Seismic Source Inversion Using Hamiltonian Monte Carlo and a 3D Earth Model for the Japanese Islands

Date: 4/25/2019

Time: 09:15 AM

Room: Cascade I

We present a database of full-waveform seismic source solutions for the Japanese Islands. Our method is based on the Bayesian inference of source parameters and a tomographically derived 3-D Earth model, used to compute Green’s strain tensor. With this approach, we infer moment tensor, location and timing of the seismic events.

To compute spatial derivatives of Green’s functions, we use a previously derived regional Earth model. The model is radially anisotropic, visco-elastic, and fully heterogeneous. It was constructed using full waveforms in the period band of 15–80 s.

Green’s strains are computed numerically with the spectral-element solver SES3D. We exploit reciprocity, and by treating seismic stations as virtual sources we compute and store the wavefield across the domain. This gives us a strain database for all potential source-receiver pairs. The displacements are then promptly obtained by linear combination of the pre-computed strains scaled by the moment tensor elements.

We infer ten model parameters – six moment tensors elements, three location parameters, and the origin time of the event. A feasible number of model parameters and the fast forward problem allow us to infer the unknowns using a Bayesian approach. The sampling is performed with a variant of Hamiltonian Monte Carlo (HMC) algorithm, which we developed previously.

HMC takes advantage of the derivatives of synthetic data with respect to the model parameters. Therefore, it converges to the posterior probability density with fewer samples compared to the derivative-free Metropolis-Hastings algorithm. Advantages of HMC become more prominent in the case of weak prior knowledge, high-quality data, and increasingly empty high-dimensional space.

We apply our method to the Izu-Bonin trench, where many events have a non-double-couple component. We expect to shed more light on the events in complicated tectonic settings by i) taking into account the complexity of the medium, ii) exploiting full-waveform information and iii) presenting the uncertainties and trade-offs between the source parameters.

Presenting Author: Saule Simute

Authors