Climatological diagnostics and subseasonal-to-seasonal predictions of Madden-Julian Oscillation events

The Madden-Julian Oscillation (MJO) is the dominant mode of tropical intraseasonal variability, which serves as a primary source of subseasonal-to-seasonal (S2S) predictability. Noticeably, MJO is not always a regularly recurring cycle but is characterized by discrete episodes. In this study, considering the quasi-consecutive actives and eastward propagating features, a standard metric is proposed to identify MJO events based on the real-time multivariate MJO (RMM) index. The re-identification of historical MJO events reveals that there were 5.4 MJO events each year since 1981, and the average duration of each event is about 31.5 days. More MJO events tend to occur in the boreal winter and spring, with stronger intensity, longer duration and faster propagating speed than those in the boreal summer. Furthermore, MJO events are more likely to initiate in Phases 2 and 5, with a longer lifetime than those initiating in other phases. The amplitude and propagation characteristics of MJO events are strongly modulated by the dominant modes of sea surface temperature interannual variability with a strong regional dependence. Based on hindcast datasets of six S2S models, the prediction skill for the MJO is evaluated in the perspective of individual events. The longest leading time of the skilful prediction for individual MJO events ranges from 11 to 17 days, far below the traditional recognition. This result could be attributed to the apparent prediction barrier of MJO initiation, that is, a rapid decrease in prediction skill when predictions are carried out before the initiation of MJO events. In addition, the prediction skills of MJO events depends on interannual variabilities, with relatively higher skills under the conditions of the El Nino and Indian Ocean basin warming. These findings may shed light on the complexity and challenges of profoundly understanding and skilfully predicting MJO events.