Predictability and prediction of the madden-julian oscillation: A review on progress and current status

The progress and current status of prediction activities on the Madden-Julian Oscillation (MJO) are reviewed. We have witnessed the significant progress in dynamical MJO prediction from 7-day limit using the (NCEP)-1 reanalysis vintage model during 1990s to about 30-day limit using the latest version of European Centre for Medium-Range Weather Forecasts (ECMWF) model during 2010s. Motivated by significant societal demands for reliable subseasonal prediction, several international efforts have been coordinated including real-time MJO and boreal summer intraseasonal oscillation (BSISO) prediction intercomparison by MJO Task Force and Working Group on Numerical Experimentation (WGNE), and Intraseasonal Variability Hindcast Experiment (ISVHE). Analysis of 12 climate models' hindcast participated in ISVHE showed that the MJO can be potentially predictable up to 35-45 days and the multi-model ensemble has a useful skill for the Realtime Multivariate MJO (RMM) index up to 26-28 days for 1989.2008 during boreal winter, demonstrating capability of current models for seamless weather-intraseasonal-seasonal prediction. It is also noted that the current MJO prediction depends on initial and target MJO phase and El Ni.no and Southern Oscillation (ENSO) phase. The dynamical models' common biases in simulating the MJO are slower eastward propagation, underestimation of MJO amplitude, and the Maritime Continent barrier. Key factors to further improve the MJO prediction include improvement of models' intrinsic MJO mode, better initialization, and proper representation of air-sea coupling.