A new interpretation of the ability of global models to simulate the MJO

Abstract Statistical diagnostics have led to a common perception that only few global models can reproduce the Madden‐Julian Oscillation (MJO). Here we demonstrate, using a method of tracking individual MJO events, that this perception is incorrect. Of 27 global model simulations diagnosed, all produced large‐scale slowly eastward propagating signals in precipitation, which are taken as manifestations of the MJO. The difference is some model produced them frequently, others infrequently. There is no statistically significant distinction between the strength and propagation speeds of MJO events produced by most of these models. A hypothesis is proposed to interpret our results: A model can produce the MJO only in a particular background state. If the background state of a model can be constantly in a condition that is conducive to its production of the MJO, this model simulates the MJO frequently. If not, this model can still produce the MJO but only infrequently when its seasonal background state occasionally migrates into a condition that is conducive to its production of the MJO. Preliminary results from testing this hypothesis are presented.