a. Data

The precipitation data used in this study are monthly amounts of 134 selected stations from Argentina, Brazil, Chile, Uruguay, and Paraguay, for the period 1956–92. The database for the circulation analysis is the National Centers for Environmental Prediction–National Center for Atmospheric Research reanalysis for the period 1963–92, because many of the raob series in the region started after 1963.

The years of EN and LN episodes included in this study were chosen following the same criteria as in Grimm et al. (1998) and indicate the beginning of the episodes (Table 1). In the case of two consecutive LN years, only the first year is used in the composites as a starting year of the episode (i.e., 1970).

b. Climatic aspects

North of 40°S, the near-surface circulation over SSA is dominated by the quasi-stationary highs of the South Pacific (west of the Andes) and the South Atlantic (east of the Andes). Most of the time, there is a low pressure center in the central–western part of Argentina and southern part of Bolivia, known as west low (WL), originated by the interaction of the Andes, upper-level westerlies, and surface heating (Lichtenstein 1982). This WL deepens before the passage of cold fronts and disappears one or two days afterward. It is a warm system, less intense in winter, affecting only the lower troposphere (up to 700 hPa) and frequently accompanied by subsidence. South of 40°S the intense and persistent westerly low-level flow is perturbed by midlatitude systems that occasionally bring easterly flow into the continent. Polar front eruptions are less frequent in summer, when the fronts remain more to the south.

The annual cycle of precipitation in central Chile, in the Andes region, and immediately to the east of the higher mountains in northern Patagonia is conditioned by the seasonal south–north displacement of the South Pacific high, with a maximum in winter and a minimum in summer (Fig. 1).

East of the Andes, there are only two possible sources of water vapor in SSA: the Atlantic Ocean and the South American tropical forest (Wang and Paegle 1996). This has an important effect on the annual cycle of precipitation. In winter, there is little chance of water vapor supply from the Atlantic, since the westerly flow reaches lower latitudes. On the other hand, relatively dry conditions prevail in the tropical forest. Thus, except on the Atlantic coastal region, the winter is dry. In the western part of the subtropical region a pronounced annual cycle with a maximum in summer prevails, when both the surface heating and the northerly water vapor advection favor the convection. In the east, the water vapor is available throughout the year, but the stronger baroclinic conditions of winter and spring favor a relative maximum in this season. Actually, all over the eastern part of the region, precipitation occurs all through the year and in fact the interannual variance is greater than that associated with the annual cycle (Gonzalez and Barros 1997).

Southern Brazil is a region of transition between two adjacent regimes: summer monsoon and midlatitude winter conditions, which are responsible for the peak rainy seasons in January (in the northern part) and July (in the southeastern part), respectively. There is also a region, in west southern Brazil, where the bimodal variation prevails, with wet seasons occurring in March–April and September–October. These features are also noticed in Paraguay and eastern Argentina and may be due to the seasonal change of the upper-tropospheric subtropical jet, which may intensify the mesoscale convective complexes in this region (Velasco and Fritsch 1987).

a. Harmonic analysis and coherent regions

Figures 2 and 3 show the vectors representing the first harmonic fitted to the composite percentile precipitation at each station. In Fig. 2 the vectors represent the maximum of this harmonic for EN events, whereas in Fig. 3 they represent its minimum for LN events. Coherent regions are characterized by the similarity of the phase angle of the vectors. Thus, some parts of a region may present smaller magnitude of the vectors than other ones, and the mean vector of the region may be diminished, in spite of the large signal in some areas.

Figure 2 suggests the existence of eight regions of coherent behavior as regards precipitation anomalies during EN events. Their description is summarized in Table 2. The vectors present a very ample range of directions, reflecting the shift of the maximum of the first harmonic from the winter of year (0) to the winter of year (+), with most frequent occurrence during the spring–summer of the year (0). That is why RH87 found a low coherence for the region they determined in SSA. It is worth pointing out that the maximum of the first harmonic does not always indicate the actual time of the maximum anomaly. This is seen in the aggregate composites for each region.

The vectors map confirms the conclusion of Grimm et al. (1998) that in SSA the area of largest impact of EN events on precipitation is southern Brazil. The amplitude of the vectorial mean is even larger in the present study, probably due to the fact that the data series are shorter (1956–92). This is consistent with the fact that the impact of EN was smaller before this period in that region, as can be seen in Fig. 9 of Grimm et al. (1998).

It should be stressed that the results for regions 4 and 7 should be viewed with caution, because the composites for these regions comprise less than five EN for some months in the period July (−) to June (+), due to missing data in the series. Region 8, comprising the extreme northwest of Southern Brazil and east of Paraguay, embraces few stations that may not be very representative of this region and therefore has a small index of coherence and a relatively small magnitude of the mean vector.

The vectors representing the minimum of the first harmonic fitted to the LN composites (Fig. 3) present a much more coherent behavior than the EN vectors. Therefore, the characteristics of the regions (Table 3) are more alike than for EN, at least as they refer to the phases. This coherence is not reproduced all over southern Brazil, but only in its southernmost part, although there are strong anomalies during the spring of LN years all over southern Brazil, as shown by Grimm et al. (1998). The minimum of the first harmonic occurs only in the range August (0) to December (0), with the exception of region 6. This region actually presents a maximum associated with LN events, and so the minimum on September (−) actually corresponds to a maximum of the first harmonic on March (+).

The EN and LN regions are approximately similar, especially region 5 (Figs. 2 and 3). The timing of the precipitation responses in these regions is roughly in opposite phase during the EN and LN events. As can be seen in Tables 2 and 3, the maximum of the first harmonic in the EN case differs by at most one month from the LN minimum. This means that in those regions, the precipitation anomalies generally have opposite signals along the EN and LN cycles, indicating a large degree of linearity in the response to these events.

The EN “aggregate” composites (average gamma percentiles) for each coherent region in Fig. 2 are shown in Fig. 4. In spite of the differences in the magnitude and timing of the anomalies between regions, it is possible to summarize a general behavior. In the year before EN there is a tendency to less than normal precipitation from April (−) to March (0) [with a discontinuity in September (−)], with the exception of region 6, where it rains more than normal during July (−) to November (−). There are positive rainfall anomalies during the spring of the warm event year [August (0) to November (0)] over all regions, except regions 6 and 7. There is a weakening (and even a sign reversal in some regions) of the anomalies in January (+), which begins in December over the coastal regions of southern Brazil. The behavior is not so coherent from April (0) to July (0) and during the year (+). While it rains more than normal in southern Brazil during the winter of the year (+), there are dry or not consistent wet anomalies in other regions.

The LN aggregate composites for the six coherent regions in Fig. 3 are shown in Fig. 5. There is a common period of negative anomalies during the spring of the year (0) in regions 1 and 2. For regions 3, 4, and 5 the negative anomalies occur mostly in winter of this year. In region 6 there is a tendency to above-normal precipitation in the beginning and winter of year (+). Generally, there is above-normal precipitation during the year before LN events and below-normal precipitation during the LN year, with the exception of region 6. The behavior in the year following LN is not so clear. A remarkable feature is the reversal of the sign of the anomalies during January (+), particularly over regions 1 and 2. This behavior is symmetric to the one observed during EN events.

In order to verify whether the average anomalies during the EN and LN events in each region occur consistently during these events, the analysis described in step 6 of section 3 is systematically applied to three-month running seasons. The results are summarized in Tables 4 and 5. The periods with the largest average anomalies in Figs. 4 and 5 are not always those with the most consistent ones. This becomes clear by comparing the levels of significance (Table 4) and average gamma percentiles (Fig. 4) for the EN wet periods September (0) to November (0) and October (0) to December (0) for region 1. This may be explained by the influence of outliers on some average anomalies.

Table 4 confirms clearly the general tendency to dryness in the year preceding EN, after April (−), except in region 6. Although half of the EN episodes of Table 1 were preceded by LN events, the significant dry anomalies in year (−) of EN and year (0) of LN do not coincide exactly. There is also a general tendency to wetness during EN year (with the exception of region 6), with higher consistency in region 1. The anomalies in the year (+) have less common aspects.

Table 5 shows a significant tendency to dryness during LN years in different regions (with the exception of region 6), with highest consistency in regions 1 and 2. In the year before LN there are significant wet anomalies, but not as spread out as the dry anomalies in the year before EN. There are few consistent anomalies in year (+).

There are other longer periods during EN and LN cycles that show a high level of significance for the consistency of the associated precipitation anomalies. For example, the period April (–) to March (0) is consistently dry in EN events in regions 1, 2, and 3, with levels of significance of 99.3, 92.6, and 95.3, respectively. The period September (0) to December (0) is consistently wet in the same regions with levels of significance of 99.8, 99.8, and 95.3.

The results obtained here are in good agreement with previous regional studies, which used a denser network of data. Southern Brazil (our region 1) presents quite similar distribution of anomalies along the EN events to that of region 1 of Grimm et al. (1998). Northwestern Uruguay, which is here included in region 1, presents strong positive anomalies during spring of the EN year and also consistent positive anomalies in the autumn–winter of the following year. Yet southeastern Uruguay, here included in region 3, has weaker positive anomalies during the spring of EN year and no consistent anomalies during autumn–winter of the following year. This general behavior agrees with the results of Pisciottano et al. (1994).

The positive significant rainfall anomalies over most of SSA during the spring of EN years and over Chile during winter and spring, shown in this study, are visible in Aceituno (1988). However, the significant positive anomalies in southern Brazil during the winter of the following year appear as a significant correlation with the SOI neither in Aceituno (1988), nor in Rao and Hada (1990), perhaps because the anomalies of this index are not consistent and strong at this phase of the EN events.

The small differences between the present results and those of RH87 and RH89 are mainly due to the fact that their coherent region in SSA comprised parts of regions 1, 2, and 3.

b. Joint seasonal analysis of precipitation and circulation

The precipitation anomalies associated with EN and LN events may be caused by changes in the water vapor availability, in the dynamics leading to vertical movements, and in the vertical stability of the air, or due to a combination of these factors. In general, the humid (dry) anomalies result from the enhancement (weakening) of already existing climatic circulation features that favor (oppose) the precipitation during a given season in specific regions. In SSA the main processes leading to EN–LN-related precipitation anomalies include

1. changes in the subtropical jet and in vorticity advection,

2. changes in the northerly advection of humidity,

3. establishment of equivalent barotropic anomalies over and near Chile, and

4. changes in the advection of humidity from the Atlantic Ocean.

For the joint analysis of precipitation and circulation anomalies, composites of low-level circulation anomalies, to identify changes of flow from moisture sources, and of upper-level anomalies, to indicate changes in the dynamic lift, are provided, with an assessment of consistency. In addition, a more detailed analysis of rainfall is presented.

The results of the detailed analysis of precipitation for each station are shown in Figs. 6 and 7. They present maps of the mean precipitation percentiles expected for every other running trimester in EN and LN events, centered on May (0) to July (+). This is a useful complement to the analysis already discussed, because it allows a greater spatial and temporal refinement. On these maps, the shadowed areas contain stations with consistently wet or dry anomalies during the indicated trimester, with a level of significance better than 90%. In regions with few stations, these shadowed areas may appear fragmented. Figures 8 and 9 show the geopotential height anomalies at 200 and 850 hPa for the same periods.

Figure 6 shows that during autumn of EN years [April (0) to June (0)], there is a broad region in subtropical Argentina and Paraguay, covering region 7 and most of regions 2 and 3, where precipitation is above the median. Also over Patagonia there are positive anomalies. Yet during winter [June (0) to August (0)] the subtropical areas have near-normal precipitation, while the positive anomalies are shifted southward in Argentina. Positive signals appear in central Chile. From late winter [August (0) to October (0)] to spring [October (0) to December (0)] the positive anomalies over central Chile and Patagonia strengthen, and, also, strong and consistent wet anomalies spread over southern Brazil and northeastern Argentina. In spring (0) there is higher than median precipitation over most of SSA, with the exception of northwest Argentina and northern Chile.

Figure 8 shows that from autumn to spring (0) of EN events the 200-hPa anomaly pattern comprises cyclonic anomalies over the South Pacific high, anticyclonic anomalies to the south, anticyclonic anomalies in the subtropical Atlantic, and cyclonic anomalies to the south, southeast of SSA. This pattern is nearly equivalent barotropic and dipole-like both in the Pacific and in the Atlantic, but with reversed polarity. Its action centers undergo some changes of magnitude and position during the EN cycle. It is stronger and more stable in the Pacific. This pattern strengthens the westerlies over the eastern Pacific in the subtropics and over the western Atlantic in the midlatitudes, causing its anomalous cyclonic curvature just west of SSA. In late winter (0), it moves to the north and to the east.

In autumn–winter (0), when the dipole pattern is more to the south, the westerlies are intensified over the southern part of SSA, south of 40°S, thus favoring precipitation over this region and the shifting to the south of migratory systems. This leads to the wet anomalies over Patagonia and part of Chile. Cold fronts are less frequent to the north and the WL can stay longer at each of its deeper phases, consistent with the results of Barros and Scasso (1994). In fact, the 850-hPa maps show a tendency to negative height anomaly over northwestern Argentina and a positive anomaly over the region of the subtropical high in the Atlantic. This pattern enhances the zonal height gradient to the east of the WL and thus the warm and sometimes humid northerly advection from the tropical forest. The April (0) to June (0) season of the EN event is an example of how this enhancement produces anomalous precipitation in Paraguay, subtropical Argentina, and parts of southern Brazil and Uruguay. Although the conditions at 200 hPa are not very favorable, with an anomalous ridge along this region and a decrease of baroclinicity to its north, there is significant above-median precipitation all over eastern Argentina and Paraguay (see Fig. 6). Though the WL enhancement is intense during the winter (0), the relatively dry conditions prevailing to the north of the region, in central Brazil, during July (0) to August (0), do not allow the advection of much additional water vapor to produce rainfall anomalies. Besides, in winter (0) the ridge in the troposphere over the subtropical region is enhanced and the positive rainfall anomalies are shifted to the south.

In late winter–spring (0), the dipole patterns move to the north and to the east, approaching the South American continent. Thus, there is anticyclonic circulation southwest of the southern tip of South America and cyclonic circulation north of 40°S. A trough axis extends southeast–northwest through southern Argentina and central Chile. Consistent with the weakening of the South Pacific high along the Chilean coast and the anticyclonic anomaly to the south, there is diversion of the migratory systems toward the latitudes of central Chile, as was described by Rutllant and Fuenzalida (1991) for case studies. This situation produces higher than median precipitation over central Chile and in most of region 3 in Argentina. At this time the northerly flux of moisture into eastern Argentina, Uruguay, and southern Brazil is again enhanced. In spring (0), when the dipoles are in their northernmost position, there is an enhancement of the subtropical jet over SSA, and a clear cyclonic vorticity advection over northeastern Argentina, southern Brazil, and Uruguay, producing strong rainfall impact on these regions, as was described by Grimm et al. (1998). This process acts jointly with the enhancement of northerly advection of moisture. At this time of the year, the baroclinicity still remains at low latitudes and its enhancement in this region favors the already frequent developments of mesoscale systems and cyclogenesis on eastern subtropical SSA (Necco 1982; Velasco and Fritsch 1987).

Figure 6 shows that in summer [December (0) to February (+)] the wet anomalies move westward to northern Argentina and Paraguay. There are no consistent wet anomalies over central Chile. Figure 8 shows that in that season the dipole patterns also move westward and weaken, and so do the cyclonic anomalies over Chile. The rainfall anomalies weaken and tend to reverse their sign during January (+), particularly in southern Brazil, Uruguay, and eastern Argentina (regions 1 and 3) as was already reported by Grimm et al. (1998) for southern Brazil (Fig. 4). Monthly analysis (not shown) indicates that also the polarity of the dipoles tend to reverse in January (+). This reduces the northerly advection of moisture, which is important for rainfall in these regions during summer. However, during summer there is another process able to impact the rainfall over Uruguay and eastern Argentina: the enhancement (weakening) of the advection from the Atlantic Ocean. During EN events, the easterly component of the near-surface flow is enhanced over Uruguay and central Argentina due to the southward shift of the South Atlantic high. In January (+) it may have a compensating effect in relation to the decrease of the northerly advection, keeping rainfall near normal in these regions.

In late summer (+) [February (+) to April (+)] the rainfall anomalies strengthen again to the east, in parts of regions 1 and 3, and in Chile, but not to the magnitude of spring (0), as circulation anomalies undergo a brief return to conditions similar to spring (0), although displaced to the south. Also the northerly advection of moisture is enhanced into these regions.

In autumn (+) [April (+) to June (+)], normal conditions prevail over most of SSA. The anomalous northerly advection decreases, the cyclonic anomaly is centered on the southern tip of South America, and the anticyclonic anomaly southeast of SSA weakens.

In winter (+) [June (+) to August (+)] the cyclonic anomaly displaces northward and significant wet anomalies reappear over southern Brazil, due to the increased subtropical jet and advection of cyclonic vorticity into this region. Dry conditions prevail over Argentina.

Figure 7 shows the maps for the LN case. In general, during year (0) the precipitation anomalies evolve in the same way as in the EN case, but with opposite sign, although there are some shifts in space and time. For instance, the dry anomalies over Chile start earlier in winter (0) and are not so strong in spring (0). The correspondence in spring (0) is very good in southern Brazil, Uruguay, and eastern Argentina. In summer [December (0) to February (+)], also in symmetry with EN events, the negative anomalies weaken over southern Brazil and shift westward over Argentina. In January (+) there is a break with respect to the preceding months, and wet anomalies appear over part of southern Brazil and Paraguay (regions 1 and 2) (Fig. 5). As in the EN episodes, after the January break, the spring (0) anomalies tend to return in late summer (+), particularly in the southernmost part of southern Brazil, Uruguay, and east Argentina. This pattern tends to weaken and become noisier during autumn (+), with more temporal and spatial variability. In winter (+) dry conditions return to the southernmost part of southern Brazil, Uruguay, and northeastern Argentina.

As for precipitation, the symmetry of the geopotential height anomaly fields with opposite signs between LN and EN cases is also remarkable during year (0), as can be seen from the comparison of Figs. 8 and 9. Thus, from autumn (0) to spring (0) of LN events there is a decreased westerly flow over the eastern Pacific in the subtropics and over the western Atlantic in the midlatitudes. An anticyclonic anomalous curvature predominates west of SSA. From winter (0) to spring (0), the dipole patterns move to the north and to the east. The symmetry is weaker in the Atlantic sector. There are some shifts with respect to the EN opposite anomalies, with a tendency to be at lower latitudes in winter and at higher in spring and summer. Zonal shifts are generally toward the west. From February (+) on, there is less symmetry with respect to the EN cycle, perhaps due to the shorter duration of the LN events.

The composites of geopotential height anomalies at 200 hPa for June (0) to August (0) and December (0) to February (+) of EN events are in reasonable agreement with those of Karoly (1989) near SSA, because they contain the same basic features. This is an indication of the stability of these composites, for Karoly used three of the nine events used here (1972, 1976, and 1982) and included the year 1977, considered a year (+) in this study.

There are some differences between the circulation features disclosed here and in Aceituno (1989). They may be due to differences in methodology and in datasets. For example, in the present study negative height anomalies prevail over SSA in spring (0) of EN events, along with a strengthened subtropical jet. In Aceituno (1989) the negative SO phase (generally associated with EN) during spring is associated with positive height anomalies and a decrease of the subtropical westerlies, as is inferred from his Figs. 3f and 6f.