The limitations of seismic vertical resolution pose significant challenges for the identification of thin beds. Improving the vertical resolution of seismic data using deep learning methods often encounters challenges related to unrealistic outputs and limited generalization. To address these challenges, we propose a novel framework that improves the fidelity and generalization of seismic super-resolution. Our approach begins with the generation of realistic synthetic training data that aligns with the structural and amplitude characteristics of field surveys. We then introduce an enhanced 2D network with 3D awareness, which builds on the 2D Swin-Transformer and 3D convolution blocks to effectively capture 3D spatial features while maintaining computational efficiency. This network addresses the limitations of traditional 2D approaches by reducing stitching artifacts and improving spatial consistency. Finally, we develop a prior-informed fine-tuning strategy using field data without the need for labels, which incorporates a self-supervised data consistency loss and a spectral matching loss based on prior knowledge. This strategy ensures that the super-resolution results preserve the original low frequency information while yielding a spectral distribution as expected. Experiments on multiple field datasets demonstrate the robustness and generalization capability of our method, making it a practical solution for seismic resolution enhancement in diverse field datasets.

Seismic migration is a tool to obtain images of underground structures; however, it requires accurate velocity models. Errors in estimated migration velocities lead to defocused and distorted migration images. We propose a deep learning method for accurate seismic imaging in the presence of velocity errors. Our idea is to correct the distorted common image gathers (CIGs) due to velocity errors by using a convolutional neural network (CNN). We design a CIG-to-CIG (CIG2CIG) CNN, in which both the inputs and outputs are CIGs. Furthermore, we apply velocity constraints to the CIG2CIG CNN, forming another velocity-constrained CIG2CIG (VC-CIG2CIG) CNN to perform the same task. To train the two CNNs, we create hundreds of true and wrong velocity models, which are applied to migration to produce true CIGs and distorted CIGs, respectively. Experiments demonstrate that the VC-CIG2CIG network is superior to the CIG2CIG network in correcting distorted CIGs and suppressing artifacts.

Sparse distributions of seismic sensors and sources pose challenges for subsurface imaging, source characterization, and ground motion modeling. While large-N arrays have shown the potential of dense observational data, their deployment over extensive areas is constrained by economic and logistical limitations. Numerical simulations offer an alternative, but modeling realistic wavefields remains computationally expensive. To address these challenges, we develop a multi-conditional diffusion transformer for generating seismic wavefields without requiring prior geological knowledge. Our method produces high-resolution wavefields that accurately capture both amplitude and phase information across diverse source and station configurations. The model first generates amplitude spectra conditioned on input attributes and subsequently refines wavefields through iterative phase optimization. We validate our approach using data from the Geysers geothermal field, demonstrating the generation of wavefields with spatial continuity and fidelity in both spectral amplitude and phase. These synthesized wavefields hold promise for advancing structural imaging and source characterization in seismology.

Seismic stratigraphic interpretation of clinothems is essential for regional basin stratigraphy analysis, hydrocarbon reservoir evaluation, and basin evolutionary history reconstruction. Traditional interpretation process remains a tedious, labor-intensive, expert-dependent, and highly subjective process, often resulting in high uncertainty and leaving large portions of seismic data uninterpreted. Data-driven AI approaches are promising to automate this process, but still face significant challenges, including the absence of training dataset, poor generalizability, weak interpretability, and physical inconsistency. To address these challenges, we develop a data-driven and geologically-informed AI workflow that integrates geological knowledge into both labeled dataset construction and AI model training. For the labeled dataset construction, we develop a geological and geophysical forward modeling workflow to simulate a massive-scale labeled synthetic dataset, encompassing diverse geological models. For the model training, we incorporate both labeled synthetic and unlabeled field datasets to jointly train a network utilizing a hybrid loss with labeled supervision and geologically-informed constraints, enabling high-precision, high-resolution, strong-robustness, and high-generalizability interpretation of all stratigraphic features within seismic data. This workflow significantly accelerates and simplifies the traditional interpretation process, achieving seismic pixel-level precision and providing detailed geological insights of the origin and architecture of sedimentary basins. Additionally, our workflow lays the groundwork for constructing a refined and quantitative sequence stratigraphic framework and can be easily extended to various geological scenarios, such as depositional system tract interpretation, shoreline or shelf-edge trajectory tracking, stratigraphic completeness analysis, sea-level curve prediction, and sedimentary evolutionary history reconstruction.

In prestack Kirchhoff depth migration, the quality of the migration profile is determined by how well common image gathers (CIGs) are flattened and corrected. The migration velocity analysis (MVA) method is proposed to flatten the offset domain CIGs (ODCIGs) by updating the migration velocities. However, conventional MVA such as the residual curvature analysis (RCA) method is typically challenging to complex structures such as lateral velocity variations or high-dip reflectors. In addition, there are structural artifacts in ODCIGs due to the multipath ray problem even if the migration velocity is accurate. To address the above problems, we developed a kernel prediction network (KPN) for ODCIG flattening and correction. Compared with conventional neural networks, the primary advantage of the KPN is that its outputs consist of a series of predicted kernels instead of pixel vectors or matrices. These predicted kernels are capable of processing the input ODCIGs pixel by pixel and slice by slice. The KPN is built by an encoder-decoder architecture, and we modified the loss function of the KPN and introduced an extra parameter associated with the migration offset to ensure that the network is more effective for the ODCIG problem.The training samples of the KPN are acquired by a random extraction algorithm, and the corresponding labels are calculated by a convolution method.Image enhancements are also applied in training samples to improve the generalization capability of the KPN. We demonstrate the effectiveness of the KPN method by comparing it with the RCA method in both synthetic and field data examples. The results show that the KPN method can flatten the events in ODCIGs, correct the depth of the improperly migrated reflectors, remove unfocused artifacts simultaneously, and further yield high-quality migration profiles in different geological examples.

In seismology, while training a specific deep learning model for each task is common, it often faces challenges such as the scarcity of labeled data and limited regional generalization. Addressing these issues, we introduce SeisCLIP: a foundation model for seismology, leveraging contrastive learning during pre-training on multi-modal data of seismic waveform spectra and the corresponding local and global event information. SeisCLIP consists of a transformer-based spectrum encoder and an MLP-based information encoder that are jointly pre-trained on massive data. During pre-training, contrastive learning aims to enhance representations by training two encoders to bring corresponding waveform spectra and event information closer in the feature space, while distancing uncorrelated pairs. Remarkably, the pre-trained spectrum encoder offers versatile features, enabling its application across diverse tasks and regions. Thus, it requires only modest datasets for fine-tuning to specific downstream tasks. Our evaluations demonstrate SeisCLIP’s superior performance over baseline methods in tasks like event classification, localization, and focal mechanism analysis, even when using distinct datasets from various regions. In essence, SeisCLIP emerges as a promising foundational model for seismology, potentially revolutionizing foundation-model-based research in the domain.

Borehole images contain the physical information and chemical properties of geological formations, which are crucial for high-resolution interpretation of subsurface stratigraphic and structural features and geological modeling of the subsurface. However, due to the special design of borehole tools and variations in borehole diameter, all kinds of borehole images (FMI, Earth-imager, OMRI, and OBMI) obtained from scanning the borehole walls exhibit varying degrees of data missing, with OBMI data missing up to 70%. We propose a deep-learning approach with a hybrid CNN and Transformer architecture to fill in the gaps in borehole images, addressing the challenges of missing training labels and filling large-scale gaps. To solve the challenge of missing labels of complete borehole images, our deep-learning model is pretrained on a vast collection of complete natural and seismic images and then fine-tuned with a partial loss function on incomplete borehole images. A multistage completion strategy is further introduced into the inference stage to enhance the continuity and textural features of the completed areas. In addition, by incorporating the circular consistency constraint between the left and right sides of the borehole image, our method can reasonably complete the gaps with highly consistent features on both sides of the image. During the tests on borehole images from multiple wells in different work areas with various geological features, our model is capable of completing any type of borehole image with masks of any size, ultimately yielding complete images free of any artifacts, while also possessing richer and more reasonable textures and semantic information. We have open-sourced the code and the fine-tuned models, which are available athttps://github.com/zgyustc/LogMAT/tree/master.

Deep learning (DL) models can exhibit powerful feature extraction ability once trained on massive, diverse labeled datasets. However, such datasets are typically lacking in geophysics, limiting the performance of DL models on solving geophysical problems. The high complexity and diversity of field data in the inference step further constrains the generalizability of DL models trained on only limited data. To address these challenges, incorporating geophysical knowledge as constraints has proven effective. In seismic interpretation, seismic attributes provide quantified geophysical insights into subsurface situation and can be used as constraints to enhance the performance of DL models. We consider the geophysically meaningful attributes as feature channels and propose optimal ways to select and incorporate suitable seismic attributes as constraints on DL models, with a focus on seismic fault interpretation. Specifically, we utilize U-Net as the DL architecture and explore three different ways of incorporating seismic attributes at the input, encoder, and decoder stages. Extensive field data applications show that seismic attributes, as features of seismic data carrying prior information, can compensate for the limitations of DL models caused by the lack of training datasets and the diversity of inference data in geophysics. This is reflected in improved accuracy, generalizability, and geological consistency of predictions by DL models with constraints. It is also discovered that simplified DL models with seismic attribute constraints outperform the complex DL models without such constraints. This further demonstrates that incorporating seismic attributes as constraints allows for the simplification of DL models while improving the prediction accuracy.

Horizon extraction and fault detection are essential in seismic interpretation and closely related to each other. Most existing methods tend to deal with these two tasks independently, and may not work well in interpreting seismic images with complex geologic structures. We propose a multi-task learning (MTL) network with two branches to extract all horizons and detect faults simultaneously by estimating a relative geologic time (RGT) map as well as computing a fault map. These two branches share training datasets, feature maps, and network parameters during the training. The RGT estimation branch, constructed with a transformer architecture, is more lightweight compared to previous CNN methods but provides a larger and structure-oriented receptive field to adaptively capture global structural information for estimating a globally optimal RGT map. The fault detection branch is a simple convolutional neural network (CNN) which merges feature maps shared by the transformer and the derivatives of the estimated RGT to compute a fault map. The fault detection branch provides boundary control for the RGT estimation branch while the later provides global constraints for the former to improve its robustness to noise. Note that our RGT estimation by globally fitting all structures in a seismic image is a volumetric horizon interpretation method with which we are able to obtain a whole volume of horizons, all at once, by simply extracting contours of the RGT map. In our method, we further enable convenient human interactions by integrating manually interpreted horizons (or horizon segments) into the network, which imposes expert knowledge on the network to estimate reasonable RGT results from seismic images with complex fault systems, unconformities, and poor data quality. Moreover, when using 3D horizons as constraints, we are able to decompose the computationaly expensive 3D RGT estimation from a seismic volume into independently parallel 2D estimations slice by slice and combine them to obtain a laterally consistent 3D result.

Reverse time migration (RTM) is a technique used to obtain high-resolution images of underground reflectors; however, this method is computationally intensive when dealing with large amounts of seismic data. Multi-source RTM can significantly reduce the computational cost by processing multiple shots simultaneously. However, multi-source-based methods frequently result in crosstalk artifacts in the migrated images, causing serious interference in the imaging signals. Plane-wave migration, as a mainstream multi-source method, can yield migrated images with plane waves in different angles by implementing phase encoding of the source and receiver wavefields; however, this method frequently requires a trade-off between computational efficiency and imaging quality. We propose a method based on deep learning for removing crosstalk artifacts and enhancing the image quality of plane-wave migration images. We designed a convolutional neural network that accepts an input of seven plane-wave images at different angles and outputs a clear and enhanced image. We built over 500 1024×256 velocity models, and employed each of them using plane-wave migration to produce raw images at 0°, ±10°, ±20°, and ±30° as input of the network. Labels are high-resolution images computed from the corresponding reflectivity models by convolving with a Ricker wavelet. Random sub-images with a size of 512×128 were used for training the network. Numerical examples demonstrated the effectiveness of the trained network in crosstalk removal and imaging enhancement. The proposed method is superior to both the conventional RTM and plane-wave RTM (PWRTM) in imaging resolution. Moreover, the proposed method requires only seven migrations, significantly improving the computational efficiency. In the numerical examples, the processing time required by our method was approximately 1.6% and 10% of that required by RTM and PWRTM, respectively.

Seismic interpretation is often limited by low resolution and strong noise data. To deal with this issue, we propose to leverage deep convolutional neural network (CNN) to achieve seismic image super-resolution and denoising simultaneously. To train the CNN, we simulate a lot of synthetic seismic images with different resolutions and noise levels to serve as training data sets. To improve the perception quality, we use a loss function that combines the$\ell _{1}$loss and multiscale structural similarity loss. Extensive experimental results on both synthetic and field seismic images demonstrate that the proposed workflow can significantly improve the perception of quality of original data. Compared to conventional methods, the network obtains better performance in enhancing detailed structural and stratigraphic features, such as thin layers and small-scale faults. From the seismic images super-sampled by our CNN method, a fault detection method can compute more accurate fault maps than from the original seismic images.

Seismic horizon extraction is important for subsurface structure interpretation and reservoir modeling. Recently developed learning-based seismic interpretation methods show great success in the case of sufficient labeled data but may fail when the labels are limited. Thus, we propose a simple but effective network for seismic horizon identification using only a limited number of labels. To avoid overfitting in the training, we introduce the mechanism of semi-supervised learning (SSL) with virtual adversarial training (VAT). With several seed points, the method can provide a good prediction and suggest regions lacking control points. By adding several seed points in these suggested regions, the performance of the network can be further improved, which can be regarded as an interactive way. In addition, iteratively retraining the network by using the previous high-confidence prediction can further refine the horizon identification. We, finally, compute a full horizon surface without holes and outliers by optimally fitting the horizon points identified by our SSL and reflection slopes estimated from the seismic amplitude image. Applications to two field datasets show our method is superior to conventional methods in picking a seismic horizon with significant waveform variations or across complex discontinuities, such as faults.

With the dramatic growth and complexity of seismic data, manual seismic facies analysis has become a significant challenge. Machine learning and deep learning (DL) models have been widely adopted to assist geophysical interpretations in recent years. Although acceptable results can be obtained, the uninterpretable nature of DL (which also has a nickname “alchemy”) does not improve the geological or geophysical understandings on the relationships between the observations and background sciences. This article proposes a noble interpretable DL model based on 3-D (spatial-spectral) attention maps of seismic facies features. Besides regular data-augmentation techniques, the high-resolution spectral analysis technique is employed to generate multispectral seismic inputs. We propose a trainable soft attention mechanism-based deep dilated convolutional neural network (ADDCNN) to improve the automatic seismic facies analysis. Furthermore, the dilated convolution operation in the ADDCNN generates accurate and high-resolution results in an efficient way. With the attention mechanism, not only the facies-segmentation accuracy is improved but also the subtle relations between the geological depositions and the seismic spectral responses are revealed by the spatial-spectral attention maps. Experiments are conducted, where all major metrics, such as classification accuracy, computational efficiency, and optimization performance, are improved while the model complexity is reduced.

The low-rankness property of the Hankel matrix formulated from the clean seismic data corresponding to a few number of linear events has been successively leveraged in many low-rank (LR) approximation methods for seismic data denoising. The common scheme in these rank-reduction methods is to compute the best LR approximation of the formulated Hankel matrix and then obtain the denoised data from the LR matrix. However, without utilizing the Hankel structure when computing the LR approximation, if we rearrange the denoised data into a Hankel matrix, it is in general not exactly LR as expected. In this paper, we propose a Hankel LR (HLR) approximation method to simultaneously exploit both the Hankel structure and the LR property underlying the clean seismic data. The formulated HLR approximation problem is solved by an alternating-minimization-based algorithm. We provide rigorously convergence analysis of the proposed algorithm. The superior performance of the proposed HLR approximation method is demonstrated on both synthetic and field seismic data.

We simultaneously estimate fault probabilities, strikes, and dips directly from a seismic image by using a single convolutional neural network (CNN). In this method, we assume a local 3-D fault is a plane defined by a single combination of strike and dip angles. We assume the fault strikes and dips, respectively, are in the ranges of [0°, 360°] and [64°, 85°], which are divided into 577 classes corresponding to the situation of no fault and 576 different combinations of strikes and dips. We construct a 7-layer CNN to classify the fault strike and dip in a local seismic cube and obtain the classification probability at the same time. With the fault probability, strike and dip estimated at some seismic pixel, we further compute a fault cube (centered at the pixel) with fault features elongated along the fault plane. By sliding the classification window within a full seismic image, we are able to obtain a lot of overlapping fault cubes which are stacked to compute three full images of enhanced and continuous fault probabilities, strikes, and dips. To train the CNN model, we propose an effective and efficient workflow to automatically create 900 000 synthetic seismic cubes and the corresponding fault class labels. Although trained with only synthetic data sets, our CNN model can be applied to accurately estimate fault probabilities, strikes, and dips within field seismic images that are acquired at totally different surveys. With the estimated three fault images, we further construct fault cells that are represented as small 3-D squares, each square is colored by fault probability and oriented by fault strike and dip. We recursively link the fault cells by following the fault strikes and dips to finally construct fault skins, which are simple linked data structures to represent fault surfaces.