Geophysical Prospecting - Volume 72, Issue 7, 2024

Multicomponent seismic technology utilizes the kinematic and dynamic characteristics of reflected P‐waves and converted S‐waves to reduce ambiguity in seismic exploration. The imaging and inversion accuracy of P‐SV‐converted waves are important in determining whether multicomponent seismic exploration can achieve higher exploration accuracy than conventional P‐wave exploration. Pre‐stack inversion of P‐SV‐converted waves requires precise input of P‐SV‐converted wave angle‐domain common‐image gathers. Consequently, the P‐SV‐converted wave angle‐domain common‐image gather extraction accuracy will significantly affect the P‐SV‐converted wave inversion accuracy. However, existing methods for extracting P‐SV‐converted wave angle‐domain common‐image gathers are constrained by issues such as the P‐ and S‐wave crosstalk artefacts, low‐frequency noises and inaccurate calculation of P‐wave incident angles, leading to poor imaging accuracy. We study an angle‐domain cross‐correlation imaging condition and address three key issues based on this condition: the decoupling of P‐ and S‐waves, the separation of up‐going and down‐going waves and the precise calculation of P‐wave incident angles. Our strategies facilitate high‐precision extraction of P‐SV‐converted wave angle‐domain common‐image gathers using elastic wave reverse‐time migration. In this paper, first, we employ the first‐order velocity‐dilatation‐rotation elastic wave equations to decouple P‐ and S‐waves automatically during source and receiver wavefield extrapolations. Second, we calculate the optical flow vectors of P‐ and S‐waves to ensure stable calculations of wave propagation directions. Based on this, we obtain up‐going and down‐going waves of P‐ and S‐waves. Meanwhile, we calculate the incident angle of the source P‐wave using geometric relations. Lastly, we apply the angle‐domain imaging condition to achieve high‐precision extraction of P‐SV‐converted wave angle‐domain common‐image gathers. Model examples demonstrate the effectiveness and advantages of the proposed method.

Transversely isotropic media with a tilted symmetry axis (TTI) exits widely underground due to tectonic movement and mineral orientation. Traditional full waveform inversion (FWI) based on isotropic media or transversely isotropic media with a vertical symmetry axis (VTI) cannot deal with such situations. To address this limitation, TTI–based FWI was developed. However, its practical application faces challenges in estimating the symmetry axis tilt angle . Previous studies have generally assumed that is equal to the strata dip angle, which is incorrect in complex structures such as salt domes and magmatic intrusion zones. Another theoretically robust way to estimate is to treat it as the parameter to be inverted, but there are still some problems unresolved. First, the parameter increases the nonlinearity of the inversion process, and its impact mechanism on inversion is not yet clear. Second, there is severe crosstalk (also known as trade‐off or coupling) between parameters, but the current parameter decoupling technique for TTI–based FWI is not mature. To address the first problem, we assess the interaction between and other parameters by analysing the radiation patterns in the TTI background. Our analysis reveals that is most coupled by S‐wave vertical velocity , and substantially affects anisotropic parameters and . Therefore, we conclude that a good inversion of velocity parameters is a prerequisite for recovering , and only after recovers can and be recovered. This conclusion provides a natural perspective for solving the second problem. We therefore propose a multi‐step and multi‐parameter joint inversion strategy to gradually improve the quality of parameter inversion and weaken their coupling. Numerical experiments demonstrate that our strategy achieves more accurate inversion results compared to previously proposed multi‐parameter inversion strategies. Finally, the application to the field OBN dataset acquired from the South China Sea verifies the practicality of our method.

We reprocessed a 50‐km long legacy 2D reflection seismic profile acquired in 1986, under a project funded by the National Geophysics Program to improve the delineation of complex geological structures that host the platinum‐bearing horizons (known as the UG2 reef; a chromitite horizon) on the south‐eastern edge of the Western limb of the Bushveld Complex and to investigate the continuity of the reef below the thick cover. The pre‐stack seismic data quality was improved through conventional processing steps. We applied standard Kirchhoff pre‐stack depth migration as well as advanced coherency migration techniques. Both imaging techniques yielded good structural images of the platinum deposits, their hanging wall and footwall rocks. In particular, the coherency migration technique has provided significantly better images in complex faulted regions, yielding a better understanding of the interrelationship between fault activity and platinum deposit distribution, and the relative chronology of tectonic events. Moreover, the regional geological structures (Crocodile River fault and Chaneng structure) that crosscut the profile are clearly defined.

The imaging condition is a crucial component of the reverse time migration. In its conventional form, it involves cross‐correlating the extrapolated source‐ and receiver‐side wavefields. Effective imaging conditions are usually developed to suppress imaging artefacts (e.g. low‐wavenumber artefacts) and enhance the image quality. For acoustic reverse time migration, not only the scalar pressure but also their spatial and/or time derivatives are used in the imaging condition, similar to the gradient terms of adjoint tomography. These operations implicitly introduce additional angle‐domain weighting factors to the image results. In this study, based on an analysis of angle‐dependent properties of the existing imaging conditions, we propose a new imaging condition tailored for acoustic reverse time migration. It can be implemented efficiently using the variables within the finite‐difference solver. Without explicitly measuring wave propagation directions, the proposed imaging condition can naturally suppress the low‐wavenumber artefacts while maintaining a relatively wider imaging aperture, thereby corresponding to a broader wavenumber sampling range. Additionally, the evolved imaging conditions for imaging elastic P–P and S–S scattering and reflections are also formulated. In the angle domain, we conduct a comparative analysis between existing imaging conditions and the newly proposed ones. Various numerical examples are provided to demonstrate the advantages of the new imaging conditions. A comprehensive understanding of their angle‐domain properties may be further beneficial to constructing reasonable inversion strategies for full waveform inversion.

Acoustic logging is an important method used to determine formation velocities near boreholes. However, in practice, determining accurate formation velocities from acoustic logging data is challenging because of the presence of various noise interferences. Accordingly, a method to increase the amplitudes of refracted waves in open boreholes is proposed herein on the basis of the directional radiation technology of pulse compression signal–driven linear phased array acoustic transmitters. The waveforms generated by a Ricker monopole acoustic transmitter, linear frequency modulation monopole acoustic transmitter and pulse compression signal–driven linear phased array acoustic transmitter in a fluid‐filled open borehole are numerically simulated by employing the finite‐difference method. The effects of the pulse compression signal–driven linear phased array parameters on the amplitudes of the refracted compressional and shear waves are studied. Results show that borehole mode waves with the same velocities and dispersion characteristics can be determined using the pulse compression signal–driven linear phased array acoustic and Ricker monopole acoustic transmitters in fluid‐filled open boreholes. Pulse compression signal–driven linear phased array acoustic transmitters leverage the advantages of pulse compression and phased array technologies, ensuring that a single element can radiate more acoustic energy, whereas pulse compression signal–driven linear phased array parameters can be modulated to further increase the amplitudes of the refracted compressional and shear waves. Compared with Ricker and linear frequency modulation monopole acoustic transmitters, pulse compression signal–driven linear phased array acoustic transmitters can provide downhole received waveforms of better quality and improved a signal‐to‐noise ratio of the mode wave dispersion curves obtained using the downhole received waveforms. Because pulse compression signal–driven linear phased array acoustic transmitters use linear frequency modulation drive signals of longer duration, the recording time required for the received waveforms is also longer and the amount of data generated is larger, presenting new challenges for downhole data processing and high‐speed data transmission.

S‐wave velocity plays a crucial role in various applications but often remains unavailable in vintage wells. To address this practical challenge, we propose a machine learning framework utilizing an enhanced bidirectional long short‐term memory algorithm for estimating S‐wave sonic logs from conventional logs, including P‐wave sonic, gamma ray, total porosity, and bulk density. These input logs are selected based on traditional rock physics models, integrating geological and geophysical relations existing in the data. Our study, encompassing 34 wells across diverse formations in the Delaware Basin, Texas, demonstrates the superiority of machine learning models over traditional methods like Greenberg–Castagna equations, without prior geological and geophysical information. Among these machine learning models, the enhanced bidirectional long short‐term memory model with self‐attention yields the highest performance, achieving an R‐squared value of 0.81. Blind tests on five wells without prior geologic information validate the reliability of our approach. The estimated S‐wave velocity values enable the creation of a basin‐scale S‐wave velocity model through interpolation and extrapolation of these prediction models. Additionally, the bidirectional long short‐term memory model excels not only in predicting S‐wave velocity but also in estimating S‐wave reflectivity for seismic amplitude variation with offset applications in exploration seismology. In conclusion, these S‐wave velocity estimates facilitate the prediction of further elastic properties, aiding in the comprehension of petrophysical and geomechanical property variations within the basin and enhancing earthquake hypocentral depth estimation. 

In recent years, self‐supervised procedures have advanced the field of seismic noise attenuation, due to not requiring a massive amount of clean labelled data in the training stage, an unobtainable requirement for seismic data. However, current self‐supervised methods usually suppress simple noise types, such as random and trace‐wise noise, instead of the complicated, aliased ground roll. Here, we propose an adaptation of a self‐supervised procedure, namely, blind‐fan networks, to remove aliased ground roll within seismic shot gathers without any requirement for clean data. The self‐supervised denoising procedure is implemented by designing a noise mask with a predefined direction to avoid the coherency of the ground roll being learned by the network while predicting one pixel's value. Numerical experiments on synthetic and field seismic data demonstrate that our method can effectively attenuate aliased ground roll.

Faults generated by seismic motion and stratigraphic lithology changes are essential research objects for seismic motion and hydrocarbon prospecting. This paper emphatically concentrates on the fault reconstruction from the existing fault probability volume. The core idea is to transform the separation of different fault sticks into a fitting and segmentation problem of point cloud data. First, we utilize the point cloud filtering algorithm to preprocess the probability volume and then complete the coarse segmentation of the fault sticks by the region growth algorithm. For the intersecting faults, we employ an enhanced random sample consensus methodology with the constraints of fault orientation and effective inliers to accomplish the detailed segmentation of different fault sticks. Finally, we take the faults identified by the region growth and the random sample consensus method as a priori to construct a random forest model to predict the fault sticks of additional data. By examining and comparing the proposed method with some other approaches with both synthetic and field data, the experimental results manifest that the novel method achieves better segmentation results than others. Moreover, the proposed method is efficient based on the fact that it can handle billions of voxels within a few minutes.

The accuracy of fault interpretation is generally influenced by the quality of seismic images. Because of the blurring effect of the migration process, faults with small throws may not be clearly imaged in seismic images, which will impose limitations on the fault detection. To address this issue, we propose a deep learning‐based method to enhance faults in poststack seismic images. We generate abundant training samples by convolving the three‐dimensional point‐spread functions with the noisy reflectivity models. The corresponding labels are synthesized using the one‐dimensional seismic wavelet convolution method, simulating conditions with perfect illumination. To train the network for optimal performance, we investigate the impact of different loss functions. Ultimately, we employ a mixed loss function combining structural similarity index measure and gradient difference loss, since the gradient difference loss focuses more on geological edge information, and the structural similarity index measure possesses excellent image perceptual capability and optimization property. Results from one synthetic seismic image and three real seismic data demonstrate that our proposed method can effectively restore the sharpness of fault surfaces, particularly for faults with small displacements. Compared to the structural smoothing method, the network we trained achieves optimal fault enhancement. Furthermore, coherence‐based fault images indicate that seismic images enhanced using our method can improve the accuracy of fault interpretation and yield more continuous fault maps.

Marine seismic multiples contain more structure information than primaries and should be considered in migrations. However, multiple migrations suffer from severe crosstalks generated by interferences among undesirable multiples. It has been proven that the water‐bottom‐related multiple migration can suppress crosstalks greatly. However, if all associated consecutive‐order multiples are considered, the computation cost is extremely high. To settle this issue, a phase‐encoding‐based multiple migration is proposed. Supergathers are first created by randomly phase‐encoding consecutive‐order multiples and stacking‐encoded multiples. By migrating supergathers, the proposed method can fulfil migrations of all order multiples simultaneously, thereby reducing the computation cost significantly. We use a three‐layer model and the Pluto 1.5 model for numerical comparisons. The results reveal that the method can retrieve high‐quality images and increase computation efficiency considerably.

Shear wave velocity is an essential parameter in reservoir characterization and evaluation, fluid identification and prestack inversion. However, conventional data‐driven or model‐driven shear wave velocity prediction methods exhibit several limitations, such as lack of training data sets, poor model generalization and weak model robustness. In this study, a model‐ and data‐driven approach is presented to facilitate the solution of these problems. We develop a theoretical rock physics model for fractured limestone reservoirs and then use the model to generate synthetic data that incorporates geological and geophysical knowledge. The synthetic data with random noise is utilized as the training data set for the artificial neural network, and a well‐trained shear wave velocity prediction model, random noise shear wave velocity prediction neural network, is established by parameter tuning, which fits the synthetic data with noise well. The neural network is applied directly to the real field area. Compared with conventional shear wave prediction methods, such as empirical formulas and the improved Xu–White model, the prediction results show that the random noise shear wave velocity prediction neural network has better prediction performance and generalization. Furthermore, the prediction results demonstrate the efficacy of the proposed approach, and the approach has the potential to perform shear wave velocity prediction in real areas where training data sets are unavailable.

The stratum can be modelled as a horizontal transversely isotropic medium when a single set of vertically parallel fractures embedded in an isotropic background medium, which facilitates efficient study for fractured reservoirs. Elastic parameters and fracture weaknesses are important parameters to describe the characteristics of fractured reservoirs, and seismic inversion plays a significant role in parameters estimation. The commonly used deterministic inversion methods do not fully utilize the prior information and fails to present the uncertainty analysis of inversion results. To address these shortcomings, we propose a Bayesian linearized amplitude variation with offset and azimuth inversion method tailored for horizontal transversely isotropic media, enabling a more robust analysis of uncertainty. Within the framework of Bayesian inversion, the proposed method successfully derives analytical expressions for the posterior mean and covariance of both elastic parameters and fracture weaknesses. The response characteristics of the anisotropic reflection coefficient are analysed, and it is found that the perturbations of elastic parameters have a greater effect on reflection coefficient compared to fracture weaknesses. Synthetic data examples confirm that the accuracy of estimated P‐ and S‐wave velocities and density surpasses that of fracture weaknesses, and the proposed method still performs well for the case of moderate noise. A field data example demonstrates that the inverted profiles agree well with the logging curve, and the estimated fracture weaknesses display significantly high values in the reservoir area. The estimated reservoir parameters not only contribute to a more accurate representation of the fractured gas‐bearing reservoir but also provide insights into the target gas reservoir through its posterior distribution. Both synthetic and field data examples demonstrate the stability and reliability of the proposed method in characterizing fractured reservoirs. We determine that the proposed method provides an available tool for nuanced evaluation of uncertainty for the inversion results, and it is helpful for the fine description of fractured hydrocarbon‐bearing reservoirs.

Since many years ago, ultrasonic velocity has been used to investigate the physical and mechanical behaviour of rocks, thereby playing an important role in reservoir characterization and seismic interpretation. In order to develop the knowledge of ultrasonic tools, I performed a noble analysis on the ultrasonic behaviour of rocks under confining stress and evaluated a distinctive property of porous media that is measured as the area under the stress–velocity curve (here defined as S^*). I further investigated its relationship with elastic and mechanical behaviours of rock. To validate the theoretical framework developed in this work, 20 core plugs from various rock units with complex microstructures were subjected to triaxial compressional tests to calculate their area under the curve. Calculations were made for crack‐closing, elastic and post‐elastic stages (e.g. pore collapse) along the ultrasonic velocity–stress curve. Moreover, the selected samples had their microstructure investigated by thin‐section studies to quantify their porosity and pore type. The results were analysed to check for the effect of pore type on S^* in different stages of the stress–velocity curve. Based on the outputs of the analysis of variance and Pearson's correlation coefficient analysis, the curve had its shape and underlying area closely related to the porosity and pore geometry. Indeed, the results showed that the shale and sandstone with micro cracks and carbonate with stiff pores correspond to smaller and larger areas under the curve in crack‐closing and inelastic stages, respectively. Cross‐correlating the results to compressibility (inverse of bulk modulus), it was figured out that the calculated area under curve was well consistent with the compressibility. In addition, S^* represents both static and dynamic behaviours of the rock, and the results revealed that the shape and curvature of the stress–velocity curve give valuable information about the rock microstructure. Another finding was the fact that the type of fluid and wave velocity seemingly affect the S^*. Our findings can help interpret wave velocity behaviour in reservoir rocks and other stressful porous media.

In seismic exploration, it is very important to consider the presence of anisotropy in order to image the subsurface correctly. The knowledge of anisotropic parameters leads to more precise characterization of reservoir, fracture density and flow paths. However, conventional geophysical methods do not directly measure these parameters, and it is useful to have a method to estimate them from seismic data. In the weak transverse isotropy approximation, the fields to be considered are the vertical and horizontal velocity components and the Thomsen parameters and . We present a method for estimating the anisotropic Thomsen parameters in the presence of weak vertical transverse isotropy using P‐wave traveltime tomography based on anisotropic ray tracing. Depending on the available information, we propose different approaches to retrieve the unknowns. A conventional three‐dimensional traveltime tomography algorithm has been extended to include anisotropic ray tracing and using the algebraic reconstruction technique or modified simultaneous iterative reconstruction technique to retrieve the unknowns. We test the method on synthetic examples for the inversion of transmitted and reflected traveltimes, and we evaluate the sensitivity of the tomographic results to the available information. Furthermore, we also consider the case of tilted transverse isotropy in a seismic reflection example.

Full waveform inversion stands at the forefront of seismic imaging technologies, pivotal in retrieving high‐resolution subsurface velocity models. Its application is especially profound when imaging complex geologies such as salt bodies, which are regions notoriously challenging, yet essential given their hydrocarbon potential. However, with the power of full waveform inversion comes the intrinsic challenge of estimating the associated uncertainties. Such uncertainties are crucial in understanding the reliability of subsurface models, particularly in terrains like subsalt regions. Addressing this, we advocate for a nuanced approach employing the Stein variational gradient descent algorithm. Through a judicious use of a limited number of velocity model particles and the integration of random field‐based perturbations, our methodology provides a local representation of the uncertainties inherent in full waveform inversion. Our evaluations, based on the Marmousi model, showcase the robustness of the proposed technique. Yet, it is our exploration into salt‐intensive terrains, leveraging data from the Sigsbee 2A synthetic model and the Gulf of Mexico, that emphasizes the method's versatility. Findings indicate pronounced uncertainties along salt boundaries and in the deeper subsalt sediments, contrasting the minimal uncertainties in non‐salt terrains. However, anomalies like salt canyons present unique challenges, potentially due to the interplay of multi‐scattering effects. Emphasizing the scalability and cost‐effectiveness of this approach, we highlight its potential for large‐scale industrial applications in full waveform inversion, while also underscoring the necessity for prudence when integrating these uncertainty insights into subsequent seismic‐driven geological and reservoir modelling.