Field experiments were conducted in arid Southern Xinjiang, Northwest China, for 3 years to evaluate sustainable irrigation regimes for cotton. The experiments involved mulched drip irrigation during the growing season and flood irrigation afterward. The drip irrigation experiments included control experiments, experiments with deficit irrigation during one crop growth stage, and alternative irrigation schemes in which freshwater was used during one growth stage and relatively saline water in the others. The average cotton yield over 3 years varied between 3,575 and 5,095 kg/ha, and the irrigation water productivity between 0.91 and 1.16 kg/m3. Crop sensitivities to water stress during the different growth stages ranged from early flowering-belling (most sensitive) > seedling > budding > late flowering-belling (least sensitive), while sensitivities to salt stress ranged from late flowering-belling > budding > seedling > early flowering-belling. Although mulched drip irrigation during the growing season caused an increase in salinity in the root zone, flood irrigation after harvesting leached the accumulated salts to below background levels. Numerical simulations, based on the 3-year experiments and extended by another 20 years, suggest that mulched drip irrigation using alternatively fresh and brackish water during the growing season and flood irrigation with freshwater after harvesting is a sustainable irrigation practice that should not lead to soil salinization. © 2013 Springer-Verlag Berlin Heidelberg.