Improved Random Vector Functional Link Network with an Enhanced Remora Optimization Algorithm for Predicting Monthly Streamflow

Precise and robust streamflow estimation is crucial for effective water resource management, particularly in mitigating extreme climatic events such as droughts and floods. This study introduces an innovative integration of the Random Vector Functional Link (RVFL) network with an Enhanced Remora Optimization Algorithm (EROA), specifically designed for monthly streamflow prediction. The RVFL-EROA is compared against standalone RVFL and RVFL models optimized using the Gorilla Troops Optimizer (GTO), Whale Optimization Algorithm (WOA), and the original Remora Optimization Algorithm (ROA). Performance is evaluated using statistical indices, including Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Coefficient of Determination (R2), and Nash-Sutcliffe Efficiency (NSE). The methodology is tested on streamflow time series data from the Kunhar River Basin in Pakistan, with input variables derived from antecedent streamflow, air temperature, and rainfall. Results indicate that temperature and streamflow-based inputs yielded higher accuracy compared to rainfall inputs. The RVFL-EROA outperformed other models, achieving improvements in mean RMSE, MAE, R2, and NSE by 8.63-1.77%, 12.08-1.58%, 16.88-3.33%, and 19.03-3.05%, respectively. Moreover, the RVFLEROA demonstrated superior performance in estimating peak streamflow values, which is critical for flood management. These findings highlight the potential of temperature-based inputs and RVFL-EROA models for streamflow prediction in data-scarce regions, particularly in developing countries. The proposed approach offers a reliable solution for enhancing hydrological forecasting and supports efficient water resource planning.