The Amazon encompasses one of the major continental regions with intense convective activity in the globe. Thus, it constitutes an important heat source for the atmosphere and contributes to the atmospheric general circulation. The seasonal cycle dominates the Amazon convective rainfall variations, with a maximum austral summer and a minimum winter (Rao and Hada, 1990). The rainfall seasonal cycle in this region is driven by the South American monsoon system (SAMS) (Zhou and Lau, 1998; Vera et al., 2006; Garcia and Kayano, 2009, 2010a). On the other hand, the Amazon rainfall exhibits marked interannual variability, part of which has been attributed to the sea surface temperature (SST) variations in the tropical Pacific manifested as an El Niño-Southern Oscillation (ENSO) mode (Kousky et al., 1984; Kayano and Moura, 1986; Aceituno, 1988; Ropelewski and Halpert, 1987, 1989; Kousky and Ropelewski, 1989; Marengo and Hastenrath, 1993; among others). The reduced (enhanced) rainfall over the Amazon, in particular in its northern and central sectors, noted in some years, is associated with El Niño (La Niña) occurrences. However, a large percentage of the interannual variance is not explained by the ENSO (Rao et al., 1996; Marengo et al., 2001).
Indications that the SST variability in the tropical Atlantic also plays an important role in modulating the Amazon rainfall variability have been shown by many papers. Among them, some use the mechanistic relationship between the tropical Atlantic SST variability and the north–south migration of the inter-tropical convergence zone (ITCZ) in this oceanic sector to explain the rainfall variability in this region. In fact, some papers have shown that the influence of the ITCZ in the South American rainfall is not limited to northeastern Brazil but extends over the Amazon (Uvo et al., 2000; Ronchail et al., 2002). Garcia and Kayano (2010a), using the outgoing long wave radiation data for the 1979–2006 period, provided observational evidence on the relationship between the Atlantic ITCZ related convection and the South American monsoon in the central Amazon. This relationship was further confirmed by Garcia and Kayano (2010b), who analysed the moisture and heat budgets associated with the South American monsoon and the ITCZ.
On the other hand, Yoon and Zeng (2010) found comparable strengths of the tropical North Atlantic (TNA) SST and ENSO-related Pacific SST influences on the Amazon rainfall, and weaker influence of the tropical South Atlantic (TSA) SST variability. For seasonally stratified analyses, they found the strongest TSA (TNA) SST variability influence over the southern Amazon during the wet-to-dry transition (dry) season, and the strongest ENSO influence over the entire basin (with a maximum in the eastern Amazon) during the wet season. They proposed that the mechanistic relations between the Amazon rainfall variability and the SST variations in the equatorial Pacific, TNA and TSA involve variations in the Walker circulation, meridional movement of the ITCZ, and Hadley circulation, respectively.
The combined effects of the tropical Pacific and Atlantic Oceans on the Amazon rainfall have also been investigated. In fact, Souza et al. (2000) showed that the El Niño and warm dipole (warmer than normal TNA and colder than normal TSA) in the tropical Atlantic have a stronger combined effect in reducing precipitation over the central-northern and western Amazon, and in increasing precipitation over the eastern Amazon during summer than each mode acting individually. Consistently, Ronchail et al. (2002) found very strong negative anomalies over the northern Amazon when El Niño and cold waters in TSA prevail.
From another point of view, evidence of the connections between the equatorial Atlantic and the eastern tropical Pacific through the SST variability associated with the Atlantic equatorial mode (AEM) and the ENSO has been found in recent studies (Wang, 2006; Kucharski et al., 2007, 2008). The AEM, similarly to the ENSO in the Pacific, features an anomalous SST pattern roughly symmetric to the equator in the eastern side of the Atlantic Basin, and varies in a 2–5-year timescale (Zebiak, 1993; Wagner and da Silva, 1994). The positive (negative) AEM is referred to as Atlantic El Niño (Atlantic La Niña) (Zebiak, 1993). When the AEM and the ENSO create an inter-basin east-west SST gradient, the Atlantic Walker circulation induces an anomalous east-west flow of air across northern South America and an atmospheric direct circulation between the two oceans. Kayano et al. (2010) examining the ENSO-related precipitation patterns over South America found substantial differences in these patterns when the ENSO extremes are associated with the AEM occurrences and AEM independent.
From the above review, it is clear that the SST variability in both the tropical Pacific and Atlantic Oceans plays an important role in the interannual rainfall variations over the Amazon (Aceituno, 1988; Marengo, 1992; Nobre and Shukla, 1996; Ronchail et al., 2002; among others). Furthermore, the seasonal differences on the effects of these two Oceanic Basins strongly suggest that a seasonal stratified analysis should be preferred rather than a whole year analysis. Here, a precipitation index derived from four raingauge stations over the central and eastern Amazon is used to classify the extreme climate conditions in this region on a seasonal basis. The causal relations between these extreme climate conditions and the large-scale SST and sea level pressure (SLP) anomalous patterns are examined from composite analyses. Data and methodology are described in Section 2, and bi-monthly composites of SST and SLP are described in Section 3. Conclusions are given in Section 4.
2. Data and methodology
The dataset used in this paper consists of reconstructed monthly SST series at 2° × 2° latitude-longitude resolution grids obtained by Smith and Reynolds (2004). This dataset consists of the version 2 of the monthly gridded SST. The SST time series are obtained in the global domain between 30°N and 30°S and for the 1931–1996 period. Monthly SST anomalies are obtained as departures from the 1931–1996 means. Monthly anomaly time series of SST at each grid point are normalized by the corresponding monthly standard deviation. In addition, the HADSLP2 archive containing monthly SLP data at 5° × 5° latitude-longitude resolution obtained from the Met Office Hadley Centre is also used (Allan and Ansell, 2006). The SLP time series in the global band between 60°N and 60°S for the same period as those of the SST are subjected to the calculations described for the SST. The choice of the 1931–1996 period for the present analysis is based on the period with precipitation data availability.
Monthly precipitation time series of Manaus, Parintins, Itaituba, and Santarém provided by the Instituto Nacional de Meteorologia (INMET) are also used. Their geographical locations and periods with available precipitation data are listed in Table I. These four raingauge stations are close to each other and located in the central and eastern Amazon. Considering all four stations, their dry and rainy seasons present a slight lag towards east. The rainy season for Santarém (the station located more to the east) lags by one month the rainy season spanning from January to April (JFMA) for the other three stations. The dry season for Parintins and Santarém lags by two months the dry period from June to September (JJAS) for the other two stations (Table I). So, considering the periods JJAS and JFMA the dry and rainy seasons, respectively, two time series representing the precipitation in the central and eastern Amazon for these seasons have been constructed based on these four raingauge stations. For each station, the seasonal rainfall value is obtained by averaging the corresponding monthly values. For each season, seasonal time series is obtained by averaging the seasonal values in space. In order to avoid unbalanced contributions of the raingauge stations, only the common period among them with available data, which spans from 1931 to 1996 is considered in the analyses. For the same reason, the years with missing data in one or more raingauge stations are not considered in the analyses. These years are 1937, 1938, 1939, 1943, 1959, 1960, 1977, and 1995. These seasonal time series represent the rainy and dry seasons in the central and eastern Amazon. Hereinafter, the references to dry and rainy seasons are for this region.
Table I. Raingauge stations in the central and eastern Amazon: geographical location, period with available data, and the months corresponding to their dry and rainy seasons
Itaituba - PA
Manaus - AM
Parintins - AM
Santarém - PA
These seasonal time series are then used to identify very dry, dry, wet, and very wet conditions in the central and eastern Amazon during its dry and rainy seasons. Considering one seasonal value per year, the time series for the period from 1931 to 1996 without the 8 mentioned missing years totalizes 58 values. So, the seasonal time series are ranked from 1 for the smallest value to 58 for the largest value. The ranked amounts are divided by 58. The resulting percentile ranks (R) varying from approximately zero to 1 are then used to classify the seasonal rainfall. Considering the ranked seasonal precipitation time series, with one seasonal value per year, a year is classified as very dry, dry, wet, and very wet if R≤0.15, 0.15 < R≤0.35, 0.65≤R < 0.85, and R ≥ 0.85, respectively. The years when the dry and rainy seasons are classified as very dry, dry, very wet, and wet categories are listed in Tables II and III, respectively.
Table II. Years when the dry season is classified as very dry, dry, very wet, and wet categories
Manaus, Parintins, Itaituba, Santarém—Dry season (JJAS)
Very dry years
Very wet years
Table III. Years when the rainy season is classified as very dry, dry, very wet, and wet categories
Manaus, Parintins, Itaituba, Santarém—Rainy season (JFMA)
Very dry years
Very wet years
It is worthwhile mentioning that if one raingauge time series is excluded from the areal averaged seasonal precipitation time series, the classification of the resulting dry and rainy seasons is not altered. However, in order to have a seasonal time series that represents better regional rainfall, we adopt the areal averaged seasonal precipitation time series for all four raingauge stations.
Two-month running mean composites of SST anomalies in the tropical sector and of SLP in the global band between 60°N and 60°S are obtained for the dry, very dry, wet, and very wet categories. The rainy season composites span from October/November to March/April, and the dry season composites, from March/April to August/September. 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 level 90% is used for both variables.
In order to facilitate the discussion, a positive (negative) inter-Pacific-Atlantic anomalous SST gradient refers to an anomalous pattern with a La Niña (El Niño) and an Atlantic El Niño (La Niña). On the other hand, in the tropical Atlantic a positive (negative) inter-hemispheric anomalous SST gradient or warm (cold) dipole refers to an anomalous pattern with positive (negative) SST anomalies in the TNA and opposite sign anomalies in the TSA.
3.1. Rainy season
The very dry rainy season seems to be more strongly modulated by the tropical Pacific than by the tropical Atlantic. For this case, a strong El Niño pattern for the SST and SLP remains well defined from October/November to March/April (Figure 1a) and b)). During these months, the largest significant positive SST anomalies move from the eastern equatorial Pacific to an equatorial band centered at 140°W, where they weaken slightly in March–April. For this bi-month in particular, non-significant negative SST anomalies are noted in the western equatorial Atlantic. So, a negative inter-Pacific-Atlantic SST gradient mode is established. The anomalous SST pattern is accompanied by an anomalous SLP pattern with significant negative anomalies centered in the eastern tropical Pacific and the positive ones almost everywhere else in the tropics featuring two centers, one over the western Pacific and another over the Atlantic-African sector. The significant positive SLP anomalies in the equatorial Atlantic contribute to weaken the Atlantic ITCZ or to maintain it to the north of its climatological position. Thus, the drying effect of the El Niño is strengthened over the Amazon. Therefore, in this case, the El Niño is the most important phenomenon modulating the rainfall over the central and eastern Amazon during its rainy season. This modulation is processed through the El Niño-related descending motion over this region, which reduces the rainfall over there, as previously shown in many papers (Ropelewski and Halpert, 1987; Aceituno, 1988; Kayano et al., 1988).
The dry rainy season shows the most significant SST anomalies in the equatorial Atlantic and TSA sectors with signs of an Atlantic La Niña (or negative AEM) pattern and significant positive SST anomalies in the western TSA during October–November (Figure 2a)). While the significant positive SST anomalies weaken, the Atlantic La Niña pattern becomes better established from November/December to December/January. Several authors have shown that an Atlantic La Niña induces the establishment of a Pacific El Niño through an east-west anomalous circulation (Wang, 2006; Kucharski et al., 2007, 2008; Kayano et al., 2010) during the subsequent months. This is the case here, as a negative inter-Pacific-Atlantic anomalous SST gradient is established from January/February to March/April (Figure 2a)). Moreover, signs of significant positive SST anomalies are also noted in the TNA during these bi-months. Although weak, a positive inter-hemispheric SST gradient in the tropical Atlantic is also established from January/February to March/April (Figure 2a)). Signs of an east-west anomalous SLP gradient between the equatorial Atlantic and central eastern Pacific, and a north-south SLP gradient between the equatorial Atlantic and TNA are apparent in the SLP composites, in particular during February–March and March–April months (Figure 2b)). Indeed, weak positive SLP anomalies prevail over the equatorial Atlantic and eastern South America, and significant negative SLP anomalies are noted over the eastern tropical Pacific, the western Amazon and part of the TNA, in particular during February–March. So, both the negative inter-Pacific-Atlantic anomalous SST gradient and the positive inter-hemispheric anomalous SST gradient in the tropical Atlantic contribute to the dry condition in the central and eastern Amazon, through anomalous east-west and north-south direct circulations. Consistent with this result, Kayano et al. (2010) found reduced rainfall over the central Amazon due to the action of a negative inter-Pacific-Atlantic anomalous SST gradient; and Souza et al. (2000) noted dry conditions over the central and northern Amazon due to the combined effect of El Niño and positive inter-hemispheric anomalous SST gradient.
The very wet rainy season case is also strongly modulated by the anomalous SST conditions in both the tropical Pacific and Atlantic Oceans (Figure 3a) and b)). In this case, a positive inter-Pacific-Atlantic anomalous SST gradient is established from October/November to March/April. The main SST anomalous centers with significant negative anomalies in the eastern Pacific and the positive ones in the equatorial Atlantic are located along the equatorial band from October/November to January/February. During the following two bi-months, although the main negative center of the equatorial Pacific is located in the central part of this Ocean, significant negative SST anomalies become more intense in the eastern equatorial Pacific due to the action of a persistent warm AEM mode. This anomalous SST pattern features a positive inter-Pacific-Atlantic anomalous SST gradient which is associated with an anomalous Atlantic Walker circulation, as shown by many authors (Wang, 2006; Kusharski et al., 2007, 2008, 2009; Losada et al., 2009; Wang et al., 2009). During February–March and March–April, a negative inter-hemispheric anomalous SST gradient is also conspicuous (Figure 3a)). The associated SLP anomalous pattern features a La Nina pattern with significant positive anomalies in the eastern tropical Pacific and the negative ones in the western tropical Pacific in most bi-months; and an inter-hemispheric anomalous SLP gradient with significant positive anomalies in the TNA and the negative ones in the TSA from December/January to March/April (Figure 3b)). Thus, the very wet condition in the central and eastern Amazon is driven by the direct east-west and north-south circulations associated with the SST anomalous patterns in the eastern equatorial Pacific and tropical Atlantic sectors.
The wet rainy season seems to be related to a weak east-west inter-Pacific-Atlantic anomalous SST gradient (Figure 4a)). Indeed, weak negative SST anomalies persist in the eastern tropical Pacific from October/November to March/April. On the other hand, a positive AEM is flanked by significant negative SST anomalies in the TNA and TSA from November/December to December/January. Gradually the negative anomalies weaken in the TNA, and a small center with significant positive SST anomalies locates in the western equatorial Atlantic, in particular during February–March and March–April. Thus, a weak positive inter-Pacific-Atlantic anomalous SST gradient is established. The associated SLP composites show significant positive anomalies in the eastern central Pacific and over most of the tropical South America from October/November to December/January (Figure 4b)). Gradually the significant positive SLP anomalies weaken over South America, in response to the weakening of the negative SST anomalies in the eastern tropical Pacific. On the other hand, significant negative SLP anomalies are established over the equatorial Atlantic and part of tropical South America in March–April, in response to the significant positive SST anomalies located in the western equatorial Atlantic, particularly during February–March and March–April. This SLP pattern is consistent with a weak positive inter-Pacific-Atlantic anomalous SST gradient during the rainy season. So, the results here indicate that the intensity of a positive inter-Pacific-Atlantic anomalous SST gradient is the factor distinguishing between a very wet rainy season and a wet rainy season. Thus, a weak positive inter-Pacific-Atlantic anomalous SST gradient is the main factor leading to the wet conditions in the central and eastern Amazon during its rainy season.
3.2. Dry season
The very dry dry season case shows a well-defined negative inter-Pacific-Atlantic anomalous SST gradient in most bi-months, in particular from May/June to August/September (Figure 5a)). Signs of non-significant positive SST anomalies noted in the TNA from June/July to August/September are indicative of a weak positive inter-hemispheric anomalous SST gradient. The SLP anomalous maps also feature a well-defined El Niño pattern, with significant negative anomalies in the eastern tropical Pacific and the positive ones in the tropical Atlantic and western tropical Pacific from May/June to August/September bi-months (Figure 5b)). The significant positive SLP anomalies extend over the eastern equatorial South America in August–September. Thus, the associated east-west anomalous Atlantic Walker circulation, with ascending motion over the eastern equatorial Pacific and descending motion extending over the western equatorial Atlantic and the surrounding central and eastern Amazon, is the main factor contributing to diminish the rainfall over this continental area. This result confirms previous findings by Kayano et al. (2010) on the relationship between the inter-Pacific-Atlantic anomalous SST gradient and the South American rainfall. It is worthwhile recalling that the weak positive inter-hemispheric anomalous SST gradient in the tropical Atlantic also contributes to dry conditions in the central and eastern Amazon during its dry season.
The SST composites for the dry dry season feature a positive inter-hemispheric anomalous SST gradient in the tropical Atlantic with significant negative anomalies in the equatorial Atlantic and in the TSA, and non-significant positive anomalies in the TNA for most of the bi-month composites (Figure 6a)). On the other hand, non-significant small-magnitude SST anomalies prevail in the tropical Pacific. An inter-hemispheric anomalous SLP gradient consistent with the positive inter-hemispheric anomalous SST gradient is noted in June–July composite (Figure 6b)). These anomalous SST and SLP patterns in the tropical Atlantic have been related to the interannual climate variations in northeastern Brazil through the dynamics due to the air-sea interactions in this oceanic sector which affect the position and intensity of the ITCZ (e.g. Moura and Shukla, 1981; Nobre and Shukla, 1996). However, more recent studies have shown that the effect of the ITCZ on the rainfall might extend over the Amazon (Uvo et al., 2000; Ronchail et al., 2002). Therefore, the dry conditions in the central and eastern Amazon during its dry season are mostly related to the tropical Atlantic SST and SLP inter-hemispheric anomalous patterns.
The SST and SLP patterns for the very wet dry season case (Figure 7a) and b)) are similar to those of the very dry dry season (Figure 5a) and b)), but with opposite signs. So, as in the analysis above, the very wet conditions in the central and eastern Amazon are determined by the combined contribution of a positive inter-Pacific-Atlantic and a negative inter-hemispheric anomalous SST gradient modes. In terms of the SST anomalies, the former is due to a La Niña and an Atlantic El Niño; and the latter, to the significant negative SST anomalies in the TNA and the opposite sign SST anomalies in the TSA. Contrasting with the very dry dry season case, the very wet dry season case shows well-defined inter-hemispheric anomalous SST and SLP gradient patterns in the tropical Atlantic, in particular from May/June to August/September. So, both the inter-Pacific-Atlantic and inter-hemispheric anomalous SST gradient modes contribute to very wet conditions over the central and eastern Amazon during its dry season.
Similar to the dry dry season case, the wet dry season case shows relatively weak non-significant SST anomalies in most of the tropical Pacific, and a negative inter-hemispheric anomalous SST gradient with significant positive anomalies in the equatorial Atlantic and the negative ones in the TNA (Figure 8a)). This inter-hemispheric anomalous SST gradient is particularly well-defined from May/June to August/September. Signs of an inter-hemispheric anomalous SLP gradient, with non-significant negative SLP anomalies in the TSA and significant positive SLP anomalies in the TNA, are apparent in July–August (Figure 8b)). In consequence, the wet condition in the central and eastern Amazon is established by the tropical Atlantic SST and SLP anomalies.
The mechanistic relations between the dry, very dry, wet, and very wet conditions in the central and eastern Amazon and the SST and SLP variability in the tropical Atlantic and Pacific Oceans are analysed for the 1931–1996 period. In order to examine the seasonal differences the dry and wet seasons are analysed separately. The tropical Pacific and Atlantic Oceans, through the ENSO and the Atlantic SST variability modes, have important roles depending on the cases and the seasons. The seasonal differences are in part determined by the seasonal phase locking of these modes. While the mature stage of the ENSO extremes from December to February (Rasmusson and Carpenter, 1982) and the period (austral autumn) when the inter-hemispheric mode occurs more frequently in the tropical Atlantic (Enfield and Mayer, 1997) overlap the central and eastern Amazon rainy season, the mature stage of the AEM from June to September (Zebiak, 1993) encompasses the dry season. Thus, one expects that the rainy season cases be mostly driven by the ENSO and by the inter-hemispheric SST mode in the tropical Atlantic and the dry season cases, by the AEM. And truly, these correspondences are noted in most cases (Table IV). For the rainy season, the very dry case is mostly driven by an intense El Niño; the dry and very wet cases coincide, respectively, with the occurrences of weak El Niño and intense La Niña patterns. Also, for the dry, very wet, and wet cases of the rainy season, the positive moderate, negative intense, and negative weak inter-hemispheric SST gradient modes are, respectively, noted (Table IV). For the dry season, the very dry and dry cases are both accompanied by an intense negative AEM, and the very wet case, by an intense positive AEM. The exception seems to be the wet dry season case which is mostly driven by a negative moderate inter-hemispheric SST gradient (Table IV).
Table IV. Summary of the results
Inter-Pacific-Atlantic SST gradient
Inter-hemispheric SST gradient
Very dry rainy season
intense El Niño
Dry rainy season
weak El Nino
Very wet rainy season
intense La Niña
Wet rainy season
Very dry dry season
Positive intense negative AEM
Dry dry season
Positive intense negative AEM
Very wet dry season
Negative intense positive AEM
Wet dry season
The association of the very dry (very wet) rainy season case with the El Niño- (La Niña)- related SST and SLP patterns confirms previous findings that the ENSO-related rainfall variations in the central Amazon region are noted mostly from June (0) to March (+1), with the symbols (0) and (+1) referring to the onset and mature ENSO phases, respectively (Ropelewski and Halpert, 1987, 1989). Yoon and Zeng (2010) also found that the strongest ENSO influence over the entire Amazon Basin is observed mostly during its rainy season.
On the other hand, the relationships of the dry season cases with the AEM confirm previous findings that the Amazon rainfall relation to the tropical Atlantic SST is mainly noted during the austral autumn (Ronchail et al., 2002). More specifically, Yoon and Zeng (2010) found that the southern Amazon rainfall is strongly influenced by the TSA SST variability during the wet-to-dry transition season which spans through the austral autumn months.
Another interesting result of the present analysis is the role of the inter-Pacific-Atlantic anomalous SST gradient mode. Except for the very dry rainy season, the other extreme cases (very wet rainy season, very dry dry season, and very wet dry season) seem to be closely related to an intense inter-Pacific-Atlantic SST gradient mode which acts to reinforce the effect of the tropical Atlantic SST anomalous mode (Table IV). The intense inter-Pacific-Atlantic SST gradient mode is also an important factor to distinguish the moderate (wet and dry) cases from the extreme cases (very wet and very dry). This is particularly noticeable for the dry season.
The inter-hemispheric and inter-Pacific-Atlantic SST gradient modes occurring simultaneously have a concordant (drying or wetting) effect on the central and eastern Amazon region through the direct anomalous circulation in the east-west (Atlantic Walker circulation) and north-south directions. The drying (wetting) effect is due to the descending (ascending) branch of the east-west and/or north-south direct anomalous circulation cells which act to impede (enhance) the tropical convection in the central and eastern Amazon. These zonal and meridional circulation cells are driven by the SST anomalies in the tropical eastern Pacific and tropical Atlantic regions with ascending (descending) motions in the areas with positive (negative) SST anomalies. The results here, especially in relation to the role of the inter-Pacific-Atlantic gradient in the central and eastern Amazon precipitation, have not been discussed before and should be taken into account in the diagnostic activities.
The authors thank the Conselho Nacional de Desenvolvimento Científico e Tecnológico of Brazil for research support (CNPq, 569749/2008-5 and INCT/CEAB-UEA). All authors were partially supported by the Conselho Nacional de Desenvolvimento Científico e Tecnológico of Brazil. The authors thank the two anonymous reviewers for their helpful suggestions.