GEOS: A Multi-Physics Simulation Framework for Subsurface Applications

GEOS is an open-source simulation framework for modeling tightly coupled multi-physics processes in subsurface reservoirs, with a primary focus on geological carbon storage. Originally developed at Lawrence Livermore National Laboratory in 2011 to study hydraulic fracture propagation in enhanced geothermal systems, the framework underwent a major redesign in 2016 as part of the U.S. Department of Energy’s Exascale Computing Project. Since then, GEOS has evolved into a collaborative effort involving TotalEnergies and Stanford University (joined in 2018), Chevron (joined in 2023), and contributors from numerous academic institutions and national laboratories.

GEOS simulates coupled thermal, hydrological, and mechanical processes associated with fluid injection and withdrawal in the subsurface. A central objective is to address challenges related to reservoir integrity and failure mechanisms through high-fidelity, scalable simulations. The codebase features a modular physics package interface that supports the development of standalone models for individual physical processes, which can then be integrated into fully coupled systems. These systems can be solved using flexible strategies, including monolithic formulations and operator-splitting techniques.

Performance portability is a core design goal, enabling GEOS to run efficiently across diverse computing platforms - from laptops to GPU-based exascale systems. In this talk, we highlight architectural innovations that support this portability, review current modeling capabilities, and present numerical results demonstrating GEOS’s performance and robustness on challenging CO? geological storage simulations.

Speaker: Nicola Castelletto, Lawrence Livermore National Labs