Abstract The influence of the tropical Atlantic on El Niño‐Southern Oscillation (ENSO) predictability remains debated, with contrasting views on whether it reflects a genuine teleconnection or ENSO's autocorrelation. Using a convolutional neural network (CNN), we revisit this issue and demonstrate the important role of summer tropical Atlantic in ENSO predictions. The CNN reveals that tropical Atlantic during boreal summer, autumn, and winter enhances ENSO predictability, whereas its predictive value is limited in spring. We employ pacemaker simulations to elucidate mechanisms underlying CNN findings. Warm summer sea surface temperature (SST) anomalies near Intertropical Convergence Zone (ITCZ) amplify atmospheric deep convection, generating atmospheric Kelvin waves, and modifying zonal atmospheric circulation that alter Pacific Walker circulation and thermocline adjustments, promoting winter La Niña. In contrast, weaker convection arises as spring SST anomalies lie farther from ITCZ, diminishing their influence. This study provides new evidence for a season‐dependent Atlantic‐Pacific linkage and highlights a summer‐initiated pathway for ENSO predictability.
Huang et al. (Thu,) studied this question.