The coordinated development of agriculture, industry and environment in inland river basins in arid areas is increasingly restricted by the shortage of water resources and the overexploitation of groundwater. To address the increasingly acute problem of limited water resources competition in the Manas River Basin of Xinjiang, a multi-objective optimization framework was constructed that couples the selection of crop planting with the coordinated allocation of water volume among different sectors. The model considers the crop structure of the six main irrigation areas and the incremental water consumption allocated to industrial and urban greening as decision variables. It is considered minimizing the total water consumption, maximizing the net economic benefit and minimizing the risk of groundwater overexploitation under the goal of water use control. Bayesian optimization algorithm based on Gaussian regression can efficiently search Pareto solutions set under the condition of small samples. The results indicate the following: (1) Analysis of the water resource allocation structure reveals that, in the planning year, total water demand exceeds the control indices by 1.99 × 10 8 m 3 and 3.12 × 10 8 m 3 , respectively. Agricultural irrigation demand was identified as the primary driver of this water deficit, highlighting the need for the reallocation of water resources to better align with regional development priorities. (2) Under the optimal scenario, in 2026 and 2030, industry receives an extra 3.5 × 10 7 m 3 and 2.0 × 10 7 m 3 , respectively, and the ecosystem receives an additional 8.0 × 10 6 m 3 and 1.0 × 10 7 m 3 ; the regional risk of groundwater over-extraction declines by 5.5% and 8.0% in those years. These results and the proposed model framework provide scientific value for the optimal allocation of water resources in arid and semi-arid regions. • Analyzed Water resource allocation of irrigation areas and cities in the Manas River Basin. • Established a water-saving redistribution model integrating crop selection, industrial and ecological water demand. • Solved the model with a Gaussian process-based Bayesian optimization algorithm under small samples. • Water redistribution increased water for industry and ecology, supporting industrial transformation and restoration. • Incorporated groundwater over-exploitation risk into the evaluation of optimal solutions.
Qin et al. (Wed,) studied this question.