MJO Prediction Skill and Associated Dynamics as Revealed by NASA GEOS Subseasonal Forecasts

Abstract:

The skill of MJO predictions in the GEOS Sub-seasonal to Seasonal (S2S) forecast model is investigated. The current operational GEOS-S2S uses a fully coupled global model, with half-degree resolution in both the atmosphere and the ocean. Every five days, four forecasts are produced using unperturbed and perturbed initial conditions based on a weakly coupled analysis. It is found that the prediction skill, in terms of bivariate correlation of the Real-time Multi-variate MJO time series between prediction and observation, exceeds 0.65, 0.55, and 0.50 at 20-, 25-, and 30-day forecast lead. Correlation at long lead times (>30 day) is noticeably higher for boreal summer initial conditions from June through September, with correlations remaining above 0.5 at 35-day forecast leads. Correlations for winter initial conditions are lower, dropping to 0.5 at ~27 day forecast lead.

The predicted MJO intensity over the Pacific is well captured in boreal summer, but is weaker than observed in boreal winter. Eastward MJO propagation is also slower than observed at long lead times in winter. Investigations of the moisture field and advection, moist static energy structure and its tendency, and tropical responses of Kelvin wave, circulation, pressure and moisture to the MJO heating reveal that, in boreal winter, those responses and premoistening processes are underestimated by the model over and to the East of the Maritime Continent when the MJO convective anomaly approaches. In contrast, this process is well represented in boreal summer. This analysis suggests that improvement of these physical interaction process could contribute to further increases in the MJO prediction skill of the GEOS-S2S system.