The Amazon Basin is the largest tropical river basin in the world, covering approximately 6.9 million km² and encompassing several countries in South America. According to Feldpausch et al. (2012), it is characterized by a vast and dense tropical forest, a complex hydrographic network, and significant spatial variability in topography and hydrological conditions, as highlighted by Sioli (1984). The region’s climate is predominantly equatorial, with high humidity and annual precipitation ranging from 1,500 to over 3,000 mm, with well-defined rainy and dry seasons depending on the sub-basin and local topography (Salati and Vose, 1984).

The Amazon basin has been divided into nine main sub-basins, each exhibiting distinct hydrological and climatic characteristics. In this study, we focus on the Japurá, Negro, Trombetas–Uatumã, Paru–Jari, Javari–Juruá, Purus, Madeira, Tapajós, and Xingu sub-basins, as shown in Figure 1. These sub-basins cover a wide latitudinal and longitudinal gradient, encompassing areas with contrasting precipitation patterns and hydrological regimes. The Central Amazon region is highlighted with a red “X” to indicate the location of the ANA hydrological station in Manaus, which serves as a key reference for river level monitoring and hydrological analysis.

This spatial framework allows for detailed investigation of oceanic climate mode influences (e.g., ENSO, EeamAM, Atlantic Dipole) on droughts and floods, capturing heterogeneous responses among sub-basins and providing operationally relevant insights for hydrological monitoring and early warning systems.

The SST data consist of a reconstructed global monthly series for each grid point, with spatial resolution of 2° latitude by 2° longitude (Smith et al., 2008). These data are available at https://psl.noaa.gov/data/gridded/data.noaa.ersst.v5.html for the period from 1950 to 2022.

The Palmer Drought Severity Index (PDSI) is a widely used climatic index designed to quantify the severity and duration of droughts and wet periods by assessing soil moisture anomalies. Developed by Wayne Palmer in 1965, the index integrates monthly temperature and precipitation data with soil water-holding capacity, enabling it to reflect deviations from normal moisture conditions relative to local climatology.

This study employs the self-calibrated Palmer Drought Severity Index (scPDSI), which improves upon the original PDSI by dynamically calibrating its parameters to local climate variability, enhancing accuracy across diverse climatic regions (Wells et al., 2004). The scPDSI accounts for regional differences in moisture availability and evapotranspiration, making it particularly suitable for heterogeneous biomes such as the Amazon.

The dataset used is provided by the NOAA Physical Sciences Laboratory (PSL) and covers the global land surface from 1948 to 2014, with a spatial resolution of 2.5° x 2.5°. This long-term monthly series enables robust trend analyses and characterization of hydroclimatic variability over decadal timescales. The self-calibrated dataset is publicly available at: https://psl.noaa.gov/data/gridded/data.pdsi.html. PDSI values are standardized so that negative values indicate drought severity (with values below −4 representing extreme drought), and positive values indicate increasingly wet conditions.

The scPDSI has been extensively applied in hydrological and climate studies worldwide, including investigations of drought trends and impacts in tropical regions, providing a reliable metric for monitoring moisture deficits and excesses relevant to ecosystem and water resource management.

The Manaus fluviometric station, identified by code 14990000, has been monitoring water levels of the Rio Negro in the Manaus region from 1903 to the present. The data collected by this station are provided by the Brazilian National Water and Basic Sanitation Agency (ANA) through the Hidroweb system, available at: https://www.snirh.gov.br/hidroweb/serieshistoricas. These historical records are essential for hydrological analyses, water resource management, and studies related to extreme events such as floods and droughts in the Central Amazon region. A station Manaus was selected due to its central geographic location within the Amazon Basin. Positioned at the confluence of major sub-basins in the lower Rio Negro region, it serves as a hydrological integrator. Typically, when low water levels are observed at this station, drought conditions are already affecting many upstream tributaries. Therefore, this central station offers a reliable proxy for broader basin-scale drought assessment.

To select the dominant modes of variability in the tropical Pacific and Atlantic, Empirical Orthogonal Functions (EOF) were applied to total SST anomalies in the tropical Pacific (180°–80°W; 20°S–20°N) and tropical Atlantic (70°W–10°E; 20°S–20°N) for the period from 1950 to 2022. EOF is a useful tool for identifying the principal modes of temporal and spatial variability in a given region. A key aspect in EOF analysis is the identification of statistically independent modes; according to the criterion proposed by North et al. (1982), a mode can be considered physically independent if its associated eigenvalue is well separated from the subsequent ones. In this study, the first two EOF modes in each basin meet this criterion. The associated principal components (PCs), also known as expansion coefficients, are normalized time series that describe the temporal evolution of the spatial patterns.

The methodology of this study adopts a statistical approach to investigate the relationship between climate patterns and drought severity in the Amazon region, using EOF analysis. The EOF technique is applied to reduce the dimensionality of the climate datasets and to identify the main modes of variability, extracting the dominant patterns that account for most of the observed variability. In this study, the first two EOF modes are associated with Pacific climate patterns—such as Central El Niño and Eastern El Niño—and Atlantic modes—such as the Equatorial Atlantic Mode (EAM) and the Atlantic Dipole.

The spectrum of each selected principal component (PC) was examined through Wavelet Transformation. The Morlet wavelet is particularly suitable for SST variability studies, as its oscillatory pattern resembles that of sea surface temperature fluctuations (Torrence and Compo, 1998). In addition to the normalized series, we present the dominant variability scales along with local and global power spectra for each pair of oceanic PCs.

Spatial correlation analysis was employed to assess the relationship between the climate modes identified by EOF and the PDSI in the Amazon. This technique helps identify areas within the Amazon Basin that are more sensitive to large-scale oceanic variability, offering a clearer picture of regional impacts associated with climate anomalies. The contributions of each oceanic mode are distinct, and the correlation between the PCs and the PDSI highlights which aspects of each mode are more closely linked to drought or increased moisture in specific regions. The PDSI integrates monthly temperature and precipitation data, along with soil water holding capacity, thereby enabling a robust assessment of how variability in oceanic modes translates into climate anomalies at the land surface.

The study also investigates the monthly PC anomaly fluctuations during years of extreme hydrological events in Central Amazonia. Historical river level data from the Manaus gauge station on the Lower Negro River sub-basin (provided by ANA) were used to represent the central region of the basin and to identify years with the most severe drought and flood events. The analysis of monthly anomalies enables the identification of SST variations during key periods leading up to the peaks of local droughts and floods.

In summary, this methodology integrates statistical techniques—including the correlation of oceanic PCs with the PDSI—to explore how and when Pacific and Atlantic climate patterns influence extreme events in the Amazon. The inclusion of hydrological data from Central Amazonia, in combination with historical climatological context, is critical for identifying local events and understanding their real-world impacts in the region.

The results presented here refer to the EOF analysis and their corresponding PCs for the Tropical Pacific region, highlighting dominant climate variability patterns. The first mode, which explains 53.9% of the total variance, shows a broad positive anomaly covering a large portion of the Tropical Pacific (Figure 2a). The most intense anomalies are centered over the central Pacific, particularly between 150°W and 120°W. This pattern is characteristic of warming events such as El Niño, marked by abnormally warm sea surface temperatures in the central Pacific basin.

The first Principal Component (PC1), associated with Mode 1 (Figure 2b), illustrates the temporal evolution of this pattern between 1950 and 2022. A clear alternation between positive and negative phases is observed, reflecting typical El Niño (positive peaks) and La Niña (negative trough) events.

Strong Central El Niño events, such as those observed in 1957–58, 1972–73, 1982–83, 1997–98, and 2015–16, stand out as positive peaks in PC1. In contrast, negative phases indicate the occurrence of La Niña events, including 1956–57, 1976–77, 1987–88, 2007–08, 2011–12, and 2021–22. This temporal pattern highlights the importance of interannual variability associated with the ENSO, the primary driver of climate anomalies in the tropical Pacific region.

The second mode (Mode 2), shown in Figure 2c and explaining 13.3% of the variance, reveals a dipole pattern with negative anomalies in the western Pacific and positive anomalies in the eastern sector, between 120°W and 90°W. This structure is characteristic of Eastern Pacific El Niño events, marked by alternating warming and cooling phases in that region. The associated PC2 exhibits smaller amplitudes over time but maintains a consistent phase alternation, with more persistent peaks compared to PC1.

Eastern El Niño events, such as those in 1965–66, 1972–73, 1982–83 and 1997–98, appear as strong positive peaks in PC2. The overlap between these peaks and the aforementioned events suggests that, in some cases, strong Central El Niños were either preceded or followed by Eastern-type episodes, which may have helped prolong or intensify the Pacific warming phase. However, not all Eastern El Niño events are associated with Central events—some occur independently, with the central Pacific region remaining cool or even neutral.

During negative phases, Eastern La Niña events are evident in the years 1954–55, 1959–60, 1967–68, 1987–88, and 2013–15. This suggests that the Niño 1 + 2 region exhibited below-average temperatures prior to the strong 2015 El Niño. In general, Mode 2 is associated with lower-amplitude variability, capturing intra-seasonal fluctuations or secondary climate patterns (Figure 2d).

Figures 3a,e display the time series of the first two PC1 and PC2, respectively in black, while the gray lines represent the normalized versions of these series. The wavelet power spectra (Figures 3b,f for PC1 and PC2, respectively) illustrate the variation in spectral power across different timescales over the study period. In these panels, the shaded gray areas denote the cone of influence, where edge effects may reduce the reliability of the results. Solid black contours highlight regions of statistically significant power at the 95% confidence level, assessed against a red-noise background spectrum.

Complementarily, the global wavelet spectra (Figures 3c,g) for PC1 and PC2, respectively, represent the average power of all periodicities across the entire time series. In these plots, the dashed horizontal lines indicate the 95% confidence threshold, with peaks above this line corresponding to statistically significant dominant periodicities. The solid gray lines show the smoothed global wavelet power spectra.

Finally, the scale-averaged power time series (Figure 3d for PC1 in the 2–8 year band, and Figure 3h for PC2 in the 1–4 year band) depict the evolution of oscillatory power within specific frequency bands over time. In these graphs, the dashed horizontal lines represent the mean background variance used in significance testing, helping identify periods of enhanced oscillatory activity.

The wavelet transform spectrum of the PC1 time series (Figure 3a) shows strong interannual variability, with pronounced peaks corresponding to major El Niño and La Niña events. The wavelet power spectrum (Morlet) (Figure 3b) reveals the highest energy concentration (in yellow shades) around 2- to 8-year periods, particularly between 1960 and 2000, and again after 2005.

These periods are directly associated with the typical interannual variability of ENSO. In Figure 3c, the global wavelet spectrum displays the average-to-high energy across different period scales for the entire time series. Two well-defined peaks are observed around the 2- to 8-year band, confirming the relevance of ENSO cycles. The dashed line indicates the statistical significance level, showing that variability around these periods is both relevant and statistically significant. In Figure 3d, the mean power in the 2–8 year band reflects the temporal evolution of energy, capturing the strength of ENSO events across decades. A particularly active phase of interannual variability is evident, with peaks occurring during 1955–56, 1964–65, 1972–73, 1982–83, 1987–88, 1997–98, and 2009–10.

The time series of PC2 (Figure 3e) reveals more frequent interannual variability compared to PC1, including peaks that correspond to some Central El Niño events. The wavelet transform applied to PC2 (Figure 3f) highlights the variability of this component at different time scales, with a notable concentration in the 1- to 4-year range, along with other frequent patterns.

The global wavelet spectrum (Figure 3g), summarizing contributions across scales, confirms that the highest power concentration is found in the 1–4 year band. The mean wavelet power (Figure 3h) highlights peaks associated with extreme climate events, such as 1965–66, 1972–73, 1982–83, 1987–88, 1991–92, 1997–98, 2008–09, and 2015–16. These peaks demonstrate that PC2 captures the direct influence of such episodes, which impact not only global climate, but also hydroclimatic variability in the Amazon Basin.

In Figure 3h, the mean power in the 1–4 year band shows the temporal evolution of energy, reflecting the increased frequency of Eastern ENSO events throughout the decades. In summary, PC2 represents dynamics characterized by shorter cycles, with direct links to extreme climate events associated with ENSO. The oscillations in PC2 occur on a higher-frequency scale, with less intense anomalies, predominantly located in the eastern Pacific. This variability has important implications for climate forecasting and for monitoring impacts on Amazonian subregions, since phase shifts in ENSO influence rainfall distribution and the occurrence of droughts and floods in the region.

Based on these analyses—including Pearson correlations, wavelet coherence and phase (Supplementary material)—the identified anomalous modes can be interpreted as distinct physical patterns widely recognized in the literature. Mode 1 represents a Central El Niño pattern, while Mode 2 corresponds to an Eastern El Niño. These results are consistent with the findings of Kao and Yu (2009) and Feng et al. (2020), who identified the existence of two ENSO types: Central Pacific or Modoki El Niño, and Eastern Pacific or Canonical El Niño (Ashok et al., 2007).

In the studies by Kao and Yu (2009), using a method that combined EOF and linear regression, the authors separated and examined the structure, evolution, and possible teleconnections with other regions of the globe due to the two modes. Studies have shown that one mode is characterized by positive SST anomalies confined to the central equatorial Pacific (Kao and Yu, 2009; Feng et al., 2020), while the other is associated with westward propagation of positive SST anomalies in the eastern equatorial Pacific near the coast of South America (Eastern El Niño) (Rasmusson and Carpenter, 1983). This spatial variability implies different convection patterns and atmospheric responses to warming, which can produce different climate teleconnections worldwide (Ashok et al., 2007); however, studies specifically focused on the Amazon region detailing the spatiotemporal variability of these effects remain quite scarce.

The EOF analyses for the Atlantic region (70°W–10°E; 20°S–20°N) for the period from 1950 to 2022 reveal modes of variability that explain 47.2% of the total variance of the data series for Mode 1 and 19.6% for Mode 2 (Figures 4a,c). The first mode shows negative anomalies throughout the ocean basin, with maximum intensity near the Equator, around 30°W, with marked positive events occurring in the years 1969, 1974, 1987, 1998, 2010, and 2020. In contrast, negative events are recorded in 1964, 1967, 1972, 1976, 1993, 1996, 2012, 2018, and 2021 (Figure 4a).

The second mode (EOF 2) (Figure 4c) displays a dipolar pattern, with positive anomalies in the Northern Tropical Atlantic region centered at 15°N, 40°W, and negative anomalies in the Southern Tropical Atlantic, with the most intense area centered at 15°S, 10°W. The main positive peaks are observed in 1957, 1970, 1981, 1996, and 2005, highlighted in the series, while the negative peaks indicate the occurrence of a Negative Dipole, such as the events in 1974, 1989, 1994, and 2018 (Figure 4d).

The Figure 5a presents the original (black line) and normalized (gray line) time series of PC1 for the Atlantic, revealing a peak in the global wavelet spectrum occurring approximately at the 2-4 year scale, with more significant peaks restricted to the 1-4 year scale (Figures 5c,d). Although the signal at this scale is not significant throughout the entire time series, periods of intensified energy for this variability are observed during the 1960s-1970s, 1995-2000 and again around 2010, indicating that interannual variability was more intense during these intervals (Figure 5b). Figure 5c shows a high-power cycle suggesting that these peaks tend to repeat approximately every 20 years, which may indicate a decadal cycle or pattern associated with the Tropical Atlantic.

Figure 5e presents the original (black line) and normalized time series (gray line) of PC2. The panel shows temporal fluctuations of PC2, with intensified variability during certain decades, such as the periods between 1958, 1970-1980 and from 2005 to recent years. These oscillations may indicate specific intervals in which the Tropical Atlantic variability was particularly active (Figure 5f). The wavelet spectral analysis of PC2 reveals a peak in the global wavelet spectrum approximately at the 6-month to 1-year scale (Figure 5f), with energy maxima occurring during 1954-55, 1968-72, 1978-79, 1982-83, 2001-02, 2009-10, 2015-16 and again in 2020 (Figure 5h). This wavelet power spectrum identifies the dominant frequencies in the series over time at intraseasonal scales, indicating a faster oscillation.

A strong pattern of variability is observed at time scales between 8 and 16 years, mainly before the 1970s (Figure 5g). This suggests that PC2 is relevant at multi-annual and decadal time scales, indicating more intense oscillations during some decades, suggesting rapid changes in the Tropical Atlantic SST. This variation is highly irregular, indicating that this temporal band may be influenced by external forcings and rapid interactions.

Results from the EOF analysis for the Atlantic region reveal that the first identified mode corresponds to the EAM, while the second mode represents the Atlantic Dipole. Spectral analyses suggest predominantly intraseasonal variability, with cycles between 1 and 4 years for the EAM and between 6 months and 1 year for the Dipole, alongside multiyear and decadal patterns. This irregular variability highlights the Atlantic Oceans influence on SST modulation and its interactions with regional climate.

The EAM impacts the global hydrological cycle by modulating precipitation over parts of South America, whereas the Atlantic Dipole can intensify or suppress African monsoons and moisture transport toward the Caribbean, southern United States, and Brazil. Both modes affect not only seasonal patterns but also the formation of tropical cyclones, severe drought events, and regional atmospheric dynamics, including shifts in the ITCZ and Walker and Hadley circulations.

This section presents the characterization of the study area and the spatial correlation patterns between oceanic climate modes and hydrological variability in the Amazon. Figure 1 highlights the sub-basins of the Amazon Basin, including Japurá, Rio Negro, Jatapu, Trombetas, Jutaí, Purus, Madeira, Tapajós, and Xingu, with the location of the ANA hydrological station in Manaus indicated on the map. This spatial delineation is essential for understanding regional hydrological responses to oceanic oscillations.

Figure 6 presents the spatial correlation patterns between the PDSI and major climate modes in the Tropical Pacific and Atlantic, considering the full time series and the months of January and July. The maps highlight how each mode influences hydrological conditions in different seasons, with blue and red areas indicating positive and negative correlations, respectively. Correlation values above 0.4 and below −0.4 are considered statistically significant at the 95% confidence level, based on the Student’s t-test adjusted for autocorrelation in the time series. This analysis allows the identification of areas within the Amazon that are more sensitive to oceanic variability, contributing to the understanding of climatic impacts on extreme hydrological events in the region.

The correlation pattern between the Central El Niño (PC1) and the PDSI across the Amazon sub-basins reveals a spatially homogeneous response. Generally, the Javari, Japurá, Purus, Negro, Trombetas/Uatumã, and Paru/Jari sub-basins exhibit negative correlations with Central El Niño events, indicating these areas experience intensified drought conditions during such events. In contrast, the southern portion of the Amazon basin shows positive correlations, suggesting enhanced moisture during Central El Niño events, both on the annual mean scale (Figure 6a) and seasonally (Figures 6e,i). The Madeira, Tapajós, and Xingu sub-basins display weaker negative signals in the north and positive correlations in the south, indicating a dipolar impact pattern associated with Central El Niño years.

In the case of the Eastern El Niño, a spatiotemporal variability with more moderate effects and less intense impacts is observed in areas mostly influenced by Central El Niño (Figures 6b,f,j). Furthermore, positive correlations between the Eastern El Niño and PDSI are found in the western portion of the international Amazon, extending to the Upper Rio Negro. This suggests increased moisture during these events in these regions, whereas during Eastern La Niña years, drought conditions may develop in the Upper Rio Negro sub-basin and in parts of northern Peru, Ecuador, and Colombia, with more pronounced effects in January (Figure 6f).

In summary, Central El Niño induces more significant and spatially coherent impacts, with a clear spatial division wherein northern and eastern basins tend to experience severe droughts, while southern basins are less affected or even benefit from increased moisture during these events. Conversely, Eastern El Niño exhibits a more heterogeneous pattern with less intense and more temporally and spatially variable effects. These differences have important implications for climate forecasting and resource management within Amazon sub-basins, highlighting the necessity for regionally tailored approaches considering the specific characteristics of each climate mode.

Regarding the EAM, a positive correlation is observed over most of the Amazon for the full time series (Figure 6c). This indicates that during warm (cold) phases of the EAM, the PDSI tends to show higher (lower) values, favoring wetter (drier) conditions in the region, with a stronger influence observed during the peak rainy season (Figure 6g). The Javari-Juruá, Purus, Madeira, Japurá, and Negro sub-basins are particularly affected by increases (decreases) in moisture.

Conversely, in July (Figure 6k), negative correlations appear in the central and northeastern Amazon basin, including the middle and lower Rio Negro, Trombetas/Uatumã, and Paru-Jari regions (Figure 6c). This suggests that during this period, drier (wetter) conditions prevail in these areas when the Equatorial Atlantic is cold (warm). However, further south, such as in the Purus, Madeira, Tapajós, and Xingu basins, Equatorial Atlantic cooling (warming) may maintain moisture (dry) conditions.

Overall, the EAM exerts a pronounced seasonal effect on the Amazon, especially in January, influencing the quality of the rainy season. In July, it creates drought conditions in the middle and lower Rio Negro, Trombetas-Uatumã, and Paru-Jari during the transition to the dry season.

For the Atlantic Dipole, negative correlations dominate the full series (Figure 6d). This implies that during positive (negative) Dipole events, there is a tendency toward drier (wetter) conditions across nearly the entire Amazon, with stronger impacts in the pre-dry season period (Figure 6l). This effect is particularly marked in the Japurá, Purus, Madeira, Trombetas-Uatumã, Tapajós, Xingu, and Paru-Jari basins, especially in July (Figure 6l). When the positive (negative) Atlantic Dipole pattern develops during the rainy season, the Japurá and Negro sub-basins experience increased (reduced) moisture, whereas other basins display opposite but generally moderate signals (Figure 6h).

These results suggest that Atlantic climate modes exert a seasonal control on the variability of moisture regimes across the Amazon basin. The positive Dipole is consistently associated with droughts, exerting greater influence in northeastern and southern sub-basins, with its impact strongly modulated by occurrence in the months preceding the dry season, acting as an amplifier of drought conditions. Conversely, the warm (cold) phase of the EAM is associated with enhanced (reduced) moisture over much of the Amazon, particularly during the rainy season, but may contribute to drought conditions in the middle and lower Rio Negro, Trombetas/Uatumã, and Paru-Jari regions during the transition to the dry season. Understanding this seasonality is crucial for climate prediction and water resource management in the region, emphasizing the need for regionally adapted strategies to mitigate the seasonal impacts of each climate mode.

The graph presented in Figure 7a provides a detailed view of the variability in water levels of the Rio Negro at the Manaus gauge station over the period from 1950 to 2024, highlighting extreme flood and drought events. The historical time series reveals well-defined patterns, characterized by pronounced annual oscillations with distinct peaks and troughs, as well as an increase in the frequency and intensity of extreme events in recent decades.

The most intense flood events, highlighted in blue, occurred in the years 1976, 2009, 2012, 2015, 2019, 2021, and 2022, whereas the most severe droughts, marked in red, include the years 1958, 1963, 1995, 1997, 2005, 2010, 2023, and potentially 2024.

The analysis of extreme flood events compared to the historical climatology shows that the most critical years exhibited water levels significantly above the long-term average, with peak stages occurring between May and July (Figure 7b). The 2022 event stands out as one of the most intense, surpassing all previously recorded floods. Following the peak months, water levels sharply decline from August onwards, gradually returning to climatological mean values. The recurrence of these floods suggests an increase in the magnitude of inundation events, which may be linked to large-scale climatic drivers such as Pacific and Atlantic oceanic oscillations.

Conversely, the analysis of extreme drought events indicates a pronounced reduction in river levels, especially starting in August, reaching the lowest values between October and November (Figure 7c). The years 2023 and 2024 exhibit exceptionally low water levels, positioning the 2024 drought among the most severe in the historical record. The progressive intensification of droughts points to a worsening hydrological condition in the region, underscoring the necessity to investigate the influence of oceanic modes, including El Niño/La Niña and Tropical Atlantic anomalies, on the hydrological variability of the Amazon basin.

Based on these data, the selected years for further analysis consider both extreme flood and drought events. This approach enables the exploration of possible links between these critical periods and the prevailing oceanic conditions on annual and seasonal timescales.

The graphs presented in Figure 8 depict the temporal evolution of anomalies in the primary tropical oceanic modes—Central El Niño, Eastern El Niño, EAM, and Atlantic Dipole—across the months corresponding to the extreme flood years in the Amazon basin: 1976, 2009, 2012, 2015, 2019, and 2021.

Each graph in Figure 8 represents a specific flood year, highlighting how these oceanic phenomena fluctuated amid the region’s hydrological conditions. For the 1976 flood (Figure 8a), a predominant influence of a Central La Niña (red) is observed, with persistent negative anomalies from January to August, coinciding with the rising flood stage in Central Amazonia. Concurrently, an Eastern El Niño (yellow) developed, characterized by positive anomalies in the Niño 1 + 2 region, while negative anomalies persisted in the central Pacific throughout the year. The negative correlation between these two tropical Pacific modes and the Palmer.

The PDSI indicates increased soil moisture during the region’s rainy season. The EAM (green) remained in a positive phase throughout the year, which, as shown by the correlation analysis, supported a regime of persistently higher humidity. The Atlantic Dipole (blue) exhibited a negative phase during the first half of the year and a positive phase during the second. Given its negative correlation with the PDSI, this configuration favors wetter conditions during the first semester. Collectively, the 1976 flood was driven by the influence of a Central La Niña and a negative Atlantic Dipole in the first half of the year, along with a positive-phase EAM and a persistent Eastern El Niño throughout the year. These oceanic patterns promoted enhanced moisture availability, contributing to the intensification of the extreme event in the region.

The oceanic anomaly pattern during the 2009 flood exhibited similarities to 1976 (Figure 8b), but with notable differences. In 2009, the Atlantic Dipole displayed a stronger negative phase in the first semester, while the Central La Niña and Eastern El Niño were less intense compared to 1976. These variations may have played a crucial role in amplifying the 2009 flood, which exceeded the magnitude of the 1976 event.

In 2012 (Figure 8c), a combined influence of a Central La Niña and an Eastern El Niño is again observed, similar to previously analyzed years. However, both the EAM and the Atlantic Dipole exhibited strong positive phases throughout the year. This pattern suggests that the warm EAM, due to its positive correlation with the PDSI, contributed to maintaining elevated moisture levels during the entire period, mitigating the usual drying effects associated with a positive Dipole phase.

In 2015 (Figure 8d), an intense Eastern La Niña combined with a Central El Niño (red) during the first semester is evident. The EAM remained in a positive phase, while the Atlantic Dipole showed a negative phase in the first semester. This combination favors increased moisture in the region, especially during this period. In 2019 (Figure 8e), a similar pattern in tropical SST anomalies is identified, although the negative Atlantic Dipole exhibited a more pronounced influence, further intensifying moisture conditions in the Amazon.

The extreme 2021 (Figure 8f) event was characterized by a strong influence of an intense Central La Niña combined with a weaker Eastern La Niña. Additionally, the EAM maintained a positive phase throughout the year, contributing to persistent moisture during this period.

These results suggest that extreme floods in the Amazon are strongly associated with the presence of a Central La Niña, a positive EAM, and a negative phase of the Atlantic Dipole, especially during the first semester. Such oceanic configurations favor a sustained increase in regional moisture, thereby intensifying hydrological extreme events.

The analysis of tropical oceanic modes contributions during drought years in Central Amazonia reveals a predominant influence of El Niño events. These events may be confined to the first semester, as observed in 1995 (Figure 9c), 2005 (Figure 9e), and 2010 (Figure 9f), persist throughout the entire year, as in 1958 (Figure 9a), or progressively intensify, as in 1963 (Figure 9b) and 1997 (Figure 9d). These patterns indicate that variability in Pacific SST anomalies plays a key role in moisture reduction and drought intensification in the region. This influence aligns with the negative correlations identified in the PDSI analysis, highlighting the strong linkage between El Niño events and decreased moisture availability in Central Amazonia.

In drought years where El Niño’s influence was more restricted to the first semester, such as 1995, 2005, and 2010, a positive phase of the Atlantic Dipole is observed either throughout the year (Figures 9e,f) or during the second semester (Figure 9c). Given the negative correlation between the Atlantic Dipole and the PDSI, this pattern corresponds to a significant reduction in moisture during this period, thereby exacerbating drought conditions in the Amazon. This mechanism compounds the drought effects already established at the end of the rainy season due to El Niño’s influence.

For El Niño events persisting throughout the year, such as in 1958 (Figure 9a), the positive Atlantic Dipole also played a crucial role in intensifying the drought by prolonging its duration and expanding the spatial extent of impacted areas. Correlation analyses with the PDSI demonstrate that El Niño primarily affects the central and northern regions, while the positive Atlantic Dipole exerts greater influence over central, southeastern, and southern areas.

In years when El Niño events progressively intensified, such as in 1963 and 1997, a simultaneous evolution of a positive Eastern Pacific El Niño was also observed, resulting in large-scale events with gradually increasing impacts throughout the year. In the second half of the year, the action of the EAM in its negative phase suggests a reduction in moisture during the period preceding the dry season in the Amazon. This is supported by the negative correlation between its July signal and the PDSI (Figures 9b,d). This pattern indicates that the negative phase of the EAM can amplify drought effects by significantly influencing the transition between seasons and intensifying hydrological deficits in the region.