| The Kuroshio Current in the East China Sea is one of the primary oceanic circulations in the northwest Pacific, exerting significant influence on China’s climate, ecology, and military activities. This study employs high-resolution monthly temperature and sea surface height data from JCOPE2M (Japan Coastal Ocean Predictability Experiment 2 Modified), along with the North Pacific Gyre Oscillation index, to explore the variability of the Kuroshio path and construct a prediction model. Through the extraction of frontal boundaries and streamline methods, the western boundary and axis of the Kuroshio are identified. Subsequently, a prediction model combining Variational Mode Decomposition (VMD) with Long Short-Term Memory (LSTM) networks is developed using a "decomposepredict-reconstruct" strategy. The results indicate that the VMD method effectively decomposes the Kuroshio path sequence into modal sub-sequences at different time scales, exhibiting an eastward (westward) shift during summer (winter) seasons and oscillations at scales ranging from 2 to 10 years in interannual variations. Furthermore, the performance and robustness of the VMD-LSTM model in predicting the Kuroshio path significantly outperform ARIMA and LSTM models, achieving the lowest mean square error of 0.064° . This study provides practical insights into understanding the dynamic variability of the Kuroshio in the East China Sea and demonstrates the potential application of the VMD-LSTM model in oceanic circulation research and prediction. |
