The mechanism of the ENSO diversity predictability paradox

• CP ENSO is more predictable than EP ENSO, but the opposite holds for El Niño. • Decadal variability in the central tropical Pacific enhances CP ENSO prediction. • Weaker thermocline feedback in the central Pacific reduces CP El Niño prediction. The predictability of the El Niño–Southern Oscillation (ENSO) has attracted increasing attention, as it has important impacts on both global ecosystems and the global climate system. However, current ENSO prediction capabilities remain limited. Previous studies have revealed an ENSO diversity predictability paradox: the predictability of Central-Pacific (CP) ENSO is higher than that of Eastern-Pacific (EP) ENSO, while the predictability of CP El Niño is lower than that of EP El Niño. The underlying mechanisms of this predictability paradox remain unclear. Here, we investigate the causes of this predictability paradox using a Linear Inverse Model and a nonlinear recharge oscillator model. Our results demonstrate that decadal variability in the central tropical Pacific is the dominant factor contributing to the higher predictability of CP ENSO relative to EP ENSO. When decadal variability in the tropical Pacific is removed using high-pass filtering, the predictability difference between CP and EP ENSO largely disappears. However, decadal variability has limited influence on the predictability of different El Niño types. Moreover, we find that weaker thermocline feedback in the central tropical Pacific is responsible for the lower predictability of CP El Niño events. In the nonlinear recharge oscillator model, strengthening the thermocline feedback parameter in the central Pacific increases the predictability of CP El Niño to that of EP El Niño. The strength of thermocline feedback is related to the mean state of thermocline depth. These results highlight the importance of decadal variability and thermocline feedback for ENSO diversity predictability, offering new insights into the simulation and prediction of ENSO diversity in climate models.