Because transpiration is often the largest component of the water budget in arid systems, the efficacy of computer simulation models as predictors of water and salt movement is predicated on their ability to predict transpiration. The objective of this study was to improve the root-sink term for water extraction and thus improve predictions of transpiration. The root-sink terms often use either a transpiration-apportioning (TA) empirical function for the plant response to the soil matric and osmotic potential or a potential-flow (PF) function. Three root-sink terms, a TA formulation, a PF formulation, and a combination of the two (PFTA, a TA function for computing water uptake in response to salinity and a PF function for computing water uptake in response to water availability) were coded into a simulation model, and model predictions were compared with field-collected data. When predicted and measured relative yields (yield/yieldmax) were compared, the PF produced the poorest agreement with data (y = 0.8741x + 0.0251), the TA gave better agreement (y = 0.9283x + 0.0476), and PFTA provided the best agreement (y = 0.9909x + 0.0430). The plant and plant–soil based formulation predicted the salinity profile at the end of the field experiment under conditions of high salinity (EC irrigation water = 6.0 dS m−1) and irrigation equal to potential evaporation. The plant–soil formulation was the better predictor under the same salinity condition with irrigation at 40% of potential evaporation. The simulated evolution of the water content and salinity profile across time was also examined.