Rapid 3D Green's Functions Using Reduced-Order Models of Physics-Based Seismic Wave Propagation Simulations
Description:
Elastodynamic Green’s functions are essential in seismology, and form a connection between direct observations of seismic waves and earthquake sources. They are key to enabling various seismological tasks, including physics-based ground motion prediction and kinematic or dynamic source inversions. In regions with comparably well-constrained 3D models of the Earth's elastic structure, such as Southern California, approximate 3D Green's functions can be computed using physics-based numerical simulations of seismic wave propagation. However, these simulations are computationally expensive, which presents a challenge for real-time ground motion prediction (e.g., ShakeMap and ShakeAlert), physics-based probabilistic seismic hazard assessment (PSHA, e.g., CyberShake), and uncertainty quantification in source inversions. Here, we address this challenge by using a reduced-order model (ROM) which enables rapid computation of approximate Green’s functions by using the proper orthogonal decomposition combined with radial basis function interpolation. We train the ROM using three-component seismograms for six elementary moment tensors, computed with SeisSol, selecting 500 source locations to calculate 1.0 Hz elastodynamic Green’s functions for approximately 10,000 sites in Southern California. Using leave-one-out cross-validation, we assess the accuracy of our Green’s functions for the SCEC CVM-S4.26-M01 velocity model in both the time domain and frequency domain. We show that the ROM can accurately and rapidly reproduce simulated seismograms for generalized moment tensor sources in our 3D region, as well as kinematic sources by using a finite fault model of the 1987 Mw 5.9 Whittier Narrows earthquake as an example. In these demonstrators, the accuracy is quantified using the mean absolute error of the velocity waveforms and the Fourier amplitude spectra. We envision that our rapid Green’s functions would be useful for physics-based PSHA, improving the uncertainty quantification in earthquake source inversions, and constructing rapid ShakeMovies with high spatial resolution.
Session: 3D Wavefield Simulations: From Seismic Imaging to Ground Motion Modelling [Poster Session]
Type: Poster
Date: 5/2/2024
Presentation Time: 08:00 AM (local time)
Presenting Author: John
Student Presenter: Yes
Invited Presentation:
Authors
John Rekoske Presenting Author Corresponding Author jrekoske@ucsd.edu University of California, San Diego |
Dave May dmay@ucsd.edu University of California, San Diego |
Alice-Agnes Gabriel algabriel@ucsd.edu University of California, San Diego |
|
|
|
|
|
|
Rapid 3D Green's Functions Using Reduced-Order Models of Physics-Based Seismic Wave Propagation Simulations
Session
3D Wavefield Simulations: From Seismic Imaging to Ground Motion Modelling