From Runoff Change Drivers Identification to Targeted Regulation: An Integrated Framework for Reservoir Group Adaptive Scheduling

Under a changing environment, the lack of precise alignment between the multi-driving mechanisms of spatiotemporal runoff evolution and the regulation of reservoir group, coupled with the frequent neglect of uncertainties in attribution analysis, results in a logical disconnect between "driver identification and regulatory response". To address this gap, this study integrates the theories of streamflow change attribution and reservoir group adaptive scheduling, proposing an integrated methodological framework of "uncertainty quantification - precise driver identification - targeted regulation design". The core of this framework comprises two interconnected modules: First, a distributed hydrological model (SWAT) is coupled with the Differential Evolution Adaptive Metropolis (DREAM) algorithm. Through Bayesian inference, the posterior distribution of model parameters is obtained, and combined with multi-route attribution analysis, the nonlinear contributions and uncertainties of climatic factors (precipitation, temperature, humidity, wind speed) and human activities (land use/cover change, LUCC) to streamflow are quantified, clarifying the positive/negative effects and spatial heterogeneity of each driving factor. Second, guided by the attribution results to target key drivers and their uncertainties, a three-dimensional adaptive scheduling system of "supply-demand-linkage" is constructed. Using a multi-objective optimization model solved by the Adaptive Hybrid Particle Swarm Optimization (AHPSO) algorithm, supply-side (cascade joint optimization, rainwater and flood resource utilization), demand-side (water-saving behavior adjustment), and supply-demand linkage regulatory measures are designed to achieve synergistic response to multi-dimensional driving forces. This framework has been applied to the Upper Yangtze River Basin, verifying its effectiveness in bridging attribution analysis and adaptive scheduling. It breaks the traditional disconnect between the two fields, providing scientific and operable methodological support for the dynamic management of water resources systems under changing environments, and can be widely extended to the collaborative optimization of reservoir group systems in complex river basins.