Spatial and temporal characteristics of water conservation services and rapid response framework for water yield in key ecological zones of the Yiluo River basin

Yiluo River basin, an important water source and ecological barrier in China. In this study, we analyzed the spatial and temporal patterns of water yield and the main influencing factors of the Yiluo River Basin based on the water yield response framework constructed by the SWAT model and the intelligent optimization algorithm. The results indicated that the average annual total of water-source containment per unit area in the district was 330.03 mm from 2019 to 2023, with a Nash-Sutcliffe Efficiency (NSE) of 0.77, based on the average of two hydrological sites in the SWAT model. The high value of water conservation goes mainly in the forested mountainous areas of the upper reaches of the Yi River and concentrated in July–October, seasonal differences in the amount of water conservation are mainly influenced by precipitation (correlation of 0.79), and potential evapotranspiration determines its lower limit value. Urban land uses and riparian areas with high levels of hydraulic erosion are areas with low water yield concentration. The artificial neural network-based prediction framework achieved high performance with Pearson correlation coefficients exceeding 0.90 across all datasets. The average relative error was 1.31 % (training), 1.39 % (validation), and 1.24 % (test), with MAPE values below 2 %. This approach allows flexible scenario modeling and has been successfully applied to seven cases, offering valuable early-warning insights for regional ecological planning and water resource management. • Built a rapid response framework for water production in a comprehensive scenario. • Improved SWAT flexibility in prediction using deep learning algorithms. • Integration of multi-factor response framework facilitates ecological decision-making.