Deep Learning Reveals ENSO's Footprint on the Indian Ocean Dipole: Insights From the Eastern Pacific (American) Coast

Abstract The Indian Ocean Dipole (IOD) significantly influences global climate and ecosystem dynamics, yet accurate forecasting remains challenging due to its complex nature. Here, we present an interpretable deep‐learning framework, STPNet, that achieves the‐state‐of‐the‐art 8‐month forecasting of fall IOD events by leveraging sea surface temperature anomalies (SSTA) and sea surface height anomalies (SSHA) from CMIP6 data. Through STPNet's interpretability features and targeted sensitivity experiments, we not only confirmed previously known IOD precursors but also identified a novel precursor along the eastern Pacific (American) Coast. Subsequent lagged regression analysis revealed this precursor's connection to ENSO‐mediated coupling between extratropical and subtropical Pacific SST patterns, completing the global map of IOD precursors. Our findings substantially advance the understanding of IOD mechanisms and provide a robust framework for operational forecasting, with direct implications for global climate adaptation strategies.