The midsummer interannual variability of the low-level tropospheric circulation and of the precipitation field in subtropical South America (SA) associated to the sea surface temperature (SST) anomalies in the western subtropical South Atlantic Ocean (WSSA) is investigated using reanalyses, regional precipitation datasets, and monthly SSTs.
The region of the WSSA where SST has the strongest relation with precipitation in subtropical SA was identified using canonical correlation analysis. This region extends from 20° to 30°S and from 30° to 50°W. Composites corresponding to extreme SSTs in this area show two well-differentiated patterns in the low-level circulation and in the precipitation fields. In the composite, corresponding to the more positive SST anomalies in this area, the mainstream of the low-level flow and of the moisture transport from the continental low latitudes starts to follow a southeastward direction at 10°S, and converges with the west flow at 35°S over the Atlantic Ocean. On the other hand, in the composite corresponding to the more negative SST anomalies, the low-level flow and the moisture transport from the continental low latitudes turn eastward toward the South Atlantic convergence zone (SACZ) at about 20°S, converging with the flow from the north driven by the South Atlantic high. In this composite, there is an anticyclonic circulation with a westward flow between 25° and 35°S, which turns southward after reaching the proximity of the Andes Mountains.
In the composite of the more positive anomalies, there are two regional maximums in the precipitation field. One maximum stretches along the continental extension of a southwardly displaced SACZ and another is centered at about 30°S and 55°W, in the path of the main stream of the low-level moisture transport. In the other composite, there is only one regional maximum in precipitation, which coincides with the continental extension of the SACZ shifted northward of its mean position, and with a relative minimum in northeastern Argentina and southern Brazil. In this composite, in western Argentina, there are positive anomalies in the precipitation field favored by the transport of moisture from the Atlantic Ocean.
The low-level patterns of the monthly composites, corresponding to the more positive and more negative SST anomalies in the WSSA, are similar to the respective patterns of each of the phases of the active and inactive SACZ. This follows from the prevalence, according to the SST in the WSSA, of one or the other of the low-level patterns associated to the seesaw of the SACZ. A positive feedback between positive (negative) SST anomalies and weak (intense) SACZ activity might enhance the low-level circulation pattern associated with the SACZ seesaw.
A monsoon circulation is usually characterized by the seasonal reversal in the lower-tropospheric large-scale circulation. Although this is not the case in South America (SA), the intensification of tropical convection during the summer and the dry winter are characteristic of a monsoon system. The large-scale convection enters the Southern Hemisphere in the austral spring, and it moves southeastward through tropical SA, initiating the rainy season in a large area of Brazil (Kousky 1999). The rainy season persists until the southern autumn, when convection shifts toward the Northern Hemisphere.
Virji (1981) described the main features of the summer mean tropospheric circulation over SA using satellite data. These features have been frequently documented since then (Zhou and Lau 1998). At low latitudes, the low-level flow from the trade winds enters the continent, and when reaching the proximity of the Andes turns to the south and it is channeled along the region between the Bolivian Plateau and the Planalto to subtropical SA. A low-level jet (LLJ) embedded in this flow enhances the heat and moisture advection toward subtropical latitudes (Wang and Paegle 1996). Thus, the mean low-level flow transports moisture from the tropical Atlantic toward the tropical continent, and from there toward the subtropical region east of the Andes. In addition, the circulation associated to the western part of the subtropical South Atlantic high also carries water vapor from the South Atlantic Ocean directly into subtropical SA.
During the austral summer, the main convective activity is over central Brazil and over an area that extends to the southeast and continues in the South Atlantic convergence zone (SACZ). In fact, the SACZ appears in mean monthly and seasonal maps as an elongation of the deep convection over central Brazil (Kodama 1992). Using a numerical model, Lenters and Cook (1995) showed that precipitation in the SACZ results from the convergence of the continental eastward branch of the low-level circulation from the contenental low latitudes and the maritime southwestward winds of the South Atlantic high circulation. Nogués-Paegle and Mo (1997) documented a seesaw pattern on the SACZ with an approximate duration of 8 days in each phase. They found that events with strong (weak) convective activity over the SACZ were associated with negative (positive) rainfall anomalies in the subtropical region to the south of the SACZ. A strong SACZ is likely associated with enhanced subsidence south of the SACZ as it was found in numerical experiments (Gandú and Silva Días 1998). In addition, the convective seesaw variability is accompanied by a change in the direction of the low-level circulation from the continental low latitudes, which flows eastward (southeastward) at about 20°S in the case of strong (weak) SACZ events, and by an eastward (westward) shift of the South Atlantic subtropical high (Nogués-Paegle and Mo 1997). In the upper troposphere, the main characteristics of the summer circulation are the stationary high over the Bolivian Plateau and the trough over northeastern Brazil (Chen et al. 1999).
Sea surface temperature (SST) in the western subtropical South Atlantic (WSSA) is correlated with the SACZ position and intensity (Barros et al. 2000); consequently, it is related to the precipitation over subtropical SA. In addition, precipitation in eastern subtropical SA between 25° and 35°S is also directly associated with the SST in the WSSA, independent of the SACZ intensity or position (Barros et al. 2000). Thus, summer precipitation in eastern SA has positive and significant correlation with SST in the WSSA, as reported by Doyle and Barros (2000) for northeastern Argentina, and by Díaz et al. (1998) for northern Uruguay and southern Brazil. In order to obtain an insight that goes beyond the statistical relation between SST in WSSA and rainfall in subtropical SA, this paper explores the link between the interannual variability of the low-level tropospheric circulation, precipitation in subtropical SA, and SST in the WSSA.
The paper is structured in seven sections. Section 2 describes tropospheric, precipitation, and SST datasets, and section 3 presents a brief description of the mean precipitation and moisture transport throughout the year. The interannual relationship between SST in the subtropical South Atlantic and precipitation in eastern subtropical SA during midsummer is discussed in section 4. The region of the WSSA where SST anomalies are more related with the interannual variability of rainfall in subtropical SA during January is identified in this section. Section 5 shows the composites of the low-level circulation fields corresponding to extreme SSTs in this region, as well as the respective composites of vertical velocity and of the upper-tropospheric circulation. Section 6 contains an integrated discussion of SST variability in the WSSA, and of the low-level atmospheric circulation and precipitation over eastern subtropical SA. Finally, section 7 summarizes the main results.
Tropospheric variables and skin surface temperatures were taken from the global National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalysis (Kalnay et al. 1996) available on a 2.5° × 2.5° latitude–longitude grid. The reanalyses corresponding to the 1979–92 period made an incorrect assimilation of the Australian sea level pressure estimates in the southern oceans. As indicated in an assessment by NCEP–NCAR, this mistake had a minor overall impact on monthly and longer-term timescales (more information available online at http://wesley.wwb.noaa.gov/paobs/). Since, in this paper, most of the analyses were done with monthly averages, this mistake would not affect the results, although as it was pointed out by Trenberth and Caron (2000), it may add some level of noise.
The water vapor transport was calculated using gridded values of the specific humidity (q), and of the zonal and meridional wind components every 6 h. Since most of the water vapor in South America is concentrated, as in many other regions, in the low levels of the troposphere (Berbery and Collini 2000), the water vapor transport was calculated integrating qV from surface to 700 hPa, using winds and specific humidity from the intermediate levels. These values were calculated every 6 h, and then averaged for every month. Mo and Higgins (1996) compared the moisture transport derived from NCEP–NCAR and the National Aeronautics and Space Administration' (NASA's) Data Assimilation Office reanalysis. They found a general agreement with some regional differences, but the moisture flux associated with the LLJ showed a large uncertainty. Doyle (2001) compared the integrated specific humidity below 700 hPa from NCEP–NCAR with data from NASA's Water Vapor Project (NVAP; Randel et al. 1996). NVAP has data of the precipitable water column integrated up to 700 hPa derived from satellite and radiosonde observations for the 1988–92 period. Over eastern subtropical South America, the monthly averages from NVAP and those calculated by Doyle (2001) differ in less than 2 mm, except in small areas and in the proximity of the Andes west of 65°W, where both datasets differ considerably. In January, the relative difference in most of the region is considerably less than 10%, except south of 30°S where it reaches 15% and west of 65°W. The correlation between daily maps of both sets over SA south of 10°S is higher than 0.8 in 98% of the days, and its mean is 0.87.
The use of independent datasets for precipitation and SST, different from the NCEP–NCAR reanalysis, reinforces the reliability of conclusions, as these are independent of the aforementioned reanalysis. Therefore, precipitation data have been taken from monthly records for Argentina, Brazil, Paraguay, and Uruguay. The Argentine National Meteorological Service and the National Direction of Meteorology of Paraguay provided the records for Argentina and Paraguay, respectively. Most of the series for Brazil were obtained from the Institute of Agricultural Research of Rio Grande do Sul and from the National Agency of Electricity. In addition, other records for Brazil and the totality of the records for Uruguay were taken from the NCAR's monthly climate data for the world. For the whole set of 107 monthly records, the missing data are less than 2%, and for each of the series, this ratio is less than 10%.
Atlantic SST data from 10° to 40°S and from the South American coast to 10°E were obtained from the Comprehensive Ocean–Atmosphere Data Set (COADS; Woodruff et al. 1987). Its initial resolution was of 2° latitude × 2° longitude, but because of missing information in many points, the data were averaged in a 10° × 10° latitude–longitude array. Even with this procedure, there were still a few missing data that were substituted with their monthly averages.
The NCEP–NCAR reanalyses use a frozen prediction model and data from a variety of sources including conventional surface and radiosonde observations and satellite data. They constitute one of the most complete and physically consistent existing datasets and have been extended back to the early 1940s. However, their quality is lower for years when data from satellites and radiosondes were not available. Since in South America radiosondes started in 1958 and satellite data were available only at a later date, the period chosen in this study starts in 1958. In addition, most of the institutions that provide precipitation data have not yet released the majority of the records following 1991. In some cases, this is due to the fact that the consistency analyses have not been completed. Therefore, because of the joint availability of adequate reanalyses and precipitation data, the period chosen for this study was 1958–91.
3. Mean precipitation and water vapor transport
The atmospheric water balance equation, vertically integrated from surface level to the top (Chen 1985) is
P and E are the precipitation and the evaporation rate, respectively, and W the precipitable water defined by
where g is the gravity acceleration, pt is the pressure at which q is practically zero, and ps is the surface pressure. The vertically integrated horizontal water vapor transport Q is defined as
where V is the horizontal wind vector.
If Eq. (1) is averaged over a long term, such as a month or more, the left side of the equation becomes negligible. Therefore, over a month or longer, the difference between precipitation and evaporation can be estimated approximately by the divergence of the horizontal water transport:
It is not the purpose of the present study to discuss the quantitative balance of water vapor in eastern subtropical SA. As discussed previously, the moisture flux associated with the LLJ, which is very important in the region, presents large discrepancies when it is calculated using global analyses from different centers (Wang and Paegle 1996). This situation persists even with the more recent reanalyses (Mo and Higgins 1996). A possible reason for these discrepancies is that some features of the LLJ, such as its sharp lateral gradient and its daily cycle are probably misrepresented with the present available data (Berbery and Collini 2000). In addition, the only evaporation dataset of subtropical SA available for long periods is the one from the NCEP–NCAR reanalysis. Due to these limitations, what is intended in this paper is to assess the circulation patterns and the qualitative features of the water vapor transport associated with the SST conditions in the WSSA. To do so, we focus on the low-level flow and on the water vapor transport Q, because as it is known, their convergence basically determines the precipitation rate.
Due to the presence of the Andes Mountains and the Bolivian Plateau, water vapor transported at low levels into subtropical SA must originate either in the Amazon basin or in the Atlantic Ocean. The annual cycle of the water vapor transport over South America was discussed in some recent papers, which considered both its stationary component (Zhou and Lau 1997), and the mean transient component of the water vapor transport and their horizontal convergence (Labraga et al. 2000).
Figure 1 shows the integrated (from surface to 700 hPa) water vapor transport Vq. In July, there is a strong southward flow that starts at about 15°S, becomes more intense at about 25°S, and turns southeastward south of 30°S (Fig. 1a). West of 60°W, coinciding with the small low-level water vapor transport, the mean precipitation is scarce in Argentina and Paraguay (Fig. 2a). The precipitation field has a maximum located in southern Brazil and northern Uruguay that extends northward as a ridge, following the path of the main meridional water vapor flux. Another region where precipitation exceeds 60 mm is the coastal area north of 20°S where there is a direct water vapor flux from the Atlantic (Fig. 1a). Halfway between these two regions, at 20°–25°S, precipitation is less than 40 mm, coinciding with the small low-level water vapor flux (Fig. 1a).
In January, the branch of the low-level water vapor flux from the continental low latitudes follows a southeasterly direction, and it converges with the southwestward transport carried by the winds of the South Atlantic high in the vicinity of the SACZ, at about 20°S (Fig. 1c). To the south, the water vapor flux has an anticyclonic circulation over subtropical Argentina. The convergence of water vapor flux in the continent near the SACZ is accompanied with abundant precipitation, more than 200 mm (Fig. 2c). Hereinafter, this area will be referred to as the region of the mean continental extension of the SACZ. To the south, the westward component of the water vapor flux reaches the Andes, and contrasting with the dry winter, it is accompanied with considerable precipitation in the region west of 60°W (Fig. 2c).
The transition seasons represented by October and April have fairly similar patterns, both in the integrated transport of moisture and in the precipitation fields. In April, the water vapor transport and the precipitation fields resemble the respective July fields rather than those of January, and vice versa in October (Figs. 1 and 2). In April, the precipitation has a maximum axis at about 55°W, coinciding approximately with the path of the greatest water vapor flux (Figs. 1 and 2). In October, the precipitation field has a relative maximum axis along the region of the mean continental extension of the SACZ. This maximum is sustained by a water vapor transport that is similar to, though weaker than, that of the summer. The rainfall west of 60°W is considerably more abundant than in April and July, but less abundant than in summer. This rainfall is fed by the westward component of the water flux between 25° and 35°S, which is, however, less intense than in summer.
The mean precipitation fields are not necessarily explained by the mean water vapor transport, since they are actually determined by the convergence of this water vapor flux. However, there is a qualitative agreement between precipitation and the low-level water vapor transport fields over eastern subtropical SA, which seems to indicate that in this region, humidity is the main limiting factor for rainfall (Figs. 1 and 2). Part of this qualitative agreement reflects the fact that over subtropical South America, the magnitude of the convergence of the transient water vapor transport is, in general, lower than that of the stationary transport (Labraga et al. 2000).
4. SST in the WSSA and precipitation in eastern subtropical SA
The following analyses are focused on January as the month that is representative of the midsummer conditions. A canonical correlation analysis (CCA; Wilks 1995) between western South Atlantic SST north of 40°S and precipitation over eastern subtropical SA was carried out. First, principal component analyses (PCA) of the SST and the precipitation field were performed separately. Only the first principal components that together represented 80% of the variance were retained to reconstruct the series of each field. This procedure filters out the small-scale features of both fields that are suspected of not being correlated. Then the CCA was applied to the reconstructed series.
The correlation of the first canonical mode with SST is positive and significant over most of the western South Atlantic Ocean, except at its southern part (Fig. 3a). Its correlation with the precipitation is weak and not significant (Fig. 3b). In the case of the second canonical mode, the correlation with SST has the same sign over practically the whole region, and it has significant values over a large area near the coast, centered at about 23°S (Fig. 3c). The pattern of the correlation with the precipitation field has a dipolar structure with opposite and significant values in eastern subtropical Argentina, Uruguay, and southern Brazil on one hand, and in eastern Brazil north of 20°S on the other (Fig. 3d). This dipolar structure is similar to that of the correlation of precipitation with the first mode, but in the second mode the southern center is displaced about 600 km to the south with respect to the first mode. According to these correlation patterns, positive SST anomalies in the WSSA tend to be associated with increased precipitation in the southern center of the dipole, and with diminished precipitation in the northern center. This northern center is located about 500 km north of the region of the continental extension of the SACZ.
When the CCA is performed for the 3-month period corresponding to the austral summer, namely, December–January–February, the first mode is similar to the second mode of the January CCA, except that the correlation with SST has a maximum at the same latitude—25°S—but farther eastward, at 25°W (Doyle and Barros 2000).
During the summer months, the outgoing longwave radiation (OLR) mean field shows a continuity of the large-scale convective activity between the continent and the ocean at about 20°S (Liebmann et al. 1999; Barros et al. 2000). Assuming that this is also the case in the precipitation field, the previous results are consistent with the behavior of the SACZ with respect to SST in WSSA during the summer. In fact, this system is enhanced and displaced to the north with negative SST anomalies to the south of 20°S, and it shifts southward and its intensity is reduced with positive SST anomalies in the same region (Barros et al. 2000).
The second mode of the CCA has a clear signal in the precipitation field. This mode has the greatest correlation with the SST in the region within 20°–30°S and 30°–50°W, approximately at the mean position of the SACZ. The rainfall composites corresponding to the warmest and coldest SST averaged on this region were constructed to explore how the rainfall varies with these extreme SSTs. We have considered the warmest cases to be those in which this SST is higher than its mean plus its interannual standard deviation. Similarly, the coldest cases were considered as those in which this SST was lower than its mean minus the interannual standard deviation. The mean SST in January averaged over this region is 25.3°C with an interannual standard deviation of 0.5°C (Fig. 4). According to this definition, the warmest cases were 1963, 1971, 1973, 1984, and 1988, whereas the coldest ones were 1964, 1965, 1970, and 1985. Hereinafter, they will be referred to as the W and the C cases.
Figure 5 shows the January composite precipitation field corresponding to the W and C cases (Figs. 5a,b) and the difference between these two composites (Fig. 5c). In the W composite, there are two zones of maximum rainfall. The first maximum encompasses an elongated area stretching from northwest to southeast, with an axis shifted a few hundred of kilometers to the southwest with respect to the axis of the region of the mean continental extension of the SACZ. The second maximum is centered farther south, at about 30°S and 55°W. In the C composite, there is only one maximum, elongated along the same direction of the first maximum of the W case, but circa 300 km to the north. In both cases, the extension of the SACZ pattern in the continent is recognizable. Its interannual variability is similar to that of the SACZ (Barros et al. 2000), being more intense and shifted to the north in the C composite, and weaker and displaced to the south in the W composite. The difference field (Fig. 5c) shows a dipole structure with centers along the first thousand kilometers from the Atlantic coast, one immediately to the north of the region of the mean continental extension of the SACZ, and the other at about 30°S. Both dipole centers are significant at the 95% level. In this case, as well as in the following figures showing composite differences, significance level was calculated using the Student's t-test considering the degrees of freedom calculated from the autocorrelation function.
5. Regional circulation and western subtropical South Atlantic SST
In the previous section, the composites of precipitation in eastern subtropical SA, stratified according to the extreme SST averaged in a zone of the WSSA near the coast of Brazil, at 20°S were shown. The corresponding composites of 850-hPa geopotential and wind vector fields for the W and C cases are presented in Fig. 6. In the W composite, the anticyclonic circulation of the South Atlantic high extends inland in tropical SA. The trade winds entering into the continent at equatorial latitudes turn southeastward near the Andes and merge with northerly winds from the west branch of the subtropical high to feed the SACZ. The strong southeastward flow is maintained by the east–west gradient of geopotential across the eastern continent, between 15° and 25°S. Part of the flow, east of the Andes, divert toward the south over subtropical Argentina (Fig. 6a).
In the C composite (Fig. 6b), the basic structure of the low-level flow in tropical latitudes is similar to that of the W case, but the main flow turns eastward toward the SACZ at about 20°S. The anticyclonic circulation of the South Atlantic high over SA is confined to low latitudes. In addition, over eastern Brazil, the northerly flow of the western part of the South Atlantic high starts farther to the north than in the W case, converging with the flow from the western branch, also at lower latitudes. This is consistent with the enhancement of precipitation and the northward and eastward displacement of the maximum precipitation over the continental extension of the SACZ, as shown in section 4. As in the W case, part of the southeastward flow diverts to the south toward subtropical Argentina, but does so only in the west, near the Andes, where the flow from the reanalyses is less reliable. There is an anticyclonic circulation over eastern subtropical SA that reflects the synoptic situations when the southwestern part of the South Atlantic high extends westward over Argentina. In these situations, the wind blows from the ocean over a considerable part of the eastern subtropical continent, and this can be interpreted as a monsoonlike circulation, since it exhibits features of a giant sea–land breeze. Indeed, this circulation is so frequent in summer that the average circulation in January has the hallmark of these cases (Fig. 1), and it is clearly dominant when the neighboring sea surface reaches the most negative temperature anomalies (C cases), as will be discussed later. Except for the subtropical continent east of the Andes, the low-level circulation in both the W and C composites is almost identical, which could indicate that the differences between both cases in this region may be caused by a regional forcing.
As might be expected, in the W and C composites, the moisture transport integrated between surface and 700 hPa (hereinafter referred to as the low-level water vapor transport) has the same features as the wind at 850 hPa, except for some minor differences (Fig. 7). The main difference between the low-level moisture transport and the 850-hPa wind is that the intensity of the former in midlatitudes is relatively smaller than it is in subtropical and tropical latitudes, because of the lower specific humidity content of the colder air.
According to the Helmholtz theorem, the low-level water vapor transport field can be decomposed into the sum of the nondivergent or rotational field and the divergent or irrotational field. As in other monsoon regions of the world (Chen 1985), the divergent component of the low-level water vapor transport is of the same order of the low-level water transport in eastern subtropical SA and in the southward flow from the tropical continent. The difference between the W and C composites of the horizontal divergence of the integrated low-level water transport roughly reproduces the dipole structure of the respective rainfall difference (Figs. 5c and 7c). Another feature, consistent with the respective difference between the rainfall fields, is that the difference between W and C composites shows the same sign in western Argentina as in the region of the continental extension of the SACZ.
Figure 8 shows the difference between the W and C composites of the meridional and the zonal components of the moisture transport. In the case of the meridional transport, the difference is negative along the path of the strong southeastward low-level flow and is positive elsewhere, especially to the north of the mean SACZ position (Figs. 7a and 8a). The opposite occurs with the difference of the zonal component of the moisture transport, which is positive to the north of the SACZ and negative to the south, especially in southern Brazil (Fig. 8b). There is a clear signal that in the C composite the transport of humidity has a strong component from the west toward the SACZ, while in the W composite there is an even stronger component from the north over southeastern SA.
The moisture transports associated with weak and strong SACZ events calculated by Nogués-Paegle and Mo (1997) are practically the same as those shown here (their Figs. 7 and 8). Besides, the dipolar structure in the precipitation field (Fig. 4c) is similar to the one described by Nogués-Paegle and Mo (1997) although in the present study the centers of the dipole are displaced to the north. In other words, the variability of the monthly patterns of rainfall and the low-level water vapor transport corresponding to extreme SSTs in the WSSA is similar to the intraseasonal variability.
Further evidence of this similarity is found when the PCA of daily low-level water vapor transport anomalies over subtropical SA and the neighboring oceans is performed. The first component explains 21% of the variance, and when it is added to or subtracted from the main field, it basically reproduces the regional features of the water vapor transport of the W and C composites (Fig. 7). An alternative way of showing this feature is to calculate the composites of the daily cases corresponding to the extreme values of the loading factor of the first principal component. Extreme values are considered as those exceeding (below) the mean plus (minus) the standard deviation. The composite of the extreme positive values of the first loading factor is similar to the mean 850-hPa flow and to the mean low-level water vapor transport of the W composite (Figs. 6a, 7a, and 9a). This is also true for the composite of the extreme negative values of the first loading factor and the W composite (Figs. 6b, 7b, and 9b). The fact that the patterns of the cases associated with the positive (negative) pattern of the first PCA mode are almost replicated by the monthly means corresponding to months with the warmest (coldest) SST in the WSSA indicates the prevalence of each of these patterns during those months. This prevalence is corroborated by the mean values of the loading factors, which are 0.2 during the W months, and −0.2 during the C months, indicating that the W and C patterns result, in each case, from the prevalence of one or the other of the two phases of the intraseasonal variability.
Figure 10 shows the difference in the mean vertical motion (ω) at 500 hPa between the W and C composites. In most of the region depicted in Fig. 10, with the exception of the SACZ and of eastern subtropical SA, there are almost no differences between the W and C cases. There is a dipole between the SACZ and eastern subtropical SA, which is consistent with the precipitation field (Fig. 5) and with the low-level convergence of water transport (Fig. 7). These fields indicate that the active phase of the SACZ prevails during the C cases and that the contrary occurs during the W cases. The dipole in the ω vertical velocity responds to the compensatory subsidence over northern Argentina and southern Brazil during the active phases of the SACZ, as was found by numerical experiments performed with an idealized heat tropical source that takes into account an additional heat source at the SACZ (Gandú and Silva Días 1998).
The SST interannual variability in the WSSA is related not only to the low-level flow but also to the upper-tropospheric circulation. In the W composite, the 300-hPa geopotential and wind fields show the Bolivian high (BH) flanked by weak easterlies to the north and by the strong flow from the west to the south (Fig. 11a). On the other hand, in the C composite, the BH is less developed, and it is displaced to the south with respect to the W composite, but the anticyclonic circulation extends to the west with a wide ridge over the continent that extends to the south to almost 40°S. Over northeastern Brazil, the flow continues with cyclonic circulation through the western part of a deep trough (Fig. 11b). It is interesting to notice that the BH, the ridge to the south of this high, and the trough over the western tropical South Atlantic are features of the mean summer upper-tropospheric circulation (Chen et al. 1999). According to the features that appear in the W and C composites, it seems that the first prevails during January months with warm surface waters in the WSSA, whereas the other two are more closely associated with cold waters in this region.
The SACZ activity appears to be the link between the SST in the WSSA (cases W and C) and the upper-tropospheric circulation over SA. The active phase of this system represents an additional heat source in the troposphere that enhances the trough over eastern SA and displaces the BH to the south, according to the numerical experiments performed by Gandú and Silva Días (1998, their Figs. 4a and 9a). However, their experiment did not show a change in the BH strength, although on the other hand, it did not account for the heating release over the Bolivian Plateau. This aspect was treated by Lenters and Cook (1999), who studied the circulation and precipitation patterns over southern SA associated to wet cases in the Bolivian Plateau. In one of these cases, the precipitation field and the circulation patterns at 850 and 200 hPa are very similar to those of the W case (their Fig. 12). Therefore, during the inactive phase of the SACZ, convection over the Bolivian Plateau is somehow enhanced, producing an additional heat source that strengthens the upper-level high.
The composites of the fields corresponding to the extreme SST in the WSSA, described in the two previous sections, show two different patterns over eastern subtropical SA. In these patterns there is evidence of qualitative consistency between the fields of the variables analyzed.
Figure 12 shows the January mean skin temperature and its respective anomalies for W and C composites. While the W minus C composite presents the greater difference in the region utilized to define the W and C composites (Fig. 12d), the W and C anomalies are not exactly antisymmetric (Figs. 12b,c). Over the ocean, the greatest anomalies in the W composite are displaced to the northeast, and those in the C composite to the south, with respect to the W minus C difference. In the W composite, at the latitude of the greatest anomalies over the ocean, namely, 10°–20°S, anomalies over land are positive, as they are over the ocean. On the other hand, the C anomalies at about 30°S, the latitude of the greatest anomalies over the ocean, have opposite signs over the ocean and land.
Cold SST anomalies generally enhance the low-level anticyclonic circulation. Thus in the C cases, there should be an enhanced northeastern component of the wind in the WSSA at about 25°–35°S. In addition, cold SST in the WSSA during midsummer is the most favorable condition for the land–sea temperature gradient to enhance the easterly low-level flow. Indeed, in January the average gradient of the skin surface temperature between the land in western Argentina and the sea at 35°W is about 3°C at 30°S, while the composite anomaly of the C cases shows an increment of this gradient of almost 1.5°C, which is about 50% of the climatological mean (Figs. 12a,c). A pressure gradient with a meridional component (Fig. 6b) consistently accompanies the easterly flow over the continent. Under these conditions, the flow from the tropical continent to the south is blocked. Consequently, this flow is diverted toward the SACZ, and converges with the southward flow driven by the South Atlantic high, thus contributing to enhance the SACZ activity (Figs. 5b, 6b, 7b, and 8b). As was shown in the previous section, this circulation associated to the active phase of the SACZ prevails during the C months, and thus, the convection along the SACZ produces compensatory subsidence over northern Argentina and southern Brazil [Gandú and Silva Días 1998; Fig. 10 (this paper)] which, in turn, favors the anticyclonic circulation in the region south of the SACZ.
During the C months, the zonal gradient of precipitation reverses its sign over subtropical Argentina, with greater precipitation in the west and south than in the east, between 25° and 40°S (Fig. 5b). This could be the result of forced upward motion in the Andes and the mountain regions in western Argentina, resulting from the low-level westward flow. In addition, this flow turns southward because of the Andes, bringing moisture toward the arid and semiarid areas of southwestern subtropical Argentina (Fig. 6b). In this region, most of the precipitation occurs in summer, when the circulation typical of the C case is so important that it is reflected in the main circulation (Figs. 1c and 2). On the other hand, over the east of SA between 25° and 40°S, the reduction of the low-level water vapor transport and the prevailing subsidence is accompanied with diminished precipitation (Figs. 5b and 7b).
The W composite SST anomalies during midsummer do not favor the development of an easterly low-level flow over the WSSA (Fig. 12b). Indeed, there is a reduction of the land–sea gradient of temperature at about 20°–30°S, though part of it could be attributed to the positive surface temperature anomalies over land probably caused by the lower cloudiness associated to the inactive phase of the SACZ. The low-level anomalous circulation associated to positive SST reduces the southern component of the wind to the west of the area with greatest positive SST anomalies (Figs. 6a,b). However, since the greatest SST anomalies in the W case are farther north, they do not influence the low-level anticyclonic circulation over southern Brazil and Uruguay. Hence, the southeastward low-level flow predominates over eastern subtropical SA, and the convergence with the southward branch of the South Atlantic high is smaller and displaced to the south with respect to mean conditions (Figs. 1a and 7a). The reduced activity over the SACZ produces less subsidence to the south of it than in the C case (Fig. 7c). In fact, there is a maximum of precipitation at 30°S of the same magnitude as the one associated with the continental extension of the SACZ (Fig. 5a). Over eastern subtropical SA, there is frequent frontal activity (Seluchi et al. 1995). Thus, the southern maximum in Fig. 5c probably results from the combination of frontal and prefrontal activity with the high availability of moisture along the path of the greatest low-level water vapor transport (Fig. 7a).
With the exception of the reduced SACZ activity observed during the W composites, the rest of the pattern predominates during the nonsummer seasons, as can be seen from the mean field of the low-level water transport and the precipitation field (Figs. 1 and 2). On the other hand, the mean summer low-level water vapor transport replicates in part the C composite, indicating that this pattern is caused by some feature of the summertime, such as the land–sea surface temperature gradient or the SACZ enhancement, or by both (Fig. 1a).
The relative predominance of the W or C patterns influences the midsummer precipitation in eastern subtropical SA, not only during those years with extreme SST conditions in the WSSA, but in every January, as indicated by the correlation between the integrated transport of water vapor from surface to 700-hPa normal to the NE–SW direction, (roughly the inland component) and monthly precipitation. Figure 13 shows such correlations for the inland component of the low-level water transport at 30°S and at the mean SACZ position over the coast. These correlation patterns indicate that the higher the inland low-level moisture flux is at 30°S (at the region of the continental extension of the SACZ), the lower the precipitation is in the nearby coastal region and the higher it is in the region of the continental extension of the SACZ (in the coastal area, about 30°S; Fig. 13). Over western Argentina and as far as 40°S, the correlation sign predominantly coincides with that of the continental extension of the SACZ (Fig. 13). These correlation patterns are consistent with the relative predominance of a C or W low-level circulation type and their associated precipitation patterns (Figs. 5, 6, and 7). Over the SACZ, this responds to the fact that most of the humidity comes from either the west or the north (Fig. 6). However, this is not the case over the subtropical region south of the SACZ, where in the C cases there is moisture transport from the Atlantic. Consequently, the negative correlation should be attributed to the prevailing subsidence that controls the possible convective activity, which would otherwise occur with the advection of maritime air over the land, heated by midsummer radiation.
In view of the close association between SST anomalies in the WSSA and the low-level atmospheric circulation over eastern subtropical SA, the question that naturally arises is whether SST forces the low-level atmospheric circulation or vice versa. In the tropical and subtropical South Atlantic Ocean, large-scale atmospheric waves may cause large SST anomalies. The similar scale of some SST anomalies and of atmospheric waves, and the correlation between low-level atmospheric cyclonic circulation and cold SST anomalies could be considered indications in this sense (Kalnay et al. 1986). Venegas et al. (1997) reached a similar conclusion analyzing coupled fields of surface pressure and SST. In the first mode of their singular value decomposition (SVD), where the pressure field roughly represents the variability of intensity of the South Atlantic high, surface pressure usually leads the SST by 1–2 months indicating an atmosphere-to-ocean forcing. On the other hand, in their second mode, which is related to the east–west displacement of the South Atlantic high, the SST frequently leads the surface pressure revealing an ocean-to-atmosphere forcing, and therefore indicating a relationship that is more complex than a simple one-way forcing.
Although in this study W and C composites were defined in accordance with the SST anomalies in a limited area, these SST anomalies were part of a broader pattern of variability (Fig. 12). In the case of Venegas et al. (1997) their analysis was limited to the Atlantic Ocean, it did not include the circulation over the continent. In spite of these differences, in both the second mode of the SVD of Venegas et al. (1997, their Fig. 19) and the low-level circulation corresponding to the W and C composites (Fig. 6), predominantly warm (cold) SST anomalies in the tropical and subtropical South Atlantic Ocean are related with a westward (eastward) displacement of the western part of the South Atlantic high. In view of the fact that in the second mode of the SVD of Venegas et al. (1997), SST leads surface pressure, this analogous behavior suggests that it is SST anomalies that force the W and C low-level circulation patterns and not the other way around. However, in view of the remote forcings of the SACZ, one should be cautious in concluding that the low-level circulation patterns in eastern subtropical SA associated with extreme SST anomalies in the WSSA are forced exclusively by these SST anomalies.
Liebmann et al. (1999) found evidence that the submonthly variations in the SACZ activity are influenced by Rossby waves propagated from the extratropical regions. This is consistent with a number of studies that found a relationship between the variability of the South Pacific convergence zone (SPCZ) and the SACZ (Kalnay et al. 1986; Nogués-Paegle and Mo 1997; Lenters and Cook 1999). For instance, according to a barotropic model using a January mean state, the upper-tropospheric divergence at SPCZ influences the region of the SACZ (Grimm and Silva Días 1995). There are also suggestions of a link between the variability at the SACZ and the Madden–Julian oscillation (Kiladis and Weickmann 1992; Nogués-Paegle and Mo 1997; Lenters and Cook 1999). Finally, the Amazon convection may play a role in the SACZ dynamics (Figueroa et al. 1995).
In view of these different forcings, which may have interannual variability as well, it is likely that the SACZ interannual variability could also be partially driven by one or more of these remote or regional forcings. In such case, the low-level atmospheric circulation variability might force the SST variability over the ocean along the SACZ. This forcing may be caused by the change in the upper-ocean heat balance due to the difference in cloudiness during the active and the inactive phases of the SACZ and by the strong winds associated to convective activity, that favor the mixing of the upper ocean and the cooling of the sea surface during the active phases of the SACZ.
Whether SST initially forces the low-level atmospheric circulation, or the SACZ variability forces the SST, a possible positive feedback mechanism can maintain the C or W patterns in the SST and in the low-level atmospheric circulation. The negative (positive) SST anomalies associated with the ocean cooling (warming) of the active (inactive) phases of the SACZ occurs locally. However, since the velocity of the Brazil Current varies from 0.2 to 0.6 m s−1 (Campos et al. 1995), these SST anomalies may be advected about 1000 km to the south in less than a month. In such a case, the negative (positive) SST anomalies advected toward the south of the SACZ enhance the westward (southeastward) component of the flow at subtropical latitudes and the humid and warm tropical flow toward the SACZ (northeastern Argentina and southern Brazil), enhancing (reducing) the SACZ activity, and then locally cooling (warming) the SST. Finally, the advection of these negative (positive) SST anomalies southward restarts the positive feedback cycle. This feedback might contribute to maintaining an abnormal period of negative (positive) SST anomalies and intense (weak) SACZ activity, as in the case of January 1970 (January 1988) (Doyle 2001). This possible feedback mechanism can explain the prevalence of the W or the C low-level circulation pattern with positive (negative) SST anomalies in the WSSA.
7. Summary and conclusions
The prevalence during midsummer of each of the phases of the low-level pattern associated to the seesaw of the SACZ depends on SST in the WSSA. This prevalence may possibly result from a positive feedback that might contribute to maintain positive (negative) SST anomalies, weak (intense) SACZ activity, and a W (C) low-level circulation pattern.
Figure 14 presents a schematic of the mean low-level flow and precipitation anomalies in subtropical SA corresponding to extreme SST in the WSSA. In the W case, the main stream of the flow from the continental low latitudes has a southeastward direction starting at 10°S and converges with the west flow at 35°S over the ocean. In the C case, the flow from the continental low latitudes turns to the east toward the SACZ, whereas to the south of it there is an anticyclonic circulation with westward flow north of 35°S. With differences in the spatial scale and in latitude, the C circulation resembles the low-level circulation of tropical South America. In both cases, the westward flow turns to the south after reaching the Andes Mountains, following an anticyclonic circulation.
In the W case, there are positive rainfall anomalies in two zones, one in the continental extension of a southward-displaced SACZ, and another centered in northeastern Argentina and southern Brazil, in the path of the mainstream of the low-level moisture transport. In the C case, the positive anomalies are in the continental extension of the SACZ, which is shifted northward of its mean position, while negative anomalies are observed in northeastern Argentina and southern Brazil. These negative anomalies are a consequence of the compensatory subsidence associated to the more active SACZ and the suppression of the important water vapor transport from the tropical continent. On the other hand, in western Argentina there are positive anomalies in precipitation favored by the transport of moisture from the Atlantic Ocean over this region.
This paper has been funded by the University of Buenos Aires under Grants UBACYT TW75 and IAI-CRN 055.
Corresponding author address: Dr. Moira E. Doyle, Department of Atmospheric Sciences, University of Buenos Aires, Cuidad Universitaria, P II 2°Piso, 1428 Buenos Aires, Argentina. Email: firstname.lastname@example.org