Exploring the freshwater provision service flows from a “Past-Future” perspective: A case study of the hilly area of South China

Against the backdrop of accelerating urbanization, freshwater provision services have become a critical area of research for sustainable development. This study develops a novel framework integrating river paths and watershed boundaries to simulate and predict freshwater provision service flows. The results revealed that from 2000 to 2020, the areas with a significant surplus water were primarily located in the northern regions of the Qiantang, Oujiang, and Minjiang River Basin, whereas severe freshwater shortages occurred predominantly in the Taihu Lake Basin. Additionally, the number of pixels with a surplus exceeding 1.5 × 10⁵ m³ per square kilometer initially increased and then declined, reaching 1.68 times the 2000 level by 2020. Based on watershed boundaries and their surplus, provision service flows were classified into three types, and a total of 34 flow paths were identified. Moreover, future projections for 2025 and 2030, have highlighted persistent spatial disparities. Machine learning models, particularly Random Forest, demonstrated superior predictive accuracy (R² > 0.8), outperforming traditional methods. Overall, this framework offers a robust approach for cross-basin water resource management, aiding in efficient allocation and sustainable governance of freshwater ecosystem services. Additionally, the proposed methodology offers a novel perspective for analyzing the spatiotemporal dynamics of freshwater provision service flows.