Water discharge forecasting in river basins using HydroESN across multiple gauging stations

Accurate River discharge forecasting is essential for effective water resource management, enabling reliable flood prevention, irrigation planning, and hydropower generation. However, traditional forecasting methods often struggle to capture the highly dynamic and nonlinear behavior of climatic variables, particularly precipitation, due to their reliance on linear assumptions and fixed modeling structures. To overcome these limitations, this study proposes a HydroESN model that integrates Echo State Networks with enhanced optimization techniques to better represent complex hydrological dynamics. The model is evaluated on benchmark datasets, including CABra and GRDC, where it demonstrates superior performance with an R² value of up to 0.98 and classification accuracy reaching 98.8%, outperforming conventional baseline models. These results highlight the model’s effectiveness and its potential to support real-world applications such as flood forecasting, reservoir operation, and irrigation management.