High-Performance Computing for Coupled Subsurface Processes: Parallelization of a Geomechanics Multiphysics FEM Code

High-performance computing (HPC) is essential for solving complex multiphysics problems in science and engineering, especially in geosciences where subsurface modeling demands high scalability and resolution. The GATOR finite element code was developed to address these challenges, supporting thermo-hydro-gas-mechanical (THGM) couplings and other geoscientific applications like CO2 storage and geothermal energy.

GATOR’s parallelization involved a full architectural redesign for distributed-memory systems. Key strategies include domain decomposition using METIS for balanced load distribution, parallel element integration, solver parallelization via PETSc for scalable performance, HDF5-based parallel I/O, and MPI for process communication.

Two benchmarks were used to verify performance: a cubic domain with regular mesh for mechanical analysis, and a hydraulic fracturing model with irregular mesh and coupled hydro-mechanical behavior. The first showed optimal solver scalability, while the second revealed performance limitations due to poor matrix conditioning.

In conclusion, parallel computing is vital for advancing geoscientific simulations. GATOR’s parallel implementation demonstrates strong potential, though further optimization—such as improved solver strategies and dynamic load balancing—is needed to handle increasingly complex models.