Improvement of ENSO simulation by the conditional multi-model ensemble method

The El Nino Southern Oscillation (ENSO) is an important interannual atmosphere-ocean interaction phenomenon in the tropical Pacific. To study ENSO, the atmosphere-ocean coupled models are frequently used to simulate ENSO behaviors and mechanisms. The capability of a single model is limited to simulate the temporal-spatial variations of ENSO; therefore, the simulations are required to be improved through the statistical ensemble method. Based on ENSO simulations of the Coupled Model Intercomparison Project Phase 5 (CMIP5), this paper applied the conditional multi-model ensemble (CMME) method, compared the MME method, investigating the ensemble results and physical reasons of improvement. The CMME can reproduce ENSO behaviors very well, including ENSO's interannual variability, temporal-spatial evolution, seasonal phase-locking, as well as the asymmetry of El Nino and La Nina. Besides, due to the reproductions of ENSO initiation, positive and negative feedback and nonlinear heating mechanism, we point out that the CMME capturing ENSO mechanisms is the reason why it can perform ENSO characteristics. Statistical results also support that this CMME method is effective, stable, and reliable.