Predictability of the Madden–Julian Oscillation in the Intraseasonal Variability Hindcast Experiment (ISVHE)*

Abstract The Madden–Julian oscillation (MJO) represents a primary source of predictability on the intraseasonal time scales and its influence extends from seasonal variations to weather and extreme events. While the last decade has witnessed marked improvement in dynamical MJO prediction, an updated estimate of MJO predictability from a contemporary suite of dynamic models, in conjunction with an estimate of their corresponding prediction skill, is crucial for guiding future research and development priorities. In this study, the predictability of the boreal winter MJO is revisited based on the Intraseasonal Variability Hindcast Experiment (ISVHE), a set of dedicated extended-range hindcasts from eight different coupled models. Two estimates of MJO predictability are made, based on single-member and ensemble-mean hindcasts, giving values of 20–30 days and 35–45 days, respectively. Exploring the dependence of predictability on the phase of MJO during hindcast initiation reveals a slightly higher predictability for hindcasts initiated from MJO phases 2, 3, 6, or 7 in three of the models with higher prediction skill. The estimated predictability of MJO initiated in phases 2 and 3 (i.e., convection in Indian Ocean with subsequent propagation across Maritime Continent) being equal to or higher than other MJO phases implies that the so-called Maritime Continent prediction barrier may not actually be an intrinsic predictability limitation. For most of the models, the skill for single-member (ensemble mean) hindcasts is less than the estimated predictability limit by about 5–10 days (15–25 days), implying that significantly more skillful MJO forecasts can be afforded through further improvements of dynamical models and ensemble prediction systems (EPS).