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

Permeability prediction in tight sandstone reservoirs remains a fundamental challenge, particularly in data-constrained frontier basins. The Mesozoic Jhuran Formation of the Kachchh Offshore Basin is characterized by low permeability (0.01–10 mD), heterogeneous facies architecture, and diagenetically modified pore systems. This study investigates the dominant geological controls governing permeability distribution and evaluates whether interpretable machine learning (ML) models can capture non-linear petrophysical interactions beyond conventional empirical transforms. Wireline log data from four wells were used to train and validate four supervised algorithms: Random Forest (RF), Support Vector Regression (SVR), Multi-Layer Perceptron (MLP), and Extreme Gradient Boosting (XGB). NMR-derived and core permeability served as calibration benchmarks. RF demonstrated favourable cross-well generalization on the single held-out blind well (R 2 = 0.77, Well-B) and the lowest prediction variance among the four models. SHAP analysis revealed that effective porosity and sonic transit time exert first-order control, while shale volume imposes a threshold-dependent permeability collapse. The results demonstrate that permeability is governed by non-linear shale–porosity–compaction interactions rather than porosity alone. This study establishes an interpretable, reproducible ML framework suitable for tight reservoir characterization in emerging basins.