Decadal modulation of ENSO dynamics emerges primarily from white-noise forcing

The dynamics of the El Niño Southern Oscillation (ENSO) vary from decade to decade in both observations and CMIP models. This complicates both ENSO prediction and the interpretation of any climate change signal. We aim to identify what processes are necessary to produce this decadal modulation, as quantified by ENSO growth rate and phase speed, using a set of recharge oscillator models with varying levels of nonlinearity.

We find that a linear recharge oscillator model with Gaussian white noise is sufficient to produce decadal modulation consistent with both observations and CMIP models. Although ENSO exhibits nonlinearity in other metrics, adding nonlinear terms to the recharge oscillator model does little to change the magnitude of decadal modulation. 

We show that the white noise forcing in recharge oscillator models impacts both ENSO emergent characteristics (e.g. amplitude) and representations of ENSO dynamical behaviour (e.g. growth rate, phase speed). Furthermore, we demonstrate that CMIP-class models do not produce ENSO predictability beyond the timescale expected from a linear recharge oscillator model. Together, these findings suggest future research should focus on the shorter-timescale processes influencing ENSO, including the sub-monthly processes typically modelled as noise, and highlight that short-timescale processes play an underappreciated role in influencing ENSO on much longer decadal timescales.