Wiley Kohler ’25
Predicting Antarctic Melt Rates Using Physics-informed Deep Learning
I studied the behavior of Antarctic ice shelves using various machine learning techniques. Understanding the role of ice shelves, the floating outlets of ice sheets into the ocean, is key to understanding global sea level rise as a result of climate change. In Antarctica in particular, ice shelves form an important buffer that stabilizes the continent’s substantial ice mass. However, many of their attributes are difficult to measure accurately, which makes predictions of future sea level change more difficult. I used neural networks to develop less noisy, continuous, and infinitely differentiable models of ice velocity, thickness, and other factors, which can be used to better understand and predict the future behavior of ice shelves. Beyond actually developing this model, I studied the properties of some of these neural networks and the effectiveness of various machine learning techniques in the context of fluid dynamics problems. I learned a lot about implementing machine learning algorithms in the context of climate science and had the opportunity to independently explore new questions. As a result of these experiences, I plan to continue to pursue coursework and independent work at the intersection of computer and climate sciences.
Extreme Weather and Impacts
Lai Group, Department of Geosciences, Princeton University - Princeton, New Jersey
Ching-Yao Lai, Assistant Professor of Geosciences