<p>Accurate statistical analysis of streamflow and sediment load plays a vital role in promoting sustainable river management, coastal engineering, and water conservancy infrastructure. Existing probabilistic models often fail to accurately reflect the complex dependence structure between streamflow and sediment load, limiting their practical application. To address this issue, this study proposes a novel bivariate probability model framework that combines nonparametric kernel density estimation (KDE) and a mixed copula function to model the joint probability distribution of streamflow diversion ratio and sediment diversion ratio. KDE is utilized to accurately fit the marginal distributions, while the mixed copula function is developed as a linear combination of Gumbel, Clayton, and Frank copulas. Through the data verification of two stations in the Pearl River network area, the results show that: (1) the framework expands the application potential of existing methods by combining multiple copula functions, and provides a more effective way to describe the dependence structure of streamflow diversion ratio and sediment diversion ratio. (2) KDE outperforms four classical parametric distributions in modeling diversion ratios; (3) the mixed copula function exhibits up to 76.08% and 43.21% relative differences in recurrence period and conditional probability estimates, respectively, compared to single copula functions under consistent evaluation settings. Moreover, spatial heterogeneity is effectively characterized through adaptive weight allocation in the mixed copula. This framework provides both a methodological innovation and a practical tool for supporting refined water-sediment management in complex river systems.</p>