Most CMIP5 models underestimate MJO amplitude, especially when outgoing longwave radiation is used in the evaluation, and exhibit too fast phase speed while lacking coherent eastward propagation between precipitation and the wind field. The RHCP-metric, indicative of the sensitivity of simulated convection to low-level environmental moisture, and the NGMS-metric, indicative of the efficiency of a convective atmosphere in exporting moist static energy out of the column, show robust correlations with a large number of MJO skill metrics. The GEF-metric, indicative of the strength of the column-integrated longwave radiative heating due to cloud-radiation interaction, is also correlated with the MJO skill metrics, but shows relatively lower correlations compared to the RHCP- and NGMS-metrics.
Our results suggest that modifications to processes associated with the moisture-convection coupling, and the gross moist stability might be the most fruitful for improving simulations of the MJO. Though the GEF-metric exhibits fewer significant correlations with the MJO skill metrics, the longwave radiation feedback is highly relevant for simulating the weak precipitation anomaly regime that may be important for the establishment of shallow convection and the transition to deep convection.