The research, conducted by a team from GEOPIC, the research arm of Oil and Natural Gas Corporation in Dehradun, along with collaborators at Pandit Deendayal Energy University and Maharashtra Institute of Technology World Peace University, focused on the gas-bearing tight sandstones of the Jhuran Formation. These rocks, deposited during the syn-rift and early post-rift phases that followed the fragmentation of Gondwana, are anything but uniform. Permeability values range from a near-impermeable 0.01 millidarcies to a still-modest 10 millidarcies, and the formation is sliced into parallel fault blocks by northwest-southeast trending structures that complicate any attempt to extrapolate reservoir quality between wells. For an emerging hydrocarbon province like the Kachchh Offshore Basin, that heterogeneity is a formidable barrier to exploration.

The conventional toolkit offers little comfort. Classic permeability estimation leans on empirical porosity-permeability transforms calibrated to laboratory core measurements. In homogeneous reservoirs such relationships perform adequately, but in tight, clay-rich sandstones the link between porosity and permeability weakens dramatically, because flow is governed by pore geometry, grain packing, clay distribution and cementation rather than by pore volume alone. Compounding the problem, core plugs are expensive and sparse; in this study, direct core measurements existed for just one well, and only fifteen of them. The team therefore turned to machine learning, training four supervised algorithms on wireline log data from four wells: Random Forest, Support Vector Regression, a Multi-Layer Perceptron neural network, and Extreme Gradient Boosting.

The input features were the workhorses of petrophysics: gamma ray, deep resistivity, bulk density, neutron porosity and sonic transit time, supplemented by engineered variables such as volume of clay, total and effective porosity, and a density-neutron-resistivity combination parameter. Nuclear magnetic resonance logging provided an independent permeability benchmark in two wells, derived through the Schlumberger-Doll Research equation, which relates permeability to porosity and the logarithmic mean of the NMR T2 relaxation distribution. Crucially, the NMR-derived permeability served only as the target variable, never as an input, a design choice that prevents data leakage and ensures the models can be applied to wells lacking NMR coverage. Spearman rank correlation analysis confirmed that no input pair approached the severe multicollinearity threshold, supporting their joint inclusion.

Validation was deliberately stringent. Wells A, C and D supplied the training data, split 70-30 into training and testing subsets, with five-fold cross-validation used for hyperparameter tuning. Well B, which contained the fifteen core measurements, was held out entirely and used only once, after model finalisation, as a blind validation well. On that single held-out well, Random Forest achieved a coefficient of determination of 0.77 against the core and NMR benchmarks, the best cross-well generalisation of the four algorithms, and posted the lowest prediction variance, with a residual standard deviation of 1.41. Support Vector Regression, by contrast, fit its training data beautifully, reaching a training R-squared of 0.92, but its validation performance collapsed, the classic signature of overfitting.

The contrast between algorithms carried a practical lesson. The Multi-Layer Perceptron and XGBoost both achieved near-perfect alignment on training data yet generalised poorly, with XGBoost showing the largest gap between training and blind-validation performance and the highest prediction uncertainty in Well A, where residual standard deviations reached 7.81. Random Forest, averaging hundreds of decision trees each trained on random subsets of the data, proved the most stable performer across all wells, with residual standard deviations as low as 0.38 in Well C. The authors suggest a heuristic for similarly data-starved settings: Random Forest as the default choice, Support Vector Regression where smooth low-variance surfaces are needed, and neural networks or gradient boosting only with strong regularisation and blind-well checks, or once more wells become available.

What elevates the study beyond a routine algorithm bake-off is its use of SHAP, or Shapley Additive Explanations, a technique borrowed from game theory that decomposes each prediction into the contributions of individual input features. The SHAP analysis revealed a clear hierarchy of geological control. Effective porosity exerted the strongest first-order influence on permeability, followed by sonic transit time, which acts as a proxy for compaction. Most strikingly, shale volume displayed strongly negative SHAP values beyond a threshold of roughly 35 to 40 percent, marking a permeability collapse in clay-rich intervals. In other words, permeability in the Jhuran sandstones is governed by nonlinear interactions among shale content, porosity and compaction, not by porosity alone, precisely the kind of behaviour that linear empirical transforms cannot capture.

The team was candid about the limits of its evidence base. Four wells with uneven log and core availability is a small dataset, and the NMR calibration constants, drawn from standard literature values for water-wet clastic systems, were checked only qualitatively against the fifteen core samples from Well B rather than formally recalibrated for the Jhuran Formation. The reported R-squared of 0.77 reflects a single blind well, and the authors explicitly caution that basin-wide generalisation requires confirmation with additional held-out wells and more core control. Well D, lacking both core and NMR data, relied on a facies-derived pseudo-benchmark, an independent consistency check but not a direct measurement. These caveats frame the work honestly as a proof of concept rather than a finished predictive product.

Even so, the implications reach well beyond one basin. Frontier exploration provinces worldwide face the same predicament: geologically complex reservoirs, a handful of wells, and sparse core data, all while drilling decisions carry enormous financial risk. The study demonstrates that a transparent, reproducible machine learning workflow, built on open-source libraries and validated against independent physical measurements, can extract usable permeability predictions from ordinary wireline logs under exactly those constraints. By pairing ensemble learning with explainability, the researchers have shown that machine learning need not be a black box; it can double as an analytical instrument that tells geoscientists which geological processes matter most. As exploration accelerates along India’s northwestern continental margin, that combination of prediction and insight may prove as valuable as the predictions themselves, offering a transferable template for characterising tight reservoirs wherever data is scarce and heterogeneity is the rule.

Subject of Research: Interpretable machine learning prediction of permeability in tight Jhuran Formation sandstones of the Kachchh Offshore Basin, India

Article Title: Data driven permeability modelling in tight Jhuran formation sandstones of the Kutch offshore Basin India

Article References: Jyotsna, G. D. D., Anees, M., Mavani, S. H., Thakur, S., Saxena, A., Singh, S. K., Desai, D. B. S. G., & Singh, D. A. (2026). Data driven permeability modelling in tight Jhuran formation sandstones of the Kutch offshore Basin India. Discover Geoscience, 4(1), Article 366. https://doi.org/10.1007/s44288-026-00724-x

Image Credits: AI Generated

DOI: 10.1007/s44288-026-00724-x

Keywords: permeability prediction, machine learning, tight sandstone, Jhuran Formation, Kachchh Offshore Basin, Random Forest, SHAP analysis, well logging, NMR logging, reservoir characterization, petrophysics, explainable AI

The central challenge is what geologists call a dual-medium reservoir. In these formations, gas lives in two very different storage systems at once: the tiny pores between sand grains, and the sparse network of natural fractures that cuts across them. Fractures act like high-speed highways for gas flow, while the matrix pores act like vast parking lots. Knowing which system dominates in any given interval of rock is essential for deciding where to drill and how to stimulate a well. But conventional classification methods rely on discrete analytical samples, mostly from core pieces pulled from a handful of wells. Most wells are never cored at all, so point-based analysis simply cannot capture how fractured and matrix-dominated zones alternate continuously down the borehole.

The research team’s first contribution is a revised classification framework, which they describe as a three-major and six-minor scheme, built on pore structure. It separates the Xujiahe sandstones into nine reservoir types spanning matrix-controlled, fractured, and hybrid matrix-fractured varieties. A key analytical device is the fracture identification factor, a quantity derived from well logs that distinguishes true fractured reservoirs from matrix-fractured ones. Because well logging curves are recorded continuously as instruments travel up the borehole, using them as classifier inputs means every centimeter of rock in every well can be typed, not just the rare intervals where core was recovered.

But continuous logging data brought a familiar artificial intelligence problem to the surface: class imbalance. Some reservoir types, particularly the highest-quality fractured varieties, are genuinely rare in the subsurface, so their examples are heavily outnumbered in any training dataset. Machine learning classifiers trained on such lopsided data tend to simply ignore the minority classes, which is precisely the wrong behavior when those rare classes are the exploration targets. To fix this, the team applied ADASYN, the adaptive synthetic sampling technique. ADASYN works by generating synthetic minority-class examples, concentrating its effort near the decision boundaries where classification is hardest. Rather than flooding the dataset with generic duplicates, it teaches the model to pay attention to the ambiguous frontier zones between reservoir types, which the authors found markedly strengthened recognition of the scarce but valuable classes.

With rebalanced data in hand, the researchers then assembled what they call a multi-ensemble learning architecture. Instead of trusting a single algorithm, they trained four different base classifiers: random forest, support vector machine, multilayer perceptron, and XGBoost. Each brings a distinct inductive bias. Random forests average many decision trees trained on random subsets of the data, support vector machines carve maximal-margin boundaries through feature space, neural networks learn flexible nonlinear mappings, and XGBoost builds boosted trees that iteratively correct earlier errors. The team optimized each base learner and then combined their votes using voting theory, testing both hard voting, where each classifier casts a single categorical vote, and soft voting, where the classifiers’ predicted class probabilities are averaged before a decision is made.

The results were striking. The best configuration, dubbed ADASYN-voting-soft, or AVS, achieved a mean accuracy of 90.25 percent across the reservoir classes. That figure represents an improvement of 12.22 to 16.84 percentage points over standalone machine learning methods. Just as important, the soft-voting ensemble proved more stable and more accurate than hard voting, especially for reservoirs whose well log signatures are ambiguous. This makes intuitive sense: when a rock sample sits near the boundary between two classes, individual classifiers may disagree, and hard voting can flip-flop from one measurement to the next. Averaging probabilities dampens that noise, producing smoother, more geologically plausible classifications along the borehole.

The AVS method also addresses a subtle risk in machine learning practice: the exclusion of small, high-quality sample sets. In geoscience, curated expert-labeled examples are expensive and scarce. A single classifier can effectively discard such precious minority data by underfitting them, but the ensemble approach compensates for the uncertainty of any individual model. Combined with ADASYN’s synthetic augmentation, the framework ensures that even thinly represented reservoir types exert real influence on the final prediction. The interpretability emphasis of the study matters here too, because geologists need to understand which logging features drive a classification before they commit drilling capital on the model’s advice.

Applied across the Xujiahe Formation’s second member, the workflow revealed the spatial distribution of reservoir types in the Xinchang area of the western Sichuan Depression, mapping where matrix-dominated zones give way to fracture-enhanced sweet spots. That spatial picture is more than an academic exercise. Accurate identification of high-quality reservoir intervals directly guides well placement, completion design, and production optimization in a basin that Chinese researchers have described as a super gas basin. Fracture-developed tight sandstone intervals hold disproportionately large shares of recoverable gas, so a tool that reliably flags them in uncored wells translates into fewer dry or underperforming wells.

The study also fits into a broader trend sweeping the geosciences: the migration of ensemble learning and imbalance-aware sampling from computer science into subsurface characterization. The authors’ bibliography traces a lineage from classic oversampling methods like SMOTE and Borderline-SMOTE through ADASYN’s adaptive variant, alongside a growing body of work applying random forests, boosting, and stacking to lithofacies identification, fracture detection, and fluid typing in basins from the Ordos to the Tarim. What distinguishes this contribution is the combination of a geologically meaningful classification scheme, an imbalance-correction step tuned to decision boundaries, and a probability-averaged ensemble, all applied to the notoriously tricky dual-medium problem.

Funded by China’s National Science and Technology Major Project and the National Natural Science Foundation of China, the work offers a template that could travel well beyond the Sichuan Basin. Any fractured tight reservoir, whether sandstone, carbonate, or volcanic, poses the same pairing of continuous well log data with rare, high-value target classes. As exploration pushes into ever deeper and more heterogeneous plays worldwide, tools like the AVS model suggest that the winning approach will not be any single clever algorithm, but disciplined engineering: classify the rocks in a way geologists understand, fix the data imbalance before it quietly biases the model, and let a well-calibrated team of classifiers vote where no individual model should be trusted alone.

Subject of Research: Machine learning classification of dual-medium tight sandstone gas reservoirs in the Sichuan Basin

Article Title: Interpretable Prediction of Dual-Medium Tight Sandstone Reservoir Types of the Sichuan Basin: A Multi-ensemble Learning Approach for Imbalanced Data Optimization

Article References: Zhang, J., Yue, D., Wang, W., Qu, L., Liu, Y., Li, W., Jin, W., & Fu, J. (2026). Interpretable Prediction of Dual-Medium Tight Sandstone Reservoir Types of the Sichuan Basin: A Multi-ensemble Learning Approach for Imbalanced Data Optimization. Natural Resources Research. https://doi.org/10.1007/s11053-026-10761-5

Image Credits: AI Generated

DOI: 10.1007/s11053-026-10761-5

Keywords: tight sandstone, Sichuan Basin, reservoir classification, machine learning, ensemble learning, ADASYN, imbalanced data, well logging, natural fractures, Xujiahe Formation, natural gas, dual-medium reservoir