The El Niño impact on Brazil's summer monsoon has not been adequately assessed through seasonal analysis because it shows significant subseasonal variations. In this study, the El Niño influence on the summer monsoon circulation, rainfall, and temperature is analyzed with monthly resolution, using data from a dense network of stations. The expected precipitation percentiles during the monsoon season of El Niño (EN) events are calculated, as well as anomalies of surface temperature and thermodynamic parameters. This information is analyzed jointly with anomaly composites of several circulation parameters. The analysis shows that some precipitation and circulation anomalies, which are consistent and important during part of the season, are smoothed out in a seasonal analysis. There are abrupt changes of anomalies within the summer monsoon season, suggesting the prevalence of regional processes over remote influences during part of the season. The probable role of remote influences and regional processes is assessed.
The anomalous heat sources associated with El Niño perturb the Walker and Hadley circulations over South America and generate Rossby wave trains that produce important effects in the subtropics and extratropics. In the early summer monsoon season, remotely produced atmospheric perturbations prevail over Brazil. Anticyclonic low-level anomalies predominate over central-east Brazil, in the Tropics and subtropics, due to the subsidence over the Amazon and to Rossby waves in the subtropics. Easterly moisture inflow from the Atlantic is favored, but diverted toward northern South America (SA) and south Brazil. There are negative precipitation anomalies in north and central-east Brazil and positive ones in south Brazil. These precipitation anomalies are favored by the perturbation in the Walker and Hadley circulation over the east Pacific and South America, and by a Rossby wave train over southern SA that originates in the eastern Pacific. In January, with the enhancement of the continental subtropical heat low by anomalous surface heating during the spring, there is anomalous low-level convergence and cyclonic circulation over southeast Brazil, while at the upper levels anomalies of divergence and anticyclonic circulation prevail. This anomalous circulation directs moisture flux toward central-east Brazil, causing moisture convergence in this region. A favorable thermodynamic structure enhances precipitation over central-east Brazil, the dry anomalies in north Brazil are displaced northward, and the anomalies in south Brazil almost disappear. In February, after the above-normal precipitation of January, the surface temperature anomalies turn negative and the precipitation diminishes in central-east Brazil. There are negative rainfall anomalies in north Brazil and in the South Atlantic convergence zone (SACZ) and positive ones in south Brazil.
Influence function analysis shows that while the anomalies of circulation over southeast Brazil in the spring of El Niño years are mostly due to remote influences from the tropical east Pacific, those in January are probably due to local influence. During this month the monsoonlike circulation is enhanced. Simultaneous and lagged correlation analysis of SST and rainfall in central-east Brazil shows that SST anomalies in the Atlantic Ocean off the southeastern coast of Brazil fluctuate on the same timescale as the circulation and precipitation anomalies.
El Niño (EN) events are very likely to affect some important components of the South America summer monsoon (SASM), in view of its impact on the tropical heat sources and global atmospheric circulation.
The summer monsoon regime in South America (SA) is responsible for the rainy season in most of Brazil (Fig. 1). Its interannual variability impacts significantly on very important economic activities such as hydroelectric power generation and agriculture. Moreover, extreme events of the South Atlantic convergence zone (SACZ), one of the important SASM features, have strong consequences in very densely populated regions in southeast Brazil.
Previous studies have cast light on some important aspects of the EN impact on summer rainfall in certain regions of SA. Ropelewski and Halpert (1987) detected precipitation deficiency in northeastern SA and enhanced precipitation in southeastern SA during summers of EN events. Rogers (1988) found that in January–February–March (JFM) the precipitation is significantly higher during La Niña (LN) than during EN events in northern SA, but did not find significant difference in the subtropics to the south. On the other hand, significantly higher precipitation during EN was found over higher subtropical latitudes in October–November–December (OND). The patterns of correlation between station rainfall and the Southern Oscillation index (SOI) in Aceituno (1988) are consistent with these findings, indicating also that the southern part of northeast Brazil tends to have above-normal precipitation during JFM of EN events. In Rao and Hada (1990) only central and central-east Brazil show significant correlation between rainfall and SOI during summer. The correlation in northern Brazil does not appear as significant in this season, but in spring it is significant in northern and southern Brazil. Marengo (1992) reported that reduced rainfall in northern Amazonia tends to coincide with strong EN events, whereas there is an observed preference for more abundant precipitation during LN events. In analyses focused on southern Brazil and southern SA, Grimm et al. (1998, 2000a) showed that the impact on rainfall in summer is much weaker than in spring, and there is even a tendency for anomalies to reverse sign in January, in relation to December and February. Zhou and Lau (2001), using a 17-yr series of rainfall product constructed from satellite estimates, station data, and predictions of forecast model, obtained an interannual ENSO-related mode of variability for DJF, in which the rainfall tends to be above normal over Uruguay–south Brazil and lower over northeast Brazil during EN years.
It is apparent that, besides showing common results, those studies also show some differences, that are worth clearing up through an analysis based on longer precipitation time series from a dense network of stations extended all over Brazil. Furthermore, it is useful to understand how EN events impact important components of the SASM as, for instance, the SACZ. Robertson and Mechoso (2000) reported that the interannual variability in the SACZ is largely independent of ENSO. Since the SASM circulation is strongly influenced by the latent heating over Amazonia (e.g., Figueroa et al. 1995; Lenters and Cook 1997), then it would be expected to be strongly impacted by the decrease of the convection there during EN events. The low-level winds that enter northern Brazil bringing warm moist air from the tropical Atlantic Ocean, and turn southeastward, are likely to be perturbed by the anomalous subsidence over Amazonia. Thus, a significant impact on the SACZ, whose moisture source is associated with these winds (Lenters and Cook 1995; Paegle and Mo 1997), would be expected. Therefore, besides assessing the EN-related rainfall anomalies, it seems worth exploring the circulation anomalies that lead to them or are produced by them. In this context, it is worth pointing out that rainfall and circulation anomalies influence soil moisture and temperature (Barros et al. 2002), that can be important factors in producing regional anomalies in a monsoon regime. This may lead to subseasonal variation of the impact of EN events on summer precipitation, reported by Grimm et al. (1998, 2000a) for southern SA. This variation may be due to competing mechanisms, related to remote and local influences, and might be able to explain some discrepancies found in seasonal analysis.
The above considerations provided the motivation to pursue the following objectives: 1) to analyze the magnitude and consistency of the EN impact on summer precipitation all over Brazil; 2) to characterize mechanisms responsible for this impact; 3) to show its subseasonal variation; 4) to show that subseasonally consistent impacts are smoothed out in a seasonal analysis; and 5) to indicate some possible reasons for the subseasonal variation.
The dataset used and some features of the SASM climate, as well as of the rainfall regimes all over Brazil, are presented in section 2. The methodology is outlined in section 3. The EN-related anomalies of seasonal summer precipitation and atmospheric circulation are presented in section 4. The inconsistencies between them lead to subseasonal analysis, carried out in section 5. In section 6 the EN-related surface temperature anomalies and the thermodynamic structure in southeastern Brazil are analyzed in connection with the circulation and precipitation anomalies. The differences between remote influences in different periods of the summer monsoon season are assessed in section 7, and some aspects of the sea surface temperature (SST) variability are presented in section 8. Section 9 presents a summary and concluding remarks.
2. Data and climatic aspects
The data used in this study include monthly precipitation totals from 1175 stations in the period 1956–92, selected from a set of 1802 stations in order to span at least five EN events. These data were supplied by various sources, mostly by Agência Nacional de Energia Elétrica (ANEEL) and Instituto Nacional de Meteorologia (INMET). They are not uniformly distributed all over Brazil, and there are areas not covered, as can be seen from the gaps in some figures. The best coverage is in the south and southeast, and the worst in the central-west and north. Also surface temperatures observed by INMET are used. Atmospheric circulation and thermodynamic structure are analyzed with National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalysis monthly data from 1958 to 1992. The reanalysis data are more reliable after 1958 in SA because the radiosonde stations started operating after this date.
The years considered as starting years of EN events (named year 0) are listed in Table 1. The austral summer, which is the period of analysis, also extends into the following year (named year +).
b. Climatic aspects of the South American summer monsoon regime
Although the seasonal reversal of the surface winds in a “classic” monsoon regime is not apparent in the SASM, there are climatic features in the region that are characteristic of a monsoon climate. When the annual mean component is removed, the seasonal reversal in surface wind, induced by the strong diabatic heating in the subtropical highlands, is obvious (Zhou and Lau 1998). In spring, heavy precipitation occurs over Central America and western Amazonia, following the surface heating by solar radiation. In austral summer, as the major heating zone migrates to the subtropics, a thermal low pressure system develops over the Chaco region, in central SA, while pressure increases over the northwestern Sahara. The southwest–northeast pressure gradient strengthens, enhancing the tropical northeasterly trade winds. Anomalous cross-equatorial flow penetrates the continent, carrying moisture. It becomes northwesterly and turns clockwise around the Chaco low. Low-level wind and moisture convergence associated with the interaction of the continental low with the South Atlantic high and the northeasterly trade winds enhance precipitation in the Amazon, and central and southeast Brazil (Lenters and Cook 1995). In these regions, austral summer is the peak rainy season (Fig. 1). The SACZ is in its most active stage, with its precipitation enhanced by southward advection of moisture between the continental low and South Atlantic high. Accordingly, from spring to summer the upper-level anticyclonic center moves from Amazonia southward, setting up the Bolivian high.
Figure 1 is a comprehensive picture of the precipitation regimes all over Brazil. It shows that, besides northeast Brazil, only the northernmost and southernmost parts of the country do not have summer peak rainy seasons. In northern Brazil, the maximum of precipitation in austral autumn and winter is related to the annual migration of deep tropical convection. The austral autumn maximum in northeast Brazil is connected to the southward penetration of the intertropical convergence zone (ITCZ) into the tropical South Atlantic Ocean during this season. In southern Brazil there is a transitional region where the peak rainy season changes from summer to spring and then to late winter. The southernmost region has a fairly uniform distribution of rainfall throughout the year, because it is also subjected to a midlatitude regime, where the rainfall is due to frontal penetration associated with migratory extratropical cyclones.
a. Impact of El Niño on precipitation and surface temperature
The EN-related median precipitation is calculated for every station for each month of the summer and for whole season (DJF), and expressed as a percentile of the gamma distribution for each station. This may be considered as the expected precipitation percentile in EN events, and gives a variance-independent measure of the impact of these events.
Surface temperature anomalies during EN events are composited for some regions, in order to study the relationship between temperature, precipitation, and circulation anomalies.
b. Consistency of the impact
The consistency of the relationship between EN and wet (dry) conditions of a population that contains r EN episodes with k of them dry (wet) is given by the probability of obtaining more than k dry (wet) cases in a sample of r episodes taken at random from this population [i.e., the cumulative probability of obtaining k + 1, k + 2, … , up to r dry (wet) cases is computed]. This probability is given by the hypergeometric distribution, and is the confidence level of that relationship. For instance, a 90% confidence level for the test of consistency means that in a random sample with the size of the number of EN events there would be 90% chance of getting more dry (wet) cases than in the sample of EN events.
This test is robust with respect to the asymmetry of the rainfall distribution, and assesses the significance not of the differences between EN and LN, but of the EN-related anomalies with respect to the whole data series. Furthermore, it is applied to precipitation at each particular station and not to an area-averaged precipitation series. As consistency of anomalies is much more easily obtained for average rainfall over a fairly homogenous region than for a single station, this is a rather stringent test.
Besides being applied to precipitation anomalies, the test based on the hypergeometric distribution is also used to assess the consistency of temperature and circulation anomalies during EN events. In the case of wind, the anomalies are considered consistent when one of the components is consistent.
c. Mechanisms of precipitation anomalies
The way the atmospheric circulation anomalies act on precipitation is assessed through the analysis of the perturbations they impose on the essential ingredients of precipitation: moisture convergence, and the force to lift the moist air to the condensation level.
The alterations in the moisture convergence are mainly related to low-level winds, though here it is analyzed through composites of anomalies of vertically integrated moisture flux, and its divergence.
The force to lift the air may be altered mainly through anomalies in the dynamic lift via upper-level winds (jet streams and advection of vorticity), or in the large-scale divergence–convergence, or in the thermal forcing. To analyze how these different factors are influencing convection during EN events, composites are constructed of anomalies of the rotational and divergent components of the wind at 850 and 200 hPa, and anomalies of the vorticity advection at 500 hPa. Changes in the thermodynamic structure that might affect convection are analyzed through vertical profiles of the anomalies of potential temperature (θ′), equivalent potential temperature (θ′e), and saturation equivalent potential temperature (θ′es), computed using Bolton's (1980) formulas.
d. Remote influences
The remote influences from anomalous tropical heat sources associated with EN events on the SASM are explored with influence functions of a vorticity equation model, as proposed in Grimm and Silva Dias (1995). The model is linearized about a realistic basic state, and includes the divergence of the basic state and the advection of the basic-state vorticity by anomalous divergent wind, as in Sardeshmukh and Hoskins (1988):
The overbar indicates basic-state quantities, and the prime indicates anomaly quantities; ζ is the absolute vorticity, D is the divergence, and Vψ and Vχ are the rotational and divergent components of the wind; F′ is the forcing term or vorticity source, R is the linear operator that relates D′ to F′, and A′ is the damping term, including linear damping and biharmonic diffusion. The model is used at the 200-hPa level, which is both near the level of maximum divergence associated with convective outflow in the Tropics, and near an equivalent barotropic level in the extratropics.
The steady version of the model (with ∂ζ′/∂t set to zero) may be expressed in abbreviated form as
where L is a linear operator, and ψ′ is the anomalous streamfunction, related to the anomalous vorticity by
The application of the linear operator R−1 yields
where M = R−1L.
Then the influence function based on divergence forcing, is defined by
where δ(λ, ϕ, λ′, ϕ′) is the delta function. From the properties of the influence functions:
Thus, the influence function GD(λ, ϕ, λ′, ϕ′) for the target point with longitude and latitude (λ, ϕ) is, at each point (λ′, ϕ′), equal to the model response at (λ, ϕ) to an upper-level divergence concentrated at (λ′, ϕ′). Therefore, a map of the influence function for a given target point will show the regions in which anomalous upper-level divergence is most efficient in producing streamfunction anomalies at the target point.
4. Results of seasonal analysis
For the presentation of the results, Brazil will be divided into some regions (which are not the official regions of the country), listed in Table 2.
The precipitation percentiles expected for December–January–February (DJF) of EN events (Fig. 2) are well under 50 in north Brazil and above 50 in central-east Brazil, although they are not consistent over many areas in this last region. The percentiles are near 50 in South Brazil (except by a small area), although there are consistent positive anomalies in central-north Argentina (not shown here, but visible in Grimm et al. 2000a). This picture is different from the north–south dipole of Zhou and Lau (2001) in that the below-median precipitation concentrates mainly in north Brazil and not in the northeast (where rainfall is normal to above normal), and the precipitation anomalies in south Brazil are not higher than in the central and eastern parts of the country. The strongest positive anomalies in DJF in southern SA are in central-north Argentina.
The divergent wind composites show anomalous upper-level convergence over the Amazon and the North Atlantic ITCZ, with corresponding subsidence and low-level divergence (Fig. 3). These anomalies are part of a strong perturbation of the Walker circulation in equatorial east Pacific and SA, associated with EN events. The upper-level westerly circulation from the ascending branch in east Pacific toward the descending branch in the Amazon, and the low-level westerly return are visible in Fig. 3. Besides this well-known anomaly in the Walker circulation component of the divergent wind, there is also a strong and consistent anomaly in the Hadley circulation component, between the Tropics and subtropics in both hemispheres. This anomalous circulation is produced by anomalous EN-related subtropical convection in North and South America. Therefore, the below-normal rainfall over north Brazil is due to subsidence caused by large-scale divergent circulation anomalies.
The rotational wind anomalies (Fig. 4) show dynamic consistency with those of the divergent wind. The rotational wind at 200 hPa features upper-level pairs of cyclonic anomalies straddling the equator, associated with suppression of convection over the Amazon and the Atlantic ITCZ. At 850 hPa, there are consistent anticyclonic anomalies over northern SA, that weaken the northeasterly trades that usually bring warm moist air through northern SA, over Colombia, Venezuela, and Guyana. On the other hand, the easterlies in the western equatorial Atlantic are strengthened, so that the low-level winds that enter Amazon are more easterly and confined to north Brazil.
The enhancement of moisture flux into north Brazil from the tropical Atlantic is clear in Fig. 5. Therefore, it is not the decrease of this moisture inflow that is responsible for the relative dryness in north Brazil. The moisture flux, however, is diverted toward northern SA (and the tropical North Atlantic), and southeast Brazil, because of the low-level anticyclonic anomalies consistent with the subsidence over this region. The positive anomalies of moisture flux divergence over north Brazil are coherent with below-median precipitation there. However, there are discrepancies between precipitation and moisture divergence anomalies in other regions. For instance, there is anomalous moisture flux convergence in the southern part of southeast Brazil, where the precipitation anomalies are predominantly negative.
These discrepancies motivated a monthly analysis, shown in the next section.
5. Results of monthly analysis
The analysis of the evolution of monthly precipitation anomalies from spring to summer (Fig. 6), along with the main characteristics of the divergent and rotational circulation anomalies at lower and upper levels (Figs. 7 and 8), as well as of the anomalies of moisture supply (Fig. 9) shows a fairly coherent picture. In November there is excess rainfall in south Brazil and deficiency in the north and central-east. Upper-level convergence and subsidence over the Amazon and central-east Brazil (Fig. 7) produce anticyclonic anomalies at lower levels over these regions (Fig. 8). A wave train emanating from the anomalous heat source in the east Pacific produces an anomalous pair cyclone–anticyclone over subtropical SA (Fig. 7). It has equivalent barotropic structure, with the largest magnitudes at upper levels (Figs. 7 and 8). At low levels it extends the anticyclonic anomalies to the subtropics, while at upper levels it strengthens the subtropical jet and shifts it southward. This pattern produces advection of cyclonic vorticity over southern Brazil at 500 hPa (not shown), favoring an ascending motion in this region. The low-level anticyclonic anomalies over central Brazil, while increasing moisture inflow from the Atlantic into the Amazon, also direct it toward northern SA and southern Brazil (Fig. 9). While convergence of moisture is increased in the south, it is decreased in central and north Brazil.
In December, negative precipitation anomalies strengthen in the northeastern Amazon and appear over the central-west (Fig. 6). The upper-level convergence and subsidence also moves toward these regions, although there is also upper-level convergence over northeast Brazil, where rainfall does not show consistent anomalies (Fig. 7). The negative precipitation anomalies in the central-east weaken and precipitation over south Brazil is near normal. The upper-level anomalous cyclonic circulation associated with convergence over the Amazon is enhanced on both sides of the equator. Over tropical SA it is extended from the southwest to northeast (Fig. 7). At low levels the circulation anomalies are still predominantly anticyclonic, but weaken over central Brazil (Fig. 8). The moisture supply into the SACZ tends to be normal (Fig. 9).
From December to January a very significant change in the precipitation percentile patterns takes place (Fig. 6). The negative precipitation anomalies that were spread over north and central-east Brazil in spring are now restricted to the northernmost part of the country, merging with the negative anomalies in the ITCZ over the Atlantic (not shown here, but visible in OLR anomalies). This distribution is coherent with the upper-level convergence over this region (Fig. 7). In south Brazil, precipitation is below normal to near normal. Strong and consistent positive precipitation anomalies appear in central-east Brazil, including the SACZ. A negative pressure anomaly (not shown) and low-level cyclonic anomaly is established over southeast Brazil (Fig. 8), while at upper levels there is an anticyclonic one (Fig. 7), enhancing a baroclinic structure characteristic of the summer monsoon. The low-level cyclonic anomaly increases the moisture flux into central-east Brazil, where strong and consistent moisture convergence anomalies take place, whereas moisture divergence prevails over the other regions (Fig. 9).
In February the low-level cyclonic anomalies and upper-level anticyclonic ones of January (+) over southeast Brazil disappear. The positive rainfall anomalies weaken in central-east Brazil and even become negative in the SACZ. Positive anomalies return to south Brazil (Fig. 6). Upper-level convergence is back over part of central-east Brazil, cyclonic anomalies dominate southwestern SA, as in previous months, while anticyclonic ones dominate southwestern South Atlantic, with equivalent barotropic structure (Figs. 7 and 8). The anomalous moisture is again directed toward south Brazil (Fig. 9).
It is worth pointing out that the circulation anomalies over southeast Brazil have a barotropic structure from spring to summer, except for January. This is the only month within the summer season in which a monsoon type of circulation is enhanced.
The precipitation analysis shown here for the period 1956–92 was also carried out for the whole period of each station. The results were qualitatively very similar.
6. Temperature and humidity analysis in southeast Brazil
The rainfall and circulation anomalies associated with EN events are likely to produce temperature anomalies. To understand the relationship between these parameters and their evolution throughout the summer of EN events, composites of surface temperature anomalies in south and southeast Brazil are presented, as well as the thermodynamic structure of the troposphere averaged over southeast Brazil. As seen previously, this region is the center of the main intraseasonal changes of circulation anomalies, because it is the center of the low-level cyclonic anomalies responsible for the excess precipitation in central-east Brazil during January.
During the spring of EN years [here represented by November(0)], the southward shift of the subtropical jet hinders the northward displacement of cold fronts. Therefore, the warm advection from north, favored by the low-level anticyclonic circulation anomalies in central Brazil, persists for longer periods of time, leading to higher surface temperature over the subtropical region, specially in southeast Brazil (Fig. 10). The intense precipitation anomalies in south Brazil counterbalance this effect and overcome it in some areas, mainly in its southernmost part, which is also more easily reached by the cold fronts. Therefore, in the southernmost part of south Brazil, there are consistent negative temperature anomalies, while in southeast Brazil there are large and consistent positive ones, between 1 and 2 K.
In December(0), as the low-level anticyclonic circulation anomalies weaken and precipitation tends toward normal in southeast Brazil, the temperature anomalies almost disappear and even tend to reverse signal in January(+) and February(+), due to the consistent precipitation anomalies in January(+) and also to the change in circulation anomalies, that now favor cold advection from the south into southeast Brazil (Fig. 10).
The surface temperature anomalies given by the reanalysis data for November(0) and February(+) are smaller than those given by the observed data by as much as 0.5 K in southeast Brazil. In December(0) and January(+) the differences between reanalysis data and observed data reverse signal but are very small. This means that the reanalysis data smooth out the strong difference between surface temperature anomalies in November(0) and January(+).
The evolution of the anomalous thermodynamic structure of the atmosphere over southeastern Brazil during the summer of EN events, is shown in Fig. 11, through the vertical profiles of θ′, θ′e, and θ′es, calculated from the reanalysis data averaged over that region. In spite of the differences between the observed surface temperatures and those given by the reanalysis, these profiles are qualitatively coherent with the previous analysis and are able to give further insight into the intraseasonal changes during EN summers.
In November the temperature anomaly is maximum, coherent with the results obtained from observations at surface. There is some increase of specific humidity at low levels (θ′e > θ′), but the relative humidity decreases (θ′es > θ′e). Both θ′ and θ′es increase with altitude, mainly from 925 to 700 hPa, enhancing an inversion at the top of the planetary boundary layer (PBL). In this same range, θ′e decreases, augmenting the difference θ′es − θ′e, suggesting that the warming aloft results from subsidence down to the top of the PBL. This is coherent with the circulation characteristics described previously. The increase of θ′es with altitude, above the increase of θ′e at the surface, signifies less buoyancy or more stability of the atmosphere.
From November to December, θ′es decreases by about 0.6 K at the surface. It also decreases aloft and the anomalous inversion between 850 and 700 hPa disappears, while a small inversion between 925 and 850 hPa is retained. The diminution of the difference θ′es − θ′e indicates that the relative humidity increases aloft with respect to November and the subsidence above 925 hPa weakens.
The situation changes dramatically in January. Differently from the previous months, θ′e > θ′ throughout the troposphere, indicating that the tropospheric column becomes more humid due to convergence of the vertically integrated humidity flux into the region under focus. The anomalies θ′es turn smaller than θ′e up to 350 hPa, signaling increase of relative humidity and destabilization of the troposphere. The inversion at the top of the PBL is much weaker.
In February the situation starts reverting to more stable and dry conditions, from the bottom. The specific humidity decreases and the temperature is lower at lower levels. There is even an anomalous inversion near the surface.
The increase of temperature in spring cannot by itself enhance convection, as can be seen from precipitation anomalies in November and December. Land warming, however, increases the gradients of land–ocean temperature and pressure, and may cause the changes of circulation more clearly observed in January, with convergence and cyclonic anomalies in the region under focus. These circulation anomalies favor convergence of humidity and facilitate destabilization of the atmosphere, leading to consistent precipitation excess in central-east Brazil. It is worth pointing out that most of this region is above 600 m and a large part of it is a chain of mountains above 1000 m. A topographic lifting effect could also help in destabilizing the low troposphere.
7. Influence function analysis
Is the upper-level anticyclonic anomaly over central-east Brazil (which includes southeast) in January(+) due to enhanced convection there or might it be ascribed to remote influence of EN-related tropical heat sources, as is the anticyclonic anomaly over southeast Brazil in November(0)? To test this possibility, and see how the remote and local influences evolve from November to January, the influence functions of November, December, and January for the same target point in southeast Brazil were computed (Fig. 12). They show clearly how the differences between the atmospheric basic states affect the wave propagation. There is not much variation in the influence function from November to December, although there is an increase of local influence (near the target point) in December. There is, however, a significant change in January: the influence function weakens in the tropical belt and changes its sign. The local influence function strengthens even more than in December.
The influence functions for November and January are only alike over the southeast Pacific. Composite divergence anomalies for November(0) and January(+) of El Niño events (not shown) are generally strong and consistent only in the Tropics and subtropics. Therefore, strong influence functions in the extratropics do not mean that these regions will influence the mean rotational circulation at the target point during these events. The tropical upper-level divergence anomalies during November(0) and January(+) of EN events have several common features, being stronger in January. One of these features is the strong and consistent anomalous divergence in the tropical central-east Pacific, associated with the anomalous convection in this region during EN events.
The comparison of the EN-related upper-level divergence anomalies in November(0) with the influence function for November suggests that the rotational response at the target point to the remote anomalies is negative (anticyclonic circulation anomaly), because the strongest divergence anomalies and the influence function have predominantly opposite signs. Yet the response to the local anomalies of divergence is positive (cyclonic circulation anomaly), because the local upper-level divergence anomaly is negative and the local influence function is negative. As the response over the target point is anticyclonic, one can conclude that this response is predominantly remote.
On the other hand, the comparison between the EN-related upper-level divergence anomalies in January(+) and the influence function for January suggests that the rotational response at the target point to the remote anomalies is weak or predominantly positive (cyclonic circulation anomaly), because the strongest divergence anomalies are in regions with weak influence or have predominantly the same sign as the influence function. Yet the response to local anomalies of divergence is strong and negative (anticyclonic circulation anomaly), because the local upper-level divergence anomaly is positive and the local influence function is negative. As the response over the target point is anticyclonic, one can conclude that this response is predominantly local.
As the tropical upper-level divergence anomalies during November(0) and January(+) of EN events have several common features, and the influence functions in November and January have opposite signs in the tropical belt, it would be expected that the influence from the remote tropical heat sources in January be contrary to that in November.
Therefore, the observed upper-level anticyclonic circulation anomaly over southeast Brazil in January(+) may be a local response to anomalous convection, produced by an enhanced monsoon circulation there. The influence functions for a point a little to the north confirm these conclusions.
8. Relationship with sea surface temperature
The correlation between January precipitation averaged over 16 stations in an area in central-east Brazil and SST in November and January shows some interesting features (Fig. 13). While in both months there is an EN-related pattern of significant positive correlation in the central-east Pacific, the correlation in the Atlantic, off the southeast coast of Brazil, fluctuates on a timescale comparable to that of the atmospheric circulation and precipitation. The period of the correlation is 1956–83, because that was the common period to all the 16 stations within our period of analysis (1956–92). It was also calculated for 1932–83 (not shown) and the main features are the same.
In spring (see November), there is significant positive correlation off the southeast coast of Brazil, indicating warmer SST in spring connected to enhanced precipitation in central-east Brazil in January. The existence of above-normal SST in that region during spring(0) of EN events is indicated in Enfield and Mayer (1997) and Diaz et al. (1998). In January the correlation turns significantly negative. As a matter of fact, while the main patterns of correlation remain the same in the Pacific and North Atlantic, the sign of correlation is reversed in the South Atlantic, off the SA coast.
The timescale of the evolution of the correlation coefficients off the southeast coast of Brazil suggests that the EN-related anomalies of precipitation and circulation set up the warmer SSTs in spring and that the enhanced convection and rainfall in January lead to negative SST anomalies. One might conjecture that the warming of SST during spring is favored by the anticyclonic wind anomalies over that region, suppressed convection and lower rainfall (and therefore, increased shortwave radiation). The warmer SST may help trigger the enhanced convection in the region in January. On the other hand, the cooling of SST in January might be related to the enhanced convection and excess rainfall.
9. Summary and conclusions
The analysis of several meteorological fields suggests that the summer monsoon circulation and precipitation during EN events is altered both by large-scale perturbations associated with these events and by the anomalous surface heating in southeast Brazil during the spring.
The impact of EN events on summer climate in Brazil shows strong subseasonal variation, so that analysis based on seasonal means does not show a consistent picture, and smoothes out anomalies that are significant during part of the season.
In the early summer monsoon season, remotely produced atmospheric perturbations prevail over Brazil. Anticyclonic low-level anomalies predominate over central-east Brazil, in the Tropics and subtropics, due to the subsidence over the Amazon and to Rossby waves of equivalent barotropic structure in the subtropics. Easterly moisture inflow from the Atlantic is favored, but diverted toward northern SA and south Brazil. There are negative precipitation anomalies in north and central-east Brazil and positive ones in south Brazil. These precipitation anomalies are favored by the perturbation in the Walker and Hadley circulation over the east Pacific and SA, and by a Rossby wave train over southern SA originated in the east Pacific. In January, probably triggered by anomalous surface heating during the spring, there is an anomalous low-level convergence and a cyclonic anomaly over southeast Brazil. This anomalous circulation directs moisture flux toward central-east Brazil, causing moisture convergence in this region. Therefore, due to a thermodynamically favorable structure, there is enhanced precipitation over central-east Brazil, the dry anomalies in north Brazil are displaced northward, and the anomalies in south Brazil almost disappear. In February, after the above-normal precipitation of January, the surface temperature anomalies in the southeast turn negative, the low-level cyclonic anomalies disappear and the precipitation diminishes in central-east. There are negative rainfall anomalies in north Brazil and in the SACZ, and positive ones in south Brazil.
The abrupt change of EN-related anomalies in January explains why seasonal analysis of EN-related circulation and precipitation anomalies over Brazil actually do not show significant impact in the SACZ region, although this impact exists in a subseasonal scale.
The evolution of the thermodynamic structure of the troposphere over southeast Brazil is coherent with the circulation changes during the summer monsoon season. In November(0), the temperature increases significantly in southeast Brazil, due to circulation anomalies produced by EN and also to the relative dryness in this region during spring(0). In spite of some inflow of moisture into this region, the relative humidity is much lower than normal due to anomalous subsidence. In December(0), the subsidence is weaker, rainfall is closer to normal, and positive temperature anomalies are lower. In January(+), the troposphere turns much more humid and unstable and the temperature anomalies are lower, due to anomalous southerly advection and positive rainfall anomalies. In February(+), the troposphere gets more stable and the temperature anomalies are mainly negative, connected to the stronger precipitation in January(+).
Also the SST anomalies off the southeast coast of Brazil are coherent with the precipitation and circulation changes. While the SST–rainfall relationship remains stable in the central-east Pacific, it fluctuates on a timescale comparable to that of the atmospheric circulation and precipitation in the southeast Atlantic. This is an indication that the SST anomalies there are mostly a response to EN-related circulation and convection anomalies, although they may help to establish some of the observed circulation anomalies.
EN events impact precipitation by enhancing or suppressing the mechanisms that produce rainfall. In the case of the summer monsoon, the driving mechanism is the establishment of a continental heat low and a thermal contrast between the continent and ocean, that brings about circulation anomalies. These anomalies provide enhanced moisture convergence that leads to moistening and destabilization of the troposphere and, thus, to enhanced convection. It seems that the surface temperature anomalies brought about in the spring(0) of the EN events in the highlands of southeast Brazil set up the conditions for circulation anomalies that enhance convection in central-east Brazil during part of the monsoon season, especially January(+), although further tests of this hypothesis are needed. This is the only month within the summer season of EN events in which a monsoon-type circulation, associated with a regional anomalous heating, is enhanced. A topographic lifting effect in southeast Brazil may also be important in enhancing ascending motion, low-level convergence, and, therefore, cyclonic anomaly in this region. The more vigorous latent heat release can also enforce the cyclonic circulation in the lower troposphere over that region.
It is important to stress that the subseasonal variation of EN impact on the summer monsoon described in this paper is due neither to intraseasonal variability connected with the 30–60-day oscillation nor to intraseasonal oscillations with higher frequency (25 days, 15 days, etc.). Intraseasonal variability is present in Brazil during the summer (Nogués-Paegle and Mo 1997; Grimm et al. 2000b), but it is not phase locked to one particular month. Therefore, it is not able to produce the same consistent anomalies at a particular month in an average over several EN events.
It is not claimed that the phases of the subseasonal variation of the EN impact coincide exactly with the calendar months. However, the consistent anomalies show that the monthly analysis is able to separate the different phases and is more adequate than the seasonal analysis. Perhaps an analysis with higher temporal resolution would show even more clearly the most probable dates of the phase changes.
As summer is the rainy season in most of Brazil, the interannual variability due to EN events may have a significant impact on agriculture, hydroelectric power generation, civil defense, as well as on several other activities. Crops of several types of cultivation depend on summer rainfall. Deficient precipitation in this season may cause shortage of available electric power. Very densely populated areas, like São Paulo, Rio de Janeiro, and Minas Gerais, in southeast Brazil, are severely impacted by strong rainfall associated with an enhanced SACZ, in terms of urban floods and landslides. Therefore, a more accurate climate prediction for EN summers, with greater spatial resolution than a season, is very useful. Numerical models do not seem to have a satisfactory performance in seasonal climate prediction for most of Brazil during DJF, except for parts of the north and northeast regions (more information available online at http://iri.columbia.edu/forecast/climate/skill/SkillMap.html). Therefore, information on the consistent impact of EN events in this season, with greater than usual temporal resolution, may be used as complementary information for climate prediction. The information formerly available on the EN impact on summer precipitation in Brazil is based on seasonal analysis and on smaller datasets than those used in this study, which discloses some different results and different aspects. For instance, although there is small impact on the SACZ region on a seasonal basis, there is significant enhancement of convection around January and weakening in February. This effect has not been detected in previous studies. This work shows that the impact can reverse in intraseasonal timescales. The detailed joint analysis of precipitation, temperature, circulation, and thermodynamic structure is also a contribution of this study to the information already available on the EN impact. This joint analysis shows good coherence between the observed precipitation and temperature anomalies on the one hand, and the circulation anomalies given by the reanalysis data on the other hand.
Although this study indicates a robust signal, it is important to point out that the temporal evolution of anomalous EN conditions from November(0) to December(+) may be different from one EN to another.
Financial support for this research was provided by the Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) and the Inter-American Institute for Global Change Research (CRN-055). The assistance of Andreas Kiefer in preparing the figures is gratefully acknowledged.
Corresponding author address: Dr. Alice M. Grimm, Department of Physics, Federal University of Parana, Caixa Postal 19044, CEP 81531-990 Curitiba, Brazil. Email: firstname.lastname@example.org