El Niño Regimes, Nonlinear Convective Feedbacks and the Predictability of Extreme Events

Abstract:

We propose that extreme El Niño (EN) events, such as the ones in 1982-1983 and 1997-1998, are qualitatively different from the other EN due to nonlinear feedbacks in the eastern Pacific, specifically the triggering of deep convection and the associated amplification of Bjerknes feedback. Evidence for this is provided by observations contrasted with the longer series from the GFDL CM2.1 global climate model (GCM), which shows a bimodal distribution of EN peaks, with one of the modes corresponding to extreme EN. We also show results from a low-order stable model that is able to reproduce this bimodality solely with the indicated threshold nonlinearity and that shows substantial sensitivity of the statistics of the extreme EN to the threshold value and the stochastic forcing characteristics.

The observational and GCM analysis indicates that positive equatorial heat content anomalies are a necessary but not sufficient condition for such events. On the other hand, the presence of strong westerly wind stress anomalies in the central equatorial Pacific in boreal summer, as seen in 1982 and 1997, appears to be both necessary and sufficient condition for triggering extreme El Niño. In 1982, external forcing of such westerly stress at that time was apparently key for kickstarting this extreme EN.

Preliminary analysis of retrospective forecasts for the 1982- 2010 period from by the GCMs participating in the US National Multimodel Ensemble (NMME) project, which includes GFDL CM2.1, specifically looking at the onset of the observed extreme El Niño, indicates that the convective nonlinear feedbacks are strongly affected by the mean surface temperature drift in the GCMs and that nonlinear bias correction might be necessary for the operational prediction of extreme EN.