principal modes of interannual and decadal variability of summer rainfall over south america

INTERNATIONAL JOURNAL OF CLIMATOLOGY

Int. J. Climatol. 21: 1623–1644 (2001)

DOI: 10.1002/joc.700

PRINCIPAL MODES OF INTERANNUAL AND DECADALVARIABILITY OF SUMMER RAINFALL OVER SOUTH AMERICA

JIAYU ZHOUa and K.-M. LAUb,*a Goddard Earth Sciences and Technology Center, Uni�ersity of Maryland, Baltimore, MD 21250, USA

b Code 913, Laboratory for Atmospheres, NASA/Goddard Space Flight Center, Greenbelt, MD 20771, USA

Recei�ed 2 October 2000Re�ised 6 June 2001

Accepted 7 June 2001

ABSTRACT

Using the Climate Prediction Center (CPC) Merged Analysis of Precipitation (CMAP) product together with theGoddard Earth Observing System (GEOS) reanalysis and the National Center for Environmental Prediction (NCEP)sea-surface temperature (SST) data, we have conducted a diagnostic study of the interannual and decadal scalevariability of principal modes of summer rainfall over South America for the period 1979–1995. By filtering theannual and short (�12 months) time-scale variations, results of empirical orthogonal function analysis show threeleading modes of rainfall variation identified with interannual, decadal and long-term variability. Together, thesemodes explain more than half the total variance of the filtered data.

The first mode is highly correlated with El Nino–Southern Oscillation (ENSO), showing a regional rainfallanomaly pattern largely consistent with previous results. This mode captures the summer season interannualvariability, not only the Northeast Brazil drought but also its connection with excessive rainfall over Southern Braziland the Ecuador coast in El Nino years. Another distinctive feature is the strengthening of the low-level flow alongthe eastern foothills of the eastern Andes, signifying an enhancement of the South American summer monsoon inresponse to an El Nino anomaly.

The decadal variation displays a meridional shift of the Inter-Tropical Convergence Zone (ITCZ), which is tied tothe anomalous cross-equatorial SST gradient over the Atlantic and the eastern Pacific. Associated with this mode isa large-scale mass swing between polar regions and the mid-latitudes. Over the South Atlantic and the South Pacific,the anomalous subtropical high and the associated anomalous surface wind are dynamically consistent with thedistribution of local SST anomalies, suggesting the importance of atmospheric forcing at the decadal time scale.

The long-term variation shows that since 1980 there has been a decrease of rainfall from the northwest coast to thesoutheast subtropical region and a southwards shift of the Atlantic ITCZ, leading to increased rainfall over northernand eastern Brazil. Possible links of this mode to the observed SST warming trend over the subtropical South Atlanticand to the interdecadal SST variation over the extratropical North Atlantic are discussed. Copyright © 2001 RoyalMeteorological Society.

KEY WORDS: decadal variability; EOF analysis; interannual variability; Lanczos filter; South America; summer monsoon; summerrainfall; SVD analysis

1. INTRODUCTION

Summertime drought and flood have had major social and economic impacts in South America. In apioneering study, Walker (1928) documented a remarkable coincidence of the anomalous warming of theeastern equatorial Pacific Ocean and drought in Northeast Brazil, where the peak of the annual rainfalloccurs during austral autumn. Since then a large number of studies have been carried out to understandthe mechanisms of droughts and floods over South America and to explore possible ways of predictingthe rainfall and sea surface temperature anomalies. A brief review of these studies will now be provided.

* Correspondence to: Code 913, Laboratory for Atmospheres, NASA/Goddard Space Flight Center, Greenbelt, MD 20771, USA;e-mail: [email protected]

Copyright © 2001 Royal Meteorological Society

J. ZHOU AND K.-M. LAU1624

There are two major hypotheses for the Nordeste drought: the shifts in the large-scale atmosphericcirculation caused by sea-surface temperature (SST) anomalies in the tropical Pacific versus the tropicalAtlantic. In a review paper, Kousky et al. (1984) suggested that in El Nino years, an anomalous eastwardsdisplacement of the Pacific Walker circulation occurs with upwards motion over anomalously warmwaters of the central and eastern Pacific and sinking motion over the eastern Amazon, Northeast Braziland the equatorial Atlantic, leading to drought conditions in Nordeste (defined as the northern part of theBrazilian states of Rio Grande do Norte, Ceara and Sao Luız). An alternative explanation was providedby Moura and Shukla (1981), who showed that the most severe Nordeste drought events are associatedwith a meridional SST dipole anomaly in the tropical Atlantic. They suggested that a possible mechanismis the intensification of the Inter-Tropical Convergence Zone (ITCZ) to the north of the equator, whichoccurs in association with local positive SST anomalies. The intensification leads to increased descendingmotion over the region of negative SST anomalies in the adjoining ocean, from which stems the rainfallreduction over Nordeste. Subsequently, a number of investigators (Houghton and Tourre, 1992; Enfieldand Mayer, 1997) demonstrated that SSTs in the regions north and south of the Atlantic ITCZ arestatistically independent at seasonal to interannual time scales. However, at a decadal time scale, Nobreand Shukla (1996) showed evidence confirming the existence of the dipole structure of SST anomalies inthe tropical Atlantic.

Rasmusson and Mo (1993) found from satellite observations that during the 1986–1987 ElNino–Southern Oscillation (ENSO) episode, anomalous equatorial convection extended northeastwardsfrom the central to eastern equatorial Pacific across Mexico into the Atlantic. The convection gave riseto a time-averaged anomalous divergent circulation whose primary subsiding branch extended fromnortheastern South America to northeastwards across the low-latitude North Atlantic. Their findingprovides new evidence of the presence of an anomalous local Hadley cell that could be responsible for theNordeste drought during El Nino years.

The regional circulation anomalies associated with ENSO were also investigated by a number ofresearchers. Aceituno (1988) showed that in austral summers of El Nino years, the near-equatorial troughshifts northwards, which leads to weakening of the northeast trades and enhancement of cross-equatorialflow from the Southern Hemisphere. Marengo (1992) found an opposite situation occurring in La Ninayears with a strong North Atlantic high, steep meridional pressure gradient on its equatorward side, andaccelerated northeast trades in the tropical North Atlantic. He suggested that the intensified northeasttrades might be associated with enhanced boundary layer moisture transport from the tropical NorthAtlantic into the Amazon basin.

Positive rainfall anomalies related to El Nino event have been found along the Ecuador–Peru coast(Hastenrath, 1978), and over the subtropical southeastern South American region, including southernBrazil, Uruguay and northeastern Argentina (Ropelewski and Halpert, 1987; Chu, 1991; Pisciottano et al.,1994; Diaz et al., 1998; Grimm et al., 1998). The former seems to be partially linked to an anomalouslysouthward-displaced near-equatorial trough over the eastern Pacific (Aceituno, 1988); while the later isjust about 15°–20° south of Nordeste and the regional precipitation is mostly concentrated in the australsummer (December–January–February; DJF), which is different from the timing of the peak rainfall overNortheast Brazil.

In addition to the influence of the tropical SST anomalies discussed above, remote influence from mid-to high-latitudes was reported by Namias (1972). He found that increased cyclonic activities in theNewfoundland area were associated with the abundant rainfall in Northeast Brazil. This implies arelationship with the North Atlantic Oscillation (NAO), which varies on the decadal time scale and isstrongest in boreal winter. Bojariu (1997) demonstrated a relationship between the low-frequencyvariability in the tropical Atlantic SST and the persistent NAO mode connected by the anomalous localHadley circulation.

In recent years the decadal and long-term variations have been found in historical hydroclimatologicalrecords of tropical South America. Dias de Paiva and Clarke (1995) revealed a negative trend ofprecipitation in two regions of western and central Amazonia and a positive trend in eastern Amazoniaover the period from the 1960s to the 1980s, during which deforestation was rapid. Wagner (1996)

Copyright © 2001 Royal Meteorological Society Int. J. Climatol. 21: 1623–1644 (2001)

SOUTH AMERICAN SUMMER RAINFALL VARIABILITY 1625

discovered a substantial warming trend in the surface waters of the tropical South Atlantic that waspositive correlated with summer rainfall in northeast Brazil. This result was also confirmed by Marengoet al. (1998), who showed a slow increase of rainfall in northeast Brazil in the historical record. Usingcentury-long station rainfall records over subtropical southeastern South America, Krepper and Sequeira(1998) demonstrated a positive trend with prominent decadal– interdecadal variations being most distinctin the summer season. A recent investigation by Robertson and Mechoso (1998, 2000) revealedinterdecadal signals in the southeastern South American streamflow records and showed the relationshipwith the long-term variation of the South Atlantic Convergence Zone (SACZ).

Due to lack of reliable rainfall data over the whole continental domain, most previous studies based onstation observations only resulted in some qualitative patterns with mixed information of the varioustime-scales of variation. To determine objectively principal modes of the interannual variability ofanomalous precipitation and the associated large-scale circulation over South America, Kousky andKayano (1994) conducted a combined empirical orthogonal function (EOF) analysis using satelliteobserved outgoing longwave radiation (OLR) and the National Meteorological Center analysis of uppertropospheric wind. They showed that the centres of action of the first and the second modes are locatedover the tropical, and the subtropical to mid-latitude regions, respectively. The ENSO-related OLRsignals, which are in common with previous findings, could be identified from both modes. Though theymade a scale-separation between the interannual and the intraseasonal variation, their results showed thatthe annual cycle of the third mode was so strong that it cannot be set apart from the second mode. Theyalso noticed the problem of including mountain areas of mid-latitudes in their analysis. The abnormalcold surface temperature of high elevation might also contribute to the negative OLR anomaly.

Until the late 1990s, most previous studies are focused on austral autumn— the rainy season ofnortheast South America, where the local rainfall characteristics are dominated by the annual cycle of theITCZ displacement over the western tropical Atlantic. Less attention was paid on the height of australsummer, when the South American summer monsoon (SASM) (Zhou and Lau, 1998; Fu et al., 1999;Kousky, 1999; Marengo et al., 2001; Nogues-Paegle and Berbery, 2000) prevails and produces annualmaximum precipitation over the subtropical continent. The impact of such a distinct climatologicalbackground between the two seasons on the characteristics of the regional rainfall variability has not beenaddressed. Besides, considerable submonthly and intraseasonal time-scale convective variations overtropical–subtropical South America were revealed by a number of investigators (Uvo et al., 1998;Liebmann et al., 1999; Zhou and Lau, 1999; Nogues-Paegle et al., 2000). Their results and the previousfinding of a strong annual cycle by Kousky and Kayano (1994) suggest the necessity of removing thesehigh frequency variability to increase the signal to noise ratio for studying interannual and longertime-scale variations.

In recent years the advancement of satellite global rainfall estimation and data assimilation has greatlyimproved the data quality especially over the areas with sparse observing stations. This provides us witha better opportunity to re-examine some of the aforementioned features. The objectives of this paper areto (1) identify distinctive features of the summer (DJF) rainfall regime over South America, (2) analysethe interannual and longer time-scale variabilities of principal modes of summer rainfall and (3) explorethe mechanism that is responsible for each mode. Section 2 introduces the data used in this study. Section3 delineates the rainfall regime in the summer monsoon season and contrasts it with the autumn regimeover South America and the nearby oceans. Section 4 shows the paper’s main results, which are composedof three parts describing the characteristics of three leading EOF modes of ENSO, the decadal and thelong-term variation, respectively. Linear regressions of the principal components (PC) with SST, andvarious atmospheric variables are performed to demonstrate the underlying mechanism. The conclusionsare presented in Section 5. This study is different from many previous studies, which focus on theMarch–May period for Northeast Brazil rainfall. We argue here that the South American summer rainfallregime defined for the period DJF is important to understand the rainfall variability over the subtropicalcontinent, and should deserve attention.



2. DATA AND PREPROCESSING

The rainfall product used in this study is the monthly mean CPC (Climate Prediction Center) MergedAnalysis of Precipitation (CMAP), which is constructed on a 2.5° latitude– longitude grid and covers a17-year period from January 1979 to December 1995 (Xie and Arkin, 1997). This dataset merged severalkinds of rainfall information sources, including gauge-based monthly analyses, satellite estimates andpredictions of the operational forecast model. It represents a state-of-the-art global precipitation productthat is well suited to studies of large-scale rainfall variability. To examine coherent variation of rainfallwith SST and circulation, the National Center for Environment Prediction (NCEP) SST analyses(Reynolds and Marsico, 1993) and the NASA/Goddard Earth Observing System (GEOS) reanalysis(Schubert et al., 1993) are used. The NCEP SST data are available on a 2° latitude– longitude grid froma blend of in situ observation, satellite data (using the observation of the Advanced Very High ResolutionRadiometer (AVHRR) instrument) and sea ice data. The GEOS reanalysis is carried out at a horizontalresolution of 2° latitude by 2.5° longitude and with 20 sigma levels in the vertical (top at 10 hPa). Itincorporates rawinsonde reports, satellite retrievals of geopotential thickness, cloud-motion winds, andaircraft, ship and rocketsonde reports. The GEOS reanalysis uses a fixed assimilation system with lowerboundary constrained by the NCEP SST and soil moisture derived by observed surface air temperatureand precipitation fields.

Madden and Jones (1999) demonstrated significant aliasing would occur from inadequate sampling (e.g.using time series of particular season) when studying low frequency variability. To relieve this problem itis necessary to filter high frequencies in all data before sampling (Bloomfield, 1976). For focusing oninterannual and longer time scale variability of summer rainfall, we remove the monthly mean climatologyfrom each monthly value and apply a low-pass Lanczos filter (details see Duchon 1979) of the form

wk=sin 2�fck

�k·sin �k/n

�k/n(k= −n, . . . , n)

to all-year data. In the formulation the cut-off frequency fc is set to 1/12 month−1 to remove thevariations of the time-scale less than or equal to twelve months and the number of weights n equals 18for using less weights but still keeping a rather sharp filter near the cut-off frequency (the more weightsused results in a sharper response profile). The DJF monthly filtered data are then used in the EOFanalysis. For consistency, the same procedure has also been applied to SST and the atmospheric data usedfor the analysis.

To ensure that the filtering does not bias the results, but rather increases the signal relative to the noise,and to confirm the relationships between rainfall and SST as revealed by EOF analysis, an analysis byway of singular value decomposition has been carried out on the DJF monthly unfiltered rainfall andSST. The results are discussed in Appendix A.

3. SUMMER RAINFALL REGIME

Before the 1970s, the knowledge of the mean state of South American climate was limited by sparse windand rainfall observations. The large-scale meteorological fields were filled by objective analysis (van Loon,1972; Schwerdtfeger, 1976). Large uncertainties exist in these analyses, because the reliability of the datais strongly dependent on the density of observations. Detailed studies of temporal and spatialcharacteristics of South American climate have become possible since satellite observations came intooperation (Aceituno, 1988; Horel et al., 1989). In recent years the data quality has been improvedsubstantially thanks to the advances of data assimilation schemes (Kousky and Ropelewski, 1997).

Figure 1 shows the migration (6 mm day−1 contour) of the climatological monthly mean precipitationover the South American continent and the tropical western Atlantic, where distinct characteristics can beclearly identified. Over the land, the heavy precipitation zone advances from the northern equatorialnorthwest towards the subtropical southeast from late spring to early summer. In January it reaches its



southernmost position. This rainfall movement is identified with the development of SASM, as recentlydocumented by Zhou and Lau (1998). We showed that from austral spring to summer, sea level pressureincreases over northwest Africa and decreases over Gran Chaco of South America. The low-level flow ofSASM originates from the northwestern African continent, travels along the northern equatorial AtlanticOcean, crosses the equator over the east of the tropical Andes and brings the rainy season to thesubtropical South America (Figure 2). In the peak of the monsoon season, the maximum rainfall axis tiltsnorthwest-southeastwards, coinciding with the SACZ, which is most active at this time of the year.

On the other hand, over the western North Atlantic Ocean the ITCZ moves southwards (restricted tothe north of 5°S) much slower than SASM over the South American continent (Figure 1(a)). In Decemberand January, the Atlantic ITCZ keeps almost 20° latitude apart from the SACZ at the east coast of Brazil.Since the ITCZ over the tropical ocean is broadly embedded in the near-equatorial trough of low pressure

Figure 1. CMAP climatology of monthly mean rainfall migration over the South American continent and the western tropicalAtlantic. The lines indicate meridional progression of the 6 mm day−1 contour during the period (a) from September to January

and (b) from January to May



Figure 2. Departure of DJF mean SLP (hPa) and the 850-hPa wind (mm day−1) from the corresponding annual mean for GEOS-1climatology. The light (heavy) shadings indicate the negative (positive) anomalies

and zone of warmest surface waters (Aceituno, 1988), the slower migration of ITCZ over the westernAtlantic could be related to the delayed response of SST to the solar heating caused by the large heatcapacity of the ocean water.

When the season changes from summer to autumn, the summer monsoon and the associated heavyprecipitation centre over subtropical South America retreat equatorwards, while the Atlantic ITCZcontinues moving towards the south (Figure 1(b)). In March and April, the two heavy convective rainfallbands join together over northeast coast of South America, leading to maximum yearly rainfall over theregion.

Figure 3(a) and (b) show the distributions of rainfall as a percentage of the annual total for DJF andMarch, April and May (MAM), respectively. In austral summer, the subtropical land receives about 50%of the local annual rainfall; while in the autumn, a large percentage of the local annual rainfallconcentrates on the northern part of Northeast Brazil and the equatorial Atlantic. Clearly, the tworainfall regimes are quite different. As stated in the introduction, in this paper we focus on the summerregime (DJF). During this period of time, the summer monsoon advances and retreats over most parts ofthe tropical–subtropical continent (from the northwest to the southeast), while the western Atlantic ITCZcontinuously moves towards Northeast Brazil. This rainfall regime should be distinguished from theautumn rainfall regime.

4. ANOMALY FEATURES

To assess the summer rainfall variability for timescales longer than 1 year, we first examine thedistribution of standard deviation of the low-pass filtered data for DJF months. It shows that the greatestvariability is located over the north-northeast coast of Brazil and the west coast of Ecuador andColombia. There are moderate variations over the subtropical regions of Bolivia and Southern Brazil.This total variance of rainfall is further decomposed by the EOF analysis for the domain of 90°–30°W



and 40°S–20°N. Results clearly show three leading modes, which are separated temporally intointerannual, decadal and long-term variations. They together explain more than half and individuallyexplain 27.2, 17.0 and 13.3% of the total variance of the low-pass filtered data. According to the EOFseparation criterion of North et al. (1982), the first and the third modes are well separated from thesecond and the fourth modes, respectively. The difference between the second and the third eigenvaluesis marginally comparable to the magnitude of their sampling errors, indicating some possible degeneracyin these modes. Nevertheless, we will show that these two modes are consistent with previous resultsderived from much longer historic records. The following analysis provides further insight into themechanism of the decadal and longer time-scale rainfall variability over tropical and subtropical SouthAmerica.

4.1. Interannual �ariability

As shown in Figure 4(a) and (c), the first mode of the rainfall EOF analysis is highly correlated withthe SST first EOF mode that reflects the ENSO variation. To facilitate comparison, the time series inFigure 4 has been normalized, such that the magnitude of each contour has the real unit and representsthe intensity of the event at one standard deviation. The spatial pattern of this mode (Figure 4(b)) showsthat the rainfall tends to be above normal over Uruguay–Southern Brazil and the west coast of Ecuadorand lower over northeast Brazil during El Nino years. These are in good agreement with previousobservational studies (Ropelewski and Halpert, 1987; Aceituno, 1988). The pattern of regression with SST(Figure 5(a)) shows a striking resemblance with that of the first mode of the independent SST EOFanalysis for the domain of 40°S–60°N (Figure 5(b)), not only in pattern but also in magnitude. In Figure5(a) shading depicts areas where the statistical significance of the correlation between rainfall PC1 andSST is at or above the 99% level. High correlation can be found in the areas where SST anomaly is largein a typical ENSO event. These clearly indicate that the interannual variability of the South Americansummer rainfall anomaly is an integral part of the global climate response to ENSO. Over the tropicalAtlantic, the moderate negative SST anomaly south of the equator is significantly correlated with the

Figure 3. CMAP climatology of seasonal rainfall concentration (percentage of the annual amount) for (a) DJF and (b) MAM,respectively



Figure 4. DJF rainfall of (a) PC1 and (b) EOF1 (mm day−1). (c) DJF SST PC1

Figure 5. (a) Regression of DJF rainfall PC1 with SST and (b) EOF1 of SST. The shaded areas indicate the correlation isstatistically significant at above 99% level. Unit: K



South American summer rainfall anomaly. There is no distinct positive correlation found over the tropicalNorth Atlantic, except for a small area confined to the west coast of Senegal. This agrees with Enfield andMayer (1997), who suggested the dipole structure of the tropical Atlantic SST anomaly is not statisticallysignificant on the interannual time scale.

To understand more completely the mechanism of the anomalous rainfall pattern over South America,linear regressions of PC1 with velocity potential and streamfunction are computed. The regression withvelocity potential shows the wave number-one pattern with two centres over the equatorial eastern andwestern Pacific, respectively, and in opposite phase between 850 and 200 hPa (Figure 6 for 200 hPa only).This pattern can be attributed to the eastwards shift of the convective heating induced by theENSO-related SST anomaly. In El Nino years, two upper-tropospheric velocity potential troughs withridges underneath can be clearly identified over the eastern Pacific, one extending northeastwards and theother southeastwards from the anomalous divergence centre over the eastern equatorial Pacific into thesubtropical North and South Atlantic, respectively. An anomalous ridge is found just above NortheastBrazil. The regression with 500-hPa vertical velocity, which is superimposed on Figure 6, shows risingmotion along the troughs and sinking motion over the ridge. The northern branch is consistent with thefinding of Rasmusson and Mo (1993), who suggest a direct link between the time-averaged ENSOprecipitation anomalies over the Northern Hemisphere subtropics and the South American–Atlantictropics. The southern branch is consistent with the finding of Aceituno (1988). Our results indicate thatENSO induced anomalous diabatic forcing over the tropical Atlantic sector is somewhat symmetric aboutthe equator. Combining the two fields of mid-tropospheric anomalous vertical velocity and anomalousdivergent wind, which follows the anomalous velocity potential gradient in both lower and upper levels;we obtain a clear picture of anomalous regional secondary circulations. Beside the east–west circulationanomaly over equatorial South America and the northern branch of the anomalous north–south

Figure 6. Regression of DJF rainfall PC1 with velocity potential (10−6 m2 s−1) at the 200-hPa (contour) and with omega verticalvelocity (hPa day−1) at 500-hPa (shading)



Figure 7. Regression of DJF rainfall PC1 (a) with the 850-hPa streamfunction (10−6 m2 s−1) and (b) with SLP (hPa) and the850-hPa wind (m s−1)

circulation between Northeast Brazil and the subtropical North Atlantic, the southern branch of themeridional secondary circulation anomaly, which was not reported in previous studies, rises overUruguay–Southern Brazil, enhancing the regional convective precipitation, and sinks over NortheastBrazil, reinforcing the subsidence induced by the other two secondary circulations and aggravating severedrought in the area.

The regression of PC1 with streamfunction exhibits the typical Pacific–Atlantic pattern, which issymmetric about the equator in the upper troposphere (not shown). At the low-level, a large cyclonic gyresituated on each side of the equator and flowing around the entire North or South Pacific basin (Figure7(a)). In the tropics, the upper and lower tropospheric circulation anomalies tend to be out of phase. The



low-level flow pattern over the tropics (between 30°S and 30°N) resembles the numerical solution of thedynamical response to the equatorial symmetric forcing illustrated by Gill (1980). The low-level cycloniccirculations are located at the northwest and the southwest flanks of the forcing region. They extend east-and polewards into the subtropical North and South Atlantic Oceans, coinciding with the two enhancedconvective bands associated with the velocity potential anomaly. In the extratropics, the signal is quitepronounced. Here, the dynamical response is mostly equivalent barotropic. Anomalous cooling of SST isconsistent with the lows over the Aleutians and southeast of New Zealand. Since the main forcing regionis over the Pacific, the circulation response over the Atlantic is likely to be induced by its Pacificcounterpart. At low-level they are anticyclonic on both sides of the equator and much weaker than thecyclonic circulation dominant over the Pacific. There are abnormal southwesterlies extending from theequatorial eastern Pacific to the tropical North Atlantic, where the northeasterly trade winds areweakened that is consistent with the observed warming tendency of SST. Over the tropical South Atlantic,anomalous southeasterlies superimpose on the southeasterly trade winds resulting in extraordinary SSTcooling.

Figure 7(b) shows the regression of rainfall PC1 with sea level pressure (SLP) and 850-hPa wind overthe square area indicated in Figure 7(a). As a result of mass redistribution during El Nino, the northwestAfrican high and the subtropical South Atlantic high are enhanced, and the subtropical highs over theSouth Pacific and the western North Atlantic are weakened. Consequently, the flow around the outskirtsof the tropical South Atlantic high and the summer monsoon flow (see Figure 2) over the equatorialAtlantic and along the eastern foothill of the Andes are substantially enhanced. These two branches ofanomalous low-level wind diverge over northeastern Brazil and converge over southern Brazil. Theabnormally strong low-level jet (LLJ) east of the Andes enhances moisture transport from the Amazonbasin, contributing anomalous wet condition over subtropical southeastern South America. Over the westcoast of Ecuador, anomalous convergence is encountered due to an above normal westerly from thedisplaced Walker circulation over the eastern equatorial Pacific and the increased easterly of enhancedsummer monsoon circulation along the equatorial Amazon Basin. The aforementioned circulationchanges imply abnormal low-level moisture convergence, which is consistent with the anomalous rainfalldistribution over South America noted in Figure 4(b).

The positive correlation between the ENSO and SASM anomaly has been confirmed in the most recentENSO event. Lau and Zhou (1999) showed substantial mid-tropospheric warming expanding from theNino 3 region of the central–eastern Pacific to the Altiplano Plateau in the austral summer of 1997–1998.Accompanied with that, the Bolivian high is much stronger than normal. The South Atlantic subtropicalhigh is enhanced and expands westwards into eastern Brazil. As a result, a strong singleupper-tropospheric jet over the subtropical continent and an intense NW LLJ along the eastern foothillof the tropical–subtropical Andes are generated. The latter was also documented by Douglas (1999), whodemonstrated a strong LLJ at Santa Cruz, Bolivia during a special pilot balloon observation period.These features characterize a strong SASM. The situation is completely reversed in 1998–1999, suggestinga weakening of the SASM.

4.2. Decadal �ariation

As shown in Figure 8(a) and (b), the second mode varies at the decadal time scale. The spatialdistribution of this mode displays a meridional shift of ITCZ over both sides of the South Americancontinent with more significant signal over the equatorial Atlantic side. During most of the 1980s, PC2remained negative, indicating above normal rainfall over northeastern Brazil and from the central Andesto Gran Chaco, and below normal rainfall north of the equator and over the southern Amazon Basin.The situation seems to reverse during the late 1980s and early 1990s.

Since the movement of the ITCZ over oceans mainly follows the progressing of the equatorial SSTridge, the relation between the principal component of this mode and SST is investigated. Figure 9 showsthe regression pattern of PC2 with SST and the 1000-hPa wind. A series of anomalous meridional SSThighs and lows are noted. These SST anomalies are roughly zonally symmetric about the South American



continent and meridionally span the Southern Hemisphere mid-latitudes to northern tropical oceans. Thepattern is highly significant south of 20°N over the Atlantic, where the cross-equatorial dipole structureand the phase variations agree with Nobre and Shukla (1996), who used longer time datasets and did acombined EOF analysis of SST and surface wind stress. They showed a similar pattern and in-phase

Figure 8. DJF rainfall (a) PC2 and (b) EOF2 (mm day−1). (c) DJF SLP (hPa) area averaged over 90°E–90°W and 55°–35°S

Figure 9. Regression of DJF rainfall PC2 with SST (K) and with the 1000-hPa wind (m s−1). The shaded areas indicate thecorrelation is statistically significant at above 95% level



Figure 10. (a) Meridional SST difference between the north (35°–20°W, 8°–20°N average) and the south (25°–5°W, 20°–8°Saverage) tropical Atlantic. (b) The same as (a) except for the difference between the north (110°–90°W, 6°–14°N) and the south(100°–80°W, 16°–8°S) tropical eastern Pacific. A three-point smoothing is applied (weights: 0.5, 1.0, and 0.5). The solid and dashlines indicate the inclusion and exclusion of the ENSO mode (represented by SST EOF1), respectively. The shadings indicate El

Nino years. Unit: K

decadal variations in the common period. Over the eastern Pacific, distinct temperature anomalies areconfined to the area close to the South American continent between 40° and 5°S. Though the SSTanomaly north of the equator in the Pacific is not significant, due to a large ENSO variability, a weakdipole pattern can still be identified across the equator. According to Hastenrath and Heller (1977) andNobre and Shukla (1996), it is the cross-equatorial SST gradient, rather than SST alone over either thetropical North or South Atlantic, which plays an important role in determining the displacement of theITCZ and hence governs the rainfall variability over Northeast Brazil. Figure 10(a) and (b) display themean SST differences between the northern and the southern centre areas over the tropical Atlantic andthe tropical eastern Pacific, respectively. On both sides of the continent the cross-equatorial SST gradientshows coherent decadal variations. Their correlations with PC2 of the rainfall anomaly are alsostatistically significant. A similar calculation was also done by Servain (1991), who used a longer timeseries and showed distinct decadal variation of SST difference between the north and the south tropicalAtlantic. The phase variations between his and our results are in agreement after December 1980. TheENSO influence is also assessed by removing the first SST EOF mode (described by the dashed lines). Itdemonstrates that the decadal variation of the cross-equatorial SST gradient over the tropical Atlantic



and the tropical eastern Pacific is not sensitive to the ENSO events. In El Nino years it shows only a slightenhancement (weakening) of the cross-equatorial SST gradient over the tropical Atlantic (the tropicaleastern Pacific), which is consistent with the local SST anomalies in EOF1.

As an additional consideration, since the South American continent separates the two oceans and theatmosphere is continuous across the continent, the rather zonally symmetric SST anomaly across thecontinent implies atmospheric forcing. Additionally, the stronger anomalous signals are found inmid-latitudes and not in the tropics, suggesting the source of perturbation could be located at higherlatitudes. Figure 11 shows the polar-stereographic view of regression pattern of SLP with the rainfall PC2.Obviously, the impact is global. We can see that a significant amount of mass is moved from Polarregions to the extratropics during the positive phase, creating an anomalous extratropical high beltaround the polar low. As a result, in the Northern Hemisphere, the climatological Aleutian low isweakened. The combination of anomalous high and low pressure centers over the North Atlantic sectorresembles the pattern of the NAO. This pattern is consistent with that reported by Watanabe and Nitta(1999), who demonstrated a very similar SLP pattern and the corresponding atmospheric structure ofdecadal change in the Northern Hemisphere winter of 1989 (see their Figure 1). In the SouthernHemisphere, the subtropical highs over the South Atlantic and the eastern South Pacific are enhanced andexpand polewards. The SST warming over the belt of 40°–30°S is consistent with the overlying anomaloussurface wind (Figure 9), which follows the contours of anomalous SLP anticyclonically and produceswind drift of warm water from lower latitudes. In the tropics of the South Atlantic between 20°S and theequator, the stronger than normal southeasterly trade wind at the northern fringe of the anomalous highsystem increases the sensible and latent heat loss from the ocean surface, hence lowering the local SST.In the tropical eastern South Pacific, the southerlies along the west coast of South America induceanomalous upwelling, creating a negative SST anomaly near the continent. Consequently, the meridionalSST gradient increases and the cross-equatorial wind intensifies on both sides of the continent. North ofthe equator the anomalous wind veers to the northeast, compelled by the Coriolis force. It weakens thenortheasterly trades; hence reduces the heat loss from the ocean surface, resulting in a positive SSTanomaly. The temporal variation of SLP averaged over the extratropical high belt is plotted underneaththe rainfall PC2 in Figure 8(c). The high correlation between the two time series is self-evident.

4.3. Long-term �ariation

The third PC (Figure 12(a)) shows a clear long-term variation from the 1980s to the early 1990s. Thespatial pattern (Figure 12(b)) shows an increase of precipitation over Northeast Brazil and the northwestcoast of Ecuador–Colombia, and a decrease of rainfall over the surrounding areas from the northequatorial Atlantic to subtropical South America. Over the equatorial western Atlantic, the pattern showsa southwards shift of the ITCZ. The magnitude of the anomaly over the north is more distinct than thatover the south.

This mode exhibits close relations with long-term SST variabilities in the subtropical and theextratropical Atlantic Ocean. Figure 13 shows the result of the regression of PC3 with SST and the1000-hPa wind. It demonstrates that significant correlations (shaded) are mostly found polewards of 20°in both the South and the North Atlantic. Over the South Atlantic a distinct warming band across thebasin at about 25°S coincides with the subtropical western South Atlantic warm tongue. Over the NorthAtlantic there is a significant dipole pattern with warming east of the USA and cooling over the northeastof Newfoundland. The 1000-hPa wind anomalies show anticyclonic (cyclonic) circulation over the warm(cold) area, which is dynamically consistent with the SST anomalies.

Though the data we used only cover a short period of time, the above findings are supported by manyobservational studies, which employed the data covered by longer time periods. Wagner (1996) usedsurface and subsurface ship observations from 1951 to 1990 and demonstrated a trend of interhemisphericSST gradient resulted from a pronounced warming of the South Atlantic SST centred at 20°–30°S in theaustral summer season. He found a positive correlation between this SST trend and the precipitationanomalies over Northeast Brazil and attributed that to the southward displacement of the lower



tropospheric wind confluence zone. Deser and Blackmon (1993) showed the North Atlantic dipolepattern, which is located near the interface of the Gulf Stream and the Labrador Current, exhibitsmulti-time scale variabilities, including the interdecadal time variation. There are many discussions on theinherent dynamics of the ocean–atmosphere coupled interdecadal variability (see recent review paper byLatif, 1998). Though Wagner (1996) gave a plausible mechanism for the local rainfall increase over

Figure 11. Regression of DJF rainfall PC2 with SLP for (a) Southern Hemisphere and (b) Northern Hemisphere. Unit: hPa. Theareas with light (heavy) shading indicate negative (positive) anomalies



Figure 12. DJF rainfall of (a) PC3 and (b) EOF3 (mm day−1)

Northeast Brazil, our result reveals a broader connection with this mode. Beside the correlation of thismode with the interdecadal Atlantic SST variation, Dias de Paiva and Clarke (1995) showed a negativetrend of rainfall over western and central Amazonia, which is in qualitative agreement with the resultpredicted by the UK Hadley Centre for Prediction and Research following 50% deforestation. In addition,Krepper and Sequeira (1998) showed a positive trend of precipitation over subtropical southeastern SouthAmerica starting from the middle of the last century with distinct interdecadal variations superimposed onit. From their result we can see the decline after 1980 is part of the interdecadal fluctuation. Due to theshort data record, we cannot provide a definitive explanation for the mechanism of the long-term summerrainfall variation.



To appreciate better the complex variability of the South American summer monsoon rainfall, Figure14 shows the time series of the area mean observed rainfall and the reconstruction by the three leadingmodes over northern Nordeste and the equatorial North Atlantic, where the rainfall anomalies reflect theAtlantic ITCZ variability. Overall, the reconstructions mimic the variations of the observed rainfall timeseries. The individual time series shows that the interannual mode varies in-phase and the other twomodes out-of-phase north and south of the equator. The interannual mode dominates the anomalous drycondition in El Nino years. However, the decadal mode has larger impact in La Nina years. We can seerainfall is significantly enhanced over northern Nordeste in the 1984–1985 case, due to the interannualand decadal modes, and over the equatorial North Atlantic in the 1988–1989 case, due to the decadalmode. The modification by the long-term variation is small over northern Nordeste but has comparableamplitude with other leading modes over the tropical North Atlantic.

5. CONCLUSIONS

In this study, we have identified the principal modes of the summer rainfall regime over South America.We demonstrate that the tropical western Atlantic ITCZ and the SASM are two major rainfall-producingsystems during summertime. The former mostly affects northeastern Brazil and the later prevails over thesubtropical regions of South America. Our results are focused on the interannual and longer time scalevariability of South American summer rainfall, using the monthly mean CMAP rainfall product alongwith GEOS reanalysis and the NCEP SST dataset. To highlight the low frequency signals, we subtract theannual cycle and filter all subannual scale variations prior to sampling and analysing the data. The resultsare summarized as follows.

South American summer (DJF) rainfall variability is dominated by three modes with distinctivetime-scales, i.e. interannual, decadal and long-term variations. Each mode is closely related to SSTvariation of the corresponding time scale, implying the significance of the atmospheric–ocean coupling tothe variability of the hydrological cycle.

The first mode shows the ENSO influence dominating the South American summer rainfall variabilityat the interannual time scale. Consistent with previous findings, the pattern of EOF1 shows a coherent

Figure 13. The same as Figure 12 except for PC3



Figure 14. Area averaged rainfall (solid) and the reconstruction by the leading three PCs (dash) over (a) northern Nordeste(50°–35°W, 10°–2.5°S) and (b) equatorial North Atlantic (60°–30°W, 2.5°–7.5°N). (c) and (d) are the leading three PCs for the

reconstruction of (a) and (b), respectively. Unit: mm day−1

picture featuring a rainfall increase over the Ecuador–Northern Peru coastal region and theUruguay–Southern Brazil area, and decrease over Northeast Brazil in El Nino years. Our diagnosticanalysis reveals that the anomalous descending motion over Northeast Brazil is not only due to theeastwards shift of Walker circulation, but also caused by anomalous local Hadley cells, which aredeveloped on both sides of the equator. The southern branch is also linked to the anomalous ascendingmotion over southern Brazil. As a result of global mass redistribution and dynamical adjustment, thesubtropical high is enhanced over the South Atlantic and northwestern Africa but weakened over theeastern South Pacific and the western North Atlantic. The South American summer monsoon isstrengthened as manifested in the increased low-level flow over the equatorial Atlantic and along theeastern foothill of the Andes, which enhances moisture transport from Amazon to subtropical SouthAmerica. Over the west coast of Ecuador, the anomalous monsoon easterlies meet with abnormalwesterlies from the displaced Walker cell, creating large convergence and a positive rainfall anomaly.



The second mode shows distinct decadal variation with a pattern of meridional shift of ITCZ on bothsides of the South American continent. The close relationship of this mode with the decadal change of thecross equatorial SST gradient is demonstrated. On the Atlantic side the signal is more significant andconsistent with previous studies (Hastenrath and Heller, 1977; Servain, 1991; Nobre and Shukla, 1996).Associated with this mode, the atmospheric forcing from mid- to high-latitudes is also revealed. TheSouthern Hemisphere subtropical highs are strengthened (weakened) and expanded polewards(equatorwards) in the positive (negative) phase. This mode may also involve a distinct decadal mass swingbetween the polar region and the extratropics in both hemispheres. The nearly zonally symmetricdistribution of the anomaly sign in SLP field is consistent with many previous studies (Rogers and vanLoon, 1982; Thompson and Wallace, 1998; Watanabe and Nitta, 1999; Gong and Wang, 1999),suggesting the possible influence of decadal variability of annual modes on the South American summerrainfall anomaly.

The temporal variation of the third mode is characterized by a long-term variation from the 1980s tothe early 1990s, during which rainfall increases over the west coast of Ecuador–Colombia and tropicaleastern Brazil, and decreases over the surrounding subtropical continent and the equatorial NorthAtlantic Ocean. The southward displacement of the Atlantic ITCZ has been correlated to the significantSST warming trend in the subtropical South Atlantic (Wagner, 1996) and remotely connected with theextratropical North Atlantic SST interdecadal variation (Deser and Blackmon, 1993). The apparentdecline of rainfall over subtropical Southeastern South America since 1980 noted here could be part of theinterdecadal variation superimposed on a long-term positive trend noted by Krepper and Sequeira (1998).

In summary, we conclude that South American summer rainfall is characterized by distinct multi-timescale variations, which include interannual, decadal and possible interdecadal variation. To gain a fullerunderstanding of climate variability over South America, all three scales and their possible interactionsneed to be considered. In particular the South American summer monsoon and the Atlantic ITCZ duringDecember–February are distinct climate subsystems that have strong regional impacts. These subsystemsmay foreshadow the development of major drought and floods during the rainy season (March–May) ofNortheast Brazil. Undoubtedly, the short data record will have an impact on the results presented here.This is particularly a problem with the second and third modes. Hence, the results should not be extendedto imply validity beyond the data periods (1979–1995). Further work with data over a longer period isrequired to substantiate the results presented in this paper.

ACKNOWLEDGEMENTS

We thank Drs Pingping Xie and Phillip A. Arkin for providing us the CMAP dataset used in this study.Special thanks also goes to the editor and an anonymous reviewer for their constructive comments, whichimproved this manuscript. This research is supported by the Earth Observing System/InterdisciplinaryScience investigation on hydrological processes and climate, and the Global Modeling and AnalysisProgram of the NASA, Mission to Planet Earth Office.

APPENDIX A

Due to the large amplitude of annual, semi-annual and intraseasonal convective variations found overtropical South America, which could conceal the signals of interannual and longer time-scale variations,we subtracted the monthly mean climatology and used the Lanczos filter to remove these variations tohighlight the signals. Mathematically, the Lanczos filter has a distinct advantage in its sharp spectralresponse near the cut-off frequency and no phase shift (Duchon, 1979; Kousky and Kayano, 1994). Sincethe calculation also uses information from neighbouring seasons, it is possible that the DJF signals couldbe contaminated. In this appendix, the robustness of the EOF modes and the relationships with SST,which are described in the main text, are examined by applying the singular value decomposition (SVD)analysis for the unfiltered DJF SST (40°S–60°N, 180°W–20°E) and the rainfall data with only removal




Fig

ure

A1.

DJF

rain

fall

(low

er)

and

SST

(upp

er)

SVD

prin

cipa

lco

mpo

nent

sof

PC

1(l

eft)

,P

C2

(mid

dle)

and

PC

3(r

ight

),re

spec

tive

ly


of the annual cycle. The SVD analysis identifies the modes that explain the co-variability between the twovariables (Bretherton et al., 1992).

The principal components of SVD analysis are shown in Figure A1. To save space, the correspondingrainfall and SST patterns have been omitted. Comparing the leading three SVD PCs with thecorresponding EOF PCs (Figures 4(a), 8(a) and 12(a)), we can see similar, albeit noisier temporal signals,indicating that the results are independent of using the filter. The first SVD mode, explaining 54.6% of thetotal co-variance, is dominated by the ENSO influence. The associated rainfall and the SST patterns arevery similar to the patterns from corresponding EOF analysis. The second and the third modes, whichexplain 12.2 and 7.8% of the total co-variance, respectively, also capture the decadal and the long-termvariations. Their rainfall spatial patterns show overall similarities between the two analyses. Differencescan mainly be seen in the third mode, which switches the anomaly sign from negative to positive insummer 1989–1990, 4 years later than that shown by the rainfall EOF PC3 (Figure 12(a)). Distinction canalso be found between the SST pattern of the third SVD mode and the corresponding pattern ofregression with rainfall EOF PC3 in the tropical oceans. Further examination shows some mixing of theinterannual and interdecadal signals in SVD PC3 and suggests the filtered data provide a much more clearseparation.

REFERENCES

Aceituno P. 1988. On the functioning of the Southern Oscillation in the South American sector. Part I: Surface climate. MonthlyWeather Re�iew 116: 505–524.

Bloomfield P. 1976. Fourier Analysis of Time Series: An Introduction. Wiley: New York.Bojariu R. 1997. Climate variability modes due to ocean–atmosphere interaction in the central Atlantic. Tellus 49A: 362–370.Bretherton CS, Smith C, Wallace JM. 1992. An intercomparison of methods for finding coupled patterns in climate data. Journal

of Climate 5: 541–560.Chu P-S. 1991. Brazil’s climate anomalies and ENSO. In Teleconnections Linking Worldwide Climate Anomalies, Glantz MH, Katz

RW, Nicholls N (eds). Cambridge University Press: Cambridge; 43–71.Deser C, Blackmon ML. 1993. Surface climate variations over the North Atlantic Ocean during winter: 1900–1989. Journal of

Climate 6: 1743–1753.Dias de Paiva EMC, Clarke RT. 1995. Time trends in rainfall records in Amazonia. Bulletin of the American Meteorological Society

76: 2203–2209.Diaz AF, Studzinski CD, Mechoso CR. 1998. Relationships between precipitation anomalies in Uruguay and southern Brazil and

sea surface temperature in the Pacific and Atlantic oceans. Journal of Climate 11: 251–271.Douglas M. 1999. The low-level jet at santa Cruz, Bolivia during January–March 1998 pilot balloon observations and model

comparisons. In AMS Preprints, 10th Symposium on Global Change Studies, 10–15 January 1999, Dallas, TX; 223–226.Duchon CE. 1979. Lanczos Filtering in one and two dimensions. Journal of Applied Meteorology 18: 1016–1022.Enfield DB, Mayer DA. 1997. Tropical Atlantic sea surface temperature variability and its relation to El Nino–Southern Oscillation.

Journal of Geophysical Research 102: 929–945.Fu R, Zhu B, Dickinson RE. 1999. How do atmosphere and land surface influence seasonal changes of convection in the tropical

Amazon? Journal of Climate 12: 1306–1321.Gill AE. 1980. Some simple solutions for heat-induced tropical circulation. Quarterly Journal of the Royal Meteorological Society

106: 447–462.Gong D, Wang S. 1999. Definition of Antarctic Oscillation index. Geophysical Research Letters 26: 459–462.Grimm AM, Ferraz SET, Gomes J. 1998. Precipitation anomalies in southern Brazil associated with El Nino and La Nina events.

Journal of Climate 11: 2863–2880.Hastenrath S. 1978. On modes of tropical circulation and climate anomalies. Journal of Atmospheric Science 35: 2222–2231.Hastenrath S, Heller L. 1977. Dynamics of climatic hazards in Northeast Brazil. Quarterly Journal of the Royal Meteorological

Society 103: 77–92.Horel JD, Hahmann AN, Geisler JE. 1989. An investigation of the annual cycle of convective activity over the tropical Americas.

Journal of Climate 2: 1388–1403.Houghton RW, Tourre YM. 1992. Characteristics of low-frequency sea surface temperature fluctuations in the tropical Atlantic.

Journal of Climate 5: 765–771.Kousky VE. 1999. The South American Monsoon system. In AMS Preprints, 10th Symposium on Global Change Studies, 10–15

January 1999, Dallas, TX; 215–218.Kousky VE, Kayano MT. 1994. Principal modes of outgoing longwave radiation and 250-mb circulation for the South American

sector. Journal of Climate 7: 1131–1143.Kousky VE, Ropelewski CF. 1997. The tropospheric seasonally varying mean climate over the Western Hemisphere (1979–1995).

NCEP/Climate Prediction Center Atlas, No. 3.Kousky VE, Kagano MT, Cavalcanti IFA. 1984. A review of the Southern Oscillation: oceanic-atmospheric circulation changes and

related rainfall anomalies. Tellus 36A: 490–504.Krepper CM, Sequeira ME. 1998. Low frequency variability of rainfall in southeastern South America. Theoretical and Applied

Climatology 61: 19–28.



Latif M. 1998. Dynamics of interdecadal variability in coupled ocean-atmosphere models. Journal of Climate 11: 602–624.Lau WK-M, Zhou J. 1999. South American summer monsoon of 1997/98 and 1998/99. In Proceedings of the 24th Annual Climate

Diagnostics and Prediction Workshop, PB00-106012, US Climate Prediction Center (ed.). US Department of Commerce:Washington, DC; 17–20.

Liebmann B, Kiladis GN, Marengo JA, Ambrizzi T, Glick JD. 1999. Submonthly convective variability over South America andthe South Atlantic convergence zone. Journal of Climate 12: 1877–1891.

Madden RA, Jones RH. 1999. Aliasing in the spectrum of time-averaged data. In AMS preprints. 8th Conference on ClimateVariations, 13–17 September 1999, Denver, CO; 229–231.

Marengo JA. 1992. Interannual variability of surface climate in the Amazon basin. International Journal of Climatology 12: 853–863.Marengo JA, Tomasella J, Uvo CR. 1998. Trends in streamflow and rainfall in tropical South America: Amazonia, eastern Brazil,

and northwestern Peru. Journal of Geophysical Research 103: 1775–1783.Marengo JA, Liebmann B, Kousky V, Filizola NP, Wainer IC. 2001. Onset and end of the rainy season in the Brazilian Amazon

Basin. Journal of Climate 14: 833–852.Moura AD, Shukla J. 1981. On the dynamics of droughts in Northeast Brazil: Observations, theory and numerical experiments with

a general circulation model. Journal of Atmospheric Science 38: 2653–2675.Namias J. 1972. Influence of Northern Hemisphere general circulation on drought in Northeast Brazil. Tellus 24: 336–343.Nobre P, Shukla J. 1996. Variations of sea surface temperature, wind stress, and rainfall over the tropical Atlantic and South

America. Journal of Climate 9: 2464–2479.Nogues-Paegle J, Berbery EH. 2000. Low-level jets over the Americas. CLIVAR Exchanges 5: 6–8.Nogues-Paegle J, Byerle LA, Mo KC. 2000. Intraseasonal modulation of South American summer precipitation. Monthly Weather

Re�iew 128: 837–850.North GR, Bell TL, Cahalan RF. 1982. Sampling errors in the estimation of empirical orthogonal functions. Monthly Weather

Re�iew 110: 699–706.Pisciottano G, Diaz A, Cazes G, Mechoso CR. 1994. El Nino–Southern Oscillation impact on rainfall in Uruguay. Journal of

Climate 7: 1286–1302.Rasmusson EM, Mo KC. 1993. Linkages between 200-mb tropical and extratropical circulation anomalies during the 1986-1989

ENSO cycle. Journal of Climate 6: 595–616.Reynolds RW, Marsico DC. 1993. An improved real-time global sea surface temperature analysis. Journal of Climate 6: 114–119.Robertson AW, Mechoso CR. 1998. Interannual and decadal cycles in river flows of southeastern South America. Journal of Climate

11: 2570–2581.Robertson AW, Mechoso CR. 2000. Interannual and decadal variability of the South Atlantic convergence zone. Monthly Weather

Re�iew 128: 2947–2957.Rogers JC, van Loon H. 1982. Spatial variability of sea level pressure and 500 mb height anomalies over the Southern Hemisphere.

Monthly Weather Re�iew 110: 1375–1392.Ropelewski CF, Halpert MS. 1987. Global and regional scale precipitation patterns associated with the El Nino/Southern

Oscillation. Monthly Weather Re�iew 115: 1606–1626.Schubert SD, Pfaendtner J, Rood R. 1993. An assimilated data set for earth science applications. Bulletin of the American

Meteorological Society 74: 2331–2342.Schwerdtfeger W. 1976. Introduction. In Climates of Central and South America, World Sur�ey of Climatology, vol. 12,

Schwerdtfeger W (ed.). Elsevier: Amsterdam; 1–12.Servain J. 1991. Simple climate indices for the tropical Atlantic Ocean and some applications. Journal of Geophysical Research 96:

15137–15146.Thompson DWJ, Wallace JM. 1998. The Arctic Oscillation signature in the wintertime geopotential height and temperature fields.

Geophysical Research Letters 25: 1297–1300.Uvo CB, Repelli CA, Zebiak SE, Kushnir Y. 1998. The relationship between tropical Pacific and Atlantic SST and Northeast Brazil

monthly precipitation. Journal of Climate 11: 551–562.van Loon H. 1972. Wind in the Southern Hemisphere. In Meteorology of the Southern Hemisphere, Newton CW (ed.). American

Meteorological Society: Washington, DC; 87–100.Wagner RG. 1996. Decadal-scale trends in mechanisms controlling meridional sea surface temperature gradients in the tropical

Atlantic. Journal of Geophysical Research 101: 16683–16694.Walker GT. 1928. Ceara (Brazil) famines and the general air movement. Beitrage zur Physik der freien Atmosphare 14: 88–93.Watanabe M, Nitta T. 1999. Decadal changes in the atmospheric circulation and associated surface variations in the Northern

Hemisphere winter. Journal of Climate 12: 494–510.Xie P, Arkin PA. 1997. Global Precipitation: A 17-year monthly analysis based on gauge observations, satellite estimates, and

numerical model outputs. Bulletin of the American Meteorological Society 78: 2539–2558.Zhou J, Lau K-M. 1998. Does a monsoon climate exist over South America? Journal of Climate 11: 1020–1040.Zhou J, Lau WK-M. 1999. Summertime intraseasonal variability over South America. In Proceedings of the 24th Annual Climate

Diagnostics and Prediction Workshop, PB00-106012, US Climate Prediction Center (ed.). US Department of Commerce:Washington, DC; 299–302.


principal modes of interannual and decadal variability of summer rainfall over south america

Documents