This study seeks to answer the questions: i) Do opposite phases of the main interdecadal modes of seasonal precipitation produce significant anomalies in the frequency of extreme events? ii) Does the interdecadal variability of the frequency of extreme events show similar spatial and temporal structure as the interdecadal variability of the seasonal precipitation? iii) Does the interdecadal variability change the daily precipitation probability distribution between opposite phases? iv) In this case, which ranges of daily precipitation are most affected? The analysis is carried out for the monsoon season (austral spring, SON, and austral summer, DJF), since it is the rainy season in most of the continent.
Besides this analysis, this study extends previous continental-scale analysis for South America in one decade. As the data spatial distribution and the period are not the same, it is an opportunity to assess the robustness of the disclosed modes of interdecadal variability, and investigate the associated anomalous oceanic and atmospheric fields.
Observed daily and monthly precipitation data from more than 10,000 stations in the 60 years period 1950-2009 are used in this analysis, after verification procedures. Series of seasonally averaged monthly precipitation and numbers of extreme events (above 90th percentile) are gridded respectively for 2.5° and 1.0° over South America. EOF analysis with varimax rotation is carried out. The first modes of precipitation in spring and in summer are both dipole-like, displaying opposite anomalies in Central-east and Southeast South America. They are significantly correlated and tend to reverse polarity from spring to summer, although the associated SST anomalies are similar. These associated SST anomalies are related with the Atlantic Mudidecadal Oscillation (AMO) and also with the Interdecadal Pacific Oscillation (IPO). In spring, this mode presents a center of large-scale divergent wind anomalies over South America that can influence directly the monsoon circulation at the beginning of the monsoon season, which is consistent with the fact that it affects most of the monsoon region. It also features anomalous Rossby wave from the central tropical Pacific into the extratropics, able to affect southeastern South America.
Contrary to the spring first mode, the divergent and rotational anomalies associated with the summer first mode do not display large-scale relevant patterns over South America, and no remote influences are evident. This reinforces the hypothesis that precipitation anomalies in spring over Central-East Brazil produce soil moisture anomalies able to trigger surface-atmosphere interactions that produce local circulation anomalies leading to the inversion of anomalies in summer.
The second modes for spring and summer present associated SST anomalies related to the IPO, but the anomalous global SST fields are not similar. While the second spring mode is associated with SST anomalies of the same sign in the tropical central-eastern Pacific and North and South Atlantic, the SST anomalies connected to the second summer mode in the eastern tropical Pacific and South Atlantic are of opposite sign to those in North Atlantic. The effects are different on the divergent and rotational circulation anomalies, leading to different precipitation anomalies in these modes.
The anomaly composites associated with each mode disclose significant changes in the frequency of extreme precipitation events, especially in regions with high factor loadings, indicating that the extreme events also exhibit interdecadal variability. Unfortunately, it is not possible to ascertain if this variability has characteristics similar to those of the seasonal precipitation, since the monthly and daily precipitation do not have the same spatial coverage during the period of analysis. Notwithstanding, the EOF analysis of extreme events frequency produces similar modes when the coverage is similar in the regions with highest factor loadings, and the composite analysis shows very similar atmospheric and oceanic anomaly fields. Therefore, the mechanisms that produce interdecadal variability of seasonal rainfall seem to be similar to those that produce interdecadal variability of extreme events frequency. This similarity seems to be greater for spring than for summer, which is consistent with the fact that local mechanisms can be more important than remote influences in summer.
The Kolmogorov-Smirnov test is applied to the daily precipitation series for positive and negative phases of the interdecadal modes in selected regions and shows, with very stringent significant level (better than 0.01), that the daily precipitation from opposite phases of interdecadal oscillations, in regions much affected by them, pertain to different probability distributions. Further analyses disclose that there is more sensitivity to the interdecadal oscillations in the extreme ranges of daily rainfall than in the ranges of moderate and light rainfall. In spring the effect on extreme events seems linear, with opposite impacts of similar magnitudes on the frequency of extreme events in the positive and negative phases of the interdecadal oscillations. In summer, although the effects are also opposite in opposite phases, the magnitudes are more frequently different, perhaps due to the stronger influence of local factors, which can produce non-linearities.