Frequency-domain reflection waveform inversion with generalized internal multiple imaging

Abstract

Full-waveform Inversion (FWI) has the potential to provide a high resolution detailed model of the earth’s subsurface, but it often fails to do so if the starting model is far from the true one. Reflection waveform inversion (RWI) is a popular method to build a sufficiently accurate initial model for FWI. In traditional RWI, the low-wavenumber updates are always computed and captured by smearing the data misfit along the reflection path with the help of migration/de-migration. However, the success of the RWI relies heavily on accurately reproducing the data in de-migration. Thus, we introduce a new generalized internal multiple imaging-based RWI implementation (GIMI-RWI), in which we avoid the Born modeling and update the primary reflection kernel directly. In the GIMI-RWI, we store one reflection kernel for each source-receiver pair, preserving the unique wave path for every single source-receiver trace. Subsequently, the convolution between the data residuals and the corresponding reflection kernel can build the tomographic velocity updates. In this situation, the long-wavelength tomographic updates are free of migration footprints, and will contribute a smoother background velocity to reduce the cycle-skipping risk and stabilize the followed full-waveform inversion process. Also, the GIMI-RWI method is source independent, as it entirely relies on the data. Using a synthetic example extracted from the Sigsbee2A model, we show the reliable performance of the GIMI-RWI technique.

Full-Wavefield Migration Using an Imaging Condition of Global Normalization Multi-Order Wavefields: Application to a Synthetic Dataset

“…Subsequently, the least-squares strategy is employed to suppress the crosstalk. The generalized internal multiple imaging (GIMI) method can simulate high-order internal scattering through interference and can subsequently develop least-squares suppression of artifacts [29,30]. Full-wavefield migration (FWM) offers various strategies for combining primary and multiple imaging [31][32][33][34][35] and uses closed-loop one-way wave operators and inversion strategies to simultaneously process data containing surface-related multiples and internal multiples.…”

Multi-Scale Acoustic Velocity Inversion Based on a Convolutional Neural Network

“…Due to the influence of the shallow tomographic component (mainly direct and diving waves), FWI mainly updates the velocity information in the shallow layers, but hardly addresses the deep layers. Reflection-waveform inversion (RWI) [7][8][9][10][11][12][13][14], developed on this basis, updates the deep layer background velocity. However, it is still difficult to update the complex structural features of the deeper layers.…”

Adaptive Feedback Convolutional‐Neural‐Network‐Based High‐Resolution Reflection‐Waveform Inversion

“…Wang et al. (2021) propose a generalized internal multiple imaging‐based RWI (GIMI‐RWI) to convolve the data residuals with the reflection kernel stored for each source‐receiver pair to update the tomographic velocity. Thus, GIMI‐RWI avoids the Born modeling for demigration and is also source independent to update the velocity along the wavepaths.…”

“…Ma et al (2012) propose to utilize image-guided interpolation and its adjoint operator to get a sparse model from the migration image and to constrain the inversion to blocky models, interpolated from the sparse model. Wang et al (2021) propose a generalized internal multiple imaging-based RWI (GIMI-RWI) to convolve the data residuals with the reflection kernel stored for each source-receiver pair to update the tomographic velocity. Thus, GIMI-RWI avoids the Born modeling for demigration and is also source independent to update the velocity along the wavepaths.…”