern United States, where winter precipitation patterns proved substantially more predictable than in northern regions.
Why Explainable AI Matters in Climate Research
For Mamalakis, the study’s most important contribution was not simply prediction accuracy. It was trust.
“We want to know whether or not the AI model we have trained predicts correctly for the right reasons,” he said.
That question sits at the center of explainable AI, an emerging field focused on opening the “black box” of AI systems to understand how they arrive at decisions.
In climate science, the stakes are especially high.
“When you’re using an AI model for a climate task, especially in high-stakes settings like forecasting the trajectory of a hurricane, you need to make sure it hasn’t learned shortcuts,” Mamalakis said. “Because if a new event falls outside the distribution of events the model was trained on, those shortcuts will not apply anymore, and the model can derive significantly wrong predictions.”
He argues that AI systems used in environmental forecasting must be evaluated not only on whether they produce accurate predictions, but whether they rely on physically meaningful climate signals to reach those predictions.
Why the Southern United States Is More Predictable
The study found that the southern United States, especially the Southeast and Gulf Coast, showed consistently higher winter precipitation predictability than northern states. Florida, Georgia, the Carolinas, and Virginia demonstrated some of the strongest forecasting skill.
That pattern aligns with decades of climate research linking winter precipitation in the South to El Niño and La Niña events in the tropical Pacific Ocean.
“The signal of El Niño and La Niña events is much stronger over the southern U.S.,” Mamalakis said. “For example, during El Niño years, the jet stream tends to intensify and shift to the south, bringing more winter storms and wetter conditions.”
Across nearly all AI systems tested, the tropical Pacific consistently emerged as the dominant source of predictive information, reinforcing the importance of tropical Pacific climate state in shaping U.S. winter weather.
The models also identified important climate signals in the tropical Atlantic Ocean, suggesting multiple ocean basins may influence seasonal precipitation patterns.
What “Meta Consensus” Reveals About AI and Scientific Discovery
One of the study’s more novel concepts is what Mamalakis calls “meta consensus.”
“If the models agree about what they agree on and where they disagree, that’s a good indicator they have learned something physical,” he said.
In the study, different AI systems independently arrived at similar conclusions about the drivers of seasonal precipitation, particularly during strong El Niño and La Niña years. The researchers found that the models demonstrated the highest agreement during periods when climate conditions became more predictable.
For Mamalakis, this represents part of a larger shift in scientific research. “We are entering a period where AI can become a scientific tool, not just a forecasting tool,” he said.
AI, Climate Science, and the “Sustainability Paradox”
While Mamalakis sees enormous promise in AI-enabled climate research, he also acknowledges the technology’s growing environmental footprint.
“On one hand, AI can help accelerate science and help us gain new knowledge,” he said. “On the other hand, at large scales, especially in massive data centers, it can require ridiculous amounts of energy. For our study, the models were relatively small and trained locally, but this becomes an important consideration as AI systems scale up.”
He describes the tension as a “sustainability paradox.” Large-scale AI systems capable of improving climate forecasting and scientific understanding also require massive data centers that consume substantial amounts of electricity and water.
Still, the potential societal benefits are significant.
By providing reliable seasonal forecasts months ahead of time, explainable AI systems could eventually help communities better manage water resources, prepare for floods and droughts, and respond more effectively to climate extremes before they occur.
