Prediction of Runoff Using Hybrid SVM-SSA in Baitarani River Basin, Odisha, India

Accurate prediction of surface runoff is essential for effective water resource management, particularly in regions susceptible to hydrological variability such as the Baitarani River Basin in Odisha, India. Traditional hydrological models often fail to capture the complex, nonlinear, and non-stationary behavior of rainfall-runoff processes, especially under the influence of climatic and anthropogenic changes. In this study, a hybrid model integrating Support Vector Machine (SVM) with Salp Swarm Algorithm (SSA) is proposed to enhance the accuracy of monthly runoff prediction. The performance of the SVM-SSA model is compared against conventional Artificial Neural Network (ANN) and standalone SVM models using datasets from Anandpur and Champua gauging stations. Inputs include rainfall, temperature, specific humidity, and relative humidity, offering a comprehensive representation of climatic influences. Model performance is evaluated using statistical metrics such as Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Coefficient of Determination (R²), and Willmott’s Index (WI). The results demonstrate that the hybrid SVM-SSA model significantly outperforms the conventional models, achieving R² values of 0.9847 and 0.9771 at Anandpur and 0.9844 and 0.9756 at Champua for training and testing phases, respectively. These findings suggest that coupling machine learning with metaheuristic optimization provides a promising approach for runoff prediction in complex river basins.