Why is the Skill of the Models in Reproducing MJO and its Impacts on the South American Monsoon Important for Subseasonal Prediction?

The Madden Julian Oscillation (MJO) influences significantly daily precipitation and the frequency of extreme events during the summer South American monsoon (SAM) in important regions of the continent. One of the main features of the SAM, the South Atlantic Convergence Zone (SACZ), extends from central South America over Southeast Brazil and into the subtropical Atlantic Ocean, affecting very densely populated areas in Southeast Brazil. During the austral summer this region is strongly affected by landslides and floods associated with active SACZ, and the extreme precipitation events receive contribution from synoptic and MJO-related intraseasonal variability. Therefore, it is important to assess the observed impacts of the MJO in its different phases and to evaluate the models' skill in reproducing these phases and their impacts on South America in order to explore extended-range predictability of those events. The MJO cycle is divided into 8 phases according to the temporal evolution of the first two observed modes of multivariate EOF analysis of tropical convection and zonal winds. The teleconnections associated with these impacts are analyzed with simulations and influence functions of a simple model. The results show that two of the MJO phases strongly enhance the extreme events in the SACZ region and indicate the responsible mechanisms, lending these events a higher degree of predictability on subseasonal time-scales. Therefore, in selecting models to build a subseasonal-range forecasting scheme for extreme precipitation events, a necessary step is the assessment of their skill in reproducing MJO and its observed impacts on South America. Well-known models of the S2S Project, among them the ECMWF and CFS-v2 models are analyzed. Their reforecasts for weeks 1, 2, 3, 4 are separately projected onto the first two modes of tropical convection and zonal wind variability in order to identify the predicted MJO phases. Although the skill of one of the models in predicting these phases extends to week 4, generally the useful skill does not extend beyond week 3. The simulation of the impacts over South America, especially on the SACZ, is also assessed for selected models.