Climate change and human activities reshape the water-food-ecology nexus and water-economy decoupling in oasis cities of the Hexi corridor, China

In arid oasis systems, water allocation links agricultural production, ecological stability, and regional economic development, making the water-food-ecology (WFE) nexus highly sensitive to hydroclimatic stress and human activities. Yet county-scale evidence remains limited on how hydroclimatic and socioeconomic changes are associated with WFE interactions and water-economy decoupling in inland oasis cities. Focusing on the Hexi Corridor, we integrate multi-decadal hydroclimatic records, remote sensing ecological indicators, and county-level socioeconomic, water use, and crop data. Tapio decoupling, temporally purged and spatially blocked random-forest validation, Grey relational analysis (GRA), exploratory two-way fixed-effects piecewise path analysis, and scenario-based reliability-resilience-vulnerability assessment were combined to characterize water-economy relationships, crop–water associations, WFE interactions, and management outcomes. During 1981–2024, annual mean 2-m air temperature increased significantly by 0.038 °C/year, annual precipitation showed a weak increase of 0.349 mm/year, and December SPEI-12 decreased significantly by 0.0476/year, indicating a deteriorating climatic water balance under pronounced warming. Strong water-economy decoupling emerged before 2020 and was associated with declining water use intensity, while industrial decoupling showed greater interannual variability during 2020–2022. Random-forest performance was moderate under temporal validation (R 2 = 0.673) and weaker under spatial validation (R 2 = 0.441), indicating stronger temporal generalization than spatial transferability. Crop composition showed pronounced spatial heterogeneity and county-specific similarity with agricultural water use. Piecewise path analysis revealed scale-dependent associations among socioeconomic activity, water use, hydroclimatic conditions, and vegetation. Within counties, land surface temperature (LST) was negatively associated with normalized difference vegetation index (NDVI), while surface water use was positively associated with vegetation condition and SPEI-12 was negatively associated with agricultural water use in the primary specification. Scenario results showed that agricultural-demand reduction and coordinated measures improved water-system reliability and resilience, whereas groundwater constraints applied alone increased unmet demand. These findings support differentiated local implementation within basin-scale governance and provide a transferable diagnostic workflow for comparable arid inland basins, although local parameters and management outcomes require region-specific recalibration.