Elizabeth A. Barnes is Boston University’s inaugural Dalton Family Chair in Environmental Data Science & Sustainability.

Convergence

Can AI Help Predict the Earth’s Climate a Decade from Now? 

For BU’s Elizabeth A. Barnes, integrating AI and Earth sciences is the next step toward solving some of the biggest, most complex challenges we face

Anyone who’s been caught in a storm without an umbrella understands the importance of having reliable weather forecasts. But, for many, this information is about more than avoiding an inconvenient soaking—it’s the foundation of their livelihood. Farmers need to know whether they’re getting one more frost before planting seeds. Emergency responders need to know when a hurricane is likely to hit, so they can evacuate people from the most dangerous zones. 

The challenge is that these predictions rely on a vast—and constantly fluctuating—set of variables, including temperature, atmospheric pressure, humidity, wind, and precipitation. A veritable fleet of surface balloons, satellites, radar, ocean buoys, ships, and aircraft are collecting and reporting a nonstop stream of information. And that’s just to get tomorrow’s forecast. Earth scientists trying to get an idea of what the weather—and the global climate—might look like a year, a decade, or a century from now, have to wrangle all of this data and more. 

So for researchers such as Elizabeth “Libby” A. Barnes, Boston University’s inaugural Dalton Family Chair in Environmental Data Science & Sustainability, the application of artificial intelligence (AI)—and its ability to crunch vast amounts of data—to Earth science is a natural fit. Earth science, the study of our planet and its many interconnected systems, encompasses fields such as geology, oceanography, meteorology, and astronomy, bringing insights into issues from earthquakes to climate change. Scientists working in either AI or Earth science have to grapple with massive datasets, and practitioners of both need to understand not only the predictions that result from their models, but also, crucially, how trustworthy they are. 

“As I tell my students, if you’re trained as an Earth scientist, you’re already trained to think about big data—to think about finding relationships, to understand that correlation is not causation. So, when you learn how to use AI tools, you immediately start to see the connections to your own work, because that’s what we do every day,” says Barnes, who joined BU in 2025 and whose research spans AI, data science, and Earth systems. She’s both a professor of computing and data sciences and of Earth and environment.

Using AI to Model the Future of Our Planet

Together with her research group, Barnes’ focus is on understanding the Earth system—that is, how our planet’s air, water, land, ice, and many life-forms might influence each other and change across time and place—all to improve our prediction of climate and extreme weather on timescales from days to decades. 

To do this, Barnes and her team use AI to make their work faster and better, and to learn new things. 

“Libby embodies what the Faculty of Computing & Data Sciences (CDS) was created to do,” says Azer Bestavros, associate provost for computing and data sciences. “CDS is really not about computer science or data science or AI on their own—it’s about how they transform disciplines. In her work, Libby doesn’t see separate disciplines; she sees these all as tools for solving broad, complex problems. That’s the definition of convergent research.”

Barnes’ own career path reflects this blending of disciplines (her BU appointment straddles CDS and the College of Arts & Sciences). Her educational degrees in physics, mathematics, and atmospheric science had an emphasis on “pure science,” she says, and came with a certain attitude that “the best kind of science [uses] paper and pencil.” It’s a posture that never quite fit for Barnes, who had loved writing code and analyzing data for as long as she can remember. Eventually, about a decade ago, she had a realization: “I’ve been a data scientist my whole life, I just didn’t know it.” She dug in.

AI tools are making huge computational advances in highly complex Earth system models that are the bread and butter of Earth science research. These models provide crucial insights for things such as climate change, food security, human health, and resource management. Scientists can run experiments on these models to test what-if scenarios, such as volcanic eruptions or carbon dioxide increases. Policymakers can also use these models to guide resource stewardship decisions for the future. 

Typically, these models, which rely on a keen understanding of physics and geosciences, have been labor-intensive and slow to make. Using AI, however, researchers can make them better and much faster—which means they can run more experiments in a shorter amount of time.

Barnes’ team, for instance, has found large-scale data models can be useful in predicting hurricanes across years and seasons. They’ve also developed a way to combine machine learning and analog forecasting to improve multiyear climate predictions. And they’ve collaborated on a project that fused climate model simulations and trade data to assess how extreme weather in one country might ripple out and affect many others.

Opening Up AI’s Black Box

AI programs, and the algorithms that power them, can be notoriously mysterious, even to the computer scientists who build them. But Barnes is especially interested in getting under the hood to understand how these algorithms make predictions, what data they use and how they learn from it, to illuminate possible pathways for new discoveries about our planet and artificial intelligence. 

Shining a light into the so-called black box of machine learning is crucial for quantifying the level of certainty associated with a weather or climate prediction, she says. One of the big focuses for Barnes’ group is explaining how an algorithm learned from the data it was given to produce a predictive model.

“Uncertainty quantification is huge in [the Earth sciences], because we do a lot of prediction. You can think of it as weather forecasting, but we go out years to decades into the future. We don’t know exactly what’s going to happen, so uncertainty has to be a part of what we produce,” Barnes says. “And to get at that uncertainty, you can’t just make a single prediction. You often have to make lots of them, because the butterfly effect takes over.” Earth scientists need to communicate how big or small that effect likely is, as well as how they determined its scale. 

For Barnes, integrating AI and Earth sciences is simply the next step toward solving some of the biggest, most complex challenges that we humans face. 

“The reason I’m in this field now is because I tried to think of the most complicated system I could study, so there’d always be questions to answer, and this was the obvious choice,” she says, only half-joking. “I’m in this field because I think it matters, and at this point, AI and Earth science don’t feel like two fields to me. It feels like I’m working in one field asking, ‘What’s next?’ And for me, what’s next is using these tools to do something hard, something that maybe we didn’t know how to do before, or didn’t even know to ask.”