Multiple driving effects and causal transmission mechanisms of watershed runoff variations based on a three-level attribution framework

Runoff variations in large connected basins are jointly affected by climate variability, human activities, reservoir regulation, and upstream–downstream hydrological connectivity. This makes cascading runoff transmission among sub-basins highly complex. Therefore, this study focuses on 11 controlled sub-basins (S1–S11) in the Yangtze River Basin and uses meteorological and hydrological data from 1965 to 2019. We developed a three-level coupled attribution framework that integrates the Mann–Kendall change-point test, interval runoff contribution identification, Budyko-based attribution, and PCMCI+ causal analysis. The results show that two main runoff change points occurred in the Yangtze River Basin, in 1985 and 2002. The mean interval runoff contribution decreased from 152.7 m 3 s −1 in the reference period (1965–1985) to 138.9 m 3 s −1 in Change Period 1 (1986–2002). It then recovered to 158.4 m 3 s −1 in Change Period 2 (2003–2019). Compared with Change Period 1, the contribution rate of human activities in the middle reaches decreased from 60.5% to 44.7% after Change Period 2. In terms of sub-basin function and transmission, S1–S2 served as upstream runoff supply zones, S3, S5, and S7 were the main midstream runoff attenuation units, and S8–S11 acted as downstream convergence recovery zones. These findings improve the understanding of runoff variation mechanisms, spatial transmission relationships, and functional differentiation in large basin–sub-basin systems. They also provide theoretical support and decision-making references for optimized water resource allocation and differentiated basin management.