<p>Effective monthly runoff prediction is crucial for reservoir scheduling, water resources management, and efficient utilization. To enhance the accuracy and stability of these predictions, this study proposes a data-driven modeling method based on Variational Mode Decomposition (VMD) and multi-strategy enhanced Artificial Rabbit Optimization algorithm (MARO), combined with Support Vector Regression (SVR) for prediction. Firstly, VMD is used to extract multi-scale features from the time series to address nonlinear and non-stationary issues. Then, MARO is employed to optimize the parameters of the SVR model. The MARO algorithm introduces adaptive weight factors, elite opposition-based learning, and a random walk strategy, significantly improving the accuracy of parameter optimization. Additionally, an error correction mechanism based on MARO-SVR is introduced to further enhance the reliability of the prediction model. The model is applied to monthly runoff predictions in the Xiajiang hydrological station in the Ganjiang River Basin and the Jiayuguan Hydrological station in the Heihe River Basin, and validated using five evaluation metrics: MAE, RMSE, MAPE, NSEC, and R. Results indicate that the VMD-MARO-SVR-EC model reduces overall errors by 75.12% in the Xiajiang hydrological station and by 80.6% in the Jiayuguan Hydrological station, significantly improving prediction accuracy and adaptability to complex runoff variations. These findings provide a new approach for monthly runoff forecasting and offer robust support for practical applications in watershed water resource management.</p>