Climate change significantly alters hydrological regimes, with direct implications for sediment transport and dam sustainability. This study introduces a novel hybrid modelling framework that combines statistical downscaling of discharge using machine learning with a probabilistic copula-based approach to assess future sediment load under changing climate conditions. The methodology is applied to the Blue Nile River at Eldeim station using projections from five CMIP6 Global Climate Models (GCMs) under three Shared Socioeconomic Pathways (SSP1 \(-\) 2.6, SSP2 \(-\) 4.5, and SSP5 \(-\) 8.5). The machine learning downscaling process is designed to ensure physical consistency and statistical robustness. First, the synoptic-scale climate variables simulated by each GCM are analyzed to identify those most correlated with observed discharge variability. The best-performing machine learning algorithm, selected through comparative performance evaluation among ANN, BNN, and Random Forest (Bootstrap and Bagging), is then applied to downscale the discharge. In parallel, individual and multi-model ensembles of the GCMs are evaluated to quantify their skill in reproducing historical discharge variability and to generate reliable future projections for 2030-2100. To cope with limited sediment data, a copula-based probabilistic model is employed to simulate the dependence structure between discharge and sediment load. Using the downscaled discharge projections, future sediment load distributions are inferred from the joint probability model. Statistical metrics indicate that the Frank copula provides the best representation of this dependency. Results reveal a projected shift in peak sediment load from August to September under high-emission scenarios, with an increase of up to \(16\%\) by 2100. These changes may reduce the storage capacity and hydropower potential of the Grand Ethiopian Renaissance Dam. The proposed hybrid data-driven framework offers a transferable and robust solution for climate-resilient water and sediment management in data-scarce basins worldwide.
