ENSO Atmospheric Feedbacks under Global Warming

Two atmospheric feedbacks play an important role in the dynamics of the El Niño/Southern Oscillation (ENSO), the amplifying zonal wind feedback and the damping heat flux feedback. Here we investigate how and why both feedbacks change under global warming in climate models of 5^th and 6^th phase of the Coupled Model Intercomparison Project (CMIP5 and CMIP6, respectively) under a “business-as-usual” scenario (RCP8.5 and SSP5-8.5, respectively). The amplifying wind feedback over the western equatorial Pacific (WEP) becomes stronger in most climate models (on average by 8 ± 8%) as well as the damping heat flux feedback over the eastern and central equatorial Pacific (EEP and CEP, respectively) (on average by 18 ± 11%). The simultaneous strengthening of both feedbacks can be explained by the stronger warming in the EEP relative to the WEP and the off-equatorial regions, which shifts the rising branch of the Pacific Walker Circulation to the east and increases mean convection and precipitation over the CEP. This in turn leads to a stronger vertical wind response during ENSO events over the CEP that strengthens both atmospheric feedbacks. Further, we separate the climate models into sub-ensembles with STRONG and WEAK ENSO atmospheric feedbacks, as 2/3 of the models underestimate both feedbacks under present day conditions by more than 40%, causing an error compensation. Despite both sub-ensembles show similar changes in the mean state and ENSO atmospheric feedbacks, the ENSO dynamics in WEAK remain weaker relative to STRONG under global warming. Due to the more realistic ENSO dynamics, we postulate that the ENSO predictions of the models in STRONG should be more reliable. Finally, we analyze the relation between changes in ENSO amplitude and ENSO atmospheric feedbacks. We find that models tending to simulate an eastward shift of the wind feedback and increasing precipitation response over the EEP during Eastern Pacific El Niño events also exhibit an increasing ENSO amplitude.