Editorial for the Special Issue “Advances in Geophysical Exploration”

Recent advances in applied geophysics have greatly enhanced our ability to accurately image and understand the Earth’s physical characteristics and geological processes. These innovations include cutting-edge technologies and methodologies that offer deeper insights into subsurface structures and materials. A key priority in modern geophysical exploration is the commitment to minimizing the environmental impacts of exploration activities. This growing awareness is shaping practices across diverse exploration domains, including mineral resources, hydrocarbon extraction, groundwater management, environmental assessments, hydrogen production, and geothermal energy exploitation.

To promote knowledge sharing and collaboration, this Special Issue provides a dedicated platform to discuss the latest advancements in geophysical exploration. Authors are invited to contribute their research and present their findings to a diverse and engaged geoscience community. The issue highlights key developments across geophysical methods, ranging from established techniques to cutting-edge innovations. We particularly welcome interpretative case studies illustrating practical applications of these methods. Contributions may also focus on developing new analytical or computational methodologies, as well as the integration of interdisciplinary strategies that offer novel insights. We propose dividing the themes into three main areas of concentration:

1

The recent advancements in forward and inverse modeling algorithms. These developments are essential for accurately interpreting geophysical data and generating reliable subsurface images.

2

Developments and case studies in mineral, hydrocarbon, hydrogen, and geothermal exploration, emphasizing successful projects, innovative techniques, and the integration of geophysical methods with other scientific disciplines.

3

Developments and case studies in hydrogeological and environmental studies, showcasing applied geophysics projects that have achieved significant results. This area also highlights innovative acquisition, processing, and interpretation techniques, as well as the integration of geophysical methods with other scientific disciplines.

We received diligent responses from the geoscience community, with fourteen papers addressing various key aspects of these three areas of concentration.

Xue et al. [1] propose an integrated approach combining multi-scale static data, including geological information, well logging, and seismic surveys, with dynamic production data to improve connectivity evaluations in fracture–cavity reservoirs. Their methodology uses structural gradient tensors and stratigraphic continuity to delineate caves and dissolution holes, RGB fusion analysis to characterize fault development, fracture simulation to map fracture zones, 3D visualization to assess spatial relationships, and dynamic data to verify interwell connectivity. Applied to the S91 unit of the Tarim Oilfield, China, this approach revealed that fracture–cavity units exhibit discontinuous patterns controlled by large-scale faults and dense fracture networks. Three-dimensional visualization identified potential connectivity channels, and production data confirmed effective interwell connectivity, highlighting residual oil potential and supporting optimized well placement and injection–production strategies.

Wang and Song [2] develop a parallel adaptive high-order finite-element framework for large-scale 3D marine controlled-source electromagnetic (MCSEM) modeling. Their approach combines coordinate transformations based on the form invariance of Maxwell’s equations with a goal-oriented adaptive mesh refinement strategy, allowing complex geometries, thin sedimentary layers, and multi-scale heterogeneities to be efficiently handled. By mapping the physical domain to a computationally favorable domain with equivalent anisotropic materials, the method preserves the mathematical structure of Maxwell’s equations while reducing computational complexity. The framework is fully parallelized and validated through synthetic models and the Marlim R3D benchmark dataset, demonstrating high accuracy and efficiency across different water depths and frequency ranges. It shows particular advantages in shallow-water settings and models with depth-dependent conductivity variations, as well as in handling realistic seafloor topographies. This work highlights the potential for large-scale, high-resolution CSEM surveys, supporting accurate imaging and inversion of offshore resources, with possible extensions to other electromagnetic modeling applications such as plane-wave EM modeling.

Wang et al. [3] propose a Kalman–FIR fusion (K-F) filtering method to overcome the challenges of noise suppression and spectral fidelity in high-dynamic airborne gravimetry. Conventional finite impulse response (FIR) filters often attenuate genuine gravity signals when suppressing platform-induced noise, while Kalman filtering lacks explicit spectral control, creating difficulties for downstream processing. To address these issues, the authors develop a dual-stage framework: first, a state-space Kalman filter that incorporates triaxial acceleration error modeling to compensate for horizontal motion coupling interference; second, a window-optimized FIR filter that ensures explicit spectral control and compatibility with terrain correction. Engineering tests on the GIPS-1A airborne gravimeter show that the K-F method achieves a repeat-line internal consistency of 0.558 mGal at 0.01 Hz—an accuracy improvement of more than 65% compared to standalone FIR filtering—and enhances spatial resolution to 2.5 km (half-wavelength). Beyond improving noise suppression under turbulent conditions, the method enables the recovery of gravity anomaly data segments that were previously lost to airflow disturbances. The method has been implemented in the GIPS-1A gravimetry system and deployed across various platforms, including UAVs, helicopters, and fixed-wing aircraft, for industrial applications.

Wang et al. [4] investigated the genesis of thermal anomalies in the southwestern A’nan Sag, Erlian Basin, using 3D magnetotelluric (MT) inversion to construct a high-resolution resistivity model. They identified an elliptical low-resistivity anomaly (0–5 $\mathrm{\Omega }$ m) at depths of 10–15 km, interpreted as hypersaline fluids serving as the primary heat source. Faults F1 and F3 act as medium–low resistivity conduits (10–40 $\mathrm{\Omega }$ m) connecting Anomaly C to mid-depth reservoirs D1/D2 (5 km, 5–10 $\mathrm{\Omega }$ m) and shallow low-resistivity layers, while an overlying high-resistivity caprock (40–100 $\mathrm{\Omega }$ m) seals the system, forming a convective “source–conduit–reservoir–cap” architecture. Integrated seismic S-wave analysis indicates that thermal energy from partial melt beneath the Abaga volcanic zone supplements Anomaly C via conduction, combining with mantle-derived heat to form a composite source. This study elucidates the coupled conduction–convection mechanisms sustaining regional thermal anomalies and provides a quantitative framework for geothermal exploration and resource assessment in the A’nan Sag.

Piauilino et al. [5] demonstrate the use of a convolutional equivalent layer for jointly processing large gravity-gradient datasets efficiently. By assuming a single fictitious physical property distribution on a planar layer, all components of the gravity-gradient tensor remain physically consistent. However, this approach becomes computationally prohibitive for large datasets due to high demands. To overcome this limitation, they exploit the block-Toeplitz Toeplitz-block structure of the sensitivity matrix for data arranged on regular grids, significantly reducing computational costs. Tests on synthetic and real datasets show that the method can process over 290,000 AGG (Falcon airborne gravity gradiometer) data points in seconds and handle up to 1,250,000 AGG points in a few minutes. The methodology was also successfully applied to real FTG (Full Tensor Gradiometry) datasets.

Wang and Zuo [6] introduce MagTCN, an Aeromagnetic compensation method based on temporal convolutional networks, developed to address the limited generalization and compensation accuracy of existing techniques, particularly as Aeromagnetic sensors reach higher sensitivity levels. Unlike earlier neural network approaches that relied on standard convolutional layers or hybrid methods such as wavelet decomposition combined with long short-term memory (LSTM) networks, MagTCN leverages depthwise separable convolutions with residual connections, which improve feature extraction efficiency, reduce computational cost, and stabilize gradient propagation. In addition, the authors design a gradient correction loss function grounded in the Tolles–Lawson (T–L) model, enabling the network to better capture the temporal dependencies between Aeromagnetic interference and auxiliary flight parameters while preventing overfitting and enhancing convergence stability. Validation on both synthetic datasets and the DAF-MIT AIA Open Flight Data shows that MagTCN achieves a significantly lower standard deviation of residual magnetic interference than traditional convolutional and wavelet–LSTM methods, thereby demonstrating superior compensation accuracy, robustness, and generalization capability for modern Aeromagnetic survey applications.

Lyrio et al. [7] propose a novel methodology to estimate magnetic susceptibility and remanent magnetization directly from the magnetic field information acquired by orientation tools in borehole imaging logs, data typically used only for spatial orientation of images. This approach addresses a critical gap in hydrocarbon exploration, where direct measurements of susceptibility and remanence are rare due to the cost and complexity of acquiring oriented core samples, despite their importance for forward magnetic modeling and risk reduction in Pre-Salt exploration, such as in the Campos Basin, Brazil. The method involves correcting the acquired magnetic field data for tool rotation and diurnal variation, followed by the application of simple equations under reasonable assumptions to derive rock magnetic parameters. Application to a Pre-Salt well demonstrated that estimated susceptibility values for carbonates and basalts closely matched laboratory measurements, while remanent magnetization of basalts was also obtained with consistent results. The methodology shows promise for providing reliable magnetic property estimates without additional acquisition costs, as imaging logs are routinely recorded in exploration wells. Moreover, it offers potential applications beyond hydrocarbon settings, notably in mining exploration, where susceptibility aids lithological identification, and possibly in magnetostratigraphy, provided improvements in resolution and corrections are achieved.

Zhang et al. [8] propose a novel inversion methodology for tilted transverse isotropic (TTI) media affected by inclined fractures, addressing key challenges in seismic parameter estimation for fractured reservoirs. The study derives a new reflection coefficient equation for TTI media with rotationally invariant inclined fractures, based on seismic scattering theory and the steady-phase method, providing improved accuracy and theoretical consistency compared to previous models. To enhance inversion stability, the authors introduce a scale normalization technique that transforms parameters into a uniform scale, mitigating errors caused by inconsistent parameter magnitudes during inversion. The methodology is validated with synthetic models and field data, showing strong agreement with well logs and geological interpretations. The results demonstrate that the proposed approach significantly improves inversion accuracy and reliability in fractured TTI media, offering practical value for reservoir characterization, well placement optimization, and fractured reservoir development.

Monroe et al. [9] present a methodology for enhanced porosity prediction in subsurface reservoirs by integrating Bayesian Linearized Inversion (BLI) of seismic and well-log data with both an empirical model (Han’s equation) and a Support Vector Regression (SVR) machine learning model. The BLI framework is employed to optimally estimate elastic parameters, including compressional and shear wave velocities ($V_P$,$V_S$) and density, mitigating uncertainties inherent in the initial seismic data and improving the reliability of subsequent porosity predictions. By incorporating sparse spike wavelets into the inversion process, the methodology refines seismic resolution and enhances the extraction of petrophysical properties. The study applies this workflow to data from seventeen wells in the Norne field, North Sea, using two primary wells for inversion and training fifteen secondary wells for predictive modeling. Evaluation metrics demonstrate that the SVR-based porosity predictions achieve superior accuracy, with a Pearson correlation coefficient of 0.90 and improved $R^2$ and RMSE values compared to Han’s empirical equation, highlighting the capacity of machine learning to capture complex nonlinear relationships in geophysical data. The authors emphasize that the workflow, including wavelet selection, hyperparameter tuning, and integration of BLI outputs, is adaptable to different geological contexts, providing a robust framework for high-precision reservoir characterization and predictive modeling in hydrocarbon exploration.

Wang et al. [10] address the limitations of conventional Transient Electromagnetic Method (TEM) modeling and inversion, which often assume non-magnetic subsurface conditions, by developing a three-dimensional (3-D) TEM framework that simultaneously accounts for variations in resistivity and magnetic permeability. Recognizing that neglecting magnetic susceptibility can lead to misinterpretation in regions with igneous rock accumulations or ferromagnetic minerals, the study employs an edge-based finite element method on unstructured grids combined with a second-order implicit backward Euler scheme for accurate forward modeling. The forward algorithm is rigorously validated against analytical solutions for homogeneous half-space models and 1D numerical solutions, demonstrating its ability to capture the influence of non-uniform magnetic permeability on TEM responses. Building on this, a 3D inversion methodology using the Limited-memory Broyden–Fletcher–Goldfarb–Shanno (L-BFGS) optimization algorithm is implemented, enabling the simultaneous recovery of resistivity and magnetic permeability anomalies. Synthetic case studies, including separate and joint inversion scenarios, show that the proposed approach accurately recovers the position and amplitude of targets, whereas inversion that neglects magnetic permeability produces spurious anomalies, underscoring the necessity of including magnetic parameters in TEM analysis. Although the current framework is limited to large-loop magnetic sources, isotropic resistivity models, and does not account for induced polarization effects, it significantly enhances the applicability and flexibility of 3D TEM inversion, offering a robust tool for quantitative interpretation of TEM data in geologically complex and magnetically active regions.

Chen and Dong [11] propose an integrated approach for improved velocity prediction in carbonate rocks, addressing the challenges posed by complex pore structures and strong matrix heterogeneity. The methodology combines a modified squirt flow model, which accounts for local fluid flow between microcracks with different aspect ratios, with a mixed random medium model representing matrix heterogeneity. Unlike David and Zimmerman’s conventional inversion method, the microcrack inversion assigns each group of microcracks a distinct equivalent medium, improving inversion accuracy by reflecting the sequential closure of microcracks. The modified squirt flow model is further integrated with the Gassmann model to capture attenuation effects caused by local flow between microcracks. The random medium model simulates inhomogeneous rock matrices by adjusting parameters such as autocorrelation lengths, rounding coefficients, and orientation angles. Validation against ultrasonic experimental data from four limestone samples and one dolomite core sample demonstrates that the new model achieves P-wave prediction errors below 5%, outperforming the Biot model, particularly in samples with uniform micropore aspect ratio distributions, smaller pore radii, and more homogeneous structures. The study highlights the effectiveness of combining microcrack-specific squirt flow modeling with heterogeneous matrix representations to enhance the accuracy and adaptability of petrophysical predictions in carbonate cores.

Correa and Regis [12] present a method for constructing 3D resistivity models from marine Controlled-Source Electromagnetic (MCSEM) data by spatially constraining 1D inversions in the Common Midpoint (CMP) domain. The methodology formulates a least-squares optimization problem where each 1D inversion column contributes to a 3D resistivity cube, and inter-column correlations are enforced via Tikhonov-style smoothing regularization applied in all three spatial directions. The iterative algorithm updates the model to minimize the misfit between the measured multi-component electric field data and synthetic responses computed from the layered model columns, while also allowing for the incorporation of hard constraints, such as well log resistivities. Synthetic tests, including scenarios with resistive targets at varying depths and resistive bodies embedded within conductive layers, demonstrate that the approach accurately reconstructs subsurface resistivity distributions. Application to real MCSEM data from a gas hydrate accumulation area off the southeast coast of Brazil shows that the method effectively generates a preliminary 3D resistivity model, which serves as both a rapid interpretive tool and a computationally efficient initial guess for subsequent full 3D inversions. The spatial constraints enhance the stability and resolution of the inversion, though the reliability of the model depends on the geological complexity; layered structures yield robust results, whereas localized features such as faults introduce lateral variability that may reduce model fidelity.

Valente et al. [13] propose a computationally efficient method for the forward modeling and inversion modeling of 3D marine Controlled-Source Electromagnetic (CSEM) data. The approach employs a time-domain finite-difference solution of Maxwell’s equations in a fictitious dielectric medium, with the correspondence principle mapping results to the conductive medium. This enables multi-frequency simulations in a single run and allows gradient computation via the adjoint-state method. Model reparameterization smoothing is used to stabilize inversion without requiring trial-and-error regularization. Two optimization strategies are tested: steepest descent with line search and a modified momentum method inspired by the Adam algorithm, using a single momentum term, and both are benchmarked against the conventional L-BFGS-B (Limited-memory Broyden–Fletcher–Goldfarb–Shanno with Bound constraints) algorithm. Validation on three synthetic datasets—including two resistive targets, a MARE-inspired model, and a turbidite reservoir analog—demonstrates that the momentum method, despite requiring more iterations, produces models comparable to L-BFGS-B, fitting observed data slightly better while consuming less memory. The study presents an effective framework for 3D CSEM inversion, yielding high-quality resistivity models with reduced computational cost, which is particularly beneficial for large-scale geophysical applications.

Rey et al. [14] integrate geological mapping and borehole data with geophysical techniques, including seismic reflection, time-domain electromagnetic (TDEM) surveys, and magnetic measurements, to characterize the northern boundary of the Bailén basin, southeastern Spain. Focusing on the Baños de la Encina fault, seismic reflection revealed lithological variations and two faults causing local subsidence. TDEM surveys provided estimates of the Palaeozoic basement depth and fault locations, exploiting resistivity contrasts between sedimentary infill and basement rocks, while magnetic data identified fault-related anomaly alignments (N30E and N60E), including previously unmapped fractures. Calibration with borehole data confirmed consistency among the datasets, allowing detailed mapping of vertical and lateral bedrock variations. The study demonstrates the complementarity of seismic, electromagnetic, and magnetic methods for resolving complex faulting, basin structure, and aquifer characteristics.