A Scalable Solution Framework for the Community Ice Sheet Model

To date, the ice sheet models used for the predictions of sea level rise have been over-simplified and, in particular, fail to reproduce the fast dynamic changes observed on the ice sheets during the last decade. Significant efforts are underway to improve the accuracy and fidelity of predictive ice sheet models for use in Earth system models via increased complexity and spatial resolution. These developments will result in a computational burden that is orders of magnitude larger than previous ice sheet models. We propose to develop a scalable, robust, and efficient solver framework for the Department of Energy Community Ice Sheet Model, which is currently under development as part of the Community Climate System Model (CCSM). This solver framework will be designed to complement corresponding CCSM component algorithmic advances by using the fully implicit Newton-Krylov method, and building around a modular solver interface with Trilinos. There are many benefits to this approach, including minimal code development, adaptability with future model improvements, and the ability to incorporate additional scalable performance enhancements.