Spatiotemporal Dynamics and Scenario Optimization of Water Resource Utilization Efficiency: A Measurement-Prediction Framework for the Yangtze River Basin

Improving water resource utilization efficiency is crucial for balancing economic development, resource conservation, and environmental protection in large river basins. This study develops a measurement prediction-optimization framework for 11 provincial-level regions in the Yangtze River Basin from 2010 to 2023. A super-efficiency epsilon-based measure model with undesirable outputs is used to measure efficiency and identify spatiotemporal patterns. An adaptive particle swarm optimization–LightGBM model is developed for efficiency prediction, with SHAP used to interpret key drivers. A constrained multi-objective particle swarm optimization model optimizes fixed asset investment, labor input, and total water use. The results show that basin-wide efficiency increased from 0.269 in 2010 to 0.732 in 2023, showing lower-reach leadership, upper-reach catch-up, and middle-reach differentiation. The APSO-LightGBM model achieved a test R2 of 0.894 and an RMSE of 0.093, with GDP, wastewater discharge, and total water use as the main contributors. Optimization generated feasible solutions for all 154 observations, increasing average predicted efficiency from 0.410 to 0.435 while reducing fixed asset investment, labor input, and water use by 4.26%, 1.81%, and 6.88%, respectively. Sensitivity analyses indicate that the resource adjustments depend on the solution-selection threshold and empirical bounds, especially for investment and employment. The results describe conditional input-saving opportunities rather than directly achievable policy targets.