Full waveform inversions and source characterization based on spectral-element simulations: innovative datasets and inversion strategies

点击翻译按钮获取中文摘要

Seismic imaging, the mapping of subsurface structures based on seismic waves traversing through the Earth's interior, has generated fascinating images of the Earth's crust and mantle which prove to be critical for understanding geodynamic processes and geological events that shaped the tectonic evolution of the Earth. On the other hand, seismology engages in understanding earthquakes, one of the most damaging natural disasters to mankind, including their locations, source mechanisms, seismic wave propagation and the intensity of induced ground shaking. The collection of this essential information offers crucial guidance on long-term seismic hazard assessment and risk mitigation, especially for earthquake-prone regions. The capability to simulate seismic wave propagation based on accurate numerical methods, such as the spectral-element method (SEM) developed in the past two decades, enabled the application of innovative, numerical-simulation-based full waveform inversion (FWI) techniques for high-resolution imaging, as well as seismic source inversions based on 3D Greens functions for complex background models. Our group continues to work at the forefront of developing capabilities for FWI based on open-source state-of-the-art SEM software packages and facilitate its adoption by the wider seismic community by improving its numerical efficiency and exploring new inversion strategies. We also apply FWI to innovative datasets, such as ambient-noise and teleseismic body waves, for array imaging. We plan to generate the first high-resolution isotropic and radially anisotropic models for the crust and upper-most mantle of the entire Alaska region based on multi-component ambient-noise FWI. we propose to develop joint ambient-noise and teleseismic-wave FWI method to take advantage of the complementary sensitivity of these two datasets, and apply the joint FWI to high-resolution imaging beneath central California plate boundary region. We also explore source inversion methods based on Greens functions computed for improved 3D regional models by SEM solvers. We propose to develop efficient workflows for near-realtime source mechanism inversions based on the storage of 3D numerical strain-Greens tensor (SGT) database that can be adopted as a regular seismic network operation. These newly developed structural imaging and source inversion capabilities will strength us as one of the leading computational seismology research groups and help train next-generation HQPs equipped with critical skills in parallel computing, large seismic dataset processing, advance imaging and source characterization methods, and machine learning.