This issue of Geophysics Free

Yin et al. introduce the WISE framework for cost-effective Bayesian variational inference for full-waveform inversion (FWI). WISE integrates generative AI technologies with physics-informed common-image gathers to quantify uncertainty in migration-velocity models and its impact on imaging.

Kouadio et al. describe how the sewage diversion strategy, blending the audiofrequency magnetotelluric method with geologic knowledge and data on waste distribution, helps pinpoint the best spots for water treatment facilities in mining regions. This approach ensures the protection of groundwater from pollution, playing a crucial role in preserving vital water supplies.

Zhang et al. describe the process of predicting reservoir lithology, petrophysical property, and gas bearing by multicomponent seismic data. The authors show that multicomponent seismic has advantages over PP seismic in tight gas reservoir property prediction.

Liu et al. show how a single-factor experiment method or numerical simulation method is not suitable for the analysis of the main controlling factors of breakdown pressure. The rock sound velocity, horizontal minimum principal stress, and fracture density are the main controlling factors of the breakdown pressure.

Taweesintananon et al. use frequencies of the S-wave resonances measured by ocean-bottom fiber-optic distributed acoustic sensing (DAS) to compute S-wave velocity models of the submarine near-surface low-velocity sedimentary layers overlying the hard-rock basement in offshore Svalbard, Norway. The lateral variation of the estimated S-wave velocity models can be used to study the distribution of glacigenic sediments and landforms deposited in the survey area.

Koyan and Tronicke develop and apply a workflow to analyze and interpret 3D ground-penetrating radar (GPR) data using the gradient structure tensor (GST), a method previously applied to only a few 2D GPR examples. Applying this workflow to synthetic and field data illustrates the efficient calculation and application of the GST to analyze and interpret modern 3D data sets.

Hao and Tsvankin find two conditions expressed in terms of the Thomsen-type parameters, which can make attenuative transversely isotropic (TI) media with a vertical symmetry axis (VTI) become elliptical. The authors use analytical and numerical methods to study P-wave propagation in attenuative elliptical VTI media.

Traveltime inversion, when applied at a single gather of seismic reflection data, is inherently ambiguous. Rommel develops an algorithm computing an infinite series of subsurface (anisotropic) velocity/elasticity models, all of which generate almost the same kinematic seismic response.

Hao and Tsvankin derive the fundamental equations of wavefield decomposition for viscoelastic anisotropic media described by a general dissipative model. It is numerically demonstrated that their approach can accurately separate wave modes in constant-Q viscoelastic VTI media.

Chao et al. describe research on high-power and high-efficiency emission of crosswell electromagnetic logging, especially through casing. Experiments associated with the research are performed in different types of wells.

Qi et al. carry out physical experiments and finite-difference numerical simulation studies on scaled physical models of wells based on acoustic logging techniques. The authors meticulously examine the interplay between different fracture attributes and the behavior of shear and Stoneley waves, providing a quantitative analysis of fracture width, density, and ductility.

Zhong and Wang develop a method that uses prior information from natural images to inpaint geologic logging images, ensuring stratigraphic continuity and providing complete features of fractures and textures. Their method achieves reliable and robust stratigraphic feature recovery in various borehole images.

Ji and Wang reveal that four key factors primarily influence S-wave excitation in slow formations for monopole logging while drilling. The authors propose that tool eccentricity, especially large eccentricity, can amplify S waves and improve their velocity measurement accuracy.

Szabó presents a novel well-log analysis method for oilfield applications, which inverts the wireline logs using a hyperparameter estimation method to estimate the volumetric and zone parameters automatedly. The author extracts the zone parameters solely from well logs to improve the efficiency of inversion and reduce the cost of laboratory measurements.

Ma et al. propose a fracture-imaging workflow by applying prestack Kirchhoff migration to coherent reflections identified on DAS-recorded microseismic data. This framework produces a reliable 3D characterization of fracture and fault lineaments and reveals fluid from the injection well in the reservoir formation.

Qi et al. show that the presence of a sonic logging tool has a significant influence on the frequency-dependent slowness and attenuation of dipole-flexural waves in a fluid-filled borehole. The authors propose a workflow of inverting the shear slowness and attenuation from dipole acoustic log data based on an equivalent tool theory.

Zhang et al. describe how frequency-domain electromagnetic (FDEM) is applicable for the detection of blockages in an insulated pipeline, in particular, when the single-ended charging method is used and the blockage has good impermeability. Owing to the conductivity of a metal pipeline, the electrical current will mainly propagate along the pipeline rather than through the fluid or blockage, so the effect of FDEM is not obvious in locating metal blockages in pipelines.

Li et al. develop an adaptive laterally constrained inversion method that can generate data-driven constraints for time-domain electromagnetic inversion. The method recovers quasilayered models with smooth lateral transitions while preserving sharp lateral interfaces supported by the available data.

Beloborodov et al. introduce a rock-physics machine-learning workflow that automates the classification of lithofluid types and models property depth trends from wireline log data, addressing the challenges of manual extraction and traditional clustering methods’ limitations due to depth-related changes in petrophysical properties. This workflow uses a maximum-likelihood approach with the expectation-maximization algorithm to link lithofluid types to specific depth-dependent rock-physics models and simultaneously estimate model parameters and associated uncertainties.

Lu et al. describe an emerging method called surface geometry inversion for the inversion of transient electromagnetic data. The inversion method can directly invert for the geometrical shape of an anomaly and its boundaries.

Ma et al. conduct electrical resistivity tomography (ERT) and induced polarization (IP) surveys to characterize landfill extent, composition, and biogeochemical activity under different geomembrane covers; a new method using synthetic landfill models of differing geomembranes was used to evaluate data reliability, which was not extensively investigated in prior studies. Findings indicate that reliable ERT and IP data obtained by ensuring adequate signal transmission paths can find landfill boundaries, capture waste distribution, and microbial degradation activity.

Li and Wang conduct a study exploring the impact of base frequency, duty cycle, and waveform repetition on transient electromagnetic responses through a 3D finite-element method applied to models of a deep-buried conductor. The numerical results show that there is a significant improvement in the detection and discrimination capability of the magnetic field for perfect conductors with a conductivity of 1000 S/m or higher as the waveform period increases, and a novel equivalence phenomenon was uncovered in the magnetic field and its time derivative data.

Qin et al. implement 3D forward modeling for the borehole-to-surface electromagnetic method using a high-order finite-element method and octree meshes. Numerical examples show excellent accuracy and highlight the ability of the method to handle complex geologic structures.

Liu et al. propose a residual convolutional neural network to invert the submarine cable position using magnetic data. The proposed method is superior to the conventional Euler method in the estimation of submarine cable positions.

You and Li develop a U-shaped transformer that enhances facies classification in Brazilian presalt acoustic image logs through transfer and confidence learning. This deep-learning approach achieves an impressive F1 score of 0.90 on test data, facilitating immediate and accurate facies analysis from image logs and substantially improving reservoir characterization.

Introduce DeepNRMS, an innovative unsupervised deep-learning approach for noise-robust CO₂ monitoring in seismic imaging, significantly enhancing detection accuracy amidst time-lapse noise by leveraging preinjection data. Through evaluations with synthetic and field data from the Aquistore project, DeepNRMS demonstrates its potential for reliable and cost-effective subsurface CO₂ monitoring, underscoring its applicability in environmental surveillance.

Sharifi proposes a novel 3D prism-shaped template for lithology and fluid discrimination, called a prism-shaped rock-physics template. Some novel triangular and rectangular ternary and binary templates were designed and connected to build a prism for fluid discrimination and lithology identification; the obtained templates were successfully validated on blind data sets (e.g., ultrasonic, well logging, and seismic data) in different reservoirs with various lithologies and fluid types.

Wang et al. describe how strong directional sources influence ambient seismic imaging using traffic-induced noise. The authors propose a new parallel observation system that can effectively mitigate this effect, yielding more consistent imaging results.

Zhao et al. describe the use of the flow-field fitting method to investigate the leakage of concrete gravity dams. Compared with traditional geophysical methods, this method successfully detects the leakage locations of the dam body, dam foundation, and reservoir area.

Trafford et al. propose an approach, using DAS of fiber-optic cables, for providing near-surface seismic geophysical information across considerable spatial scales (multikilometer) at high resolution, which is beneficial for the design of subsea cable routes and landfall locations. The authors consider appropriate selection and directional sensitivity of fiber-optic cables and validate the measured data with respect to conventional seismic data acquisition approaches along with independent borehole and seismic cone penetration test data.

Shragge et al. use a dip-based method to separate transmitted and backscattered surface waves to extend an elastic time-reverse imaging (E-TRI) framework originally developed for elastic microseismic imaging applications to the analysis of backscattered surface-wave data, yielding high-resolution images of laterally discontinuous velocity structure. The comparison of E-TRI results with the FWI S-wave model suggests a strong correlation in the location of near-surface short-wavelength structure.

Xu et al. propose a 3D sequence-to-sequence (STS) prediction paradigm to enable machine learning to extract 3D seismic data features for porosity prediction. Building upon the 3D STS prediction model, three strategies from different perspectives, including data augmentation, spatial constraints, and geologic constraints, are introduced to further enhance the performance of seismic porosity estimation.

Chi et al. study the permeability anisotropy of conglomeratic sandstone using a multiscale digital rock fusion method. The anisotropy arises from the layering and directionality of the grains.

Yang and Chen describe a rock-physics model of organic-rich shale, in which they consider the continuous process of thermal maturity. The authors apply this model to investigate how sweet spot parameters (thermal maturity, total organic carbon content, and clay content), overpressure, and diagenesis affect the overall elastic properties and anisotropy of shale during thermal maturation.

Fei et al. introduce an innovative neural network that directly disentangles and characterizes individual geologic factors from seismic data, overcoming the challenges posed by the intertwined nature of traditional seismic attributes. Their method, validated with synthetic and field data, outperforms established approaches by providing clearer geophysical insights into seismic attribute extraction.

Zappalá et al. present an innovative seismic acquisition system in onshore and transition zone environments, applying a total of four recording sensors’ typologies in a carbon capture and storage interest area. The authors improve the land sensor processing with respect to previous studies and apply it to the new transition zone acquisition setup.

By combining waveform analysis with frequency-wavenumber spectra to distinguish between body waves and surface waves, Jin et al. introduce a technique that significantly enhances the interpretation accuracy and signal-to-noise ratio of passive seismic data. It not only improves the quality of subsurface imaging but also serves as a validation tool, aiding in the identification and elimination of misleading geologic interpretations.

Bustamante et al. present a multiinput neural network to estimate seismic velocities from marine seismic data collected on subsea permafrost zones. The authors show that deep-learning-based seismic inversion could become a cost-effective technology to map, on a large scale, the lateral and vertical distribution of subsea permafrost and, more generally, high-velocity rocks in shallow water.

Liu et al. develop a Laplace-domain source encoding strategy for elastic FWI with time-domain solvers that significantly reduces elastic FWIs computation and input/output time by an impressive 80 times for 8 s recording time at a frequency band of 10 Hz. The proposed Laplace-domain approach empowers source-encoded elastic FWI to mitigate the risk of getting stuck in a local minimum.

Saadallah and Buland present a new Bayesian joint residual moveout (RMO) and amplitude-variation-with-offset (AVO) inversion method that estimates the optimal AVO parameters and corrects the seismic data for RMO simultaneously. The inversion results include uncertainty, which is crucial given the nonuniqueness of the inverse problem.

Lin et al. present a new joint inversion method based on Bayesian inference and the assumption of a Gaussian distribution for processing vertical cable seismic data. The authors first accurately reconstruct the primary and multiple wavefields, obtaining the corresponding reflectivity updates, then approximate the model-domain Hessian matrices with their diagonal components, and finally approximate the covariance matrix to reduce computational costs.

Heir et al. introduce the stratigraphy-guided deep-learning method, a seismic inversion method that accounts for the discrete nature of stratigraphic units. The authors show how the method estimates properties more reliably than traditional seismic inversion methods and alternative deep-learning methods.

Fu et al. propose data-driven double-focusing resolution analyses to measure the resolution characteristics of migrated images by incorporating weighted focal beams into the prestack migration process. Numerical experiments with a wedge model and field data show the performance of the data-driven method, demonstrating the effects of propagation attenuation, incorrect velocity, and noise contamination.

Liu et al. present a new technique called adaptive merging migration (AMM) to improve the migration quality in the presence of velocity errors and data noise. Numerical tests on two complex synthetic data sets and one field data set validate that AMM can effectively improve the image quality in the presence of different types of velocity errors and data noise.

Xie et al. propose an elastic reverse time migration (RTM) based on first-order velocity-dilatation-rotation equations using the optical flow vector. The proposed method can better eliminate the migration artifacts and improve the imaging accuracy of the elastic RTM than conventional methods, resulting in more accurate wavefield decomposition and better S-wave polarity reversal correction.

Zhang et al. propose a method for calculating the optical flow vector of elastic waves in TI media to obtain the propagation direction. The optical flow vector can overcome the Poynting vector limitations, get more accurate and reliable information regarding the direction of elastic wave propagation in TI media, and precisely separate the wavefields.

Ma et al. present two cost-efficient parallel iterative solvers for large-scale 3D frequency-domain multisource seismic wave modeling in complex viscoelastic anisotropic media. These solvers offer performance comparable to direct solvers but with lower memory requirements, making them viable options for scenarios with limited computational resources.

Wang et al. investigate the seismic wave propagation across rock masses with thin-layer joints. The authors propose a modified displacement discontinuity method by introducing effective joint stiffness to accurately predict the effect of joint thickness on seismic wave propagation.

Zhang et al. derive a new viscoacoustic wave equation with decoupled fractional Laplacians to simulate the power-law frequency-dependent Q effect. Numerical analysis and experiments demonstrate the performance of the proposed method in power-law frequency-dependent Q simulations.

Caunt et al. present a generalized immersed boundary method for including topography in finite-difference acoustic wavefield simulations. This approach enables accurate representation of irregular topography in seismic simulations and imaging workflows without the need for irregular or curvilinear meshes.

Yang et al. derive stress-dependent reflection and transmission of elastic waves under confining, uniaxial, and pure-shear prestresses from the theory of acoustoelasticity with third-order acoustoelastic constants. Prestress effects on the critical angles, gradients, converted waves, and reflection and transmission energy ratios provide new insights for monitoring geopressure and tectonic stress using sound-logging data or seismic data.

Chang et al. present the deep Lax-Wendroff correction (DeepLWC) method, a deep-learning-based numerical format for solving 2D hyperbolic wave equations, and the numerical results indicate that the DeepLWC significantly improves calculation speed (by more than 10 times) and reduces storage space by more than 10,000 times compared to traditional numerical methods. DeepLWC introduces a novel research paradigm for numerical equation solving, which can be combined with various traditional numerical methods, enabling acceleration and reduction in storage requirements of conventional approaches.

Wang et al. incorporate the high-frequency information obtained from downsampling the data into multitrace deconvolution, aiming to constrain high-frequency information in the results that exceed the initial frequency bandwidth.

Brandolin et al. propose a physics-informed neural network framework that uses the local plane-wave differential equation to reconstruct the seismic data while simultaneously estimating the local slopes. Results on synthetic and field data validate the effectiveness of the proposed method in handling aliased (coarsely sampled) data and data with large gaps.

Harsuko and Alkhalifah propose a unique deep-learning-based seismic processing workflow based on an optimized transformer network. The workflow is demonstrated on a sequential processing of realistic Marmousi and challenging real field data.

Saad et al. propose an unsupervised framework to simultaneously interpolate and denoise 5D seismic data using a deep-learning model and the projection onto convex sets method. The proposed framework shows a robust reconstruction performance by reconstruction of 5D seismic data with an 80% missing rate.