Advanced numerical simulation models can potentially help improve guidelines for irrigation and salinity management. Many simulation model parameters have considerable uncertainty, and ideally that uncertainty should be reflected in model predictions and recommendations. In this work, we investigate solute transport predication intervals that can be generated by propagating model parameter uncertainty using Monte Carlo techniques. Flow and transport is simulated with a standard numerical model, while soil parameters and their uncertainty are estimated with pedotransfer functions. Generalized global sensitivity coefficients are computed to determine the parameters having the greatest impact on transport prediction and uncertainty. Simulations are compared with Br transport measured under unsaturated conditions in large lysimeters packed with clayey soil materials. In a 48 cm tall, homogeneous soil profile, model prediction intervals provided a reasonably good description of a single, relatively “noisy” breakthrough curve. In replicated 180 cm tall, layered soil profiles, model structural errors limited the accuracy of the prediction intervals under one irrigation water treatment, whereas under another treatment the predictions tracked the time course of the data reasonably well but tended to overestimate solute concentrations. The width of the prediction intervals tended to be small relative to the range of transport variability that existed across replicated lysimeters, particularly at shallow depths. Additional work aimed at operational field testing of model prediction uncertainty is needed if advanced water management models are to reach their full potential.