A large-scale hydrological model (MIKE HYDRO) was established for the purpose of sustainable agricultural water management in the main stem Tarim River, located in northwest China. In this arid region, agricultural water consumption and allocation management are crucial to address the conflicts among irrigation water users from upstream to downstream. The results of model calibration indicated a close correlation between simulated and observed values. Scenarios with the change on irrigation strategies and land use distributions were investigated. Irrigation scenarios revealed that the available irrigation water has significant and varying effects on the yields of different crops. Irrigation water saving could reach up to 40% in the water-saving irrigation scenario. Land use scenarios illustrated that an increase of farmland area in the lower reach gravely aggravated the water deficit, while a decrease of farmland in the upper reaches resulted in considerable benefits for all sub-catchments. A substitution of crops was also investigated, which demonstrated the potential for saving considerable amounts of irrigation water in upper and middle reaches. Overall, the results of this study provide a scientific basis for decision-making on agricultural water consumption and allocation in the study area.