Under intensifying climate change and anthropogenic pressures, extreme low-flow events increasingly jeopardize water security in the Yellow River water supply region. This study develops the Inter-basin Multi-source Water Joint Allocation and Interconnected Routes Regulation System (IMWA-IRRS) to optimize spatiotemporal allocation of multi-source water and simulate topological relationships in complex water networks. The model integrates system dynamics simulation with multi-objective optimization, validated through multi-criteria calibration using three performance indicators: correlation coefficient (R), Nash-Sutcliffe Efficiency (E_ns), and percent bias (PBIAS). Application results demonstrated exceptional predictive performance in the study area: Monthly runoff simulations at four hydrological stations yielded R > 0.98 and E_ns > 0.98 between simulated and observed data during both calibration and validation periods, with |PBIAS| < 10%; human-impacted runoff simulations at four hydrological stations achieved R > 0.8 between simulated and observed values, accompanied by PBIAS within ±10%; sectoral water consumption across the Yellow River Basin exhibited PBIAS < 5%, while source-specific water supply simulations maintained PBIAS generally within 10%. Comparative analysis revealed the IMWA-IRRS model achieves simulation performance comparable to the WEAP model for natural runoff, human-impacted runoff, water consumption, and water supply dynamics in the Yellow River Basin. The 2035 water allocation scheme for Yellow River water supply region projects total water supply of 59.691 billion m³ with an unmet water demand of 3.462 billion m³ under 75% low-flow conditions and 58.746 billion m³ with 4.407 billion m³ unmet demand under 95% low-flow conditions. Limited coverage of the South-to-North Water Diversion Project’s Middle and Eastern Routes constrains water supply security, necessitating future expansion of their service areas to leverage inter-route complementarity while implementing demand-side management strategies. Collectively, the IMWA-IRRS model provides a robust decision-support tool for refined water resources management in complex inter-basin diversion systems.

Mid–long-term streamflow prediction (MLSP) plays a critical role in water resource planning amid growing hydroclimatic and anthropogenic uncertainties. Although AI-based models have demonstrated strong performance in MLSP, their capacity to quantify predictive uncertainty remains limited. To address this challenge, a DeepAR-based probabilistic modeling framework is developed, enabling direct estimation of streamflow distribution parameters and flexible selection of output distributions. The framework is applied to two case studies with distinct hydrological characteristics, where combinations of recurrent model structures (GRU and LSTM) and output distributions (Normal, Student’s t, and Gamma) are systematically evaluated. The results indicate that the choice of output distribution is the most critical factor for predictive performance. The Gamma distribution consistently outperformed those using Normal and Student’s t distributions, due to its ability to better capture the skewed, non-negative nature of streamflow data. Notably, the magnitude of performance gain from using the Gamma distribution is itself region-dependent, proving more significant in the basin with higher streamflow skewness. For instance, in the more skewed Upper Wudongde Reservoir area, the model using LSTM structure and Gamma distribution reduces RMSE by over 27% compared to its Normal-distribution counterpart (from 1407.77 m³/s to 1016.54 m³/s). Furthermore, the Gamma-based models yield superior probabilistic forecasts, achieving not only lower CRPS values but also a more effective balance between high reliability (PICP) and forecast sharpness (MPIW). In contrast, the relative performance between GRU and LSTM architectures was found to be less significant and inconsistent across the different basins. These findings highlight that the DeepAR-based framework delivers consistent enhancement in forecasting accuracy by prioritizing the selection of a physically plausible output distribution, thereby providing stronger and more reliable support for practical applications.

Mid–long term streamflow prediction (MLSP) plays a critical role in water resource planning amid growing hydroclimatic and anthropogenic uncertainties. Although AI-based models have demonstrated strong performance in MLSP, their capacity to quantify predictive uncertainty remains limited. To address this challenge, a DeepAR-based probabilistic modeling framework is developed, enabling direct estimation of streamflow distribution parameters and flexible selection of output distributions. The framework is applied to two case studies with distinct hydrological characteristics, where combinations of recurrent model structures (GRU and LSTM) and output distributions (Normal, Student’s t, and Gamma) are systematically evaluated. Results indicate that models employing the Gamma distribution consistently outperform those using Normal and Student’s t distributions. In the Upper Wudongde Reservoir area, the model using LSTM structure and Gamma distribution reduces RMSE from 1407.77 m³/s to 1016.54 m³/s. As the forecast horizon extends, the Gamma-based models demonstrate more reliable probabilistic predictions, reflected by sharper and better-calibrated prediction intervals. This is evidenced by substantially reduced CRPS values at the 18th forecast horizon (521.4 and 1746.6 m³/s), compared to Normal-based models (747.6 and 1877.5 m³/s). Although the improvements in predictive performance achieved by the proposed modeling framework vary depending on the RNN model architecture used and the specific application region, it generally delivers consistent enhancement in forecasting accuracy, thereby providing stronger support for practical applications.

Improving water resource use efficiency (WRUE) is essential for achieving sustainable water management under increasing socioeconomic and environmental pressures. This study investigates the spatiotemporal evolution, associated factors, and spatial transition characteristics of WRUE across 11 provincial-level administrative regions in the Yangtze River Basin during 2010–2024. An integrated framework combining the super-efficiency SBM-window DEA model, Malmquist–Luenberger index, GeoDetector, and conventional and spatial Markov chain models was developed to characterize efficiency dynamics, productivity changes, explanatory factors, and state-transition pathways. The results showed that WRUE exhibited an overall fluctuating upward trend with a clear spatial gradient of lower reaches > middle reaches > upper reaches. The mean ML index was 1.004, indicating that technological change (TC) was the main contributor to productivity improvement. Urbanization rate, water use per CNY 10,000 of GDP, industrial water-use share, and primary-industry share exhibited relatively high explanatory power, and their interactions enhanced explanatory power. Markov analysis revealed strong persistence in WRUE states, while transition probabilities differed across spatial neighborhood conditions. Assuming stable transition probabilities, the high-efficiency state would reach a steady-state probability of 0.8155. These findings provide insights for differentiated water resource management and coordinated regional development.