Deep learning-based automatic horizon tracking has achieved promising results but still faces serious cross-horizon phenomena. Therefore, based on multitask learning, seismic data characteristics, and the seismic horizon sequence (or position) relationship, we have developed a sequence-constrained multitask horizon tracking (SMHT) method for high-precision automatic horizon tracking. SMHT contains the horizon label automatic enrichment, the multitask horizon tracking network (MHTN), and the horizon sequence-constrained loss function. Horizon label automatic enrichment aims to automatically generate the upper and lower auxiliary horizon labels of the target horizon label, and then the horizon region labels corresponding to the auxiliary and target horizon labels for MHTN. MHTN contains the shared layer, the auxiliary task, the main task, and the horizon sequence-constrained horizon correction. In MHTN, the shared layer generates multiscale feature maps of the input seismic data. The auxiliary task uses the concept of object detection to extract the horizon region (or probability), and the main task extracts the high-precision horizon within the extracted horizon region. The horizon sequence-constrained horizon correction with the horizon sequence-constrained loss function aims to avoid the cross-horizon phenomenon and finally obtain precise horizon tracking results. Application of MHTN to two field 3D seismic data sets finds that SMHT performs well in automatic horizon tracking.

High-precision seismic fault detection and fault surface extraction (or grouping) are critical steps in reservoir characterization. In this article, basing on the gradually changing characteristics of the target faults in the adjacent 2-D seismic fault profiles, we propose an efficient and high-precision automatic fault surface grouping (EHFG) method for complex 3-D seismic fault data; and it is realized using pairs of adjacent 2-D seismic fault profiles without human intervention. EHFG comprises of high-precision fault separation (HFS) and high-precision fault labeling (or naming) (HFL), in which HFS aims to separate the positive-slope and negative-slope faults in the seismic fault detection result and ultimately obtain the high-precision positive-slope and negative-slope faults; HFL aims to label the separated positive-slope and negative-slope faults and ultimately obtain the corresponding high-precision fault surface grouping results. An actual 3-D seismic fault data example demonstrates that EHFG is a good potential technique for fault surface grouping.

A great number of high-quality fractured-vuggy reservoirs (FVRs) are buried in the ultradeep carbonate strata in the Tarim Basin, Western China. Drilling confirms that some FVRs with strong beaded reflections are partially or completely filled with solids, resulting in oil and gas exploration failures. In this paper, we propose a well data (the drilling information and the well-logging data) constrained seismic waveform component decomposition and reconstruction method for ultradeep FVR characteristic identification (WSFI). WSFI contains the well data (the drilling information and the well-logging data) constrained seismic waveform component linear discriminant analysis and reconstruction and seismic attribute (i.e., root-mean-square amplitude)-based FVR characteristic identification. A forward-modeling example indicates that WSFI can effectively identify the FVR characteristics. Field data example in the ZG block in the Tarim Basin demonstrates that WSFI can be widely used for FVR characteristic identification.

Fractures are widely distributed in the subsurface and are crucial for hydrocarbon, CCS, offshore infrastructure (windfarms), and geothermal seismic surveys. Seismic anisotropy has been widely used to characterize fractures and has been shown to be sensitive to background matrix porosities in theoretical studies. An understanding of the effects of background porosity on seismic anisotropy could improve seismic characterization in different fractured reservoirs. Based on synthetic rocks with controlled fractures, we conducted laboratory experiments to investigate the influence that background porosity has on P-wave anisotropy and shear wave splitting. A set of rocks containing the same fracture density (0.06) with varying porosities of 15.3%, 22.1%, 26.1% and 30.8% were constructed. The P- and S-wave velocities were measured at 0.5 MHz as the rocks were water saturated. The results show that when porosity increased from 15.3% to 22.1%, P-wave anisotropy and shear wave splitting exhibited slight fluctuations. However, when porosity continued to increase to 30.8%, P-wave anisotropy declined sharply, whereas shear wave splitting stayed nearly constant. The measured results were compared with predictions from equivalent medium theories. Qualitative agreements were found between the theoretical predictions and the measured results. In the Eshelby–Cheng model, an increase in porosity reduces fracture-induced perturbation in the normal direction of the fracture, resulting in lower P-wave anisotropy. In the Gurevich model, an increase in porosity can reduce the compressional stiffness in parallel directions to a larger extent than that in perpendicular directions, thus leading to lower P-wave anisotropy.

Stratigraphic forward modeling (SFM) is a numerical approach used to reconstruct sedimentary basin evolution by simulating the infilling and tectonic evolution process of strata. The challenge is that existing approaches inevitably require trade-offs among modeling fidelity and computational cost. We present a novel clastic stratigraphic forward modeling (CSFM) approach to reducing computational cost while retaining key flow and transport behaviors relevant to stratigraphic architecture. In CSFM, Lagrangian water particles affect momentum and sediment, while a fixed Eulerian grid stores topographic elevation and lithologic fractions. A simplified form of the Navier–Stokes equations is proposed to compute the trajectories of fluid particles, which can greatly reduce the computational cost. Sediment dynamics are represented by coupled suspended load and bedload modules. To validate CSFM, we constructed a synthetic alluvial fan model and performed stratigraphic forward modeling on it. Five lake-level cycles were imposed and results showed that cyclic sand–clay couplets and isolated channel sand bodies were formed during repeated progradation and backstepping. These results are consistent with established sedimentological knowledge, confirming the geological plausibility of CSFM.

The popular deep learning-based horizon tracking methods heavily relies on large volumes of well-labeled horizon data, which face significant challenges in achieving high-precision horizon tracking with limited-label (or few-sample), especially when encountering complex seismic data and geological structures with one-dimensional limited-label. In this paper, we propose a limited-label multi-scale multi-horizon tracking method (LMMT) based on the multi-modal deep learning and one-dimensional limited-label. In this method, the horizon is characterized in the seismic trace (one-dimensional), the seismic profile (two-dimensional), and the horizon slice(two-dimensional). Limited-label multi-dimensional multi-horizon tracking method (LMMT) is comprised of the flowchart of limited-label multi-scale multi-horizon tracking method (LMMT), the one-dimensional convolution kernel single-modal multi-horizon tracking method (OMT), the high-precision high-continuity horizon and strata optimization (HHO), and the two (or three)-dimensional convolution kernel multi-modal multi-horizon tracking method (TMT). One-dimensional convolution kernel single-modal multi-horizon tracking method (OMT) takes the input one-dimensional limited horizon labels as its labels and utilizes a one-dimensional convolution kernel for strata division and multi-horizon tracking. High-precision high-continuity horizon and strata optimization (HHO) aims to generate the two(or three)-dimensional high-precision and high-continuity horizon and strata based on the three-dimensional horizon tracking results obtained by one-dimensional convolution kernel single-modal multi-horizon tracking method (OMT) or two(or three)-dimensional convolution kernel multi-modal multi-horizon tracking method (TMT), thereby providing high-precision high-continuity horizon labels and strata for two(or three)-dimensional convolution kernel multi-modal multi-horizon tracking method (TMT). Two(or three)-dimensional convolution kernel multi-modal multi-horizon tracking method (TMT) incorporates the two(or three)-dimensional high-precision high-continuity horizon obtained by high-precision high-continuity horizon and strata optimization (HHO) as its labels, integrates the random masking result of the high-precision high-continuity strata obtained by high-precision high-continuity horizon and strata optimization (HHO) as the reference information, and utilizes a two(or three)-dimensional convolution kernel for high-precision multi-horizon tracking. Two three-dimensional seismic dataset applications demonstrate that limited-label multi-scale multi-horizon tracking method (LMMT) achieves high-precision multi-horizon tracking results with limited-labels.