Integrated evaluation and forecasting of water resource spatial equilibrium in arid Xinjiang using machine learning and multi-dimensional indicators

Study region: Xinjiang, located in northwestern China, is a typical arid region with scarce and unevenly distributed water resources. The fragile ecological environment and strong spatial mismatch between water and land resources make water allocation a critical challenge for sustainable development. Study focus: This study constructs an integrated framework to evaluate and forecast Water Resource Spatial Equilibrium (WRSE) in arid regions. The framework combines variable set and partial correlation coefficient methods with three key indicators—water resource load, soil–water matching, and water use efficiency. A bridge transition module was developed to address data gaps and inconsistencies, while a hybrid prediction model integrating the Prophet algorithm, gradient boosting regression, and rule-based correction was applied to capture both long-term and short-term variations in WRSE.New hydrological insights for the region: Results show that since 2007, WRSE in Xinjiang has shifted from deterioration to relative stability, maintaining a spatial pattern of “stronger in the north, weaker in the south; better in the east, weaker in the west.” Forecasts for 2025–2030 indicate persistent disparities, with Urumqi maintaining relatively high equilibrium levels and southern areas such as Kizilsu and Hotan remaining in disequilibrium. To enhance regional water balance, ecological compensation, efficient irrigation, and construction of the “Xinjiang Water Network” are recommended. The framework offers new hydrological insights and decision support for sustainable water management in arid regions.