Full Waveform Inversion Research Articles

Seismic full-waveform inversion regularized with a migration image

Seismic full-waveform inversion (FWI) aims to reconstruct high-resolution velocity models of the subsurface from seismic data. However, it is known to face the problem of ill-posedness and has difficulties in reconstructing deep, low-illumination regions. To overcome these challenges, we use a migration image to regularize the FWI problem. In contrast to conventional approaches where the model regularization is added to the data misfit term, our method uses a U-net to introduce this prior information. We show that the prior information about the structure contained in a migrated image can be naturally and effectively incorporated into the generated output, and therefore use the migration image as input to the U-net. We reparametrize the velocity model as a combination of the U-net output and a background velocity model. The learnable parameters of the U-net are optimized against the FWI data misfit by using the gradient calculated from automatic differentiation. Our method requires no network pre-training and is time efficient due to the use of automatic differentiation accelerated by Graphic Processing Units (GPUs). We implement this method on the synthetic Marmousi and Sigsbee2a models, as well as on field data from the Gulf of Mexico. The results show a regularization effect for mitigating local minima issues, improving convergence speed, and preserving geological structures. Such a regularization effect is well suited to improve velocity reconstruction under poor illumination conditions, such as subsalt structures.

WISER: Multimodal variational inference for full-waveform inversion without dimensionality reduction

We develop a semiamortized variational inference (VI) framework designed for computationally feasible uncertainty quantification in full-waveform inversion to explore the multimodal posterior distribution without dimensionality reduction. The framework is called full-waveform VI via subsurface extensions with refinements (WISER). WISER builds on top of a supervised generative artificial intelligence method that performs approximate amortized inference that is low-cost albeit showing an amortization gap. This gap is closed through nonamortized refinements that make frugal use of wave physics. Case studies illustrate that WISER is capable of full-resolution, computationally feasible, and reliable uncertainty estimates of velocity models and imaged reflectivities.

Radially Anisotropic 3D Velocity Model of the Central Apennines Lithosphere: The CI23 $C{I}_{23}$ Model

AbstractWe present the first application of Full‐Waveform Inversion (FWI) for a radially anisotropic 3D velocity model of the lithosphere beneath central Italy. The retrieved model constrains P‐wave (, ) and S‐wave velocities (, ) in the period range 8–50 s. model correlates well with regional lithological formations and highlights a negative radial anisotropy anomaly localized beneath the epicentral region of the 2016–2017 Amatrice‐Visso‐Norcia (AVN) seismic sequence. An opposite trend (positive radial anisotropy anomaly, i.e. ) is observed in the area affected by the 2009 L’Aquila (AQ) seismic sequence. We interpret the velocity anomalies in terms of local tectonic structures, particularly the Olevano‐Antrodoco‐Sibillini (OAS) thrust—Gran Sasso Thrust (GST) systems, while also considering possible evidence of overpressured fluids and fluid migration. Additionally, the observed anomalies may reflect transient velocity variations induced by the ANV and AQ seismic sequences.

Unbalanced L1 optimal transport for vector valued measures and application to Full Waveform Inversion

Unbalanced L1 optimal transport for vector valued measures and application to Full Waveform Inversion

Unraveling the Mantle Dynamics in Central Asia With Joint Full Waveform Inversion

AbstractWe use the full waveform inversion method to study the crustal‐mantle seismic structure beneath Central Asia. By combining earthquake waveforms and ambient noise cross‐correlations, we construct a 3D model of Vp and Vs down to a depth of 220 km. This model reveals a complex Indian‐Asian plate configuration and interaction, resulting from the plate subduction, indentation, and break‐off. Beneath the Hindu Kush, the marginal Indian slab with its lower crust is successfully imaged, the latter of which hosts vigorous intermediate‐depth seismicity. The subducted marginal Indian slab can be traced further east to the Kohistan Arc, which is a previously undetected structure. We first imaged a flat cratonic Indian plate beneath the Pamir. The indentation of the cratonic Indian plate forces the Asian plate to delaminate, indicated by the south‐eastwards dipping high‐velocity anomalies, atop which a south‐dipping low‐velocity zone is observed with higher resolution than previous studies, which we interpret as the delaminated Asian lower crust. In addition, a sharp velocity transition at lithospheric depth is newly discovered and coincides with the Talas‐Ferghana fault, delineating the boundary of the Ferghana basin with the Central Tian Shan. Low‐velocity anomalies mainly focus beneath the south and northern part of the Central Tian Shan with deep Moho, indicating the lithosphere is possibly delaminated and the deformation of the Central Tian Shan is probably concentrated at the north and south margins by the Tarim basin and Kazakh Shield, respectively. In contrast, West Tian Shan displays a simpler lithospheric structure with a single deep Moho.

GPR Full-Waveform Inversion Through Adaptive Filtering of Model Parameters and Gradients Using CNN

GPR Full-Waveform Inversion Through Adaptive Filtering of Model Parameters and Gradients Using CNN

Initial gradient optimization for elastic full-waveform inversion by means of spectral recomposition

Initial gradient optimization for elastic full-waveform inversion by means of spectral recomposition

Corrigendum to “Elastic full waveform inversion for tilted transverse isotropic media: A multi‐step strategy accounting for a symmetry axis tilt angle”

Corrigendum to “Elastic full waveform inversion for tilted transverse isotropic media: A multi‐step strategy accounting for a symmetry axis tilt angle”

Detection of moisture content profile of concrete using ground penetrating radar

Abstract Moisture content in concrete is essential, as it influences durability, strength, and cracking susceptibility, critical for structural longevity and safety. This paper aims to explore the potential of GPR in detecting the distribution of moisture content in concrete. By utilizing the approximate linear relationship between the dielectric constant and the moisture content of concrete, this study realizes the detection of concrete moisture content profile by GPR. The concrete moisture content profile is derived from radar echo data using full waveform inversion (FWI). The simulations show that different initial values converge to the true values with a maximum error of 0.33%, demonstrating the feasibility of the method. Experimental results demonstrate a strong correlation between GPR measurements and the gravimetric method, with an average error of 2.5%, affirming the accuracy of GPR in detecting moisture content profiles for engineering applications.

Optimal transport-based full-waveform inversion for shallow seismic data

Optimal transport-based full-waveform inversion for shallow seismic data

Retraction Note: Imaging Enhancement of Gas Reservoir via Full-Waveform Inversion Imaging: Case Study, Gas Field

Retraction Note: Imaging Enhancement of Gas Reservoir via Full-Waveform Inversion Imaging: Case Study, Gas Field

Gabor-wavelet-activation implicit neural learning for full waveform inversion

Seismic full waveform inversion (FWI) is an efficient imaging method for estimating subsurface physical parameters. However, when the initial models are inaccurate or seismic data lack low frequencies, most traditional discrete grid-based FWI algorithms often encounter local minimum problems. Additionally, the inversion performance and computational cost are closely related to spatial resolution. Reparameterizing velocity models using neural networks (NNs) effectively mitigates local minimum issues. However, most NN-FWI approaches perform well when overparameterized and are limited to fixed grid-like representations. In contrast, implicit neural learning (INL) enables the representation of models at any resolution using a multilayer perception (MLP) network that learns a continuous function from discrete coordinates. To enhance the capability of INL, we developed a general Gabor-wavelet-activation INL approach and applied it to FWI, referred to as wavelet implicit neural learning FWI (WinFWI), using only the vertical particle-velocity. The constructed MLP was lightweight, which reduced the computational overhead. Numerical experiments on a synthetic block model and Marmousi2 model demonstrate the robustness of the proposed method to challenges such as parameter crosstalk. This was further validated using the Chevron 2014 blind test data. All comparisons show that our method is more general and robust than traditional and INL-based FWI algorithms. Moreover, additional feature visualization and numerical analysis illustrate the potential advantages of WinFWI in balancing computational cost and inversion accuracy.

Time-lapse full-waveform inversion with a new gradient-weighting strategy applied to marine seismic data

Time-lapse full-waveform inversion (FWI) has the potential to reveal high-resolution reservoir production-related velocity changes. However, time-lapse FWI can introduce inversion artifacts, which might mask the true time-lapse signature within the reservoir zone, especially for the inversion of field data examples. We implement sequential bootstrap time-lapse FWI on marine seismic data from North West Australia and observe noticeable inversion artifacts. To better understand these time-lapse FWI artifacts, we implement time-lapse FWI using seismic data with different frequency bands and observe the distribution of the potential true time-lapse signature and inversion artifacts. On the inverted [Formula: see text] (P-wave velocity change) models using data with different frequency bands, the distribution of inversion artifacts exhibits inconsistencies, whereas the inverted [Formula: see text] at the reservoir remains consistent. To comprehend these observations, we analyze the origin of these time-lapse artifacts from a theoretical perspective and find that a portion of these artifacts may arise from inversion null space and differences in data residuals between the baseline and monitoring inversions. Building on these insights, we develop a novel time-lapse FWI method to suppress these inversion artifacts. We use the energy of the inverted [Formula: see text] using data with a lower dominant frequency band as a gradient-weighting term for time-lapse FWI using data with a higher dominant frequency band. The test of the novel time-lapse FWI method on marine data demonstrates its capability to suppress inversion artifacts effectively. The inverted [Formula: see text] primarily reflects the 4D response, which is interpreted to be the softening of the reservoir caused by gas coming out of the solution within the oil column due to a pressure drop within the reservoir. To validate this data-driven research solution, we conduct tests on the SEG Advanced Modeling 4D model and synthetic 4D data for which we know the true underlying earth models and time-lapse reservoir changes.

Full-Waveform Inversion with Denoising Priors Based on Graph Space Sinkhorn Distance

Full-Waveform Inversion with Denoising Priors Based on Graph Space Sinkhorn Distance

Frequency-differencing strategy to kickstart full-waveform inversion without cycle skipping.

Ultrasound tomography fundamentally relies on low-frequency data to avoid cycle skipping in full-waveform inversion (FWI). In the absence of sufficiently low-frequency data, we can extrapolate low-frequency content from existing high-frequency signals by using the same approach used in frequency-difference beamforming. This low-frequency content is then used to kickstart FWI and avoid cycle skipping at higher frequencies. In simulations, the structural similarity index measure and peak signal-to-noise ratio of the reconstructed image improve by 0.28 and 8.6 dB, respectively, as a result of frequency differencing. Experiments show that internal structures can be seen with greater clarity because of frequency differencing.

Obtaining an initial model for acoustic full waveform inversion using generalized regression neural networks

Obtaining an initial model for acoustic full waveform inversion using generalized regression neural networks

Full Waveform Inversion Method for Roadways Based on Wave Velocity Structure Correction and Regularization Constraints

Full Waveform Inversion Method for Roadways Based on Wave Velocity Structure Correction and Regularization Constraints

A Full-Waveform Inversion Method Based on Structural Tensor Constraints

A Full-Waveform Inversion Method Based on Structural Tensor Constraints

A Feature-Boosted Convolutional Neural Network for Full-Waveform Inversion

A Feature-Boosted Convolutional Neural Network for Full-Waveform Inversion

Acoustic anisotropic time-lapse full-waveform inversion based on elliptical anisotropic assumption at the Sleipner site

Acoustic anisotropic time-lapse full-waveform inversion based on elliptical anisotropic assumption at the Sleipner site