Modeling the dynamics of Greenland and Antarctic ice sheets is critical for computing projections of sea-level rise. In this talk, I will provide a partial overview of ice-sheet modeling and highlight some of the related computational challenges. I will also present the work we performed over the years as part of the DOE's efforts on earth system modeling. Specifically, I will discuss our efforts to enhance the performance and portability of the ice-sheet code, MALI. Additionally, I will describe our PDE-constrained optimization approach to estimate the initial thermo-mechanical state of the ice sheets. Finally, I will touch on our recent work to reduce the computational cost of uncertainty quantification, including the development of neural network surrogates for replacing expensive parts of the model.
Benedum 1045