Figure 14. Prediction skill of the MJO’s RMM indices in S-SHiELD with (solid) and without (dotted) the interactive MLO for 92 40-day predictions initialized during the 2011-2012 DYNAMO period. Top: Correlation coefficient; bottom: RMSE.
The behavior of the MJO in GFDL’s CMIP6-generation climate models (Zhao et al. 2018) suggests that the two keys for a good MJO simulation are an appropriate convection scheme and some form of interactive ocean, a result found also by DeMott et al. (2019) and others. A second set of S-SHiELD experiments was performed using specified climatological SSTs plus frozen anomalies. These simulations without the interactive MLO had much smaller RMM correlations, with predictions no longer useful after day 20, and larger errors. The effect of the interactive ocean is made clear in Figure 15, in which S-SHiELD with the MLO correctly predicted the formation of all three strong MJO events during this period 10–15 days in advance, and correctly propagated all events through the Maritime Continent (near 120 E longitude), although the propagation speed is slower than observed and there is some disruption near the Maritime Continent. However, S-SHiELD with prescribed SSTs has difficulty propagating the MJO through the Maritime Continent and for the November event creates no MJO whatsoever. The November event proves particularly challenging for S-SHiELD without the MLO as it performs poorly at a range of lead times (Supplemental Figure S2) but poses no problem for S-SHiELD with the MLO. It is clear that the simple, inexpensive MLO used in S-SHiELD is sufficient to significantly extend the predictability of the MJO.
DeMott et al. (2019) did not describe any deficiencies of the MJO from models using a 1D column ocean instead of a 3D dynamical ocean, which suggests a limited role for direct feedbacks between ocean circulations and the MJO. However they did not rule out indirect effects of the MJO on ocean circulation that could impact other S2S-timescale phenomena or MJO teleconnections. Other investigators have found that the MJO does alter ocean circulations on intraseasonal timescales, notably the result of Moum et al. (2014) found during DYNAMO. It remains to be seen whether these ocean dynamical effects of the MJO are of sufficient impact to affect S2S prediction skill. One advantage of the MLO is that we can nudge to climatological SSTs and so do not have climate drifts that challenge fully coupled models.