The dominant role of the atmospheric component of coupled models in ENSO amplitude simulations

To explore the relative contributions of the atmospheric and oceanic components of coupled models to ENSO amplitude simulations, we innovatively assembled four coupled models and performed analyses on their ENSO simulations. Specifically, the atmospheric and oceanic components of two commonly used coupled models are cross-coupled to construct four parent models. Based on the simulated ENSO amplitude, the four parent models are classified into two groups: Grid-point Atmospheric Model of IAP LASG Version 2 (GAMIL2)-based models whose ENSO amplitudes are comparable to (although slightly weaker than) observations, and Community Atmosphere Model Version 4.0 (CAM4)-based models whose ENSO amplitudes reach up to twice those of observations. The BJ-index analysis reveals that the atmospheric components modulate ENSO amplitude by affecting the atmospheric thermodynamic (TD) feedback and the oceanic thermocline (TH) feedback. The TD feedback biases in the CAM4-based models are attributable to an overly negative low-cloud fraction feedback and low-cloud liquid water feedback in the Nino-3 region. The underestimated TH feedback in the GAMIL2-based models is due to an underestimated mean upwelling (w) over bar), while the seemingly accurate TH term in the CAM4-based models is the result of compensation by an overestimated regression of zonal tilt of the thermocline on the equatorial zonal wind stress (h) and an underestimated (w) over bar. Furthermore, h dominates TH differences in the two atmospheric groups, and is mainly associated with the normalized wind stress anomaly over the Nino-4 region and the vertical ocean subsurface temperature structure.