Mechanisms of Intraseasonal Oscillation in Equatorial Surface Currents in the Pacific Ocean Identified by Neural Network Models

Abstract The characteristics and origins of intraseasonal oscillations (ISOs) in surface currents over the equatorial Pacific are yet to be detailed due to the deficiency of observational data. This study constructs the Pacific surface currents along the equator (every 0.25° of longitude) from 1993 to 2017 (at daily intervals) using a feedforward neural network and multiple sea surface variables, showing superior correlations and root mean square errors with in situ measurements. Based on this product, the ISOs explain ∼10%–30% and ∼20%–50% of the zonal and meridional current variance, respectively, exhibiting disparate characteristics in the western and eastern Pacific. The Madden‐Julian Oscillation (MJO) governs the western basin, where the ISOs are more intense during El Niños. Particularly, significant ISOs in zonal currents (−0.25–0.28 m s −1 ) span nearly the entire basin during EP‐El Niño summers and winters; the intensity becomes even stronger (−0.27–0.32 m s −1 ) for CP‐El Niño winters, as the MJO convection center thrives throughout the life cycle. The intraseasonal meridional currents, though much weaker, extend eastward up to ∼150°W during EP‐El Niño winters. As for the eastern basin, the ISOs arise primarily from baroclinic instability and propagate westward with the phase speed and the domain being fastest and most widespread for La Niñas and vice versa for El Niños. Both the temperature and salinity effects play an essential role. This study introduces an efficient approach to construct equatorial currents using machine learning, facilitating a deeper diagnosis of the tropical ocean circulation dynamics.