Advances in computational resources from current and planned HPC systems have enabled Earth System models with finer spatial resolution and additional complexity than was supported by past systems. However, obtaining physically accurate yet efficient results requires both new model formulations (e.g. non-hydrostatic equations, realistic process coupling, or complex convection processes) and new methodologies (e.g. machine learning for improving/emulating parameterization schemes or domain specific languages for code refactoring/optimization). These new model approaches and methodologies in turn require new numerical techniques for the discretization of time, physical space, and parameter space to address issues of scalability and data storage on target HPC ecosystems, which include both CPUs and GPUs. This session will present progress on the new modeling approaches, numerical techniques, and data management strategies that will enable the Earth System models of the future.

• NG - Nonlinear Geophysics