Surfactant-polymer flooding is an effective process in extracting most of the original oil in place remained after conventional water flooding process. However, this technique is complicated and involves extensive screening and numerous experiments to find the optimum chemical composition, salinity, etc. Surfactant-polymer flood modeling can facilitate the optimization of the process, however, the inherently large parameter space results in great uncertainty and poor predictive capability. Here, by means of a novel approach using global sensitivity analysis, we reduce the parameter space of a typical surfactant-polymer flood model to facilitate model calibration and history matching process. To inform our analysis, we performed three Berea coreflood experiments with different slug designs and salinity profiles. The results from our coreflood experiments revealed and quantified the high sensitivity to salinity, underlying the importance of accurate phase behavior modeling. In addition to coreflood experimental data, we used an extensive set of laboratory data including polymer rheology, surfactant phase behavior, polymer permeability reduction, and capillary desaturation along with results from sensitivity analysis to build a mechanistic surfactant-polymer flood model. After modeling of sub-processes such as polymer flood model or phase behavior of our surfactant/oil/water system, through a multi-stage calibration algorithm, coreflood experimental data was used to build a thorough surfactant-polymer flood model where cumulative oil production and pressure profile were history matched simultaneously. Finally, we showed that our surfactant-polymer flood model has predictive capabilities with no need for ad hoc tuning of the model parameters by modeling two additional coreflood experiments where cumulative oil production and pressure profile matched those of experiments.