The geometry and topology of shale pore-fracture systems govern hydrocarbon migration and control the feasibility of geological carbon dioxide storage in shale reservoirs. This study examines lacustrine shale across a range of maturities by integrating (ultra) small-angle neutron scattering, repeated mercury intrusion capillary pressure, field-emission scanning electron microscopy, and computed tomography following Wood's metal impregnation. The pore system is divided into four pore-size classes, and their volumes and connectivity are tracked with increasing thermal maturity. At low maturity, mechanical compaction and early cementation reduce the total pore volume and concentrate connected porosity in fractures. As maturity increases, newly formed organic-matter pores lead to a modest increase in total pore volume, while liquid hydrocarbons generated within the oil window occupy part of the pore space and weaken pore-fracture connectivity. At high maturity, the secondary cracking of liquid hydrocarbons to gas raises pore pressure, partially reopens previously sealed pores and fractures, and enhances both total pore volume and pore-fracture connectivity. These results indicate that mature to high-mature lacustrine shales provide more pore surface area, storage space, and connected pathways for the long-term storage of carbon dioxide than low-maturity shales.

Tuff-rich mixed shales,representing an important class of unconventional hydrocarbon resources,exhibit diverse components,rapid lithofacies variations,complex diagenetic evolution,and strong heterogeneity.For these shales,there remains a lack of a systematic understanding of the diagenetic evolutionary processes and reservoir formation mechanisms of different lithofacies,restricting target area selection and evaluation for shale oil exploration.In this study,we investigate the diagenesis,reservoir formation,and favorable exploration targets of mixed shales with unique components in the 2nd Member of the Lucaogou Formation(also referred to as the Lu 2 Member)in the Santanghu Basin.To this end,a range of test methods are employed,including core characterization,thin section observation,scanning electron microscopy(SEM),whole-rock X-ray diffraction(XRD)analysis,electron probe microanalysis,high-pressure mercury injection(HPMI),and nuclear magnetic resonance(NMR),along with measurements of porosity,permeability,and oil saturation.The results indicate that the mixed shales of the Lu 2 Member are composed primarily of tuffaceous materials,carbonates,and organic matter and can be classified into 10 lithofacies,which are frequently interbedded.The mixed shales mainly contain nano-to micro-scale intercrystalline pores in dolomites,devitrification-induced pores in volcanic ash,and dissolution pores,suggesting complex pore structures.Distinct lithofacies exhibit significant differences in physical and oil-bearing properties.Among these,massive lithofacies featuring low organic matter abundance and composed primarily of a single component display the most favorable properties,followed by lamellar transitional lithofacies dominated by dolomites,while lamellar transitional lithofacies composed primarily of tuffaceous materials show the poorest physical and oil-bearing properties.The shale component types and their differential diagenetic evolution govern reservoir quality.Rapid deep burial and compaction in the early stage represent primary factors responsible for the deterioration of reservoir physical properties.However,they occurred earlier than other diagenetic processes.Furthermore,dolomitization and devitrification occurred before organic acid-induced dissolution.This diagenetic evolutionary sequence provides effective spaces for organic acid migration while also offering a material basis for dissolution,serving as the key mechanism behind the formation of high-quality reservoirs.A comprehensive analysis reveals that favorable exploration targets in the Lu 2 Member include the basin margin zone and the slope zone near the basin margin.

Conventional macroscopic approaches to reserve estimation and productivity forecasting are inadequate for assessing shale oil systems characterized by heterogeneous wall-surface wettability and nanoscale confinement effects. As such, a deeper understanding of the microscale mechanisms governing shale oil occurrence, along with robust quantitative characterization methods, is essential. This study focuses on clay-laminated shale oil from the Cretaceous Qingshankou Formation in the Songliao Basin, employing a combination of swelling-extraction experiments, SEM-AFM pore structure analysis, and advanced molecular dynamics (MD) simulations to explore how multi-component shale oil is distributed within slit-like pores of clay minerals and various kerogen types. The results demonstrate that the kerogen-based model provides a more precise representation than the graphene model in capturing the impacts of organic matter composition, oil component polarity, and varying pressure-temperature conditions on adsorption behavior. Based on a spherical-pore model and fractal theory, we establish a specific surface area evaluation model that couples porosity, pore size distribution, and surface roughness. By integrating MD-derived swelling parameters with their evolution across thermal maturity stages, a ternary quantitative occurrence model was developed, encompassing swelling-bound, adsorbed, and free oil states. The model delineates distinct maturity-dependent transitions: kerogen-swelling oil predominates at low maturity; free oil accumulates rapidly during the mature stage; and compaction at high maturity leads to a moderate decline in free oil content. Clay-laminated Qingshankou shale oil, characterized by high clay mineral content and abundant organic matter, serves as a representative case. Based on above new methodology, the free oil resource in Member 1 of the Qingshankou Formation is estimated at 0.94×109 t. In the northern upper sub-member (Q8–Q9 reservoirs) of the Gulong Sag, free oil is predominantly enriched in inorganic pores (up to 9.0 mg/g), whereas in the lower sub-member (Q1–Q4 reservoirs), it is mainly hosted within organic pores (up to 6.0 mg/g). The upper sub-member of the Qijia Sag shows development potential (7.0–10.7 mg/g inorganic-pore free oil). In contrast, the Sanzhao Sag and Changyuan Anticline currently lack conditions for large-scale economic development because free-oil contents are <4.0 mg/g. The occurrence-state framework and quantitative characterization method proposed here provide a multiscale quantitative tool for sweet-spot selection and development scheme design in shale oil plays.

Abstract Chemometrics analysis integrated with biological markers and stable isotopic data offers an effective approach for unravelling crude oil mixing in complex multilayered unconventional plays. In this study, we investigated petroleum fractionation during primary migration and later production by comparing biomarker compositions and stable carbon isotope signatures of extractable organic matter (EOM) and produced crude oils from five shale–sandstone interbeds within a multilayered, organic-and liquid-rich shale system. Biomarkers and isotopic parameters showing minimal fractionation were selected for chemometrics analysis to explore genetic relationships between produced oils and EOM from individual reservoirs and to quantify their relative contributions over four months of production. The results show that hydrocarbon fractionation primarily occurs during oil migration and production from shale to sandstone, leading to an increase in the relative abundance of saturate and aromatic fractions compared to their relative proportions in the source material. Stable carbon isotopic ratios exhibit negligible variation during upward vertical migration but progressively become lighter during lateral migration in horizontal wells. Although the concentrations of tricyclic terpanes, hopanes and steranes were notably varied relative to the saturate fraction, the relative distributions of individual homologues remained largely unchanged, which justifies the applicability of ratio-based chemometric approaches. In contrast, n-alkanes, alkylcyclohexane and naphthalenes display clear fractionation, with enrichment of lighter-molecular-weight components. Further comparison of biomarkers revealed that long-distance lateral flow in horizontal wells enhances compositional fractionation compared to vertical wells. Chemometric results clearly indicated that the sandstone layers are the primary reservoir, which contributes more than 50% to the produced oil, with a gradual increase in quantities during the subsequent four months of production. Conversely, following fracturing of sandstone layers in horizontal wells, oil contributions from the adjacent shale layers are limited in volume and unsteady. The integrated chemometric framework presented here offers a reliable tool for effectively resolving oil mixing from multiple sources in hydraulically fractured complex unconventional shale plays. These insights can support more informed decisions related to future drilling strategies, fracturing design, and overall field development.

The Late Ordovician-early Silurian transition witnessed intense tectono-thermal activity that generated widespread fracture systems in sedimentary basins worldwide. Those fractures are not only structural features but also preserve critical records of fluid flow and tectonic evolution. Despite extensive research on the organic-rich shales of the Wufeng-Longmaxi Formation and their importance as a major unconventional reservoir, the timing of fracture development, the sources of fracture-filling fluids, the nature of fluid-rock interactions, and their linkage to tectonothermal events remain poorly constrained. To address this gap, we investigated vein-filling fractures in the Wufeng-Longmaxi shales from a weakly deformed zone using a multipronged approach, including petrographic observation, fluid inclusion microthermometry, micro-laser Raman spectroscopy, and isotopic analysis. The results show a decrease in fracture vein fill along the vertical section, with four distinct fracture types (types I-IV) identified, including two subtypes within type III (type III1 and type III2). Most fractures are filled with calcite, followed by quartz, or a mixture thereof. Type I and type II fractures, which are nearly horizontal and filled with calcite and pyrite, respectively, represent early-stage mineralization associated with synsedimentary processes and initial basin subsidence. Nearly vertical type III1 fractures, filled with delta 13C-depleted fibrous calcite and often solid bitumen, reflect organic fluid migration driven by tectonic compression and fluid overpressure during the Indosinian orogeny. Oblique type III2 fractures contain twin-crystal and delta 18O-depleted calcite, along with occasional siliceous veins and mafic bands, suggesting hydrothermal fluid activity linked to volcanic processes. The dominance of CH4 vapor inclusions, with minor CO2, further supports a volcanic degassing origin. These features indicate that type III2 fractures formed within the Wufeng-Longmaxi shales during postdepositional tectonic uplift along the Yangtze margin associated with the Himalayan orogeny. Ongoing postdepositional tectonic uplift related to the Himalayan orogeny resulted in the development of low-angle type IV fractures, characterized by late-stage mineral precipitation. These findings offer refined constraints on the diagenetic evolution of the shales and provide new insights into fluid-rock interaction driven by multistage postdepositional tectono-thermal events from the Late Ordovician to the Cenozoic in South China.

The distribution of in-situ stress in sedimentary basins worldwide is complex and heterogeneous, yet its fundamental control on reservoir quality and permeability anisotropy remains inadequately understood, posing a universal challenge for unconventional resource exploration. This study investigates this critical mechanism within the continental shale oil system of the southern Songliao Basin, which typically exhibits ultra-low porosity and permeability. We focus on the southern Songliao Basin as a case study, integrating experimental data (Kaiser tests and dynamic permeability measurements under variable confining pressures) with field observations to decipher how in-situ stress governs multiscale reservoir properties and identifies sweet spots. The basin is divided into three structural domains based on relative overpressure gradients: (1) a northern transition zone (high to normal pressure), (2) a central high-pressure zone, and (3) a southern transition zone (under to normal pressure). Our results reveal distinct geomechanical regimes: the north is characterized by strike-slip and thrust faulting (compressional), whereas the central domain exhibits a normal faulting regime (extensional). Permeability sensitivity to confining pressure (10–50 MPa) is strongly stress-dependent. In the shallow Member 1 of the Cretaceous Qingshankou Formation, horizontal permeability increases significantly with confining pressure, while vertical permeability at greater depths shows high variability. Notably, lower relative overpressure in the northern and central domains correlates with higher daily oil production. The central domain emerges as the most favorable target, exhibiting superior hydraulic fracability, a lateral pressure coefficient below 1.00, and optimal permeability potential within the Changling Sag. The insights gained here provide a transferable framework for evaluating in-situ stress controls on reservoir quality, enhancing predictive models for shale oil exploitation in analogous continental basins globally.

Characterization of microfracture systems and nanopore networks in organic-rich shales is critical for understanding fluid transport, hydrocarbon storage capacity, and the feasibility of carbon dioxide (CO₂) sequestration. However, the inherent heterogeneity, complex morphology, and nano-scale features of shale microstructures present significant challenges for conventional image segmentation techniques. This study reviews and evaluates the performance of state-of-the-art deep learning architectures for automated high-precision segmentation of microfractures and pore systems in FIB-SEM images of organic-rich shales. To enhance the generalizability of each model, a hybrid training dataset comprising 5,000 real and 5,000 synthetically generated GAN-based FIB-SEM images is evaluated. Quantitative analysis reveals that Kite-Net (KiU-Net) outperforms both Swin UNET Transformers (Swin-UNETR) and Attention U-Net, achieving an overall segmentation accuracy of 94%, precision of 94%, and recall of 93%. Notably, KiU-Net excels in accurately delineating microfractures and complex pore geometries within kerogen-rich matrices. Based on KiU-Net's superior validation performance compared to the two other deep learning models, we employed it to segment 3D FIB-SEM image stacks, enabling volumetric reconstruction and analysis of pore connectivity. Results revealed marked morphological distinctions between organic and inorganic pores, with over 94% of pores existing as isolated, non-percolating clusters, a finding consistent with prior geological investigations. Cross-validation considering various shales and coals further validates the model's effectiveness across a range of lithofacies. Our study presents a scalable deep learning framework for analyzing nanoscale shale microstructures from images.

Shale oil and gas, as important unconventional resources in China, play a crucial role in optimizing the energy structure and enhancing energy self-sufficiency. Shale reservoirs are typically characterized by low porosity and low permeability, and the development degree and spatial architecture of natural fracture systems directly control the occurrence state and flow capacity of oil and gas. Therefore, conducting detailed research on reservoir fracture systems is a core geological issue for predicting shale oil and gas "sweet spots" and achieving economical development. Based on the latest domestic and international research findings, an integrated macro‑ and micro‑scale approach with multi‑technology collaboration was adopted to systematically investigate the types of shale fractures, their integrated characterization, main controlling factors, and their roles in reservoir formation and hydrocarbon accumulation. The results indicate that natural fractures in shale form a complex network system characterized by multi‑genesis and multi‑scale ordered development. Geologically, they can be classified into three major categories: tectonic fractures, diagenetic fractures, and overpressure‑related fractures. Their scales span from nanometers to kilometers, requiring integrated characterization through seismic prediction, imaging logging, core observation, thin‑section analysis, scanning electron microscopy, and other techniques. Fracture development is controlled by the coupling of multiple geological factors: tectonic fractures are jointly influenced by tectonic activity, rock brittleness, and mechanical layer thickness; diagenetic fractures are mainly controlled by lamina combinations, total organic carbon content and thermal maturity; while the development of overpressure‑related fractures critically depends on the formation and preservation conditions of overpressure mechanisms such as hydrocarbon‑generation pressurization and under‑compaction. The study further reveals that fracture development exerts a significant “double‑edged sword” effect on shale oil and gas accumulation: constructively, a moderately developed fracture network can effectively increase storage space, significantly enhance flow capacity, connect isolated pores, form dominant migration pathways, and control the distribution of high‑productivity “sweet spots”； destructively, excessively developed fractures (especially late‑stage tectonic fractures) can compromise the self‑sealing capacity of shale layers, leading to the escape of accumulated hydrocarbons, and may trigger strong hydrocarbon expulsion fractionation, resulting in poorer composition and reduced mobility of residual oil. Therefore, the optimal configuration of fracture type, scale, density, and current stress state to form a “moderately developed” fracture network is key to the efficient enrichment and high production of shale oil and gas. The above research findings provide an important theoretical basis for the detailed characterization and efficient exploration and development of shale oil and gas reservoirs in China.

Abstract A comparative evaluation of depositional controls and pore evolution in marine shales is essential for understanding differential shale-gas enrichment in South China. This study investigates the Lower Silurian Longmaxi Formation and the Lower Cambrian Niutitang Formation in northwestern Guizhou to clarify the controls on depositional environment, organic matter (OM) enrichment, and reservoir quality. Integrated analyses, including X-ray fluorescence elemental geochemistry, field-emission scanning electron microscopy, low-temperature nitrogen adsorption, and thermal maturity assessment, were conducted on 210 samples from wells XY1 and RY2. Results show that the Niutitang Formation was deposited under more strongly restricted and reducing conditions with higher paleoproductivity and higher total organic carbon (TOC) contents (1.28–11.8 wt %, average 5.58 wt %), but it also exhibits significantly higher thermal maturity, with equivalent vitrinite reflectance (EqRo) values of 2.13%–3.41% (average 2.92%). In contrast, the Longmaxi Formation formed in a semi-restricted reducing setting with moderate paleoproductivity, lower TOC, and relatively lower maturity (EqRo 2.01%–2.65%, average 2.24%). Despite lower TOC abundance, the Longmaxi Formation displays superior reservoir properties: higher Brunauer–Emmett–Teller (BET) specific surface area (average 16.95 m2/g), and much larger average pore diameter (50.77 nm versus 4.80 nm), together with well-developed honeycomb-like organic pores and microfractures. Conversely, the Niutitang Formation is characterized by excessive thermal evolution, severe deterioration of organic pores, and pore space dominated by inorganic intraparticle pores. These results indicate that thermal maturity exerts a stronger control than OM abundance on reservoir quality in overmature marine shales. Accordingly, the Longmaxi shale is more favorable for free-gas enrichment, whereas Niutitang shale exploration should target relatively low-maturity and shallowly buried areas around paleo-uplifts. This framework provides practical guidance for sweet-spot prediction in overmature marine shale-gas systems within structurally complex regions.

The sealing capacity of fault zones, which fundamentally influences hydrocarbon migration and entrapment, is predominantly governed by their internal architecture. In sand-clay sequences, this capacity typically correlates positively with clay content. However, in the study area-the X492 trap bounding fault in the Huimin depression, Bohai Bay Basin of China, demonstrates effective sealing despite occurring in sand-rich sequences and having limited displacement. The reservoir description results showed that the height of the oil column sealed by this bounding fault reaches 30 m, and the clay content within the fault zone, as indicated by the Shale Gouge Ratio, is generally below 15%. To determine the cause, we studied the subcore of the fault zone surrounding the trap through observation and description. Core-based analysis of fault rocks reveals a sand-mud mixture with a mud-encased texture. Dense networks of deformation bands and laminated phyllosilicate fabrics, approximately 1 mm thick, display high continuity and no evidence of hydrocarbon invasion. Laboratory measurements show these fault rocks are clay-dominated with minimal carbonate cement (<5%). Thin-section and micro-CT analyses indicate that deformation bands and phyllosilicate layers reduce grain size by 0-2 orders of magnitude and porosity by 10-20% relative to the host rock, significantly degrading petrophysical properties. This millimeter-scale microstructure is interpreted as the key mechanism for subsurface fluid sealing. This study conducts an analysis based on the above and presents a discussion on the potential sealing mechanism of the bounding fault of a trap, proposing a sealing model for low-displacement faults developed in high net/gross sand ratio sequences. This study demonstrates that in specific scenarios, low-displacement faults possess sealing capabilities and exploration potential.

The temperature-pressure history of the organic-rich shale in the Cretaceous Qingshankou Formation in the northern Songliao Basin was reconstructed through comprehensive analyses, including field tests, paleo-heat flow reconstruction, overpressure evolution and geochemistry. The formation and evolution process of the Gulong shale oil was reproduced, and its enrichment patterns were clarified. Influenced by tectothermal events and tectonic movements at the end of the Cretaceous Mingshui Formation deposition, the evolution of organic matter thermal maturity in the first member of Qingshankou Formation (Qing-1 Member) exhibited distinct stages, which can be divided into the Cretaceous rapid evolution stage and the Paleogene-Neogene slow evolution and stabilization stage. High paleogeotemperature drove secondary cracking of retained oil in the Qing-1 Member, forming light shale oil in the Gulong Sag. This sag experienced three phases of overpressure during the late depositional stage of the Nenjiang Formation and late depositional stage of the Mingshui Formation of the Cretaceous, and the Neogene. The first two phases were related to the oil generation peak and secondary cracking in the sag, respectively, while the third phase resulted from the inheritance of earlier overpressure, as well as sustained hydrocarbon cracking and heat-induced fluid volume expansion. Crude oil is distributed orderly in the northern Songliao Basin. Conventional oil reservoirs such as Saertu and Putaohua contain high contents of non-hydrocarbon compounds, and they are believed to have formed by hydrocarbon charging as a result of the first phase of overpressure. Tight oils in the Fuyu and Gaotaizi reservoirs, most similar to shale oil in the Qing-1 Member in terms of composition and physical properties, are characterized by high content of saturated hydrocarbons, with their hydrocarbon charging and accumulation related to the second phase of overpressure. High paleo-heat flow generated by tectothermal events is determined to be the main driving factor for the staged hydrocarbon generation of organic matter in the Qingshankou Formation. The shale of Qing-1 Member with high thermal conductivity and the Cretaceous Nenjiang shale with low thermal conductivity constitute a thermal structure with lower conducting and upper sealing. This structure has prolonged secondary cracking of hydrocarbons, widened the liquid hydrocarbon window, and helped self-sealing enrichment of the Gulong light shale oil by virtue of the third phase of overpressure.

Permeability is a key parameter characterizing the fluid flow capacity in porous media and has significant application value in fields such as oil and gas exploration and development, and carbon dioxide geological storage. However, the complexity of the pore structure in reservoir rocks poses a huge challenge to the accurate prediction of permeability.Taking the Winland model as the core research object, this paper systematically reviews its development process, analyzes the differences between it and models such as Pittman and Swanson, explores the key factors affecting its prediction accuracy, and proposes targeted improvement methods and verifies them through experiments and data. The research results show that: ① The core parameter of the Winland model is the throat radius r35 corresponding to the mercury injection saturation of 35%, which later developed into a multiple linear regression model. The optimal throat radius parameters vary for different reservoirs. For example, r30 corresponding to the mercury injection saturation of 30% is more suitable for tight sandstone reservoirs, while r40 and r45 corresponding to the mercury injection saturation of 40% and 45% are more suitable for carbonate reservoirs. ② The prediction accuracy of the model is affected by the experimental dependence of the throat parameters, the heterogeneity of the reservoir, and diagenetic processes. In the low-permeability range of 0.1 to 1.0 mD, there is a tendency for the calculated values to be lower than the measured values. ③ By introducing percolation theory and combining CT scanning and nuclear magnetic resonance technology, a multi-modal seepage and diagenetic correction system for fractures and cavities can be constructed, which can effectively improve the model's adaptability. ④ An improved model based on the random forest algorithm and the fusion of multi-source logging data has a determination coefficient R2 of 0.76 for permeability prediction in the test set, which is significantly better than the traditional Winland model (R2=0.48). The conclusion is that the Winland model characterizes rock permeability through the statistical characteristics of throat size, and its improvement should focus on the dynamic optimization of core parameters, the correction of complex reservoir conditions, and the fusion of multi-source data. In the future, it should deeply integrate multiple regression and percolation theory to construct a "data-physical" dual-driven high-precision permeability prediction system.

The first member of the Qingshankou Formation (referred to in this paper as the First Member) in the Songliao Basin represents a typical shale sequence deposited in a sulfate-depleted, anoxic lacustrine environment. It is considered one of the most important terrestrial source rocks of the Cretaceous. Although previous studies have attributed organic matter enrichment in the First Member to lake anoxia, the specific types and degrees of anoxia remain poorly constrained. In this study, we investigated the First Member from the SYY3 well in the central Sanzhao Depression. Based on redox-sensitive elements (RSEs) and iron speciation data, we identified three distinct redox conditions: suboxic, ferruginous, and ferruginous-euxinic. Ternary C-S-Fe analysis indicates that relative supply of terrestrial iron was a key factor governing transitions between ferruginous and ferruginous-euxinic conditions. Due to the limited external sulfate input, widespread water-column euxinia was unlikely in the First Member. Using pyrite iron (Fepy) and carbonate-associated iron (Fecarb), we estimate that bacterial sulfate reduction (BSR) consumed generally <1% of organic carbon (OC), whereas dissimilatory iron reduction (DIR) consumed OC of >3% (up to 7.98%). Meanwhile, BSR in intermittently euxinic bottom water can efficiently promote phosphorus recycling and offset OC loss due to anaerobic processes by elevating primary productivity. Thus, in low-sulfate lacustrine systems, ferruginous-euxinic conditions are more favorable for developing source rocks with high TOC, S1 and HI. These findings provide important insights for oil and gas exploration in freshwater basins, particularly for identifying potential “sweet spots” of shale oil and gas.

Reliable identification of faults and fractures in sedimentary rocks is essential for hydrocarbon exploration and production. This study presents a comprehensive evaluation of Generative Adversarial Network (GAN) approaches for fracture segmentation in Formation MicroImager (FMI) images of tight sandstone reservoirs in the Southern basins of China. The GAN architecture was systematically compared with state-of-the-art models, including SegFormer, U-Net, and DeepLab3+, across two data availability scenarios: low-data (200 images) and high-data (1600 images). Under low-data conditions, the GAN model achieved superior performance with Intersection over Union (IoU) = 0.75, Dice score = 0.78, and recall = 0.85 (p<0.001 versus all baseline models). The approach demonstrated enhanced recovery of thin and discontinuous fractures and produced geologically plausible fracture segments that were not consistently present in the original annotations. However, further independent verification is required. In high-data scenarios, SegFormer demonstrated optimal performance (IoU = 0.85, Dice = 0.91), though GAN maintained competitive results (IoU = 0.80, Dice = 0.88). Cross-basin validation using Qingshankou Formation (shale oil reservoirs in the northeastern basin of China) data revealed GAN’s superior generalization capacity (cross-domain IoU = 0.68 versus 0.61 for SegFormer). The comparison of results demonstrates that GAN is suitable for early-stage reservoir evaluations with limited labeled FMI data for modeling fracture-controlled permeability. This study demonstrates that GAN models could provide robust solutions for fracture detection in data-scarce environments, while transformer architectures excel with abundant training data. This research offers insights into model selection based on data availability and provides scalable approaches for automated structural analysis in unconventional reservoirs.

Abstract The limited availability of shear-wave velocity (Vs) logs in deep-water reservoirs presents a significant challenge for reliable reservoir characterization and elastic property analysis. This study aims to develop and validate a robust machine learning (ML)-based workflow for predicting Vs from conventional well logs and integrating the results into rock physics–driven reservoir characterization. The dataset comprises five wells from two geologically analogous gas-bearing marine reservoirs: four wells from the West Offshore Nile Delta, Egypt, and one well from the Scarborough Gas Field, Australia. All wells include measured Vs data, enabling a structured training and validation strategy in which three wells are used for training, and two wells are reserved for blind validation across different fields. Input logs include gamma-ray (GR), bulk density (RHOB), deep resistivity (LLD), neutron porosity (APLC), and compressional velocity (Vp). Four ensemble ML models, gradient boosting (GB), extreme gradient boosting (XGB), light gradient boosting (LGB), and categorical boosting (CB), were evaluated. The CB model achieved the best performance, with R 2 values of 0.912 for testing and 0.904 for blind validation, along with the lowest RMSE and MAPE. The predicted Vs was subsequently used to derive elastic attributes, including Lambda-Rho and Mu-Rho, and to construct rock physics templates for lithology and fluid discrimination. These elastic properties were further integrated into a three-dimensional geostatistical modelling framework using Gaussian Random Function Simulation (GRFS), constrained by seismic-derived structural surfaces, to generate volumetric distributions of porosity, acoustic impedance, and Lambda-Rho across the reservoir. The results demonstrate that the ML-predicted elastic properties consistently identify gas-bearing sands, brine sands, and shale intervals. Furthermore, the successful application of the workflow across two distinct but analogous reservoir settings highlights its robustness and transferability. This integrated ML–rock physics framework offers an effective solution to enhance reservoir characterization while reducing reliance on costly and limited Vs measurements.

The Bakken Formation is one of the largest unconventional shale oil reserves, where hydraulic fracturing and horizontal drilling are key methods for enhancing recovery. A thorough understanding of its mechanical properties is crucial for designing these processes, especially since kerogen within the shale undergoes chemical changes during thermal evolution that alter its mechanical behavior. This communication primarily concentrates on evaluating how the mechanical behavior of Bakken shale, rich in Type II kerogen, evolves during controlled thermal maturation under anhydrous and hydrous pyrolysis conditions. The immature samples are pyrolyzed at 300, 325, 350, 365, 400, and 450 degrees C. To extensively investigate the mechanical characteristics of shale samples, a sum of 775 nanoindentation tests is performed on their surface. The results indicated that during the AHP, the initial increase of Young's modulus of the samples to more than 34 GPa was followed by a relatively large decline and it starts fluctuating between 20 and 25 GPa from 350 to 450 degrees C. The hardness value during the AHP, on the other hand, increased to almost 0.77 GPa at 300 degrees C and then started a steady decrease to approximately 0.25 GPa at 365 degrees C. It then experienced an uprise to 0.39 GPa at 400 degrees C and finally, fell to 0.32 GPa in the last temperature step. During the HP, the trends were quite different, which reflects the impact of the presence of water on kerogen maturation. The hardness values reduced constantly during the HP process reaching approximately 0.15 GPa at 450 degrees C. Young's modulus, on the other hand, showed a reduction to 21.6 GPa at 300 degrees C and then increased to almost 25 GPa at 325 degrees C. This increase was followed by a large decrease to nearly 11 GPa at 365 degrees C. Then, after a small increase to more than 12 GPa at 400 degrees C, it fell to 10.5 GPa at 450 degrees C. The existence of water during hydrous pyrolysis enhances thermal cracking phenomenon. This leads to the generation of more oil and bitumen and making the kerogen more porous. This was considered the main cause of the lower mechanical properties of HP samples. The deconvolution approach also showed that the distribution curves of the mechanical properties of samples were a combination of three different mechanical phases. This study provides new insights into the micromechanical evolution of shale during thermal maturation by explicitly distinguishing the effects of pyrolysis on kerogen-controlled mechanical behavior which establishes a direct link between mechanical properties and maturation pathways. This is basis for predicting changes in stiffness, brittleness, and deformation response in organic-rich shales in optimizing hydraulic fracturing strategies and improving geomechanical modelling of unconventional reservoirs.

Three-dimensional (3D) digital rock reconstruction, image resolution enhancement, pore network segmentation, and porosity prediction are fundamental to advancing digital rock physics. This study introduces a deep learning framework, Swin-UNETR, based on the Swin Transformer. It is a hybrid architecture that integrates Swin Transformer attention with the U-Net decoder for automated pore network segmentation and connectivity analysis from computed tomography (CT) images. The framework addresses key limitations of traditional Vision Transformers, particularly for heterogeneous low-permeability rocks. Furthermore, to address dataset scarcity, a generative adversarial network (GAN) was used to generate synthetic digital cores. This enhancement contributed to improved class balance and diversity. A comparative evaluation reveals that Swin-UNETR achieved a higher Dice score and improved porosity estimation compared to TransUNet on the same dataset. Cross-domain validation using coal and shale samples from the Longmaxi Formation in the Sichuan Basin of China confirmed strong generalization across lithologies and imaging conditions. In addition, the workflow was extended and applied to predict water distribution and wetting domains in shale for remaining gas recovery. This leveraged grayscale features from two-dimensional (2D) CT images to map the presence of fluid. The results correlated with cryo-SEM measurements, validating its physical interpretability. The 3D volumetric reconstructions revealed complex pore connectivity patterns that were invisible in 2D analysis, enabling a more robust characterization of fluid transport pathways. The proposed Swin-UNETR framework established a new paradigm for digital rock physics, with applications to coal and hydrocarbon flow modeling.

FE-SEM (Field Emission Scanning Electron Microscopy) is a leading imaging technique for investigating complex rock properties of shale and similar fine-grained sedimentary rocks. Although FE-SEM imaging can capture very fine-scale and high-resolution mineral compositions by QEMSCAN (Quantitative Evaluation of Minerals by FE-SEM), it is expensive and time-consuming. The CNN configured with U-Net architecture has recently demonstrated promising results for extracting multimineral information from grayscale images. However, minerals present in very small quantities are challenging to differentiate using such CNN architectures. This study develops an automated workflow using KiU-Net architecture to analyze multimineral compositions in lacustrine shale oil formations from FE-SEM images. Initially, the KiU-Net model was trained using labels from QEMSCAN images of samples segmented into five components: pore space ('pores'), calcite, quartz, clay, and pyrite. Additionally, the KiU-Net was trained using the same labels and segmented into 22 mineral classes (including pores + 4 dominant minerals). Once the models had been trained, additional images of lacustrine shale oil samples were segmented from 5-component (simplified) to 22-component (complex) mineral classification. KiU-Net outperformed traditional CNN/U-Net in detecting mineral compositions at the nanoscale by precisely defining their edges. KiU-Net's dual architecture enables a balance between edge sensitivity (Ki-Net) and contextual shape recognition (U-Net). Furthermore, a microstructure model of lacustrine shale oil indicates that illite, pyrite, biotite, quartz, and anhydrite contribute significantly to the high probability of pores and micro-fractures. This study establishes a cost-effective, automated alternative to QEMSCAN for shale oil exploration, combining advanced deep learning with practical geological insights.

Fine-grained sedimentary rocks exhibit significant textural heterogeneity, often obscured by conventional grain size analysis techniques that require sample disaggregation. We propose a non-destructive, image-based grain size characterization workflow, utilizing stitched polarized thin-section photomicrographs, k-means clustering, and watershed segmentation algorithms. Validation against laser granulometry data indicates strong methodological reliability (absolute errors ranging from −5% to 3%), especially for particle sizes greater than 0.039 mm. The methodology reveals substantial internal heterogeneity within Es3 laminated shale samples from the Shahejie Formation (Bohai Bay Basin), distinctly identifying coarser siliceous laminae (grain size >0.039 mm, Φ < 8 based on Udden-Wentworth classification) indicative of high-energy depositional environments, and finer-grained clay-rich laminae (grain size <0.039 mm, Φ > 8) representing low-energy conditions. Conversely, massive mudstones exhibit comparatively homogeneous grain size distributions. Additionally, a multifractal analysis (Multifractal method) based on the S50bi/S50si ratio further quantifies spatial heterogeneity and pore-structure complexity, significantly enhancing facies differentiation and reservoir characterization capabilities. This method significantly improves facies differentiation ability, provides reliable constraints for shale oil reservoir characterization, and has important reference value for the exploration and development of the Bohai Bay Basin and similar petroliferous basins.