Seasonality in Prediction Skill of the Madden-Julian Oscillation and Associated Dynamics in Version 2 of NASA's GEOS-S2S Forecast System

The prediction skill of the Madden-Julian Oscillation (MJO) in Version 2 of NASA's Global Earth Observing System Subseasonal to Seasonal (GEOS-S2S) forecast system is investigated for winter and summer focusing on moistening-related processes crucial for eastward propagating MJO activity. It is found that the annual bivariate correlation of the Real-time Multivariate MJO time series between prediction and observation is similar to 0.70, similar to 0.57, and similar to 0.50 at 20-, 25-, and 30-day forecast leads. Correlation at long-leads (>30 days) is noticeably higher for boreal summer initial conditions (June-September [JJAS]), with correlations remaining above 0.5 at 35-40 days leads. Correlations are lower for boreal winter initial conditions from January through March (JFM), dropping to similar to 0.5 at 25-day lead, still comparable to the skills in the other reliable S2S forecast systems. The predicted eastward MJO propagation across the Indo-western Pacific sector is well captured in JJAS, but is slower than observed in JFM. Investigations of the moisture field and advection and moisture sink, moist static energy (MSE) budget, and tropical circulation/pressure responses to the MJO convective heating reveal that, in JFM, those responses and moistening processes, especially the vertical MSE advection, are underestimated over and to the east of the Maritime Continent when the MJO anomaly approaches from the west. In contrast, those processes are well represented in JJAS, although moistening is overestimated due to large surface evaporation. This study suggests that improvement of the moistening tendency over that region in the boreal winter could contribute to further increases in MJO prediction skill of the GEOS-S2S system.