AI reveals new insights into Antarctic ice flow

AI reveals new insights into Antarctic ice flow
by Clarence Oxford
Los Angeles CA (SPX) Mar 14, 2025

As climate change accelerates, the Antarctic ice sheet continues to shrink, contributing to rising sea levels worldwide. Holding enough frozen water to raise sea levels by approximately 190 feet, understanding how Antarctic ice moves and melts is crucial for forecasting future coastal impacts. However, existing climate models face significant challenges in accurately simulating ice dynamics due to the region's complexity and limited observational data.

A study published on March 13 in *Science* by researchers at Stanford University leverages machine learning to analyze high-resolution satellite and radar data, offering fresh insights into the fundamental physics governing Antarctic ice movement. This novel approach could enhance predictions of the continent's evolution amid global warming.

"A vast amount of observational data has become widely available in the satellite age," said Ching-Yao Lai, an assistant professor of geophysics at the Stanford Doerr School of Sustainability and senior author of the study. "We combined that extensive observational dataset with physics-informed deep learning to gain new insights about the behavior of ice in its natural environment."

Ice Sheet Dynamics

The Antarctic ice sheet, nearly twice the size of Australia, plays a critical role in stabilizing global sea levels by locking away freshwater. Traditional models have relied on assumptions about ice mechanics derived from controlled laboratory experiments. However, Antarctica's ice is far more complex. Ice properties differ depending on formation processes-whether from seawater or compacted snow-and structural variations such as cracks or air pockets further complicate movement dynamics.

"These differences influence the overall mechanical behavior, the so-called constitutive model, of the ice sheet in ways that are not captured in existing models or in a lab setting," Lai explained.

Rather than attempting to model each variable individually, the researchers developed a machine learning framework that analyzed large-scale ice movement and thickness recorded from remote-sensing data collected between 2007 and 2018. The AI-driven approach adhered to known physical laws governing ice flow, allowing researchers to refine constitutive models and more accurately describe the ice's viscosity-its resistance to deformation.

Compression Versus Strain

The study examined five Antarctic ice shelves-floating extensions of land-based glaciers that act as barriers against the loss of continental ice. Findings revealed that ice near the continent undergoes compression, with mechanical properties aligning closely with laboratory predictions. However, as ice moves further from land, it experiences increased tensile strain, pulling outward toward the ocean. This change in force causes the ice to behave anisotropically, meaning its properties vary in different directions-similar to how wood splits more easily along the grain.

"Our study uncovers that most of the ice shelf is anisotropic," said first author Yongji Wang, who conducted the research as a postdoctoral scientist in Lai's lab. "The compression zone-the part near the grounded ice-only accounts for less than 5% of the ice shelf. The other 95% is the extension zone and doesn't follow the same law."

With sea levels already rising and intensifying coastal flooding, erosion, and storm damage, understanding ice sheet behavior becomes increasingly vital. Previous models largely assumed that Antarctic ice possessed uniform mechanical properties, a simplification researchers had long suspected was flawed. The work by Lai, Wang, and colleagues confirms these limitations and underscores the necessity of updating predictive models to incorporate anisotropy.

"People thought about this before, but it had never been validated," said Wang, now a postdoctoral researcher at New York University. "Now, based on this new method and the rigorous mathematical thinking behind it, we know that models predicting the future evolution of Antarctica should be anisotropic."

AI for Earth Science

Although the researchers have yet to determine the precise cause of anisotropy in the extension zone, they plan to refine their models with additional Antarctic data as it becomes available. Their findings also offer insights into structural weaknesses that may contribute to ice shelf rifting or calving events, where massive icebergs break away from the main shelf. This research represents a crucial step toward developing more realistic simulations of future ice dynamics.

Beyond glaciology, Lai and her team believe that combining AI-driven data analysis with established physical principles could unlock new discoveries in Earth sciences. This method may be applied to study a variety of natural processes where large observational datasets exist.

"We are trying to show that you can actually use AI to learn something new," Lai said. "It still needs to be bound by some physical laws, but this combined approach allowed us to uncover ice physics beyond what was previously known and could really drive new understanding of Earth and planetary processes in a natural setting."

Research Report:Deep learning the flow law of Antarctic ice shelves