Semi-Supervised Salt Segmentation Using Mean Teacher

Detecting subsurface salt structures from seismic images is important for seismic structural analysis and subsurface modeling. Recently, deep learning has been successfully applied in solving salt segmentation problems. However, most of the studies focus on supervised salt segmentation and require numerous accurately labeled data, which is usually laborious and time-consuming to collect, especially for the geophysics community. In this paper, we propose a semi-supervised framework for salt segmentation, which requires only a small amount of labeled data. In our method, adopting the mean teacher method, we train two models sharing the same network architecture. The student model is optimized using a combination of supervised loss and unsupervised consistency loss, whereas the teacher model is the exponential moving average (EMA) of the student model. We introduce the unsupervised consistency loss to better extract information from unlabeled data by constraining the network to give consistent predictions for the input data and its perturbed version. We train and test our novel semi-supervised method on both synthetic and real datasets. Results demonstrate that our proposed semi-supervised salt segmentation method outperforms the supervised baseline when there is a lack of labeled training data.

Application Of Finite Difference Eikonal Solver For Traveltime Computation In Forward Modeling And Migration

Traveltime is one of the propagating wave’s components. As the wave propagates further, the traveltime increases. It can be computed by solving wave equation of the ray path or the eikonal wave equation. Accurate method of computing traveltimes will give a significant impact on enhancing the output of seismic forward modeling and migration. In seismic forward modeling, computation of the wave’s traveltime locally by ray tracing method leads to low resolution of the resulting seismic image, especially when the subsurface is having a complex geology. However, computing the wave’s traveltime with a gridding scheme by finite difference methods able to overcomes the problem. This paper aims to discuss the ability of ray tracing and fast marching method of finite difference in obtaining a seismic image that have more similarity with its subsurface model. We illustrated the results of the traveltime computation by both methods in form of ray path projection and wavefront. We employed these methods in forward modeling and compared both resulting seismic images. Seismic migration is executed as a part of quality control (QC). We used a synthetic velocity model which based on a part of Malay Basin geology structure. Our findings shows that the seismic images produced by the application of fast marching finite difference method has better resolution than ray tracing method especially on deeper part of subsurface model.

Lateral migration of large sedimentary bodies in a deep-marine system offshore of Argentina

Contourite features are increasingly identified in seismic data, but the mechanisms controlling their evolution remain poorly understood. Using 2D multichannel reflection seismic and well data, this study describes large Oligocene- to middle Miocene-aged sedimentary bodies that show prominent lateral migration along the base of the Argentine slope. These form part of a contourite depositional system with four morphological elements: a plastered drift, a contourite channel, an asymmetric mounded drift, and an erosive surface. The features appear within four seismic units (SU1–SU4) bounded by discontinuities. Their sedimentary stacking patterns indicate three evolutionary stages: an onset stage (I) (~ 34–25 Ma), a growth stage (II) (~ 25–14 Ma), and (III) a burial stage (< 14 Ma). The system reveals that lateral migration of large sedimentary bodies is not only confined to shallow or littoral marine environments and demonstrates how bottom currents and secondary oceanographic processes influence contourite morphologies. Two cores of a single water mass, in this case, the Antarctic Bottom Water and its upper interface, may drive upslope migration of asymmetric mounded drifts. Seismic images also show evidence of recirculating bottom currents which have modulated the system’s evolution. Elucidation of these novel processes will enhance basin analysis and palaeoceanographic reconstructions.

Automatic fault detection on seismic images using a multiscale attention convolutional neural network

High-fidelity fault detection on seismic images is one of the most important and challenging topics in the field of automatic seismic interpretation. Conventional hand-picking-based and semi-human-intervened fault detection approaches are being replaced by fully automatic methods thanks to the development of machine learning. We develop a novel multiscale attention convolutional neural network (MACNN for short) to improve machine-learning-based automatic end-to-end fault detection on seismic images. The most important characteristics of our MACNN fault detection method is that it employs a multiscale spatial-channel attention mechanism to merge and refine encoder feature maps of different spatial resolutions. The new architecture enables our MACNN to more effectively learn and exploit contextual information embedded in the encoder feature maps. We demonstrate through several synthetic data and field data examples that our MACNN tends to produce higher-resolution, higher-fidelity fault maps from complex seismic images compared with the conventional fault-detection convolutional neural network, thus leading to improved geological fidelity and interpretability of detected faults.

Non-stationary local slope estimation via forward-backward space derivative calculation

The local slope estimated from seismic images has a variety of meaningful applications. Slope estimation based on the plane-wave destruction (PWD) method is one of the widely accepted techniques in the seismic community. However, the PWD method suffers from its sensitivity to noise in the seismic data. We propose an improved slope estimation method based on the PWD theory that is more robust in the presence of strong random noise. The PWD operator derived in the Z-transform domain contains a phase-shift operator in space corresponding to the calculation of the first-order derivative of the wavefield in the space domain. The first-order derivative is discretized based on a forward finite difference in the traditional PWD method, which lacks the constraint from the backward direction. We propose an improved method by discretizing the first-order space derivative based on an averaged forward-backward finite-difference calculation. The forward-backward space derivative calculation makes the space-domain first-order derivative more accurate and better anti-noise since it takes more space grids for the derivative calculation. In addition, we introduce non-stationary smoothing to regularize the slope estimation and to make it even more robust to noise. We demonstrate the performance of the new slope estimation method by several synthetic and field data examples in different applications, including 2D/3D structural filtering, structure-oriented deblending, and horizon tracking.

Recognition and characterization of small-scale sand injectites in seismic data: Implications for reservoir development

Small-scale (< 20 m), non-resolvable sand injectites can constitute a large part of the net-to-gross volume and affect fluid flow in the reservoir. However, they may also cause challenges for well placement and reservoir development because they are too small to be reliably constrained by reflection seismic data. It is therefore important to better understand how small-scale injectites influence seismic images and may be recognized and characterized above reservoirs. The Grane Field (North Sea) hosts numerous small-scale sand injectites above the main reservoir unit, causing challenges for well placement, volume estimates and seismic interpretation. Here, we investigate how such small-scale sand injectites influence seismic images and may be characterized by (1) using well-, 3D seismic- and outcrop data to investigate geometries of small-scale sand injectites (0-15 m) and creating conceptual models of injectite geometries, (2) performing seismic convolution modelling to investigate how these would be imaged in seismic data, and (3) compare these synthetic seismic images to actual 3D seismic from the well-investigate Grane Field.Our results show that despite injectites being below seismic resolution, small-scale sand injectites can be detected in seismic data. They are more likely to be detected with high thickness (> 5 m), steep dip (> 30°), densely spaced sand injectites, and homogeneous background stratigraphy. Furthermore, as fraction of sand injectites increases the top reservoir amplitude will decrease. Moreover, comparison of the synthetic seismic images with real seismic data from the Grane Field indicates that the low-amplitude anomalies and irregularities observed above the reservoir may be a result of the overlying sand injectites. Additionally, the comparison strongly suggests that the Grane Field hosts sand injectites that are thicker and located further away from the top reservoir than what is indicated by well observations. These results may be used to improve well planning and develop reservoirs with overlying sand injectites.Supplementary material: A PDF file containing all the seismic modelling results allowing the reader to flip back and forth between the different models is available at https://www.doi.org/10.6084/m9.figshare.14333102 . Well logs from well 25/11-18 T2 are available at https://factpages.npd.no/pbl/wellbore_documents/2358_25_1_18_COMPLETION_REPORT_AND_LOG.pdf