This research evaluates how well two empirical and one semi-mechanistic model predict root water uptake under optimal and deficit-irrigation conditions in two soils (loess and fine sand), and with three crops (wheat, tomato, and sorghum), each characterized by a unique root distribution. Models were calibrated by optimizing unknown soil hydraulic and root water uptake parameters using a global search algorithm in conjunction with a numerical model. System variables, including the spatial and temporal soil water potential measured in cropped, soil columns, served as input data. Model calibration was validated by comparing the measured transpiration of deficit-irrigated plants cultivated in a rotating lysimeter system, with modeled transpiration values. Model performance under varying degrees of compensation was evaluated, along with the effect of initial soil water potential conditions. We conclude that model performance was affected by soil texture, root distribution, and irrigation regimes. Therefore, a more mechanistic approach, able to account for such factors, should be incorporated into existing models.