Skip to Main content Skip to Navigation
Journal articles

Bessel smoothing filter for spectral-element mesh

Abstract : Smoothing filters are extremely important tools in seismic imaging and inversion, such as for traveltime tomography, migration and waveform inversion. For efficiency, and as they can be used a number of times during inversion, it is important that these filters can easily incorporate prior information on the geological structure of the investigated medium, through variable coherent lengths and orientation. In this study, we promote the use of the Bessel filter to achieve these purposes. Instead of considering the direct application of the filter, we demonstrate that we can rely on the equation associated with its inverse filter, which amounts to the solution of an elliptic partial differential equation. This enhances the efficiency of the filter application, and also its flexibility. We apply this strategy within a spectral-element-based elastic full waveform inversion framework. Taking advantage of this formulation, we apply the Bessel filter by solving the associated partial differential equation directly on the spectral-element mesh through the standard weak formulation. This avoids cumbersome projection operators between the spectral-element mesh and a regular Cartesian grid, or expensive explicit windowed convolution on the finite-element mesh, which is often used for applying smoothing operators. The associated linear system is solved efficiently through a parallel conjugate gradient algorithm, in which the matrix vector product is factorized and highly optimized with vectorized computation. Significant scaling behaviour is obtained when comparing this strategy with the explicit convolution method. The theoretical numerical complexity of this approach increases linearly with the coherent length, whereas a sublinear relationship is observed practically. Numerical illustrations are provided here for schematic examples, and for a more realistic elastic full waveform inversion gradient smoothing on the SEAM II benchmark model. These examples illustrate well the efficiency and flexibility of the approach proposed.
Complete list of metadata

https://hal.archives-ouvertes.fr/hal-02009705
Contributor : Ludovic Métivier Connect in order to contact the contributor
Submitted on : Thursday, October 14, 2021 - 12:47:03 PM
Last modification on : Wednesday, November 3, 2021 - 7:49:02 AM

File

ggx103.pdf
Publisher files allowed on an open archive

Licence


Distributed under a Creative Commons Attribution 4.0 International License

Identifiers

Citation

Phuong-Thu Trinh, Romain Brossier, Ludovic Métivier, Jean Virieux, Paul Wellington. Bessel smoothing filter for spectral-element mesh. Geophysical Journal International, Oxford University Press (OUP), 2017, 209 (3), pp.1489-1512. ⟨10.1093/gji/ggx103⟩. ⟨hal-02009705⟩

Share

Metrics

Record views

592

Files downloads

94