Evaluating water saturation models for the Middle Bakken

Water saturation ( S w ) estimation in the Middle Bakken is affected by low porosity, hypersaline formation water, mixed siliciclastic–carbonate mineralogy, clay-bound water, and mixed wettability. Archie, Modified Simandoux, Dual-Water (DW), and the Interlocked Lattice Model (ILM) were evaluated against Dean–Stark core measurements from eight wells. Separate upper (UMB) and lower (LMB) Middle Bakken calibrations were performed within leave-one-well-out cross-validation. A binary-coded genetic algorithm minimized mean absolute error (MAE), and every model predicted effective water saturation on the same pore-volume basis. Mean blind-well MAE ranged from 0.104 to 0.111 in the UMB and from 0.102 to 0.115 in the LMB. These intermodel ranges, 0.0069 and 0.0136, were much smaller than the average within-model standard deviations among held-out wells, 0.0419 and 0.0562. Interwell variation was therefore 6.1 times the intermodel range in the UMB and 4.1 times the range in the LMB. The model with the lowest error also changed among held-out wells. Sensitivity tests further show that input uncertainty, especially porosity in the LMB, can change MAE more than the observed differences among model means. These results support multi-well calibration and explicit reporting of interwell variability. They do not establish statistical equivalence among the equations or a common high-salinity asymptote.