The sea surface temperature (SST) fields in the eastern equatorial sector of the Pacific and Atlantic Oceans oscillate in an interannual time scale between two extreme conditions characterized by anomalous warming and cooling of the surface waters. These extreme conditions are reinforced through a positive ocean–atmosphere feedback mechanism proposed by Bjerknes (1969). In each basin, an anomalous warming in the eastern equatorial side alters the climatological atmospheric Walker circulation by weakening the trade winds along the equatorial band. The weak trade winds yield weak upwelling, which favour the warming in the eastern equatorial side. The anomalous warm conditions in the eastern equatorial side of the Pacific and Atlantic basins are referred to as the Pacific El Niño and Atlantic El Niño (e.g. Zebiak, 1993, Neelin et al., 1998, Wang and Picaut, 2004), respectively. The opposite conditions in these basins are referred to as the Pacific La Niña and Atlantic La Niña. The anomalous oceanic conditions in the tropical Pacific are also linked dynamically to a near global-scale anomalous standing wave in the sea level pressure (SLP) with action centres over Indonesia and the southeastern Pacific (e.g. Rasmusson and Carpenter, 1982). The coupled ocean–atmosphere system in the Pacific is known as the El Niño-South Oscillation (ENSO). On the other hand, the interannual Atlantic mode is referred to as the Atlantic equatorial mode (AEM). Hereinafter, the terms El Niño and La Niña refer to the Pacific events, and the terms warm AEM and cold AEM refer to the Atlantic events.
Handoh et al. (2006b), using the SST data for the 1948–1999 period, analysed the influence of the ENSO on the AEM. They found that some AEM events are ENSO dependent, and others ENSO independent. Those which are ENSO dependent show a positive linkage between ENSO- and AEM-related SST anomalies and occur preferably in the austral summer. Their explanation for this positive connection involves the atmospheric Rossby wave train emanating from the tropical Pacific and travelling in the Southern Hemisphere across southern South America, and changes in the Walker circulation. They also showed that the ENSO-independent AEM events occur preferably from April to July.
From another optic, Lu and Dong (2005) using an atmospheric general circulation model investigate the role of the Atlantic SST anomalies on the tropical circulation during the 1997–1998 period. In particular for the 1998 boreal summer, they found that the Atlantic, characterized by the dominance of positive SST anomalies in most of its tropical sector, induced low-level equatorial westerlies to the west, and easterlies to the east, thus triggering equatorial Rossby and Kelvin waves. In a similar context, Wang (2006) proposed that the ENSO and AEM might be related through the inter-Pacific-Atlantic east–west SST gradient. This author suggested a feedback mechanism for the climate variability of the tropical Pacific and Atlantic basins. This mechanism involves the inter-Pacific-Atlantic SST gradient formed due to the equatorial modes in the eastern side of both basins and the induced variations in the surface zonal winds over equatorial South America, which bridge both basins. In addition, Wang (2006), using the SST indices for the Niño3 (4°N–4°S, 150°W–90°W) and Atl3 (4°N–4°S, 20°–0°W) areas, pointed out that the eastern equatorial Pacific and Atlantic are not simultaneously correlated. However, a negative relationship between these two basins has been noted since the 1970s by some authors (Mélice and Servain, 2003). In this relation, the AEM leads by some months the establishment of the ENSO mode.
Kucharski et al. (2007) found coexisting cold AEM and El Niño events during the boreal summer (June–September) in the 1975–1999 period, not noted in the 1950–1974 period. More recently, Losada et al. (2010), using four different atmospheric global circulation models, found that the AEM in May–June influences the next ENSO mode in July–August–September. They showed that a warm AEM in May–June favours the establishment of La Niña conditions in the tropical Pacific during the July–August–September period. Their physical explanation is based on the fact that the equatorial heating in the Atlantic induces a Gill–Matsuno-type response in the atmosphere (Matsuno, 1966; Gill, 1980). This atmospheric response involves a Kelvin wave to the east, an equatorial Rossby wave to the west, and an east–west circulation with anomalous upward motion in the Atlantic sector and descending motion in the central Pacific, where the ENSO extremes start to develop. As part of this equatorial Atlantic induced circulation, the Indian summer monsoon might also be modulated. As a response to the anomalous warming (cooling) in the equatorial Atlantic, a low-level anticyclone (cyclone) is formed over India, thus weakening (strengthening) the Indian monsoon (Kucharski et al., 2009; Wang et al., 2009). In addition, Rodríguez-Fonseca et al. (2009), based on observational and numerical results, provided a physical explanation for a La Niña establishment during the austral summer following a warm AEM during the previous austral winter. Their explanation is that the descending branch over the central Pacific related to Pacific–Atlantic Walker circulation enhances in this region the surface divergence, which in turn shallows the equatorial thermocline, thus triggering the Pacific La Ni na.
To summarize, previous studies support the notion that the tropical Atlantic can influence the tropical Pacific Ocean via the inter-basin SST gradient variability associated with the Atlantic Walker circulation. Based on these studies, the present work examines the ENSO-related evolving SST anomalies in the tropical basins which follow the occurrences of AEM events and those that are independent of the AEM. The associated rainfall anomalies over South America are also examined. The present analyses differ from the ones previously mentioned because we focus on the equatorial Atlantic conditions during the austral summer and the tropical Pacific conditions one year later. Furthermore, the analyses here are not restricted to the decades after the 1977 climate shift, rather long-term SST time series from 1900 to 2006 are used. The next sections are organized as follows: description of data and methodology in Section 2; analyses of the composites for ENSO modes that are dependent and independent on the AEM in Section 3. Conclusions are given in Section 4.
2. Data and Methodology
The datasets used in this paper consist of reconstructed monthly SST series obtained by Smith and Reynolds (2004), and monthly SLP data obtained from the Met Office Hadley Centre (Allan and Ansell, 2006). The SST and the SLP data are at 2° × 2° and at 5° × 5° latitude–longitude resolution grids, respectively. The SST dataset consists of the version 2 of the monthly gridded SST data and the SLP dataset is the HADSLP2 archive. Both variables are obtained in the global domain between 60°N and 60°S and for the periods 1900–2006 for SST, and 1900–2004 for SLP.
Monthly gridded precipitation series used here are at 1° resolution in the South American area bounded at 15°N, 50°S, 90°W and 30°W for the 1901–2006 period. These data are derived from the gauge-based reconstructions of the Global Precipitation Climatology Center (GPCC) (Beck et al., 2005; Rudolf and Rubel, 2005). Prior to any calculation, the linear trend is removed from the SST time series at each grid point. Monthly anomalies of SST, SLP and precipitation are obtained as departures from the 1900–2006, 1900–2004 and 1901–2006 means, respectively. The anomaly series at each grid point is normalized by the corresponding standard deviation of the anomaly time series.
The SST anomaly time series at each grid point of the tropical Atlantic is subjected to the filtering procedure to isolate the interannual oscillations from the time series. This procedure is done using the Morlet wavelet as a bandpass filter, whose calculation is that described by Torrence and Compo (1998). The Morlet wavelet is a complex exponential modulated by a Gaussian, , with , where t is the time, s is the wavelet scale and ω0 is a non-dimensional frequency. The tropical Atlantic is considered the sector bounded at 70°W, 20°E, 30°N and 30°S, and the interannual scale is the 1–6-year band, which is the same used by Handoh et al. (2006a, 2006b) to define the SST interannual variations in the tropical Atlantic sector.
The empirical orthogonal functions (EOFs) of the interannual SST in the tropical Atlantic are calculated using the covariance matrix. The physically meaningful identity of the modes is investigated using the method proposed by North et al. (1982). The eigenvectors are presented as correlation patterns, whose statistical significance is examined using 37 degrees of freedom which is the record length of 1284 (107 years times 12 months) divided by 35 months. The time interval of 35 months represents the time interval for two independent realizations and is the lag needed to obtain autocorrelation coefficients of the time series close to zero in each grid point. The Student's t-test for 37 degrees of freedom gives the thresholds of 0.32, for the correlations to be significant at the 95% confidence level. The principal component (PC01) of this mode is used to determine the periods dominated by anomalously warm or cold conditions in the equatorial Atlantic. The PC01 is standardized. It will become clear in the results that this mode describes mostly the SST variations in the equatorial and tropical southern sectors of the Atlantic. Warm (cold) AEM is determined using the criterion that the PC01 exceeds (is less than) 1 (−1) standard deviation for at least six consecutive months including at least two summer (December–February) months.
Because of the data resolution, the Niño-3 SST index is calculated here as the five-month running mean of the averaged linearly detrended SST anomalies in the area bounded at 4°N, 4°S, 150°W and 90°W. Using the criterion suggested by Trenberth (1997), an El Niño (La Niña) event is identified when the detrended Niño-3 SST index exceeds (is less than) the threshold of 0.5 °C (−0.5 °C) for at least six consecutive months.
Monthly composites of non-filtered monthly SST and SLP anomalies in the global domain between 60°N and 60°S are obtained for the possible combinations of the AEM preceding an ENSO extreme in the following year. Composites of SST and SLP anomalies are obtained every month of a 19-month period from November (−1) to May (+1). The symbols (−1), (0) and (+1) are in relation to the ENSO extreme year and refer to the previous year, the onset year, and the mature stage year, respectively. The cases, warm AEM followed by an El Niño and cold AEM followed by a La Niña, did not show significant anomalous patterns. So, these composites are not shown here. However, the other cases show interesting significant anomalous patterns.
The effects of the ENSO extremes on the rainfall over South America are also examined from the three-month composites of rainfall anomalies. The three-month composites refer to three-month running means and are indicated by the initials of the months, such as ASO (0) for August–September–October (0), SON (0) for September–October–November (0) and so on up to MJJ (+1) for May–June–July (+1). These composites are obtained for three-month running means to smooth the patterns. Because the interest is on the ENSO extreme effects, the precipitation composites span from the onset to decay stages of the ENSO extremes.
In order to assess the statistical significance of the composites, the number of degrees of freedom is the number of years. It is assumed that a variable X with n values and S standard deviation shows a Student-t distribution. So, only composites with absolute values exceeding are statistically significant (Panofsky and Brier, 1968). The confidence levels used are of 95% for SST and precipitation, and of 90% for SLP.
3.1. EOF analysis
The first EOF mode pattern of the interannual SST anomalies in the tropical Atlantic and the corresponding amplitude time are displaced in Figure 1. The first and the second EOF modes explain, respectively, 30.6% and 18.7% of the total interannual variance for the SST in the tropical Atlantic. According to the North et al. (1982) method, the first eigenvalue is well separated from the higher ones. So, it represents a physically meaningful identity. Thus, only this mode is discussed. The first mode, featuring the largest loadings in the eastern equatorial Atlantic and tropical South Atlantic (TSA), resembles the equatorial anomalous SST mode previously documented by Zebiak (1993). The PC01 exhibits temporal variations with periods spanning from 2 to 6 years (lower panel in Figure 1). For positive (negative) amplitudes, anomalous warming (cooling) in the eastern equatorial Atlantic extends westward and southeastward reaching the eastern and northern coast of Brazil.
3.2. Classification of the AEM according to the ENSO phases
The year for which PC01 exceeds (is less than) 1 (−1) standard deviation for six or more consecutive months, with at least two of them in the austral summer, is referred to as warm (cold) AEM. This criterion captures only the persistent AEM. This is desired because one expects that the persistent conditions in the eastern equatorial Atlantic and TSA might influence the tropical Pacific. The La Niña event preceded by a warm AEM is called AEM-dependent La Niña, otherwise AEM-independent La Niña. Conversely, the El Niño event preceded by a cold AEM is called AEM-dependent El Niño, or else AEM-independent El Niño. Table I lists the years classified as warm AEM, AEM-dependent La Niña, AEM-independent La Niña, cold AEM, AEM-dependent El Niño and AEM-independent El Niño, and the corresponding number of occurrences of these events. In this table, each ENSO extreme event is indicated by two sequential years: the first one corresponds to the year of the ENSO extreme onset, and the second year is that of the ENSO extreme mature stage. These years are indicated by the symbols (0) and (+1), respectively. So, the year before the ENSO extreme onset is indicated by the symbol (−1). In Table I, each warm (cold) AEM event is also indicated by two sequential years: the first year corresponds to the year (−1), and the second year is the year (0).
Analyses of Table I in percentage terms show unexpected results. Nine of the twenty-four La Niña episodes were preceded by warm AEM, meaning that 37.5% of the total La Niña episodes were AEM dependent. Eleven of the twenty-four El Niño episodes were preceded by cold AEM, so 46% of the total El Niño episodes were AEM dependent. On the other hand, it is remarkable that of the 12 (15) years with warm (cold) AEM, 9 (12) are followed by the La Niña (El Niño) occurrences. Thus, 75% (73%) of the years with warm (cold) AEM are followed by a La Niña (an El Niño) occurrence in the next year. These percentages are too high to be only a coincidence. So, it seems that the SST conditions in eastern equatorial Atlantic and TSA have precursor potential for ENSO extremes.
In order to examine the differences of the SST and SLP anomalous evolving patterns between the AEM-dependent and AEM-independent ENSO extremes, monthly composites of the SST and SLP anomalies are obtained for AEM-dependent La Niña (or warm AEM followed by La Niña), AEM-independent La Niña, AEM-dependent El Niño (or cold AEM followed by El Niño) and AEM-independent El Niño cases. Differences are expected in the SST and SLP evolving patterns. These composites are analysed in the next section.
3.3. SST and SLP composites
Although composite maps have been obtained for every month, only those of alternate months from November (−1) to May (+1) are shown. The SST and SLP maps for cold AEM followed by El Niño occurrence in the next year are displayed in Figures 2 and 3. These maps show a weak La Niña in November (−1) evolving into an El Niño in September (0) which remains quite strong until January (+1), and weakens from March (+1) to May (+1). Within this cycle, the El Niño onset in May (0) features an east–west anomalous SST (SLP) gradient with significant negative (positive) anomalies in the western tropical Pacific (tropical Indian Ocean-Indonesian region) and opposite sign anomalies in the eastern tropical Pacific. As these gradients grow stronger, an El Niño settles by September (0). The SLP positive centre splits into two centres in November (0) and January (+1): one extending eastward from northern South America to central equatorial Africa, and another over tropical Indian Ocean and western Pacific. In May (+1), although small positive SST anomalies remain in the central and eastern tropical Pacific, the SLP no longer features an El Niño pattern. Another important aspect is the absence of an anomalous warming in the tropical North Atlantic (TNA) in response to the El Niño. Connections between the SST variability in the TNA and ENSO have been previously demonstrated in observational and modelling studies (Hastenrath et al., 1987; Hameed et al., 1993 among others). Enfield and Mayer (1997) showed that the warming in the TNA lags the Pacific warming by one–two seasons, being strongest during the boreal spring and early summer. The absence of warming in the TNA might be due to the persistent negative SST anomalies in this Atlantic sector from November (−1) to September (0), which create an unfavourable pre-condition. The SST patterns in the tropical Atlantic evolve into a cross-equatorial negative SST gradient with neutral or negative anomalies in the TNA and positive ones in the TSA from November (0) to March (+1).
The AEM-independent El Niño composites are displayed in Figures 4 and 5. The SST and SLP patterns of a weak La Niña in November (−1) evolve into the El Niño mature stage from November (0) to January (+1). As part of this cycle, an east–west anomalous SLP gradient with its positive centre in the tropical Indian Ocean-Australia region and the negative one in the eastern tropical Pacific is established in July (0), and remains quite strong until May (+1) (Figure 5). In this case, the SLP positive centre splits into two centres in November (0) and March (+1): one extending over the tropical Indian Ocean and western Pacific, and the other over the tropical Atlantic in November (0) and over TSA in March (+1). The east–west SST gradient in the tropical Pacific becomes well defined from July (0) onwards. It is important to note positive SST (negative SLP) anomalies in the TNA in March (+1) and May (+1). This results from a nearly neutral pre-condition in most of the tropical Atlantic including the TNA from November (−1) to March (0) and from September (0) to January (+1) allowing the El Niño related warming to occur in the TNA in March (+1) and May (+1). Concordant with this, Enfield and Mayer (1997) suggested that the SST dipolar structure in the tropical Atlantic develops during the boreal spring, when SST anomalies with opposite signs occur in the TNA and to the west of Angola. This seems to be the case of the analysis here. A positive SST dipolar structure in the tropical Atlantic with positive SST anomalies in the TNA and negative SST anomalies in the TSA is also evident in May (+1). So, both the El Niño and the tropical Atlantic positive SST dipolar structure contribute to dry conditions in northeastern Brazil.
The composite maps for the warm AEM followed by a La Niña in the next year are shown in Figures 6 and 7. These maps show a weak El Niño in November (−1) evolving into a La Niña, whose onset occurs in May (0) and the mature stage, from November (0) to January (+1). The La Niña onset features an east–west SST (SLP) gradient with significant positive (negative) anomalies in the western tropical Pacific (Atlantic and Indian Oceans) and negative (positive) ones in the central and eastern equatorial Pacific (tropical Pacific). The east–west SST gradient remains strong until January (+1), and the La Niña SST pattern remains well defined until May (+1). In addition, the east–west SLP gradient remains quite well defined until November (0), but less organized afterwards. Positive SLP anomalies extend over southern and southeastern South America in September (0) and November (0). In the tropical Atlantic, a well-defined warm AEM with the largest anomalies along the equator and in the TSA contrasts with neutral conditions in most of the TNA forming a cross-equatorial negative SST gradient mode from November (−1) to July (0). The cold pre-condition in the TNA during this period seems to trigger an early cooling in this area which starts almost simultaneously to the enhancement of the Pacific cooling in September (0). The TNA cooling strengthens in March (+1) and remains strong in May (+1), when a positive SST anomalous centre is noted in the TSA. The Pacific and Atlantic SST pattern in January (+1) is very similar to the one previously discussed by Andreoli and Kayano (2006) for a combination of La Niña and neutral TSA conditions.
The sequential maps leading to the establishment of AEM-independent La Niña are displayed in Figures 8 and 9. An interesting aspect in these maps is the early and slow La Niña onset. This is indicated by significant negative SST anomalies off the coast of Chile between 20°S and 40°S from January (0) to March (0). The east–west SST (SLP) gradient starts to be established by May (0), when significant positive (negative) anomalies occur in the western tropical Pacific (tropical Indian and western Pacific Oceans), and the opposite signs anomalies occur to the east of 120°W off the Chilean coast and along the equator. As the negative SST anomalies along the equator grow strong and spread westward occupying most of the eastern equatorial Pacific, the east–west SLP gradient intensifies. So, a La Niña pattern is established from September (0) onwards, when the east–west SLP gradient is well established. It is noticeable that positive SLP anomalies are found over southern and southeastern South America in November (0) and January (+1). Concerning the Atlantic, almost neutral SST conditions are noted in its tropical sector most of the times. So, this allows the establishment of a La Niña related cooling in the TNA from March (+1) to May (+1).
3.4. Precipitation composites
The precipitation composites for the cold AEM followed by El Niño are shown in Figure 10. The dry condition found over northwestern/northern South America in ASO (0) intensifies gradually, and remains quite strong in particular over the northwestern region in MAM (+1). This condition is consistent with the positive SLP anomalies noted over northern South America from November (0) to January (+1). The SLP anomalous pattern over the eastern Pacific, northern South America and the surrounding TNA areas is related to an anomalous Walker circulation associated with the inter-Pacific-Atlantic east–west SST gradient associated with the El Niño. Indications of neutral and positive precipitation anomalies during FMA (+1) and MAM (+1) over northeastern South America are contrary to the dry conditions normally expected there during an El Niño episode. This result is due to the presence of cross-equatorial negative SST gradient in the tropical Atlantic, which favours the rainfall over northeastern South America. Concordantly, Andreoli and Kayano (2006) found that the El Niño effect on the rainfall over northeastern Brazil might be reduced due to a cross-equatorial negative SST gradient in the tropical Atlantic. In addition, Kayano and Andreoli (2006) showed that the warming in the TSA occurs a few months before the rainy season in northeastern Brazil, as is the case here. On the other hand, the El Niño related positive anomalies found over southeastern South America from SON (0) to DJF (0) grow smaller over this area from JFM (+1) onwards. This result is in concordance with Barreiro and Tippmann (2008) who found a weaker El Niño effect in southeastern South America for the warmed equatorial Atlantic. Thus, the results show a combined El Niño and AEM effect on the precipitation over South America.
For the AEM-independent El Niño, the precipitation composite maps over South America show quite well-defined precipitation anomalies dipole featuring dry conditions over northwestern/northern and wet conditions southeastern in most of the three-month periods (Figure 11). The precipitation anomalies in these regions might be associated with the persistent and well-defined east–west SLP gradient associated with the El Niño prevailing from July (0) to May (+1). The significant negative precipitation anomalies over northeastern Brazil during FMA (+1), MAM (+1) and AMJ (+1) result from the cross-equatorial positive SST gradient noted in March (+1) and May (+1).
The precipitation composites for the warm AEM followed by La Niña are displayed in Figure 12. The significant positive anomalies scattered in small areas of the South American northwestern/northern region from ASO (0) to DJF (0) are related to the enhanced Walker circulation associated with a well-established La Niña during these periods. The non-significant quite well-organized positive anomalies over northeastern South America from JFM (+1) to AMJ (+1) result from the effects of the La Niña and the cross-equatorial negative SST gradient in the tropical Atlantic. In fact, Andreoli and Kayano (2006) showed that this precipitation pattern is related to the occurrence of a La Niña and neutral TSA. On the other hand, the significant negative precipitation anomalies occupy a large area of southeastern South America, in particular during SON (0), OND (0) and NDJ (0). From DJF (0) onwards, these anomalies are quite reduced, possibly due to the weakening of La Niña circulation as shown by the less-organized east–west SLP gradient from January (+1) onwards.
Precipitation composites for the AEM-independent La Niña are shown in Figure 13. The main feature at most times is the presence of negative anomalies over southeastern South America. In some cases this pattern is due to the occurrence of positive SLP anomalies over southeastern South America as during the period from November (0) to January (+1). In other cases, the dry conditions over southeastern South America might be caused by the persistent La Niña related circulation, as shown by the persistence of well-organized east–west SLP gradient from September (0) to May (+1).
The analyses above show important differences in the precipitation composites over South America for El Niño and La Niña episodes which are AEM dependent and AEM independent.
4. Concluding remarks
Recent studies have provided observational and numerical evidence that the tropical Pacific and Indian Oceans are influenced by the tropical Atlantic within a one season time scale (Wang, 2006; Kucharski et al., 2007, 2008, 2009; Losada et al., 2010). Wang et al. (2009), in a comprehensive review of these papers, showed that anomalous warming or cooling associated with the AEM induces teleconnections through a Gill–Matsuno-type circulation or through variations in the Atlantic Walker circulation. Here, the relationships between the AEM and ENSO modes are re-examined for the 1900–2006 period through the ENSO-related evolving SST and SLP anomalies in the tropics that follow the occurrences of AEM and those that are AEM independent. The corresponding anomalous rainfall patterns over South America are also investigated. Substantial differences are noted in the sequences of SST and SLP composites of ENSO extremes between the AEM-dependent and -independent cases.
The cold (warm) AEM followed by an El Niño (a La Niña) shows negative (positive) SST anomalies in the eastern side of the tropical Pacific and Atlantic basins, and in part of the tropical Indian Ocean in the first months. This result is consistent with a positive linkage between the tropical Pacific and Atlantic previously documented for the austral summer (Handoh et al., 2006b). The weakening of the SST anomalies in the eastern tropical Pacific and Indian Oceans in March (0) might be explained by the previously outlined mechanism on the relationship of the tropical Atlantic and the other tropical Oceans (Wang, 2006; Kucharski et al., 2007, 2008, 2009; Wang et al., 2009). The persistent negative (positive) AEM from November (−1) to January (0) induces an anomalous Atlantic Walker circulation and a Gill–Matsuno-type atmospheric response which weaken the negative (positive) SST anomalies in the central and eastern equatorial Pacific and Indian Oceans. As these anomalies weaken further and the opposite sign SST anomalies appear in the western equatorial Pacific, east–west SST and SLP anomalous gradients are established in the tropical Pacific in May (0). The east–west SST and SLP gradients and the persistent negative (positive) AEM act to strengthen the positive (negative) SST anomalies in the central and eastern tropical Pacific, leading to an El Niño (a La Niña) in July (0). The negative (positive) SST anomalies in the western Pacific maintain the El Niño (La Niña) during the following months. This mechanism justifying the results is the same as that previously used to explain the relationship between the tropical Atlantic and the other basins, in particular the Indian Ocean. However, the seasonal timing of the relationship studied here differs from that of the previous studies. Kucharski et al. (2007) noted that the interannual variability of the Indian summer monsoon could be realistically simulated if the SST conditions in the Atlantic during austral winter were considered. We noted that the Atlantic SST anomalous conditions are persistent and might be noted 5–6 months before. Furthermore, the ENSO extreme conditions are reinforced and maintained by the east–west SST anomalous gradient in the tropical Pacific.
For the AEM-dependent ENSO extreme composites, an interesting difference for the SST composite is noted in the TNA. For the El Niño composites, no warming is noted in the TNA during the boreal spring and early summer months. Conversely for the La Niña composites, a cooling is observed in the TNA during these months. The main difference is the presence of a persistent unfavourable SST pre-condition in the TNA for the El Niño composite contrasting with neutral or favourable SST pre-condition in the TNA for the La Niña composite.
For the AEM-independent ENSO extremes, an important feature is the absence of significant SST anomalies in the tropical Indian Ocean. In these cases, the first signs of significant SST anomalies appear along 20°S and the equator in the eastern Pacific in May (0). In the subsequent months, the east–west SST and SLP anomalous gradients in the tropical Pacific play an important role in establishing an AEM-independent La Niña (El Niño) pattern from July (0) to January (+1). For the AEM-independent ENSO extreme composites, an El Niño related warming and a La Niña related cooling are noted in the TNA during the boreal spring and early summer months. This feature results from a nearly neutral SST pre-condition in the TNA for these composites.
The precipitation composites over South America show that the AEM-dependent ENSO extremes have the combined effects of the tropical Pacific and Atlantic. The cold AEM followed by an El Niño case evolves into a pattern such that a warm AEM coexists with an El Niño from November (0) to January (+1). Under these conditions, the El Niño effects on the precipitation over South America are weakened in its northeastern (Andreoli and Kayano, 2006) and southeastern (Barreiro and Tippmann, 2008) regions. The warm AEM followed by a La Niña evolves into a pattern with negative SST anomalies in the eastern tropical Pacific and in the TNA by January (+1). Andreoli and Kayano (2006) showed that this pattern is related to the presence of a La Niña and neutral TSA, and causes the increase of positive precipitation anomalies over northeastern Brazil, as shown here.
The common characteristic of precipitation composites over South America of the AEM-independent ENSO extremes is the persistent anomalies over southeastern South America for the period from SON (0) to AMJ (+1), with positive anomalies for the El Niño composites and negative anomalies for the La Niña composites. A possible explanation is that for these composites the corresponding east–west SLP gradient remains well defined until May (+1), thus affecting the rainfall over southern South America.
The results here, as far as the authors know, have not been discussed before and might represent a potential for long lead predictability of the climate variations in the tropical Pacific.
The authors thank the two anonymous reviewers for their helpful comments and suggestions. The first and second authors were partially supported by the Conselho Nacional de Desenvolvimento Científico e Tecnológico of Brazil. Wavelet software was kindly provided by C. Torrence and G. Compo and is available at URL: http://paos.colorado.edu/research/wavelets.