Water availability is a major factor limiting crop yields on the Loess Plateau of China. As competition for water intensifies, it is essential to develop alternative irrigation schedules that maximise crop yield and water use efficiency (WUE) for a given level of water supply. A field experiment with winter wheat (Triticum aestivum L.) from 1995 to 1998 calibrated and tested a biophysically based model (WAVES) in terms of grain yield and WUE prediction. The data collected include water and energy balance components, biomass and grain yield. Comparisons between the measurements and the model predictions were made with three years of field data. Modelled grain yield and WUE based on biomass and harvest index were in better agreement with the measurements than those based on transpiration and harvest index. The model was sensitive to different irrigation treatments, and in reasonable agreement with field measured data. The highest irrigation treatment resulted in the greatest evapotranspiration but not the highest yield, so WUE was relatively low. Appropriately limited irrigation could improve the grain yield and WUE. Aiming only for maximum grain yield or for maximum WUE could lead to uneconomical irrigation management.