What Is Geophysics Computing?

Geophysics computing is the application of computational methods, high-performance computing, and numerical algorithms to the simulation, processing, and interpretation of geophysical data. It provides the tools that allow geophysicists to solve the large-scale mathematical problems at the core of subsurface imaging, Earth structure modeling, and geophysical inversion, problems whose scale and complexity make analytical solutions impossible. The field draws from numerical analysis, parallel computing, computer-aided analysis, and scientific software engineering, and it has grown in scope as the volume of seismic, satellite, and sensor data has expanded beyond what conventional workstation processing can handle.

The computing requirements of modern geophysics are substantial. A single 3D seismic survey covering several hundred square kilometers may generate terabytes of raw data, and forward-modeling the wave propagation through a realistic Earth model at full bandwidth requires billions of arithmetic operations per time step. These demands have driven geophysics computing toward cluster computing, GPU acceleration, and cloud-based processing pipelines.

Numerical Simulation and Forward Modeling

Forward modeling, computing the predicted response of a geophysical instrument to a hypothetical subsurface model, is the computational foundation of geophysics. Seismic wave propagation is typically simulated using finite-difference or finite-element discretizations of the elastic wave equation. Spectral methods, including the spectral element method developed at institutions such as Princeton and the Institut de Physique du Globe de Paris, offer high accuracy for global and regional scale problems. The Computational Infrastructure for Geodynamics distributes open-source codes for mantle convection, seismic wave propagation, and crustal deformation used by research groups worldwide. These forward codes also serve as the engine inside iterative inversion algorithms.

Inversion and Imaging Algorithms

Geophysical inversion estimates subsurface properties from surface measurements by iteratively updating a model until its simulated response matches the observed data. Full waveform inversion, applied to seismic data, minimizes the misfit between the synthetic and recorded wavefield and can recover velocity models at spatial resolutions approaching the dominant wavelength. Reverse time migration, a two-way wave-equation imaging method, uses the same computational machinery as forward modeling but runs the wavefield backward in time. Both methods are computationally demanding, and their practical adoption has depended on GPU hardware capable of sustaining the required floating-point throughput. Research on advanced digital signal processing of seismic data describes the evolution from early Fourier-domain methods to the current wave-equation approaches.

High-Performance and Cloud Computing

The parallelism inherent in geophysical forward modeling, time steps advance independently in space given boundary conditions, makes these algorithms well-suited to distributed-memory clusters using MPI-based decomposition. More recently, GPU clusters and cloud computing platforms have become the preferred infrastructure for large-scale production imaging, with some oil and gas operators running 3D full waveform inversion on thousands of cores simultaneously. Machine learning methods are also entering the geophysics computing toolkit, with neural networks applied to velocity model building, noise suppression, and fault detection in seismic images. The SEG geophysics computing literature covers the methods that underpin these workflows.

Applications

Geophysics computing has applications in a range of fields, including:

• Seismic imaging for petroleum reservoir characterization and carbon capture site assessment
• Global earthquake simulation for seismic hazard modeling and early-warning system testing
• Crustal deformation modeling, combining GPS, InSAR, and gravity data to study tectonic processes
• Mineral exploration, processing large-scale airborne electromagnetic survey datasets
• Climate modeling, running coupled Earth system simulations with geophysical components