Influence of the Madden Julian Oscillation on precipitation and surface air temperature in South America

Abstract

The regional influence of the Madden–Julian oscillation (MJO) on South America is described. Maps of probability of weekly-averaged rainfall exceeding the upper tercile were computed for all seasons and related statistically with the phase of the MJO as characterized by the Wheeler–Hendon real-time multivariate MJO (RMM) index and with the OLR MJO Index. The accompanying surface air temperature and circulation anomalies were also calculated. The influence of the MJO on regional scales along with their marked seasonal variations was documented. During December–February when the South American monsoon system is active, chances of enhanced rainfall are observed in southeastern South America (SESA) region mainly during RMM phases 3 and 4, accompanied by cold anomalies in the extratropics, while enhanced rainfall in the South Atlantic Convergence Zone (SACZ) region is observed in phases 8 and 1. The SESA (SACZ) signal is characterized by upper-level convergence (divergence) over tropical South America and a cyclonic (anticyclonic) anomaly near the southern tip of the continent. Impacts during March–May are similar, but attenuated in the extratropics. Conversely, in June–November, reduced rainfall and cold anomalies are observed near the coast of the SACZ region during phases 4 and 5, favored by upper-level convergence over tropical South America and an anticyclonic anomaly over southern South America. In September–November, enhanced rainfall and upper-level divergence are observed in the SACZ region during phases 7 and 8. These signals are generated primarily through the propagation of Rossby wave energy generated in the region of anomalous heating associated with the MJO.

Appendix

The phase of the OMI was made directly comparable to RMM Index following Kiladis et al. (2014), and the calculations in this work were then repeated using that index. As the RMM Index is constructed using OLR and circulation, it was expected for the RMM composites to show more signal overall, particularly in the geopotential height anomalies composites. As an example, Figs. 15 and 16 show the large-scale anomalies of geopotential height at 250 hPa and velocity potential at 0.21 sigma-level and for MJO phases 1–4, which, with the opposite sign, are to be interpreted as MJO phases 5–8, for DJF and SON respectively.

Fig. 15

As in Fig. 1 but using the OMI to generate the composite

Fig. 16

As in Fig. 4 but using the OMI to generate the composite

The OMI-based (Fig. 15) and the RMM index-based (Fig. 1) composites for the DJF season are remarkably similar. The extent of the statistically significant signals and the amplitude of the anomalies are generally reduced, although the locations of the centers agree well. The OMI-based composites for MAM and JJA (not shown) also resulted comparable to the RMM index-based, though some centers over the southern Indian Ocean presented larger amplitude when using the OMI during MAM. During JJA, significance spans slightly larger regions using RMM index, and is particularly strong to the southwest of South America on phases 1 and 2 (and 5 and 6, Fig. 3), which was not the case when using OMI.

Table 2 of Kiladis et al. (2014) shows that the OMI leads the RMM Index by 4 days in SON, being the greatest lag of the four seasons. It was therefore expected that we would find the greatest differences in the composites during SON. Figure 16 shows that phases 2, 3 and 4 of the OMI-based composites may be associated to phases 1, 2 and 3 of the RMM Index-based composites respectively (Fig. 4), consistent with the lag between the indexes. The OMI-based composites (Fig. 16) did not show a significant signal to the southwest of South America, and significance and amplitude of the centers were reduced compared to RMM index-based composites (Fig. 4).

Figures 17 and 18 show the rainfall probability composites for DJF and SON respectively, analogous to Figs. 11 and 14 but using the OMI as opposed to the RMM index. Again, during DJF both composites results are quite alike (Figs. 11, 17), although the OMI-based probability composites seem to have significant signal over more regions of South America, with a few exceptions, as eastern Brazil in MJO phase 8 and during neutral MJO cases, when OMI-based composites do not show a coherent signal (Fig. 17). During MAM, the OMI-based composites (not shown) are overall similar to RMM index-based ones, however, the most remarkable discrepancies are that the former showed a stronger signal with ratios lower than 1.0 over SESA (eastern Argentina) on MJO phase 1 (3) and greater than 1.0 ratios over SESA and west of the South Brazilian Bight on MJO phase 8. During JJA, OMI-based composites are seen to have an overall larger signal (not shown), though these signals are, as in RMM-based composites, less coherent than in the other seasons.

Fig. 17

DJF season composites of weekly rainfall probabilities for OMI-defined MJO phases 1–8 and neutral MJO cases. The probabilities refer to the chance of weekly averaged rainfall exceeding the upper tercile, expressed as a ratio with the mean probability (nominally 33 %) for every grid point. Only 10 % significant values are shown and those gridpoints with missing values were masked out in grey

Fig. 18

As in Fig. 17 but for SON

As previously seen with the circulation composites, during SON, OMI-based rainfall probability composites (Fig. 18) lead RMM index-based ones (Fig. 14) by one MJO-phase, that is, OMI-based composite on MJO phase 1 resembles the RMM index-based composite during MJO phase 8, and so on. The most remarkable differences when comparing both composites were observed over eastern Argentina and SESA, though the general pattern resulted similar.

The OMI-based temperature composites (not shown) presented overall less regions with statistically significant signal. The regional pattern observed in DJF was quite similar to the RMM index-based composites (Fig. 7), the latter presenting higher values over northern Patagonia. During MAM, RMM index-based composites showed significance over SESA and slightly higher anomalies over the continent compared to the OMI-based composites, while during JJA the extratropical significant signal using OMI was considerably reduced. Finally, during SON, the lag observed in the comparison with the other variables was also present in temperature composites. In particular, the signal over northwestern Argentina and in the South Atlantic Ocean was noticeably reduced when the OMI was used. The fact that OMI signals are comparable to those using RMM is reassuring, although it is perhaps not surprising that in some cases there is less significance to the OMI results, since a significant portion of the RMM signal is likely related to circulation anomalies over the South America sector (Straub 2013; Kiladis et al. 2014).