The most important necessary condition of NEWS is the ENSO skewness (El Niños are stronger than La Niñas). Most CMIP5 models do not reproduce the observed ENSO skewness, while GFDL-ESM2M exhibits the realistic ENSO skewness, which suggests that, despite being in the minority, the La Niña-like trend of GFDL-ESM2M could be a plausible equatorial Pacific response to transient warming. In this study, we introduce another interesting outlier, MIROC5, which reproduces the observed skewness, yet exhibits an El Niño-like response.
We decompose the source of the ENSO nonlinearity into the following three components: "SST anomalies modulate winds", "winds excite oceanic waves", and "oceanic waves modulate the subsurface temperature", and show that the large inter-model spread of the third component appears to explain most important cause of the poor reproducibility of the ENSO nonlinearity in CMIP5 models. It is concluded that the change in the relationship of subsurface temperature to oceanic waves is the primary explanation for the different warming response of GFDL-ESM2M and MIROC5. Our analyses suggest that the differences of the warming response are caused by difference in the climatological thermal stratification.
This study may shed new light on the fundamental question of why observed ENSO has a strong skewness and on the implications of this skewed ENSO for the mean-state sea surface temperature response to global warming.