Geophysical Prospecting最新文献

Surface-wave full-waveform inversion (FWI) has emerged as a powerful technique for resolving high-resolution S-wave velocity (V_s) structures in shallow seismic investigations. However, strong surface-wave dispersion often exacerbates the cycle-skipping problem, and field-recorded wavefields are frequently affected by amplitude errors. To mitigate these challenges, we propose a misfit function that integrates an amplitude-unbiased instantaneous-phase coherency (IPC) measure with sigmoid modulation. The sigmoid function reshapes the objective function landscape by suppressing spurious local minima while preserving the robustness of phase-based metrics against amplitude variations. Notably, this approach is straightforward to implement, as it directly builds upon the existing IPC framework and can be readily adapted to other misfit functions. Synthetic tests demonstrate that the proposed sigmoid-modulated IPC misfit function significantly improves inversion convergence and enhances the recovery of deeper subsurface anomalies. Compared with conventional IPC-based FWI, the modulated approach yields sharper anomaly boundaries and exhibits greater resilience to random noise, achieving lower relative percentage error (RPE) in noisy data scenarios. Application to a published field dataset from an archaeological site in Southern Germany further validates the method's robustness. Inversions with and without sigmoid modulation produce broadly consistent V_s structures that align well with features independently confirmed by archaeological excavations. Although the high quality of the initial model limits the observable advantage of the modulated objective in this specific case, the sigmoid-modulated IPC approach demonstrates faster convergence, reaching a lower misfit value within fewer iterations due to its improved objective function landscape. These findings confirm that the proposed method is a practical and reliable alternative for field-scale surface-wave FWI applications, particularly in environments prone to amplitude errors and cycle skipping.

Full-waveform inversion (FWI) is a powerful technique for constructing high-resolution subsurface models, which are essential for hydrocarbon exploration, field development and the interpretation of deep Earth structures. However, seismic signals are often subject to amplitude decay and phase distortion due to intrinsic attenuation, which can significantly degrade inversion accuracy, particularly when attenuation and the associated velocity dispersion effects are not accounted for. To address these challenges, there is an increasing demand for multiparameter viscoelastic FWI methods that can simultaneously invert for both velocity and attenuation models. In this study, we present a novel P- and S-wave separation approach based on the nearly constant-Q viscoelastic model. The proposed method incorporates a gradient preconditioning technique designed to improve inversion accuracy by mitigating the interference between the gradients of P- and S-wave velocities—a persistent issue in multiparameter inversions, particularly in the presence of anomalous geological bodies. We derive the necessary formulations for wavefield separation and gradient preprocessing using the pseudo-Hessian operator and validate the methodology through numerical simulations based on the Marmousi-2 model and real ocean bottom node (OBN) data. The results demonstrate that the proposed method significantly enhances inversion accuracy, particularly in geologically complex regions. This approach offers a promising solution for achieving more precise subsurface imaging in viscoelastic media.

The Born approximation has been extensively used to approximate scattered waves in a variety of areas in physics. The Born approximation utilizes specific terms of a series expansion and the general issues of convergence of the series and accuracy of the approximations are difficult to ascertain. I use a simple one-interface model to examine these issues for both acoustic and elastic wavefields. The series converges in the normal incidence case, although to a value that is less accurate than that of the first-order approximation. Also, for normal incidence and two-dimensional acoustic waves, the series contains only odd terms. At non-normal incidence angles the inclusion of higher order terms does increase accuracy over a reasonable range of incidence angles, but ultimately diverges from the exact reflection coefficient at large reflection angles. The series for elastic waves contains both odd and even terms, and the effects of mode conversions begin with the even second-order term. Adding the second-order term for the elastic case provides a more accurate prediction than the first-order term alone, yet it is unclear if additional terms will continue to improve the accuracy of the approximation for non-normal incidence.

We aim to explore the distribution characteristics of coal mine goaf and water-rich areas and improve the accuracy of underground hidden structure identification. We take the Renjiazhuang Coal Mine in Ningxia as a case study and deploy a linear station array to acquire microtremor data. Surface–wave dispersion curves are extracted using the spatial autocorrelation (SPAC) and extended spatial autocorrelation (ESPAC) methods. We then invert the S-wave velocity structure using a genetic algorithm (GA) and a simulated annealing (SA) algorithm. The results show that ESPAC provides better noise resistance in the low-frequency band and yields more continuous and higher resolution dispersion curves than SPAC. In addition, the GA inversion outperforms the SA inversion in terms of convergence accuracy and stability. The low-velocity anomalies identified from the GA inversion are highly consistent with the actual goaf and water-rich zones, demonstrating the accuracy and applicability of the combined ESPAC–GA approach under complex geological conditions. We provide a technical reference for the application of microtremor technology in complex geological environments and lay a foundation for the optimisation of subsequent multimodal fusion and inversion algorithms.

Traditional partial differential equation (PDE)-based numerical methods for modelling three-dimensional (3D) magnetic potential require solving a large, sparse linear system using either matrix inversion or iterative solvers. The corresponding computational cost can present considerable challenges for large-scale 3D magnetic inversions. Additionally, the accuracy of the computed magnetic anomalies, such as the magnetic field vector and magnetic gradient tensor, derived through finite-element or finite-difference approximations, may degrade due to the inherent limitation of the numerical differentiations of the magnetic potential. To address these challenges, we present a fast Fourier transform (FFT)-based spectral scheme aimed at effectively simulating the boundary value problem associated with the 3D magnetic potential. By applying the 3D FFT technique along three Cartesian axes to Poisson's equation, the corresponding PDE in the spatial domain is converted to an algebraic equation in the wavenumber domain, allowing for the straightforward computation of the magnetic potential. Additionally, the vector and tensor magnetic fields can be computed with high precision by utilizing the differential operators of the Fourier transform for spatial derivatives. Through applications to two synthetic and SEG/EAGE salt models, we validate the numerical accuracy and computational efficiency of our proposed method. Compared with other approaches in similar classes, such as the hybrid wavenumber-domain finite-element method, the proposed FFT-based spectral method has a higher accuracy and requires lower computational cost and time. Thus, the new method is superior in the forward and inverse modelling of large-scale magnetic problems.

$�{\mathrm{CO}}_2$ sequestration has shown an important potential to reduce greenhouse gases in the atmosphere. One of the key aspects of $�{\mathrm{CO}}_2$ sequestration is geological storage, which aims to store $�{\mathrm{CO}}_2$ safely over a long period with low risks. Geophysical monitoring of the reservoir is required for this objective, with microearthquakes from injection serving as indicators for assessing changes in seismic velocity over time. We develop a geometry optimization algorithm that can significantly reduce the number of required microearthquake events for velocity estimation while balancing the subsurface illumination. A modified source-independent waveform inversion algorithm is proposed to eliminate the negative effects of the usually unknown origin time and source wavelets. However, the non-linearity of the objective function also increases, requiring at least 94% of the starting model's accuracy to avoid convergence to local minima in a synthetic experiment. Furthermore, we design a neural network that takes the traveltime difference of P- and S-wave arrivals to relocate the microearthquakes. We test various scenarios, including noisy data, parameter crosstalk, inaccurate source location and unknown source wavelet, based on the Illinois Basin Decatur $�{\mathrm{CO}}_2$ storage project. Finally, we apply the developed algorithms to the field data and discuss their advantages and weaknesses. In practice, the poor azimuthal coverage of microearthquakes and borehole sensors may limit the quality of final images. The developed algorithms are also applicable to seismological imaging of the Earth using ambient or active seismic sources.

Characterization of cracks is a key issue in shale oil and gas that have become increasingly important in the hydrocarbon industry. Seismic exploration is frequently employed for the characterization of cracks in shale reservoirs. However, the accurate interpretation of seismic data for characterizing cracks in shale reservoirs remains a significant challenge, primarily due to an insufficient understanding of how subsurface pressure affects the anisotropic elastic properties of cracked shales. To address this knowledge gap, this study systematically investigates the effects of confining pressure on the anisotropic elastic properties of cracked artificial shales, with a specific focus on decoupling the distinct roles of background porosity and crack porosity. The five anisotropic elastic velocities were measured on manufactured shale samples with varying crack and background porosity, respectively, and the corresponding anisotropic parameters, Young's moduli and Poisson's ratios were derived as a function of confining pressure. The results demonstrate that the influence of crack porosity on reducing the velocities and on enhancing the elastic anisotropy is significantly more pronounced than that of background porosity. Notably, the velocities across the cracks, V_p(0°) and V_sh(0°), exhibit the greatest sensitivity to pressure changes, especially in samples with high crack porosity. Consequently, all the anisotropic parameters reduce exponentially with increasing confining pressure, with the reduction being most significant in shales with either the lowest background porosity or the highest crack porosity. The pressure-dependent geomechanical properties (Young's moduli and Poisson's ratios) reveal that the direction parallel to cracks remains the most favourable path for hydraulic fracturing, particularly under low confining pressure and in rocks with high crack porosity. These findings provide critical insights for improving the quantitative interpretation of seismic data for characterizing cracks and for optimizing hydraulic fracturing design in shale reservoirs.

Enhancing seismic data resolution is a crucial step for geological interpretation and imaging. Deep learning-driven resolution enhancement primarily depends on sophisticated network architectures and extensive datasets. A lightweight seismic super-resolution model based on contrastive learning and knowledge distillation is proposed. Knowledge distillation is implemented by training a compact student network to mimic a powerful teacher model, thereby reducing reliance on extensive datasets and complex architectures. Contrastive learning is leveraged to align the bottleneck features encoded from the teacher network with the ones from the student network across different noisy inputs. The student network's total loss comprises a supervised loss with ground-truth labels, a distillation loss with the teacher's pseudo-labels and a feature-matching loss derived from the bottleneck features of both networks. The comparative experiments were conducted on four field datasets and 3200 pairs of slices extracted from 800 pairs of synthetic three-dimensional seismic cubes. Experimental results demonstrate that the proposed model achieves similar to or better performance than the comparison models for noise suppression and weak signal recovery, even with only $��6.8�\%�$ parameters and $��37.5�\%�$ training data compared to the reference model.

The solution of magnetotelluric equations is used to determine the apparent resistivity and to model the electromagnetic field's behaviour within the Earth. In this paper, we extend the scaled boundary finite element method (SBFEM) to compute the solutions of magnetotelluric equations. The salient features of the proposed framework are that internal features and boundaries are captured through a quadtree decomposition. The SBFEM handles the resulting hanging nodes as a part of local refinement efficiently without needing additional constraints or shape functions. Further, we employ patterns to speed up the computations of the essential matrices without compromising accuracy. The results from the present approach are compared with other approaches, and it is seen that the SBFEM framework is not only efficient but also accurate. The efficacy and robustness are demonstrated with a few examples.