On 15 May 2021, China's Tianwen-1 mission successfully delivered the Zhurong rover to southern Utopia Planitia. During sols 244-266, the rover identified a shallow pockmarked terrain and acquired subsurface structural data at its edge using the Rover Penetrating Radar (RoPeR). Analysis of the radar data reveals a distinct reflection interface at a depth of approximately 2-4 m, indicating physical differences between the ejected deposits and the original surface materials. The inverted dielectric permittivity ranges from 2.9 to 6.3, suggesting that the subsurface is dominated by dry, fine-grained sediments rather than ice-rich materials. Based on HiRISE imagery, we identified more than 50 pockmark-like features and classified them into two morphological types: blocky and linear. Morphological analysis shows that these features are small-scale, shallow depressions exhibiting significant orientation along the NEE-SWW and WNW-ESE directions, implying a structural control by regional stress fields. Integrating topographic, geomorphological, and radar-derived evidence, we infer that these pockmarks most likely formed through the release of subsurface volatiles followed by surface collapse. This study reveals shallow geological processes in the Zhurong landing region possibly associated with late Amazonian water-related volatile activity, providing new evidence for the recent surface evolution of Mars.

The low-frequency absence in seismic data is likely to influence the vertical resolution, the imaging for complex structures and the convergence of full-waveform inversion (FWI). The extrapolation task of absent low-frequency components is a strong non-linear problem, which can be solved by deep learning (DL) owing to its strong capability of non-linear mapping. Meanwhile, low-frequency extrapolation can be regarded as a global waveform transformation, in which global features are beneficial for improving the accuracy and generalization of DL-based low-frequency extrapolation methods. Therefore, we propose a Transformer-based network oriented for capturing more global features, called low-frequency extrapolation Transformer (LFET), to extrapolate the absent low-frequency components of seismic data. To our best knowledge, this is the first attempt to apply the Transformer model to the task of low-frequency extrapolation. This proposed LFET is composed of eight successively connected Transformer blocks (TBs) which can extract contextual and global information by using the core operation: multi-head self-attention (MSA). Moreover, we introduce the shifted window partitioning strategy into LFET, thus alleviating the computational burden almost without influencing the capture of global features. In the experimental part, we use several synthetic and field records to test the effectiveness of LFET. Our method exhibits better extrapolation performance than three convolution-based competitive methods as well as lower training and testing time-cost. Furthermore, in the discussion part, a benchmark model is used to preliminarily demonstrate the potential application of LFET in improving the convergence of FWI.

The application of full waveform inversion (FWI) on field data is currently a widely concerned challenge in seismic exploration. FWI has been proven on synthetic data to have the ability to obtain high-accuracy subsurface velocity and improve the imaging effect of the pre-stack depth migration (PSDM). However, its practical application remains constrained by challenges such as cycle-skipping. Meanwhile, the method’s capability to recover detailed subsurface properties retains considerable, yet underexploited, potential for advanced geological interpretation. In this study, we integrated FWI with amplitude variation with angle (AVA) analysis to identify potential gas reservoirs using 2D OBC data from the South China Sea. Our workflow began with 0–9 Hz FWI processing to obtain an updated velocity model. Comparative analysis demonstrated the FWI model’s superiority over the initial model through: (1) improved data fitting, (2) enhanced PSDM imaging with better structural alignment, and (3) higher-quality ADCIGs. The FWI results revealed a distinct high-velocity anomaly, which was validated through integration with PSDM images and well data. Subsequent AVA analysis suggested this anomaly may represent gas-bearing sandstone, though additional verification would be required to fully characterize the formation. Nonetheless, our workflow offers a reference for subsequent interpretations, which narrows down the scope for targeted objects. Our findings demonstrate that FWI and ADCIGs can obtain more geological information than traditional velocity modeling methods, thereby becoming a valuable tool for geological interpretation.

Seismic imaging plays an important role in the discovery of deep discontinuities, mineral exploration, oil and gas prospecting and development, as well as geological surveys. Over the past half-century, with the rapid advancement of high-performance computing and advanced data acquisition, seismic imaging methods have undergone an evolution from traditional ray-based migration to wave-equation imaging, and further to least-squares migration (LSM) and full-waveform inversion (FWI) imaging. By solving a linear or nonlinear inversion problem, inversion-based migration estimates a generalized inverse of the subsurface reflectivity model, which overcomes the limitations of conventional adjoint-operator-based migration methods in irregular acquisition, limited-bandwidth data, and unbalanced illumination. It can significantly enhance imaging resolution and amplitude fidelity. We systematically review the progress and cutting-edge developments of inversion-based migration in exploration seismology, especially focusing on the theory and methodology of data-domain, image-domain and intelligent LSMs. Additionally, we describe various regularization and preconditioning strategies for LSM in terms of their mathematical principles and practical effectiveness. Finally, we discuss the latest development in nonlinear FWI imaging, offering theoretical and methodological references for high-precision seismic imaging studies.

In marine seismic exploration, towed-streamer seismic data acquisition is a critical strategy, yet it often faces significant challenges due to sparse spatial sampling along the crossline direction. This limitation hampers seismic imaging quality, particularly in multiple-wave imaging and the resolution of fine subsurface structures. While Fourier interpolation is commonly used to mitigate these issues, traditional methods struggle with the large crossline spacing and limited cable numbers in narrow-azimuth marine towed-streamer surveys. The advent of deep learning offers a promising solution to these challenges. Therefore, we propose a novel self-supervised learning approach to enhance crossline spatial sampling for towed-streamer seismic data. Our method combines a partial convolutional neural network (PCNN) with a Transformer architecture and incorporates a jittered sampling algorithm optimized by a genetic algorithm (GA). This innovative integration enables the model to effectively capture latent patterns and extract implicit supervisory information from unlabeled seismic data, overcoming the limitations of traditional interpolation techniques. The proposed method was rigorously validated using both a 3-D salt model and field data. Results demonstrate its ability to significantly improve crossline spatial sampling, leading to enhanced seismic imaging resolution and accuracy. This advancement not only addresses the inherent challenges of towed-streamer surveys but also opens new possibilities for high-quality subsurface characterization in marine seismic exploration.

In recent years, deep learning methods, especially convolutional neural networks (CNNs), have been widely applied in the field of seismic record denoising. The supervised training approach depends on crafted training datasets which contain the pairs of synthetic clean and noisy data. The generation of complete and sufficient datasets is costly. In addition, there are inherent generalization issues of supervised learning due to the synthetic training datasets cannot encompass all of the features of events and noises in field seismic records. To address these challenges, we propose a novel correction-denoising-inverse (CDI) strategy combined with a lightweight multiscale network (LiteMSNet), termed CDI-LiteMSNet. First, a linear invertible flattening transformation (LIFT) is applied to the original seismic records with low signal-to-noise ratios (SNRs) to fully enhance the horizontal coherence of their effective events. Subsequently, these transformed records are fed into the proposed LiteMSNet, which is trained using a few-shot learning approach. Finally, the denoised records are obtained through an inverse flattening transformation (IFT). During this denoising process, we introduce a dynamic edge-aware (DEA) regularization and employ the alternating direction method of multipliers (ADMM) algorithm to suppress false reflections induced by few-shot learning. Thus, our method combines seismic denoising and deep learning to solve the problem of insufficient generalization ability on field seismic records, while ensuring that CNNs can be trained using datasets with pairs of few pure and noisy data. The numerical experiments demonstrate that CDI-LiteMSNet exhibits excellent denoising performance with few-shot training, while also requiring shorter training time.

ABSTRACT Due to certain environmental limitations and economic considerations, seismic interpolation is occasionally used to compensate for sparse and incomplete geometries. In recent years, many deep learning (DL)-based methods have shown great promise in seismic data interpolation. Most DL-based interpolation methods adopt architectures that rely on convolutional neural networks, which are inherently limited in capturing global features, owing to their local receptive fields. Transformer-based architectures have been recently proposed in several attempts to capture such global features and thus improve the interpolation performance. However, the core operation, multihead self-attention (MSA), imposes a substantial computational burden. The purpose of this study was to explore a lightweight and globally-oriented DL architecture for seismic data interpolation that achieves performance comparable to, or even surpassing, transformers while substantially reducing the heavy computational cost. To this end, the Mamba module was, for the first time, adopted in the context of seismic data interpolation. Different from transformers, the Mamba uses the selective state space model (SSM) to capture the global features instead of the MSA. The selection mechanism and hardware-aware algorithm in SSM could significantly alleviate the heavy computational burden by compressing features and reducing memory swaps. Therefore, several Mamba modules were incorporated into a feed-forward neural network, with dense connections introduced to enhance the feature interactions among different layers. In synthetic and field examples, the proposed dense Mamba-based interpolation network achieved a superior trade-off between the computational cost and interpolation performance, compared with two existing architectures: U-Net and an encoder-stacked transformer.

One of the commonly used methods for correcting timing errors in Ocean Bottom Seismograph (OBS) data is linear correction based on the skew value obtained by synchronizing with the Global Positioning System (GPS) after recovery. However, when the timing information recorded by the OBS is incomplete, linear correction may fail to effectively remove the time errors and can even increase the clock drift. Accurate time correction is therefore essential for OBS data processing. In this study, we analyzed Chinese broadband OBS data collected in the South China Sea (SCS) basin between 2019 and 2020, together with data from several surrounding land stations. Depending on the available timing information, we applied a grid search method and ambient noise cross-correlation to perform precise time correction. The results show that the skew values recorded by Chinese broadband OBSs may contain integer-second timing errors, resulting in noticeable linear clock drift even after conventional correction. The grid search method can accurately identify and compensate for these integer-second errors, thereby recovering the complete skew value and enabling effective linear correction. Moreover, this approach is also applicable to OBS data without timing information. After linear correction, some OBS records still exhibited nonlinear clock drift. By performing iterative ambient noise cross-correlation analysis, we retrieved the full clock drift values, reducing the average timing error to 26 ms. This study improves our understanding of the timing behavior of Chinese OBS instruments and refines their time correction procedures, yielding high-accuracy data suitable for studies such as seismic tomography and earthquake location that require precise arrival times.

This paper proposes modified angle-domain dot-product imaging conditions for vector-based elastic reverse time migration, which yields PP and PS angle gathers that capture improved amplitude-versus-angle (AVA) information. The vector P-wavefields and converted S-wavefields can be constructed by separating the coupled elastic wavefields. A simple and stable dot-product imaging condition can be employed to generate scalar PP and PS angle gathers. However, estimating amplitudes via migrated angle gathers using the conventional angle-domain dot-product imaging conditions leads to inaccurate AVA information. The inaccuracy arises because the dot-product imaging condition does not correctly represent seismic PP and PS reflectivity. Hence, the relationship between the dot-product imaging conditions and the reflection coefficients is explicitly derived in this paper. Then, the modified angle-domain dot-product imaging conditions for calculating PP and PS angle gathers are presented based on the kinematic characteristics of the vector P- and S-wavefields, along with an explanation of their amplitude correction mechanism. Finally, simple layer models with positive and negative reflection polarities are used to compute angle gathers to demonstrate the accuracy of the proposed imaging conditions in calculating AVA imaging. Numerical results from the Marmousi2 model validate the effectiveness and correctness of the proposed imaging conditions in complex media.

The Yangtze Craton in South China is generally regarded as an Archean‐Paleoproterozoic continental nucleus with a coherent basement that collided with the Cathaysia Block in the Early Neoproterozoic. Here we reprocessed a 300‐km‐long NW‐trending seismic reflection profile in the eastern Yangtze Craton. High amplitude events reveal the nearly flat Moho at depth of ∼40 km and SE‐dipping mantle reflections extending from ∼54 km to at least 100 km beneath the eastern Sichuan fold‐and‐thrust belt. We interpret these mantle reflections as a fossil subduction zone within the eastern Yangtze Craton, which was probably linked with a buried Paleoproterozoic orogen beneath the Xuefengshan belt. Hence the Yangtze Craton should be formed by the collision or accretion of microcontinents prior to the Neoproterozoic. During NW‐ward crustal shortening in the Middle Jurassic, this mantle suture zone controlled the thick‐to thin‐skinned structural transition from east to west due to tectonic inheritance.

Intermediate-depth earthquakes (IDEs), i.e., earthquakes at depths of 70 to 300 km, have been observed in subduction zones globally and extensively investigated. However, the seismogenic mechanism of IDEs is still controversial, especially in the southern end of the Mariana Trench, where near-field observations are lacking. By using machine-learning-based methods in three sets of near-field Ocean Bottom Seismogram (OBS) network data, we detected and located more than 1,000 intraplate and interplate earthquakes. The seismogenic volumes in different regions of the subducted plate are different, showing the character of double seismogenic zones (DSZ) and single seismicity layer (SSZ). The seismicity features coincide well with the regional landform, development of outer-rise faults, and hydration scenarios, suggesting a dehydration-related mechanism for the generation of IDEs. The subducted slabs experience different degrees of slab hydration, leading to various seismic behaviors.

Compared with regularly and irregularly missing traces, consecutively missing traces (also called large gaps) create a more challenging task in seismic data interpolation. Recently, deep-learning-based methods have proven to be an effective tool for seismic data interpolation. However, widely adopted convolutional neural network (CNN) archi-tectures, which are constrained by their limited receptive fields, emphasize local features while overlooking global features. This inherent characteristic limits their ability to re-construct the more challenging large gaps. Compared with CNN-based architectures, transformers can capture more global features using a multihead self-attention mechanism. Therefore, we develop a novel transformer-based architecture, the dense double-branch attention transformer (D2AT), to in-terpolate large gaps in seismic data when the model is prop-erly trained. The backbone of D2AT is a double-branch structure based on the Swin transformer, which enables it to learn low-and high-resolution features. Moreover, in the high-resolution upper branch, we use dense connections to facilitate feature interactions at different stages, and a spatial attention block is used to refine the extracted features. Exper-imental results on synthetic and field seismic records demon-strate that D2AT is a potential solution for reconstructing the consecutively missing traces in 2D seismic data.

Impact craters represent the most prominent geological features on the Moon. Current methodologies for investigating impact craters primarily involve morphological analysis and numerical simulations. However, direct subsurface structural data of lunar impact craters remain scarce. Utilizing the Lunar Penetrating Radar (LPR) aboard the Yutu-2 rover, we conducted comprehensive subsurface exploration of the Chang'e-4 landing area on the lunar far-side over 65 lunar days, acquiring unprecedented dual-channel radar profiles of a buried Eratosthenian crater, characterized by a rim-to-floor diameter of 484 m and subsurface structural features extending to 215 m depth. Our comprehensive analysis delineates a 12-layer stratigraphic architecture comprising lunar regolith, paleoregolith horizons, ejecta deposits, and basalt units, while simultaneously reveals characteristic subsurface features of lunar impact structures, including reworked zones, radial collapses, central rebounds, and fracture systems. These findings establish a critical foundation for understanding crater formation mechanisms from surface expressions to subsurface responses, while providing new constraints on the impact chronology and origin of impactors within the South Pole-Aitken (SPA) basin.

Post-stack seismic profiles are images reflecting containing geological structures which provides a critical foundation for understanding the distribution of oil and gas resources. However, due to the limitations of seismic acquisition equipment and data collecting geometry, the post-stack profiles suffer from low resolution and strong noise issues, which severely affects subsequent seismic interpretation. To better enhance the spatial resolution and signal-to-noise ratio of post-seismic profiles, a multi-scale attention encoder-decoder network based on generative adversarial network (MAE-GAN) is proposed. This method improves the resolution of post-stack profiles, and effectively suppresses noises and recovers weak signals as well. A multi-scale residual module is proposed to extract geological features under different receptive fields. At the same time, an attention module is designed to further guide the network to focus on important feature information. Additionally, to better recover the global and local information of post-stack profiles, an adversarial network based on a Markov discriminator is proposed. Finally, by introducing an edge information preservation loss function, the conventional loss function of the Generative Adversarial Network is improved, which enables better recovery of the edge information of the original post-stack profiles. Experimental results on simulated and field post-stack profiles demonstrate that the proposed MAE-GAN method outperforms two advanced convolutional neural network-based methods in noise suppression and weak signal recovery. Furthermore, the profiles reconstructed by the MAE-GAN method preserve more geological structures.

With the development of fiber-optic seismology, distributed acoustic sensing (DAS) has made significant progress in vertical seismic profiling (VSP). The integration of full-waveform inversion (FWI) with DAS data acquired through well-bore optical fibers presents a promising frontier for subsurface characterization. Although the successful implementation of FWI can yield precise velocity models essential for reservoir monitoring and imaging, its application to DAS data has been limited predominantly to vertical wells. In such conventional approaches, the strain rate measurements from DAS are typically converted to vertical particle velocities at corresponding channel locations before applying standard FWI algorithms. However, this methodology faces significant limitations when extended to deviated wells, wherein the conversion to vertical particle velocity becomes inapplicable. Addressing this challenge, our study introduces a novel FWI strategy that enables the processing of DAS VSP data across various well configurations. The core innovation lies in the conversion of DAS measurements into scalar particle displacement, an approach that maintains the essential phase information while accommodating amplitude and frequency variations. Through the rigorous derivation of the relationship between DAS data and particle displacement in wavefields, we establish that the fundamental distinction between these measurements resides solely in their frequency components and relative amplitudes, with phase characteristics remaining intact. This displacement-based conversion method offers unprecedented flexibility, allowing for the application of conventional FWI to DAS data from deviated wells through appropriate amplitude adjustments based on wave velocity and incident angles at channel positions. We demonstrate this approach with a comprehensive set of DAS VSP data collected from an offshore inclined well, where the DAS data are directly fed into the FWI algorithm after amplitude scaling and simple preprocessing. FWI reduces data misfit and enhances velocity updates, and the inverted model provides an improved prestack depth migration image along with angle-domain common-image gathers.

Ground-penetrating radar (GPR) has become a pivotal tool in lunar geological exploration, providing direct insight into the Moon's subsurface structure. This review provides a focused summary of GPR applications in the Apollo 17, Kaguya, and Chang'e lunar missions, synthesizing their key scientific findings and technological advancements. In terms of scientific discoveries, GPR observations have revealed a lunar subsurface far more complex than previously thought, characterized by a multi-layered stratigraphy of volcanic deposits, impact ejecta, and ancient buried regolith. Furthermore, GPR has identified numerous unique subsurface features, including lava tubes, buried impact craters, and bifurcated structures. Notably, radar data reveal that the lunar farside's regolith exhibits significantly lower attenuation of radio waves than the nearside's, due to its lower iron and titanium oxide content, which allows for greater penetration depths. The paper also highlights recent advances in lunar GPR data processing, including forward modeling, improved dielectric property estimation through methods such as hyperbolic fitting and direct comparison with returned samples, and the application of advanced migration and machine learning-based imaging methodologies. These techniques have been crucial for converting radar travel times to depth, enhancing image resolution, and accurately characterizing lunar materials. Different GPR systems offer a trade-off between penetration depth and resolution. Orbital sounders can see deep into the subsurface with low resolution, whereas surface rovers provide high-resolution data limited to shallow depths. Therefore, a comprehensive understanding of the lunar subsurface requires a multi-frequency, data-fusion approach. Looking ahead, GPR will be indispensable for future resource-prospecting missions, such as mapping water ice deposits, and for the selection of sites for lunar bases.

Subbottom profilers (SBPs) using shipboard sonar can acquire massive amounts of data during exploration missions. Some geological intrusions and artificially buried objects show hyperbola-shaped signatures in the SBP images. These signatures are valuable information, but their detection is time-consuming. In addition, noise and geometric spread further increase the difficulty of detection. This article proposes an automated detection method of hyperbola-shaped signatures in SBP images by utilizing morphological processing. The proposed method can be summarized into four steps: preprocessing, segmentation, morphological processing, and fitting. The morphological processing is the critical technology in the proposed method, including opening, dilation, and skeletonization. Trend curves of signatures can be outlined without a priori knowledge by exploiting morphological processing. The fitting algorithm can refine the curves further into an analytical curve. We validate the feasibility and effectiveness of the proposed method in field data acquired from the Marianas region. Meanwhile, we demonstrate that the proposed method better detects ill-shaped and large curvature hyperbola-shaped signatures. Compared with the template matching and the column-connection clustering (C3) methods, the proposed method can provide better precision and recall using an optimized threshold. In addition, the proposed method is a general detection methodology that can be applied to any SBP images with proper parameters. In conclusion, the morphological processing presented in this article can be employed as a generic hyperbola-shaped signature detection module in SBP image processing.

Velocity and reflectivity models are fundamental outcomes obtained from seismic exploration, enabling the identification of subsurface structures and physical properties. Two primary methods for obtaining these high-resolution results are two-way wave equation-based full waveform inversion (FWI) and least-squares reverse time migration (LSRTM). Despite sharing a similar least-squares inversion framework, FWI and LSRTM present challenges in effectively combining them to generate both velocity and reflectivity models simultaneously due to their different modeling engines and dependencies on data components. In this study, we propose a novel joint migration inversion (JMI) approach by incorporating a two-way full-wavefield modeling engine parameterized by the subsurface velocity and reflectivity. By providing adjoint velocity and reflectivity sensitive kernels, this JMI can produce high-quality velocity and reflectivity models concurrently. The method developed in this study has the potential to directly process seismic data containing full-wavefield information, reducing data preprocessing complexity and avoiding potential damage to valid signals during the processing. Through a synthetic data test and a benchmark test, we demonstrate the effectiveness of our JMI approach. Moreover, when compared with conventional approaches for velocity building and imaging, our method yields velocity and reflectivity models with higher resolution and improved consistency.

As the strongest component in the worldwide ambient noise field, secondary microseisms are dominantly excited by cyclonic storms. Whether unidirectionally propagating monsoon winds can induce seismically observable microseisms remains unclear. In this study, we use broadband seismic data recorded by seven ocean-bottom seismometers (OBSs) to systematically examine the possible excitation mechanism associated with the summer monsoon 2020 in the South China Sea. Spectrograms of the OBS data reveal remarkable power characterized by two notable peaks during May 18-31 in the short-period secondary microseism band of 0.25-0.5 Hz. In this time period, a monsoon onset occurred (May 23) and no tropical cyclones were recorded either in the South China Sea or Western Pacific. Around the summer monsoon onset, wind fields are transformed from northeasterly to southwesterly, driving ocean waves to propagate northeastward and exciting secondary microseisms by interactions with opposing wave trains driven by the pre-existing northeasterlies, which corresponds to the first power peak observed during May 18-20. After the dissipation of the northeasterlies-driven swell, a cyclone associated with the monsoon developed around Taiwan produces northeasterlies and excites secondary micro- seisms, resulting in the second peak emerged during May 22-26. Polarization analyses indicate that the microseisms are probably amplified by wave shoaling when ocean waves passing from deep to shallow water. Overall, this study provides evidence of a causal relation between monsoon onset and secondary microseisms, highlighting the potential of collecting ocean-bottom seismic data for the early detection and monitoring of monsoons.

Magmatism and crustal structural characteristics are well-known phenomena that support understanding the marginal sea's tectonic evolution. The South China Sea (SCS) has become a natural laboratory for researchers to investigate the evolution of the marginal sea because of its complex tectonic background and multiple phases of magmatic activities. However, due to limitations in resolution and depth of observations, the evolutionary process of the SCS is still controversial. We use the latest processed multi-channel seismic reflection profile Line AB to evaluate the sedimentary sequences, crustal tectonics, and the Cenozoic magmatism in the southern SCS. The gravity modeling results indicate a reduction in crustal thickness from 18 km in the proximal domain to 7 km in the Continent-Ocean Transition (COT). The COT is roughly 35 km wide, representing a rapid break up process of the SCS. A high-velocity layer (HVL) of about 3 km thickness was formed in the continental slope crust, interpreted as post-spreading magmatic underplating. In addition, four sedimentary sequences, post-spreading magmatic intrusion, and the possible magma conduits were identified in the profile. Combining with previous ocean bottom seismometer (OBS) data, multi-channel seismic data, fault system data, and petrologic observations, we indicate the deep mantle magmatic material upwelled to the crust and further intruded into sediments along the hyper-extended crust or detachment faults during the post-spreading stage. We contend that cooling and heat shrinking within the SCS oceanic lithosphere have the potential to initiate decompression melt deformation in the continental slope. In summary, we construct a model of the origin and migration mechanism of extensive post-spreading magmatism in the SCS. Our research reveals the Cenozoic evolution and the HVL of the southern SCS, and provides a possible genesis for the anomalously young magmatism observed at the continental margin.