Natural Resources Research最新文献

This paper presents a thorough investigation into the evolutionary patterns of pore–fissure networks and their anisotropic connectivity characteristics within oil shale. We utilized CT digital core analysis after steam heating at varying temperatures. The study revealed that untreated oil shale has a densely compacted internal structure without distinguishable pore–fissure networks. However, steam exposure at temperatures exceeding 314 °C induced penetrating cracks along the bedding plane. This significantly modifies the mass transfer properties in the parallel bedding direction. Beyond 382 °C, continuous thermal cracking occurred, leading to numerous fissures along sedimentary bedding planes. This was accompanied by clustered pores formed through organic matter pyrolysis. These aggregated pores gradually interconnected adjacent parallel fissures, forming distinctive pore–crack clusters. Notably, as the temperature surpassed 500 °C, these pore–crack clusters continued to expand perpendicular to the lamination plane, profoundly influencing the mass transfer performance in this orientation. This phenomenon underscores the fundamental mechanism altering oil shale's mass transfer behavior perpendicular to the layer plane. From the perspective of percolation theory, the percolation threshold parallel to the lamination orientation was approximately 3%, with the transition around 300 °C. Conversely, the percolation threshold vertical to the sedimentary rock layers was approximately 14%, with the transition at temperatures surpassing 500 °C.

Three-dimensional (3D) geological modeling is a process of interpretation that integrates multiple source inputs and knowledge into geometry to represent the understanding of geologists. When geologists build a high-quality 3D geological model, this process still involves some issues such as sparse drillhole data, imperfect prior knowledge, and sensitive modeling algorithms. Therefore, taking uncertainty as the measurement criterion for the variation extent of the posterior likelihood of the 3D geological model and assisting in increasing the quality of the model are crucial issues in this domain. This paper proposes a novel method based on a (1 + ε)-approximation global optimum strategy, which is a type of big data and machine learning technique, to determine and present the uncertainty hidden in geometry. Compared with previous approaches, our strategy made the following new contributions: (1) the global optimum solution calculated by potential models is utilized to represent the uncertainty at each location; (2) the strategy offers a quantifiable reliability to each model that is involved in the evaluation process, and values of reliability are unknown before the commencement, meaning that they do not depend on expert experience; moreover, they can also be verified by comparing prior knowledge with information that such 3D models possess; (3) compared with previous studies, the number of perturbing models is no longer a key prerequisite for this kind of study to evaluate the quality of one geological model, thereby greatly reducing the computational complexity and improving the practicability. Finally, a case study was conducted to assess the uncertainty of a real 3D geological model in northwest Hunan Province, China.

In the early stages of a coalfield fire, CO₂ emissions are relatively low, and it is challenging to detect CO₂ concentrations in the soil surface due to the impact of surface temperature and wind. Investigating CO₂ concentration changes under surface temperature and wind conditions can provide experimental evidence and theoretical foundation for selecting optimal sampling locations and time. Using an automated monitoring platform for shallow soil CO₂, this study analyzed how surface wind speed and temperature affect the diffusion of CO₂ gas of surface sands. The effects of surface wind and temperature on CO₂ concentrations growth at different depths of the shallow surface were studied experimentally. When the surface temperature was 40 ℃ higher than the ambient temperature, the decrease of CO₂ concentrations for coarse sands with permeability of 2.13 × 10^-9 m² was most significant under high surface temperature conditions. However, the effect of high surface temperature on fine sands with permeability of 1.1 × 10^-12 m² was insignificant. Coarse sand with high medium permeability was most vulnerable to the fluctuation of surface wind speed. The surface CO₂ concentrations decreased by 93% at a depth of 22 cm in the coarse sands on the downwind side of the surface compared to natural convection conditions. In comparison, the CO₂ concentrations decreased by 37.5% on the upwind sides under small wind speeds. The coupling effect of high temperature and wind speed on the surface had a greater disturbance depth on fine and medium sands than low windy conditions. In addition, detecting shallow surface concentrations of CO₂ for the localization of fire sources was more advantageous during low temperature detection periods. In order to describe gas diffusion at the surface, mathematical and physical equations were developed. A combination of experimental and simulation theory was used to predict the depth of penetration of shallow surface gas by wind speed and temperature. The critical Darcy–Rayleigh number for temperature disturbance to shallow surface gas was approximately 6.3 when using medium and coarse sands with high permeability. Simulations show that the wind-induced penetration depth was 40.8 cm for coarse sand and 23.5 cm for medium sand at a surface wind speed of about 0.4 m/s combined with the experiments. It is necessary to detect CO₂ concentrations at least at depth of 23.5 cm in conditions of low surface wind speed, particularly in the overlying medium with high porosity.

During the coal mining process, fractures generated can lead to crude oil infiltrating into coal seams, forming coal–oil symbiosis (COS). The complex three-phase interaction of coal–oil–oxygen makes the mechanism of COS spontaneous combustion filled with uncertainties. This study utilized synchronous thermal analysis to analyze the physico-chemical behavior of raw coal and COS at different heating rates. Additionally, detailed characterization of their surface morphology and functional groups was conducted using scanning electron microscopy (SEM) and in situ FTIR technology. The findings suggest that the coverage of crude oil on the surface of coal inhibits the adsorption of oxygen by the coal, leading to the disappearance of the stage where COS absorbs oxygen and gains weight. Moreover, the continuous decline of –OH groups and aliphatic hydrocarbons in the later stages suggests that crude oil acts as a catalyst for combustion during the latter stages of the reaction. The Kissinger–Akahira–Sunose, Starink, and Flynn–Wall–Ozawa methods showed that the apparent activation energy of COS is 23.3 and 19.7% lower than that of raw coal in thermal decomposition and combustion stages, respectively.

Permeability and porosity are key parameters in reservoir characterization for understanding hydrocarbon flow behavior. While traditional laboratory core analysis is time-consuming, machine learning has emerged as a valuable tool for more efficient and accurate estimation. This paper proposes an ensemble technique called adaptive boosting (AdaBoost) for porosity and permeability estimation, utilizing methods such as support vector machine (SVM), Gaussian process regression (GPR), multivariate analysis, and backpropagation neural network (BPNN) for prediction based on well logs. Performance evaluation metrics including root mean square error, mean square error, and coefficient of determination (R²) were used to compare the models. The results demonstrate that AdaBoost outperformed GPR, SVM, and BPNN models in terms of processing time and accuracy, achieving R² values of 0.980 and 0.962 for permeability and porosity during training, respectively, and 0.960 and 0.951 during testing, respectively. This study highlights AdaBoost as a robust and accurate technique that can enhance reservoir characterization.

China’s deepest coalbed methane (CBM) exploratory well was drilled to over 4000 m in the Sichuan Basin, providing an effective database for analyzing the depth effect on methane accumulation. The results indicate that the in situ gas content generally continues to increase with depth, from 11.61 m³/t at 453.75–1829.6 m to 23 m³/t at 2467.98–4324.29 m. For coals at depths less than 2000 m, poor preservation conditions that are unable to seal free gas and adsorbed gas are dominant, with gas saturation ranging 45.79–97.61%. At depths deeper than 2000 m, the total gas content tends to be greater than in situ gas adsorption capacity, with gas saturation of 137.9–150.72%, indicating the coexistence of free gas (4.54–10.22 m³/t) and adsorbed gas (11.04–21.98 m³/t) as better preservation conditions. Similar to shale gas, deep coal seams exhibit excellent productivity potential as free gas can be extracted without the necessity of a drainage-pressure reduction stage for desorption, while a large amount of adsorbed gas can support long lifecycle of production wells. This research fills the gap in understanding the gas-bearing characteristics of ultra-deep CBM and is of great significance in guiding the exploration and development of deep CBM.

Mineralization temperature determination is fundamental to economic geology research, yet quantifying it across mineralization remains a challenge. Sphalerite is ubiquitous in various types of mineral deposits and particularly abundant in Pb–Zn deposits, and its trace element composition is temperature-dependent, making it an ideal candidate for geothermometry. Here, we first compiled a global sphalerite trace element composition dataset (n = 1416, T = 75–430 °C), encompassing different Pb–Zn deposit types (Mississippi Valley-type, epithermal, sedimentary-exhalative, skarn-type, and volcanic massive sulfide deposits). After data processing following statistical norms, the different machine learning algorithms (random forest (RF), gradient boosted decision trees, artificial neural networks, least absolute shrinkage and selection operator, support vector machine, k-nearest neighbors, and linear regression) were employed to train different models to explore the potential link between the sphalerite-forming temperature and trace element geochemistry. Each of the model’s performance was evaluated using the leave-one-out cross-validation approach, which revealed the RF (R² = 0.88, RMSE = 26 °C) as the best-performing algorithm. Meanwhile, five-fold cross-validation results indicated that the RF model (R² = 0.87, RMSE = 25 °C) outperformed the GGIMFis thermometer (R² = 0.53, RMSE = 50 °C). Meanwhile, the feature importance analysis revealed that Ge and Mn displayed significant impacts on temperature prediction as the high temperature generally favors Mn, but not Ge, incorporation into the sphalerite structure. Finally, a model was trained with the entire dataset, generating a reliable sphalerite thermometer (SPRFT software, freely provided here) suitable for low to moderate temperature (75–430 °C) hydrothermal environments. This SPRFT thermometer was applied to evaluate the temperature of Pb–Zn mineralization in the Sichuan–Yunnan–Guizhou Pb–Zn metallogenic belt (SW China) and it provides an innovative perspective into the ore-fluid evolution. This study demonstrated a robust approach for calculating mineralization temperatures using machine learning. This novel methodology opens new avenues for investigating and recalculating more mineral geothermometers.

In the realm of deep resource exploration, risk is a factor that cannot be neglected. This study innovated existing quantitative mineralization prediction models with consideration of risk. Different from conventional approaches, which primarily focus on the conditional means and show obvious limitations in handling enriched or barren mineralization that deviate significantly from the mean, quantile regression (QR), as a method to predict the conditional distribution instead of conditional means, is expected to break through these limitations and to be used further for risk prediction. Drawing upon data from the Xiadian gold deposit, five geological factors were extracted as explanatory variables and gold grade was used as response variable. Four QR-based regression models were employed to predict the conditional distributions of gold grade. The comprehensive performance evaluation and comparison of these models focus on reliability, clarity, and their combination. The results unequivocally indicate that the quantile regression forest (QRF) model outperformed the other QR-based prediction models. Subsequently, a detailed performance analysis was conducted on the optimal QRF model, followed by a comparison with the RF model to validate its effectiveness. Building upon this foundation, by analyzing the predictive results of QRF models at certain quantiles in unknown regions, several discernible targets were delineated at different risk levels. Overall, this paper introduces the consideration of risk in mineral prospectivity prediction and attempts to predict the conditional distribution of mineralization in deep mineral prospectivity mapping contexts. These insights can offer valuable guidance in identifying optimal targets and in reducing exploration risks.

The three-dimensional (3D) stress waves of coal samples were studied using a true triaxial split Hopkinson pressure bar compression rod. The results indicate that the 3D strain of the coal samples increased gradually under vibration load. The 3D stress wave of coal samples showed attenuation characteristics, and the change amplitude of the stress wave of coal samples along the direction of dynamic load was the most obvious. The amplitude of stress wave was the largest in the axial direction constrained by pre-stressing 3 MPa, while the amplitude of stress wave in the lateral 2 MPa pre-stressing was smaller than that under the constraint of 1 MPa. The results showed that the main deformation of coal samples was along the impact direction, while the larger horizontal and vertical lateral binding forces limited the deformation of coal samples. The Fourier transform was performed on the 3D stress wave of the coal samples, and the change in the amplitude of the stress wave spectrum was correlated positively with the vibration. The spectrum amplitude of the coal samples under the pre-stressed 3 MPa constraint (axial) direction was the largest, while the spectrum amplitude of the coal samples under the lateral 2 MPa pre-stressed constraint was smaller than that under the binding 1 MPa. However, the main frequency of the three-way stress wave was distributed in 0–10 kHz. By calculating the energy consumption rate and wave velocity decay rate, it was verified that the damage of coal samples increased with increase in dynamic load. This experimental testing provides an effective testing method for studying the 3D stress waves of coal samples under complex stress medium conditions. In addition, a dynamic constitutive model of coal was constructed according to the mechanical behavior of coal and rock mass and the measured data.

Recent advances in geological exploration and oil and gas development have highlighted the critical need for accurate classification and prediction of subterranean lithologies. To address this, we introduce the Meta-Vision Transformer (Meta-ViT) method, a novel approach. This technique synergistically combines the adaptability of meta-learning with the analytical prowess of ViT. Meta-learning excels in identifying nuanced similarities across tasks, significantly enhancing learning efficiency. Simultaneously, the ViT leverages these meta-learning insights to navigate the complex landscape of geological exploration, improving lithology identification accuracy. The Meta-ViT model employs a support set to extract crucial insights through meta-learning, and a query set to test the generalizability of these insights. This dual-framework setup enables the ViT to detect various underground rock types with unprecedented precision. Additionally, by simulating diverse tasks and data scenarios, meta-learning broadens the model's applicational scope. Integrating the SHAP (SHapley Additive exPlanations) model, rooted in Shapley values from cooperative game theory, greatly enhances the interpretability of rock type classifications. We also utilized the ADASYN (Adaptive Synthetic Sampling) technique to optimize sample representation, generating new samples based on existing densities to ensure uniform distribution. Our extensive testing across various training and testing set ratios showed that the Meta-ViT model outperforms dramatically traditional machine learning models. This approach not only refines learning processes but it also adeptly addresses the inherent challenges of geological data analysis.