Inter-basin water diversion projects (IWDPs) reshape the trade-offs among water supply, hydropower generation, and environment conservation (SHE) in both source and receiving regions. However, dynamic feedbacks between subsystem performance and management decisions remain difficult to represent in SHE nexus modelling and optimization. This study develops a dynamic feedback-driven modelling framework integrating hydrological simulation, ecological flow assessment, water quality modelling, reservoir operation, and system dynamics. Iteratively updated shape factors, defined as dimensionless priority coefficients, are introduced to transfer subsystem performance and socioeconomic changes into multi-objective synergistic optimization, enabling adaptive adjustment of priorities among the three subsystems. Applied to the Hanjiang River Basin during 1985–2024, the framework increases high-synergy months by 65 and substantially improves ecological flow provision, while maintaining high water-supply and diversion performance. The framework reveals ecological pressure redistribution caused by IWDPs and provides a transferable approach for adaptive management of the regulated river basins worldwide.
Accurate characterization of distribution tails is crucial for quantifying hydrometeorological extremes and estimating design quantiles. Existing Decision Support Systems (DSSs) often depend on subjective visual interpretation, limiting consistency and efficiency. This study introduces an automated Deep Learning (DL)-based DSS for objective tail characterization, classifying input datasets into ten distinct probability distributions. The framework leverages DL's pattern recognition capabilities applied to graphical representations of tail behavior, using seven specialized deep neural network models trained on concentration profiles, concentration-adjusted expected shortfall, Zenga curves, Hill ratio and Zipf plots. Extensive simulation experiments across multiple sample sizes demonstrate high accuracy, exceeding 90% and 98% for datasets larger than 5000 and 25,000 samples, respectively. Application of the framework to hydrometeorological extremes yielded tail characterizations that were broadly consistent with findings reported in previous studies. The proposed framework enables robust, scalable, and reproducible inference of distributional tails, supporting data-driven decision-making in hydrological and environmental modelling contexts.
