Yaxin Zhu, Guilin Han, Di Wang, Yuchun Wang
As the world's largest hydropower project, the Three Gorges Dam has created a relatively static water environment in the backwater area, which favors algal bloom occurrence and may negatively affect the regional ecosystem. However, how mainstream-tributary mixing shapes distinct water-mass states and their relative susceptibility to algal blooms remains unclear. In this study, stratified water samples were collected from the Three Gorges Reservoir mainstream Yangtze River (YZR) and from the Pengxi River (PXR), Daning River (DNR), and Xiangxi River (XXR), three representative tributaries. Stable hydrogen (δ2H) and oxygen (δ17O and δ18O) isotope results showed that the YZR had depleted isotopic compositions, with mean δ2H, δ18O, and δ17O values of -75.86‰, -10.85‰, and -5.81‰, respectively. Bayesian mixing analysis quantified distinct mainstream recharge proportions among the tributaries, averaging 29.84% in the PXR, 49.45% in the DNR, and 61.06% in the XXR. These proportions reflected stratified mixing in the PXR, downstream-increasing mainstream intrusion in the DNR, and persistent mainstream control in the XXR. K-means clustering, an unsupervised machine learning method, further distinguished four hydro-isotopic water-mass types based on mainstream contribution and isotope-derived hydrodynamic signatures. Historical bloom records, together with chi-square tests and a binomial generalized linear model revealed significant differences in bloom susceptibility among these types. Among them, Cluster 2 showed the highest predicted bloom probability, which was associated with its strong mainstream intrusion and limited water exchange. Overall, stable isotopes sensitively capture hidden hydrodynamic structure and provide a practical basis for bloom-susceptibility assessment and adaptive management in large regulated reservoirs.