Oasis systems in Xinjiang face severe stress from extreme weather, water scarcity, and ecological decline, with water shortages driving ecological changes; however, the coupled hydro-climatic and anthropogenic drivers of this problem remain poorly quantified at fine spatial scales. Here, we develop an integrated framework combining a 1-km flash-drought detection workflow, a deep-learning groundwater inversion model, structural equation modeling (SEM), and a lag analysis to characterize flash drought dynamics, subsurface water stress, and ecosystem responses across Xinjiang. The findings show a decline in annual mean precipitation (-0.90 mm yr- 1) and groundwater levels (the water table deepened by +0.044 m yr- 1) within the study region, indicating worsening water availability. There has also been a steady increase in flash drought events and escalating water stress. Overall ecological quality has marginally improved, but this trend varies: non-farmland zones are deteriorating, whereas irrigated farmland areas are benefiting from improved irrigation security. SEM indicates that extreme temperatures and precipitation patterns indirectly impact the ecological quality by intensifying water stress. Additionally, the ecological quality responds to high temperatures with a noticeable delay and is particularly sensitive to changes in moisture levels. In oasis regions, sustained growth in agricultural water demand exacerbates the ecosystem's vulnerability to extreme climatic events. Our findings provide actionable targets for water management and oasis ecosystem resilience under ongoing warming and increasing extreme climate risk.