Abstract:Abstract Precisely constraining the source parameters of large earthquakes is one of the primary objectives of seismology. However, the quality of the results relies on the quality of synthetic earth response. Although earth structure is laterally heterogeneous, particularly at shallow depth, most earthquake source studies at the global scale rely on the Green's functions calculated with radially symmetric (1D) earth structure. To avoid the impact of inaccurate Green's functions, these conventional source studies use a limited set of seismic phases, such as long period seismic waves, broadband P and S waves in teleseismic distances (30 degrees < ∆ < 90 degrees), and strong ground motion records at close-fault stations. The enriched information embedded in the broadband seismograms recorded by global and regional networks is largely ignored, limiting the spatiotemporal resolution. Here we calculate 3D strain Green's functions at 30 GSN stations for source regions of 10 selected global earthquakes, with frequency up to 67 mill hertz (15 s), using SPECFEM3D_GLOBE and the reciprocity theorem. The 3D SEM mesh model is composed of mantle model S40RTS, crustal model CRUST2.0, and surface topography ETOPO2. We surround each target event with grids in horizontal spacing of 5 km and vertical spacing of 2.0∼3.0 km, allowing us to investigate not only the mainshock but also the background seismicity. In total, the response at over 210,000 source points is calculated in simulation. The number of earthquakes, including different focal mechanisms, centroid depth range, and tectonic background, could further increase without additional computational cost if they were properly selected to avoid overloading individual CPUs. The storage requirement can be reduced by two orders of magnitude if the output strain Green's functions are stored for periods over 15 s. We quantitatively evaluate the quality of these 3D synthetic seismograms, which are frequency and phase dependent, for each source region using nearby aftershocks, before using them to constrain the focal mechanisms and slip distribution. Case studies show that using a 3D earth model significantly improves the waveform similarity, agreement in amplitude and arrival time of seismic phases with the observations. The limitations of current 3D models are still notable, dependent on seismic phases and frequency range. The 3D synthetic seismograms cannot well match the high frequency (>40mHz) S wave and (>20mHz) Rayleigh wave yet. Though the mean time-shifts are close to zero, the standard deviations are notable. Careful calibration using the records of nearby better located earthquakes is still recommended to take full advantage of better waveform similarity due to the use of 3D models. Our results indicate that it is now feasible to systematically study global large earthquakes using full 3D earth response in a global scale.

Synthetic ground motions in heterogeneous geologies from various sources: the HEMEWS-3D database

Machine learning opportunities to conduct high-fidelity earthquake simulations in multi-scale heterogeneous geology

Improvement of Stochastic Green's Function Method for 3D Broadband Ground-Motion Simulation

SEM3D: A 3D High-Fidelity Numerical Earthquake Simulator for Broadband (0–10 Hz) Seismic Response Prediction at a Regional Scale

Regional seismic ground‐motion simulation and observation with engineering applications

Fourier Neural Operator Surrogate Model to Predict 3D Seismic Waves Propagation

Source parameters inversion of the global large earthquakes using 3D SEM Green's functions: Strain Green's function calculation and validation

Simulating Multicomponent Elastic Seismic Wavefield Using Deep Learning

Learning Physics for Unveiling Hidden Earthquake Ground Motions via Conditional Generative Modeling

High Resolution Seismic Waveform Generation using Denoising Diffusion

Multiple-Input Fourier Neural Operator (MIFNO) for source-dependent 3D elastodynamics

Predicting the seismic ground-motion parameters: 3D physics-based numerical simulations combined with artificial neural networks

Deep-learning seismology

Local earthquakes detection: A benchmark dataset of 3-component seismograms built on a global scale

Recent Advances in Machine Learning-Enhanced Joint Inversion of Seismic and Electromagnetic Data

Validation of (not‐historical) large‐event near‐fault ground‐motion simulations for use in civil engineering applications

Deep-neural-network model for predicting ground motion parameters using earthquake horizontal-to-vertical spectral ratios

Wavelet-Based Stochastic Model for Jointly Simulating Three-Component Ground Motions

An Unsupervised Machine Learning Approach for Ground-Motion Spectra Clustering and Selection

Recent advances in earthquake seismology using machine learning

Deep learning for geological hazards analysis: Data, models, applications, and opportunities