Estimation of Surface Zonal Current in the Equatorial Indian Ocean Based on Mechanism-Data Dual-Driven Model

Introduction: This study proposes a novel mechanism–data dual-driven algorithm for estimating surface zonal currents in the Equatorial Indian Ocean (EIO). Method: A feature extraction framework is first developed, integrating mechanistic analysis with data-driven approaches to predict surface zonal currents. Remote sensing data, including sea level anomalies and sea surface winds in the EIO, are preprocessed and analyzed to generate a preliminary feature set for model training. Principal Component Analysis (PCA) is then applied to reduce dimensionality and extract the most informative features. Subsequently, a Bi-directional Long Short-Term Memory (BiLSTM) network is designed for model training, resulting in the final computational model. Results: Comparative experiments demonstrate that the proposed mechanism–data dual-driven algorithm closely reproduces measured surface zonal current data, achieving high consistency in trend fitting and predictive accuracy. The model yields a mean absolute error (MAE) of 0.0159 m/s and a root mean square error (RMSE) of 0.0196 m/s. Compared to a BiLSTM model without feature extraction, these errors are reduced by 16.8% and 12.1%, respectively. Discussion: The model’s validity is further confirmed using data from an independent buoy observation point in the EIO, demonstrating strong generalization capability. Conclusion: The proposed mechanism–data dual-driven algorithm provides accurate estimation of surface zonal currents in the EIO, offering robust support for the analysis and prediction of surface current dynamics in this region.