South American Monsoon: Variability of Extreme Events and Subseasonal Prediction

The South American monsoon (SAM) during the warm season is responsible for large fraction of the annual precipitation over most of South America (SA). The monsoon region includes the most populated areas and those that have the largest contribution to agricultural production and hydroeletric power generation. In the core of the monsoon region it rains above 10 times more in austral summer than in winter. Accordingly, most of the precipitation extremes with great social and economic consequences occur during the monsoon season. One of the worst natural disasters in the continent, associated with heavy rainfall and widespread landslides, happened on January 2011 in the Rio de Janeiro state, associated with active South Atlantic Convergence Zone (SACZ) over southeastern Brazil, when more than 900 people perished and other 35,000 were displaced.
This presentation seeks to provide 1) an overview of the way climate variability affects the frequency of extreme events, 2) an assessment of the Madden-Julian Oscillation (MJO) influence on the active and break periods of the SAM and of the skill of models of the S2S Project in simulating this influence and predicting these periods, and 3) an example of how prediction of extreme rainfall can be improved and the leading time extended with calibration of the model output.
Local and remote influences contribute to produce rich variability of the SAM, from both temporal and spatial points of view. Climate variability in different time scales (intraseasonal, interannual and interdecadal) affects and modulates the synoptic and mesoscale features responsible for extreme events, influencing the frequency of extreme events during the monsoon season. The climate variability impact is often stronger in the extreme ranges of daily precipitation, which is an important aspect, since its most dramatic consequences are due to extreme events.
The skillful prediction of active and break periods of the monsoon with lead time beyond one week has great economic and social importance. To diagnose these periods a monsoon index is defined as the standardized daily rainfall anomaly over (10º-20º S, 45º-55º W), within the core monsoon region. This index is equal or greater (equal or less) than +1 (-1) for active (break) monsoon days. Composite anomalies for active (break) monsoon rainfall episodes show that they are associated with large-scale cyclonic (anticyclonic ) low-level wind anomalies over central SA and tropical westerlies (easterlies), for which the MJO makes a significant contribution and is therefore a source of predictability. Thus, it is important that the models selected for subseasonal forecast of the SA monsoon variability are able to simulate and forecast the MJO and its impacts over SA. Among the models participating in the S2S Project one shows skill in predicting the right phase of the MJO for lead times beyond 3 weeks, while most of them are limited to 2-3 weeks in advance. The skill of the S2S models in reproducing the proportion of SAM active (or break) days in each MJO phase is assessed, and results are shown for two selected models (NCEP-CFSv2 and ECMWF). The most skillful models represent well the observed modulation by the MJO up to week 3. These models show a good skill for predicting the MJO related precipitation anomalies over SA and associated teleconnections, but tend to shift a little the phase in which they occur within the MJO cycle. The correlation coefficient between the predicted and observed monsoon indexes shows small forecast skill after week 2, although better results can be achieved with a wind based monsoon index.
Improved prediction with longer lead time of heavy rainfall during the SAM is possible by applying calibration procedures on model output. An example is shown for the case of January 2011 mentioned above. When the precipitation rate hindcasts from the NCEP CFSv2 are submitted to a two-step calibration, they were able to depict both active and break phases of the South Atlantic Convergence Zone with up to two weeks in advance.