Does the MJO make ENSO more, not less, predictable?

Madden Julian Oscillation (MJO), being a largely stochastic phenomenon, is often considered a limitation on the predictability of El Niño—Southern Oscillation (ENSO). However, many characteristics of MJO events, such as amplitude and geographical location are not entirely stochastic; they are modulated by sea surface temperature (SST) that itself contains ENSO signature. Idealized surface wind-stress representing MJO forcing is added to a Coupled [ocean-atmosphere] General Circulation Model (CGCM) that does not contain any inherent MJO. Experiments are performed with and without idealized MJO being a function of model SST, to test whether the SST-modulated part of MJO dominates the stochastic part in terms of its ability to influence ENSO. Ensemble simulations are performed with a common initial state. It is found, that for lead times of up to one year, uncertainty of simulated ENSO with SST-modulation of MJO is lower than if the MJO is SST-independent and purely stochastic. Predictability, measured in terms of the ensemble spread of simulated ENSO indices, tends to be higher if the MJO is allowed to be influenced by ENSO SST. The predictability is found sensitive to calendar month of initialization, and in a related manner to the starting phase of ENSO. Consistent with previous studies, ENSO events exhibit much more realistic amplitude if the SST-modulation is switched on. Even some subtler features of a developing El Niño appear to be explainable upon considering the mutual interaction between MJO and ENSO.