- A research team has built a near real-time system for the simulation of Earth’s interior and earthquakes.
- To do this, they used the world’s fastest supercomputer, Summit.
- They are currently addressing issues in helioseismology, exploration geophysics, and global and regional seismology.
Recent advances in high-performance hardware and efficient algorithms have enabled researchers to develop high-resolution, three-dimensional simulations of regional and global seismic wave propagation with remarkable preciseness.
For over two decades, spectral-element method (SEM) has been used in computational fluid dynamics. Now it’s gaining popularity in seismology due to its incredible capability of combining the flexibility of the finite-element method with the global pseudospectral method.
A research team at Princeton University is using the same technique to image Earth’s interior. They are currently addressing issues in helioseismology, exploration geophysics, and global and regional seismology.
Simulating Earth’s Interior In 3D
For this project, the team used Earth’s subsurface and years’ worth of kinematic representations of earthquakes. The aim is to minimize the differences between synthetics and data, for example, amplitude anomalies, waveform differences, and cross-correlation travel-time.
To image the opaque solar interior, they used properties of near-surface supersonic turbulence. Also, local helioseismology methods make it possible to obtain seismic data from the observed noisy solar wavefield.
They studied time-lapse experiments in which characteristics of the model’s specific region change between 2 succeeding surveys. The time-lapse migration is also known as 4D seismic imaging because it represents flow-induced temporal change (the 4th dimension) along with the spatial distribution of reflection coefficients. It’s a crucial tool for monitoring both carbon sequestration and fluid injection in reservoirs.
Researchers have used the world’s fastest supercomputer, Summit to precisely simulate the 3D acoustic, anelastic and poroelastic wave propagation. The supercomputer is equipped with NVIDIA Tesla V100 Tensor Core GPUs, which is significantly boosting the simulation’s performance, reported by authors. For better results, they have started adopting deep learning techniques where large amounts of complex data need to be interpreted.
Tools and Research Papers
The researchers have also built a machine learning tool to detect appropriate earthquake measurements that can be used in the project. Also, they maintain a set of open source software to generate comprehensive datasets for simulating acoustic and poroelastic wave propagation.
The team has published numerous papers describing their methods and strategies. The most recent describes the full-waveform inversion method for determining Earth’s material properties.