Benchmarking Water saturation Models for the Mishrif Formation with Dean–Stark Ground Truth, southern Iraq

Accurate estimation of water saturation (Sw) is essential for reservoir characterization and calculating hydrocarbons in place. This study delivers the first systematic benchmarking of 81 water‑saturation models for the Mishrif Formation in southern Iraq, integrating Archie‑based (including Indonesian, Double‑Layer, Waxman–Smits, Simandoux, and Borai dual‑water) and data‑driven formulations (including neural‑network and clustering methods) with Dean–Stark ground‑truth calibration. The performance of each model is evaluated by different metrics including mean square error (MSE), concordance correlation coefficient (CCC), Pearson correlation (R) and coefficient of determination (R²). By combining four evaluation metrics (MSE, CCC, R², R) into dual ranking algorithms, the workflow resolves disagreement among individual metrics, ensuring robust model selection. Archie-based Borai‑Dual‑Water‑Linear outperformed all others, while the Self-Organization Map 9 Minimum (SOM-9-MIN) ranked first among data-driven methods.Archie-based models consistently outperformed data-driven counterparts, occupying all top-ten ranks and demonstrating superior accuracy for the Mishrif carbonate reservoir.Uncertainty bounds (|error| ≤ 0.20) confirm reliability for volumetric use. The approach reduces model‑selection ambiguity and can be transferred to shaly carbonate settings worldwide.