Emma Liu (UChicago): Schmidt AI in Science Speaker Series

Organized by the University of Chicago’s Eric and Wendy Schmidt AI in Science Fellowship Program.

Agenda
4:00pm – 4:45pm:  Presentation
4:45pm – 5:00pm:  Q&A
5:00pm – 5:30pm: Reception

Meeting location
William Eckhardt Research Center. Room 401
5640 S Ellis Avenue, Chicago, IL 60637
Map It

Title: Probabilistic Inference of Ice Flow Law Using Simulation-Based Machine Learning

Abstract: Glacial mass loss has contributed approximately 21% of global sea-level rise since 2019. However, projections of future contributions from ice sheets remain highly uncertain, with estimates spanning up to ~1 meter over the next century. A key source of this uncertainty lies in our limited understanding of small-scale ice dynamics that control large-scale flow behavior. In particular, the constitutive (rheology) relationship that governs how ice deforms under various stress and thermal conditions, is often oversimplified and its parameters remain poorly constrained in large-scale ice-sheet models.

In this talk, I will present a data-informed framework that integrates observational data, forward modeling, and machine learning to better constrain ice rheology. The approach aims to infer probability distributions of rheological parameters directly from observations, rather than relying on fixed or poorly constrained values. Specifically, I employ a normalizing flow model to represent complex parameter distributions via a sequence of invertible transformations applied to a base distribution. During the training, samples drawn from the parameter distributions are propagated through a forward glacier model to generate predictions of glacier velocity and thickness. These predictions are then compared with observations, and the resulting misfit is the training loss. This probabilistic, simulation-based inference framework enables systematic quantification of uncertainty in rheological parameters and provides a pathway toward reducing uncertainty in ice-sheet projections of sea-level rise.

Bio: Weijia (Emma) Liu is an Eric and Wendy Schmidt AI in Science Fellow in the Department of Geophysical Science at the University of Chicago where she works at the intersection of AI/ML and glacier modeling. Prior to this, she obtained her Ph.D. in the Geophysics Department at Stanford University in 2025 where she studied shear localizations in glaciers using high-performance computing and ML.