Fourth-Order Staggered-Grid Finite-Difference Seismic Wavefield Estimation Using a Discontinuous Mesh Interface (WEDMI)

TitleFourth-Order Staggered-Grid Finite-Difference Seismic Wavefield Estimation Using a Discontinuous Mesh Interface (WEDMI)
Publication TypeJournal Article
Year of Publication2017
AuthorsNie S.Y, Wang Y.F, Olsen K.B, Day S.M
JournalBulletin of the Seismological Society of America
Volume107
Pagination2183-2193
Date Published2017/10
Type of ArticleArticle
ISBN Number0037-1106
Accession NumberWOS:000412920500016
Keywordsbasins; boundary-condition; model; motion; san-andreas fault; simulation; southern california
Abstract

In a realistic geological structure with a large contrast in seismic wavespeed between shallow and deep regions, simulation of seismic wave propagation using a spatially uniform grid can be computationally very demanding, due to over-discretization of the high-speed material. Thus, numerical methods that allow for coarser discretization of the faster regions have the potential to be much more efficient. Discontinuous mesh (DM) methods, operating by exchanging wavefield information between media partitions discretized with two different grid spacings, provide a convenient way to improve such efficiency issues. Unfortunately, discontinuous staggered-grid finite-difference (FD) methods typically suffer from inherent stability problems, in particular in strongly heterogeneous media, arising from numerical noise generated at the overlap of the two regions with different grid spacing. We have developed a 3D fourth-order velocity-stress staggered-grid FD DM anelastic wave propagation method (AWP-DM) for seismic wavefield estimation using a discontinuous mesh interface (WEDMI) between fine and coarse meshes. Benchmarks in models with realistic 3D velocity variations and finite-fault sources across the grid interface show stable results for a number of timesteps, exceeding the need dictated by current high-frequency ground-motion simulations. In the case of a factor-of-three ratio between the coarse and fine grid sizes, this method is capable of producing a level of accuracy comparable to that from the uniform fine-grid scheme, using at least 7-8 grid points per minimum S wavelength inside the mesh overlap zone.

DOI10.1785/0120170077
Student Publication: 
No
sharknado