The discrete fracture model is a mainstream tool for shale gas productivity prediction but it often simplifies reservoir heterogeneity and fracture roughness, compromising accuracy. To address this limitation, this study developed a numerical productivity model for shale gas that incorporates fracture roughness and reservoir heterogeneity based on fractal theory and the Weibull distribution. History matching was conducted using data from a specific gas well in the Weiyuan shale block of the Sichuan Basin, China, with focus on analyzing the impacts of reservoir heterogeneity and fracture roughness. The results indicated that although enhanced reservoir heterogeneity reduces the average effective normal stress on natural fractures, which in turn induces shear slip and enhances their conductivity, it simultaneously impairs matrix porosity and permeability, leading to a ultimate decline in cumulative production. After two years of production, as the homogeneity index increased from 6 to 12 (indicating weakened heterogeneity), the cumulative gas production increased by 6.57
CO2-EGS has drawn attention for its advantage in clean energy extraction and CO2 sequestration. However, the evolution of reservoir transport characteristics under fluid–rock interaction effect (FIE) remains unclear. Its control mechanism on the performance of CO2-EGS is also not well understood. Therefore, this study systematically investigated the evolution of reservoir transport properties under FIE and evaluated response characteristics of the performance of CO2-EGS. Experimental results revealed that FIE enhances fluid transport properties while deteriorating heat transport capacity. The changes in rock transport properties are primarily attributed to microstructural and mineralogical alterations induced by FIE. A thermal–hydro coupled numerical model was established to analyze how long-term heat extraction behavior of CO2-EGS responds to the evolution of the reservoir transport properties. The simulation results indicated that the evolution of transport properties has a substantial impact on the system performance, and this impact shows a strong temperature dependence. Under high-temperature conditions (>350 °C), the transport properties changes induced by FIE deteriorated the key performance indicators, including production temperature, heat extraction power, and cumulative heat output, but reduced flow resistance. These findings emphasize the importance of incorporating dynamic rock property evolution into predictive modeling and design optimization for reliable and efficient CO2-EGS operation.
Mineral prospectivity mapping (MPM) supports exploration targeting, yet most machine learning approaches emphasize point probabilities with limited uncertainty quantification or coverage guarantees. Regional MPM faces incomplete deposit labels from biased exploration, spatial autocorrelation invalidating random cross-validation, and limited decision-relevant guidance linking probability surfaces to exploration priorities. This study develops a framework that combines positive-unlabeled learning, nested spatial block cross-validation, bootstrap ensemble uncertainty, and cross-conformal calibration, and converts the resulting signals into five priority zones defined by lift-based percentile thresholds and a 30 ρ _IQR ) cut-off. On the Yilgarn Craton, Western Australia (70,200 grid cells at 5 km resolution, 818 cells containing at least one known Ni deposit location, a 1.17 ρ _IQR of 18
Machine-learning mineral prospectivity mapping commonly reports high predictive skill under random cross-validation, but such scores do not show whether learned relationships transfer to a geologically distinct mineralized cluster. We examined this issue at the Sitaihaiquan carbonate-hosted Pb–Zn deposit in the Western Tianshan, NW China, by withholding the entire MZ-II mineralized zone from model training. A gravity–magnetic model achieved area under the curve (AUC) = 0.955 and average precision (AP) = 0.717 under random five-fold cross-validation, but its performance decreased to AUC = 0.388 and AP = 0.017 under leave-MZ-II-out validation. MZ-I and MZ-II occur in the same principal host unit and have broadly comparable mineralization, but they are controlled by the south-dipping F2 and north-dipping F7 faults, respectively. The withheld-zone failure coincides with a strong reversal in the magnetic signature between the two zones (Cohen’s d = 3.78), an approximately 166° contrast in the first-order local anomaly-asymmetry azimuth, and an approximately 180° reversal in mapped orebody attitudes. Replacing the original magnetic predictor with analytic signal or analytic signal plus total horizontal gradient did not restore transferability (AUC = 0.511 and 0.489, respectively). Adding mapped fault distance variables increased the leave-MZ-II-out AUC to 0.719 and AP to 0.046, improving regional ranking while retaining limited cell-scale precision. Under a stricter spatial hold-out that removes the entire MZ-II domain together with a 500 m buffer, this apparent gain does not persist: The fault distance predictors are strongly spatially autocorrelated (Moran’s I = 0.959–0.999), and their advantage under looser partitioning reflects near-neighbor information leakage rather than transferable skill. These results show that the magnetic–mineralization relationship learned in MZ-I is not stable across the reversed structural setting of MZ-II. Validation by geological cluster, followed by feature tests tied to the diagnosed geological mechanism, provides a direct means of separating high within-domain fit from transferable prediction.
The meaning of targets in data-driven mineral prospectivity mapping (DD-MPM) is foremost dependent on the quality of data. The resolving power of annotated data depends on the ratio of annotated area or volume and target footprint. At sufficiently large-scales of DD-MPM, it is impossible to match annotated area and covariate resolution to the footprint of mineral deposits. Consequently, information exposed to data modelling is unrestricted to between a target and its features. By definition, this type of annotation is a form of weak annotation, whose predictor–predictand relationship is weakened geostatistically through smoothing (a high ratio of annotated area or volume to target footprint). Annotation strength in DD-MPM is unstudied, although it is a research frontier in adjacent domains such as medical image segmentation. In this study, we present experimental results that reveal the spatial and aspatial significance of smoothing-induced weak annotation. Major findings include: (a) model performance improves with increasing smoothing, which is consistent with the linearization of the predictor–predictand relationship; (b) prospective area decreases with increasing smoothing, as smoothing weakens the signature of positive samples; (c) multifocal workflow uncertainty increases, caused by the same mechanism as in (b); and (d) the advantage of nonlinear and more expressive models is diminished with increasing smoothing. Therefore, although annotation strategy in DD-MPM seems innocuous, it has profound effects on the meaning of target area and exploration models. These effects impact: (a) the in-field validation of targets or model deployment expectations; (b) the mathematical meaning of trained models and targets; (c) the choice of model complexity; and (d) comparative analysis of workflow design (e.g., algorithms).
