In arid areas, agricultural development can occur at the expense of ecosystem conservation and water quality. To balance the conflicting interests, the water-ecosystem-agriculture (WEA) nexus in such areas needs to be systematically addressed. In this study, a multi objective linear fractional programming (MOLFP) approach is proposed to model the synergies and tradeoffs in the WEA nexus. Using this approach, the impacts of changes in hydrological conditions, land use, and the water management regime on the economic, ecological, and environmental efficiencies of the agricultural system in Zhangye Basin (ZB) (a typical arid area in northwestern China) were assessed. The major study findings are as follows: First, the MOLFP approach can effectively identify solutions for optimization problems with multiple interrelated ratio-type objectives, and the optimization results intrinsically reflect the behavior of the socioeconomic system. Second, in wet conditions, the agricultural system can be more profitable and environmentally friendly but less efficient in saving water for the ecosystems in the downstream area. Third, the potential land-use conversion from croplands into fallow or built-up areas, as well as the new water quality regulation, may lead to co benefits of economic, ecological and environmental efficiencies. Finally, the flexibility to adjust the spatial pattern of crop planting makes the agricultural system in ZB exhibit good resilience to water, land, and food stresses. However, the resilience is seriously impaired under very dry conditions and dramatic land-use changes. Overall, this study lays out a new framework for addressing the complex WEA nexus through system modeling and provides insights into agriculture planning and water resources management in arid areas.