Seasonal and Interannual Variations in Hydrological Dynamics of the Amazon Basin: Insights from Geodetic Observations

Abstract

The Amazon Basin plays a crucial role in the global hydrological cycle, where seasonal and interannual variations in terrestrial water storage (TWS) are essential for understanding climate–hydrology coupling mechanisms. This study utilizes data from the Gravity Recovery and Climate Experiment (GRACE) satellite mission and its follow-on mission (GRACE-FO, collectively referred to as GRACE) to investigate the spatiotemporal dynamics of hydrological mass changes in the Amazon Basin from 2002 to 2021. Results reveal pronounced spatial heterogeneity in the annual amplitude of TWS, exceeding 65 cm near the Amazon River and decreasing to less than 25 cm in peripheral mountainous regions. This distribution likely reflects the interplay between precipitation and topography. Vertical displacement measurements from the Global Navigation Satellite System (GNSS) show strong correlations with GRACE-derived hydrological load deformation (mean Pearson correlation coefficient = 0.72) and reduce its root mean square (RMS) by 35%. Furthermore, the study demonstrates that existing hydrological models, which neglect groundwater dynamics, underestimate hydrological load deformation. Principal component analysis (PCA) of the Amazon GNSS network demonstrates that the first principal component (PC) of GNSS vertical displacement aligns with abrupt interannual TWS fluctuations identified by GRACE during 2010–2011, 2011–2012, 2013–2014, 2015–2016, and 2020–2021. These fluctuations coincide with extreme precipitation events associated with the El Niño–Southern Oscillation (ENSO), confirming that ENSO modulates basin-scale interannual hydrological variability primarily through precipitation anomalies. This study provides new insights for predicting extreme hydrological events under climate warming and offers a methodological framework applicable to other critical global hydrological regions.

1. Introduction

Terrestrial water storage (TWS) changes constitute a crucial component of the global water cycle, reflecting processes such as precipitation, evapotranspiration, runoff, and groundwater dynamics [1,2,3,4]. Variations in TWS perturb the local gravity field, enabling their monitoring via the Gravity Recovery and Climate Experiment (GRACE) satellite mission. Additionally, TWS changes induce elastic deformation of the Earth’s surface, predominantly manifesting as vertical displacements [5,6], which can be directly observed using Global Navigation Satellite System (GNSS) data. However, GRACE observations are limited by their spatial resolution (approximately 300 km) and increasing noise levels at higher spherical harmonic (SH) degrees [7,8]. Simultaneously, GNSS observations are susceptible to various environmental factors, making it challenging for single-site data to fully capture regional-scale dynamics. Consequently, the synergistic integration of GRACE and GNSS technologies offers a novel approach to overcome these individual limitations, providing a more comprehensive perspective for three-dimensional surface deformation monitoring.

Launched in March 2002, the GRACE satellite mission provided 15 years of continuous observations until a one-year data gap, succeeded by its successor mission, GRACE Follow-On (GRACE-FO, hereinafter collectively referred to as GRACE), deployed in 2018 [9]. GRACE has significantly advanced the monitoring of global gravity field variations and climate observations [2,8]. Wahr et al. (1998) [10] established the theoretical foundation for extracting TWS changes from GRACE data, enabling its application to studies of hydrological and oceanic processes. Building upon this framework, numerous studies have utilized GRACE-derived products to investigate diverse geophysical phenomena, including regional TWS variations [11,12], ice sheet dynamics [13,14], tectonic activity [15,16], and sea-level change [17,18].

Extensive research has investigated hydrological loading deformation within the Amazon Basin. Davis et al. (2004) [19] pioneered the combination of GRACE and Global Navigation Satellite System (GNSS) data to assess the impact of climate-driven hydrological processes on solid Earth deformation, revealing a strong correlation between GRACE-observed seasonal gravity variations and GNSS-recorded vertical surface deformations in the Amazon Basin. Bevis et al. (2005) [20] provided the first evidence of the elastic crustal response to seasonal mass fluctuations in the Amazon River system, elucidating the mechanisms of instantaneous surface deformation induced by large-scale hydrological loads. GNSS observations from the Manaus station (central Amazon) revealed annual vertical displacements of 50–75 mm, significantly exceeding model predictions. Fang et al. (2021) [21] utilized 71 GNSS stations and GRACE data to examine nearly two decades of vertical crustal deformation in the Amazon Basin, reporting a high correlation (average correlation coefficient = 0.75) between GNSS-observed vertical deformations and GRACE-derived mass load changes. Youm et al. (2023) [22] developed a high-resolution surface mass load distribution method for the Amazon Basin by integrating GRACE data with hydrological models. The refined mass load distribution demonstrated improved agreement with GNSS vertical deformation on both seasonal and interannual timescales, particularly near the main Amazon River, where enhancements over direct GRACE predictions were significant. Wang et al. (2023) [23] compared SH and mass concentration (mascon) products from GRACE data, revealing spatiotemporal characteristics of GNSS vertical load deformations and their consistency with GRACE. Mascon products provided higher spatial resolution and signal-to-noise ratio than SH products, especially in the central Amazon Basin, where they more accurately captured significant deformation signals caused by seasonal water mass variations.

Overall, existing studies on the Amazon basin primarily focus on seasonal hydrological fluctuations, with limited understanding of interannual variations and their potential links to the El Niño–Southern Oscillation (ENSO). Xavier et al. (2010) [24] investigated the relationship between interannual changes in TWS and ENSO in the Amazon basin. They found that, during 2003–2008, the derivative of basin-averaged water storage from GRACE data was highly correlated with the Southern Oscillation Index, confirming that the spatiotemporal variations in the basin’s hydrology are partially driven by ENSO. However, their study was based on fewer than 10 years of data, which limits its scope. Additionally, no comparison was made with GNSS displacement observations.

This study examines seasonal and interannual hydrological dynamics of the Amazon Basin by integrating GNSS and GRACE data. First, we analyze the spatial distribution and temporal evolution of TWS using GRACE observations. Next, we quantify the elastic displacements caused by hydrological mass changes by combining GRACE and GNSS data and compare these with estimates from hydrological models. Finally, we identify oscillation patterns and interannual signals in hydrological fluctuations and explore their relationship with ENSO.

2. Materials and Methods

2.1. GNSS Data and Processing

The GNSS data utilized in this study were sourced from the Nevada Geodetic Laboratory (NGL) (http://geodesy.unr.edu) (accessed on 1 May 2025). NGL integrates core stations from the International GNSS Service (IGS) and regional densification sites, encompassing over 20,000 continuously operating GPS/GNSS reference stations globally. This dataset provides high spatial resolution with relatively uniform land-based station distribution. It includes raw observation files (RINEX format), precise satellite orbit/clock products, and time series solutions. Precise Point Positioning (PPP) processing was performed using GIPSY/OASIS software (version 6.2) incorporating ambiguity resolution techniques to enhance accuracy. The positioning precision achieved is better than 2 mm horizontally and 5 mm vertically at daily resolution. The dataset spans from the 1990s to the present, supporting extensive studies on long-term crustal deformation, tectonic motion, glacial isostatic adjustment (GIA), and sea-level change.

This study collected GNSS observation data from 113 sites across the Amazon Basin and its surrounding regions. All station coordinates are aligned with the ITRF2014 terrestrial reference frame (station locations shown in Figure 1). Notably, most GNSS stations are located along the periphery of the Amazon rainforest, with fewer stations in the central region. The raw GNSS time series exhibited step changes and outliers, primarily attributable to equipment replacements and seismic events. These artifacts were addressed using the TSAnalyzer tool (version V2) [25,26]. Figure 2 presents the data record lengths for the 113 GNSS stations. Specifically, “Sites Number” denotes the number of GNSS measurements from 113 stations at each time point along the x-axis, representing the monthly count of stations with available data. The height of the black region indicates the measurement frequency at each time point. Analysis reveals that data from these 113 stations are primarily concentrated between 2008 and 2021. Consequently, this time period was selected for subsequent principal component analysis (PCA) [27] to ensure the accuracy and robustness of the results. Additionally, twelve representative GNSS stations were randomly selected for subsequent analysis, as indicated by the blue markers in Figure 1.

2.2. GRACE Mascon Data

GRACE (2002–2017) and its successor mission, GRACE Follow-On (GRACE-FO; 2018–2021), measured temporal variations in Earth’s gravity field to infer surface mass changes. The corresponding Level-2 data products were released by the Center for Space Research (CSR) at the University of Texas at Austin, the Jet Propulsion Laboratory (JPL), and Goddard Space Flight Center (GSFC) [28,29,30]. These institutions provide global mascon solutions (gridded data) representing TWS changes with a spatial resolution of 0.25° or 0.5° and a temporal coverage exceeding 20 years.

The mascon approach partitions the Earth’s surface into mass concentration blocks and incorporates prior geophysical constraints (e.g., GIA models), effectively mitigating signal leakage and noise amplification inherent in traditional SH solutions. This study utilized GRACE/GRACE-FO mascon datasets from CSR, JPL, and GSFC, covering the period April 2002 to December 2021.

To calculate the elastic surface displacement induced by hydrological mass loading, we employed the load Green’s function of Wang et al. (2012) [31], based on the Preliminary Reference Earth Model (PREM) [32], within the center of surface figure (CF) reference frame [33]. The detailed methodology is described in Chanard et al. (2014) [34] and Jiao et al. (2024) [35]. During the convolution process, the equivalent water height (EWH) mass change fields from three GRACE mascon datasets were averaged before calculating the displacements [36,37].

2.3. Surface Elastic Loading Models

To more accurately investigate hydrological loading changes, we utilized surface deformation products from the Earth System Modeling group at the German Research Centre for Geosciences (ESMGFZ) in Potsdam. These products were used to remove displacements caused by non-tidal atmospheric loading (NTAL) and non-tidal oceanic loading (NTOL) within the CF reference frame [38].

The ESMGFZ deformation products represent surface displacements resulting from Earth’s rotation effects and mass loads associated with geophysical fluids. These displacements are computed using an Earth system model that assimilates ocean circulation, TWS, and atmospheric data from the European Centre for Medium-Range Weather Forecasts (ECMWF). Specifically:

• NTAL displacements are calculated using ECMWF’s 3-hourly atmospheric surface pressure data. Tidal atmospheric effects are removed through harmonic analysis of the 12 primary tidal constituents.
• NTOL displacements are derived from 3-hourly ocean-bottom pressure data generated by the Max-Planck-Institute Ocean Model (MPIOM).
• Hydrological loading (HYDL) displacements are computed from 24-hourly hydrological variables using the Land Surface Discharge Model (LSDM).

2.4. Meteorological Data

ERA5-Land is a high-resolution global reanalysis dataset that provides a consistent, long-term record of land surface variables [39]. Compared to its predecessor ERA5, ERA5-Land offers enhanced spatial resolution. The reanalysis methodology combines model simulations with global observational data to generate a physically consistent global dataset, thereby accurately reconstructing past climate variations. ERA5-Land utilizes atmospheric variables from ERA5 (e.g., air temperature, humidity) as boundary conditions to constrain the simulated land surface fields—a process termed atmospheric forcing. Without this constraint, model estimates would rapidly diverge from reality. Although observational data are not directly assimilated into ERA5-Land, they indirectly influence the product through the atmospheric forcing provided by ERA5. Additionally, atmospheric variables (temperature, humidity, pressure) are adjusted for elevation differences between the ERA5 forcing grid and ERA5-Land’s higher-resolution grid—a procedure known as bias correction.

The high spatiotemporal resolution of ERA5-Land renders it particularly valuable for land surface applications, including flood and drought forecasting. Its consistent high resolution, extensive temporal coverage, and fixed grid structure provide accurate land-state information valuable to policymakers, businesses, and researchers. In this study, we utilized monthly precipitation and 2 m air temperature data from ERA5-Land, available at: https://cds.climate.copernicus.eu/cdsapp#!/dataset/reanalysis-era5-land-monthly-means?tab=overview (accessed on 1 May 2025).

We also employed the Multivariate ENSO Index (MEI), Version 2, to investigate potential linkages between ENSO and hydrological variability in the Amazon basin. MEI data are accessible at: https://psl.noaa.gov/enso/mei (accessed on 1 May 2025).

2.5. Method

Since GNSS coordinate time series have been corrected for displacements caused by NTAL and NTOL, the seasonal and interannual oscillation signals observed in the Amazon basin GNSS data are most likely attributed to hydrological load changes. In this study, we first processed the daily vertical time series from GNSS stations into monthly data, aligning it with the GRACE and hydrological models. Subsequently, we assessed the Pearson correlation coefficient (PCC) and root mean square (RMS) reduction between the monthly GNSS vertical time series and the elastic load displacements derived from GRACE, as well as with HYDL displacements, using the following formula [40,41]:

$$PCC={\displaystyle \sum _{i=1}^n\left(y_i-\overline{y}\right)\left(x_i-\overline{x}\right)/\sqrt{{\displaystyle \sum _{i=1}^n{\left(y_i-\overline{y}\right)}^2{\displaystyle \sum _{i=1}^n{\left(x_i-\overline{x}\right)}^2}}}}$$

(1)

$$RM{S}_{reduction}={\displaystyle \frac{RM{S}_{GNSS}-RM{S}_{GNSS-GRACE(HYDL)}}{RM{S}_{GNSS}}}$$

(2)

where x_i and y_i represent the two time series corresponding to GNSS-GRACE and GNSS-HYDL at each of the 113 GNSS stations, with means $\overline{x}$ and $\overline{y }$, respectively; n is the number of observations in each time series; RMS_GNSS denotes the RMS value of the monthly GNSS time series, while RMS_{GNSS−GRACE(HYDL)} represents the RMS value of the monthly GNSS time series after subtracting the elastic load displacements derived from GRACE or HYDL.

Given the high dimensionality of the dataset, the large number of GNSS stations, and the presence of noise, spurious signals, and redundant information, dimensionality reduction was applied to the GNSS vertical coordinate time series to extract dominant spatiotemporal patterns. PCA was employed to extract the common-mode components (CMCs) from the GNSS vertical displacement time series across the Amazon basin network. PCA decomposes the spatiotemporal variability of the GNSS vertical displacement field into a set of orthogonal principal components (PCs) representing the dominant modes of variation. The algorithm implementation followed Dong et al. (2006) [27] and Pan et al. (2023) [42].

Additionally, wavelet time–frequency analysis [43] was applied to examine the temporal evolution of spectral characteristics within the hydrological signals of the Amazon basin. This technique facilitates the identification of the timing and magnitude of transient fluctuations, enabling the analysis of interannual hydrological variability in the basin.

Furthermore, to address gaps in the GRACE and GRACE-FO records, as well as the intervals between missions, missing values were interpolated using a non-parametric adaptive method based on Singular Spectrum Analysis (SSA) [44]. This method reconstructs time series by leveraging temporal correlations derived from lagged covariance matrices [45]. It offers significant advantages in filling data gaps and has been successfully applied in global TWS studies, yielding reliable results [44,46,47,48,49].

3. Results

3.1. Spatiotemporal Dynamics of TWS in the Amazon Basin Based on GRACE Data

The TWS variations in the Amazon basin have a significant impact on the surrounding hydrological environment. To investigate this, we applied least-squares fitting to assess the spatial patterns of TWS changes, focusing on trends, annual amplitudes, and phases, using three GRACE Mascon solutions (CSR, JPL, and GSFC) (see Figure 3). The trend analysis reveals mass gain concentrated along the Amazon River and its vicinity, while localized mass loss is evident in the eastern and northern basin. This suggests a long-term redistribution of water resources, benefiting the midstream sections of the Amazon. The annual amplitude patterns show substantial surface-mass variability in the central basin, exceeding 65 cm. Moving outward from the center, the amplitude decreases, particularly in the mountainous western, northern, and southern margins, where variations fall below 25 cm, reflecting relatively stable hydrological conditions. The annual phase analysis shows distinct phase shifts between the northern and southern basin, likely linked to variations in monsoon dynamics and spatial precipitation distribution [21].

Additionally, we computed the latitude-weighted regional average TWS time series for the Amazon basin using CSR, JPL, and GSFC Mascon datasets (see Figure 4a). The results demonstrate strong consistency in seasonal and interannual variations across all three solutions. Over the past two decades, TWS in the Amazon basin has shown an overall increasing trend, with change rates of 0.25 ± 0.10 cm/yr (CSR), 0.23 ± 0.11 cm/yr (JPL), 0.22 ± 0.11 cm/yr (GSFC), and 0.23 ± 0.11 cm/yr (averaged). Furthermore, Figure 4b presents the wavelet time–frequency spectrum of the GRACE Mascon average time series. While the Fourier spectrum decomposes signals into the frequency domain to reveal overall characteristics, the wavelet time–frequency spectrum captures singularities and abrupt changes, precisely identifying the timing, location, and magnitude of fluctuations. The analysis reveals non-stationary random signals with 2–6-year periods in the wavelet spectrum, along with exceptionally strong signals during specific intervals. These anomalies may be associated with hydrological variations linked to extreme climate events and will be further investigated in subsequent sections.

3.2. Comparative Analysis of GNSS, GRACE, and HYDL in Quantifying Seasonal Hydrology-Induced Load Effects in the Amazon Basin

GNSS and GRACE observations provide complementary measurements of surface deformation and gravity field variations caused by hydrological mass loading. Vertical elastic loading displacements at GNSS station locations were calculated from GRACE-derived mass change fields using Green’s function theory. These GRACE-derived displacements were then compared with (1) GNSS vertical displacement time series corrected for non-tidal atmospheric and oceanic loading (hereinafter referred to as GNSS), and (2) displacements predicted by the HYDL model.

Figure 5 presents a temporal comparison of displacement datasets (GNSS, GRACE-derived, and HYDL-predicted) from a representative subset of twelve GNSS stations across the Amazon basin, demonstrating their strong temporal alignment. The displacement time series for all three datasets exhibit consistent seasonal and interannual variations, reflecting the basin’s hydrological response to climate variability. Figure 6 presents linear regression analyses of seasonal variations among the three datasets. The results indicate that GNSS annual amplitudes exceed those of GRACE and HYDL, primarily due to GNSS’s greater sensitivity to local short-wavelength signals [50,51]. Additionally, HYDL annual amplitudes exceed those of GRACE, primarily due to the spatial resolution limitations of GRACE observations.

Figure 7 quantifies the spatial distribution of the agreement between these datasets through correlation analysis at all 113 GNSS stations. As can be seen, AO-corrected GNSS displacements exhibit larger magnitudes than those derived from GRACE or predicted by HYDL, yet show stronger agreement with GRACE-derived signals than with HYDL predictions. Quantitative analysis across the network confirms this: the average PCC between monthly AO-corrected GNSS and GRACE-derived displacements is 0.72, with a 35% reduction in RMS. In contrast, the average PCC and RMS reduction between AO-corrected GNSS and HYDL-predicted displacements are lower, at 0.64 and 24%, respectively. The higher correlation and greater RMS reduction with GRACE-derived displacements indicate a stronger alignment between observed GNSS vertical deformations and total TWS changes measured by GRACE. The superior agreement between GNSS and GRACE, compared to GNSS and HYDL, highlights limitations in the HYDL model, likely due to its inadequate representation of groundwater dynamics [52], a significant component of TWS variability in the Amazon basin.

Spatially, correlations are weaker in the southern Amazon basin and coastal mountainous regions (Figure 7). This spatial heterogeneity may result from several factors: (1) the limited spatial resolution of GRACE (~300 km), which constrains its ability to resolve fine-scale hydrological signals, especially in topographically complex areas; (2) the dominance of more stable, less hydrologically responsive soil and bedrock types in mountainous areas; and (3) potentially lower signal-to-noise ratios in GNSS time series within these regions [21]. Notably, the lowest correlations are observed at stations in northern Chile, the Peruvian coast, and Ecuador. These arid to semi-arid regions, influenced by the cold, dry Peru Current, exhibit smaller amplitude hydrological variations and consequently weaker vertical loading signals. This inherent signal weakness likely contributes to the reduced correlations observed in these locations.

Previous studies indicate that CMCs derived from dense GNSS networks primarily capture interannual signals originating from surface mass loading variations [53]. Following the removal of NTAL and NTOL effects during preprocessing, the residual GNSS vertical displacement time series is predominantly governed by surface deformation induced by hydrological loading changes. To extract the spatially coherent CMC signal across the Amazon basin GNSS network, we applied PCA. Prior to PCA, long-term trends, annual, and semiannual signals were removed from each GNSS station’s time series using least-squares fitting. This procedure isolates low-frequency variations (primarily at interannual or longer timescales) linked to global climate cycles and geophysical processes. The first principal component (PC), representing the dominant mode of spatially coherent variance and exhibiting the most uniform response across the network, was scaled and designated as the representative CMC for the region [53,54,55].

Figure 8 depicts the resulting CMC characteristics: Figure 8a illustrates the spatial response pattern, Figure 8b presents the temporal evolution, and Figure 8c displays the corresponding wavelet time–frequency spectrum. The spatial distribution reveals markedly elevated deformation amplitudes along the Amazon River corridor compared to surrounding regions. Temporally, abrupt amplitude excursions during 2010–2011, 2015–2016, and 2020–2021 correspond to TWS anomalies identified in Figure 4a, indicating strong coupling between hydrological loading and climate variability. Furthermore, the wavelet spectrum identifies interannual oscillatory modes within the 2–6 years band, reflecting elastic crustal responses to hydrological mass fluctuations. These spectral characteristics exhibit subtle discrepancies from those of the GRACE-derived TWS spectrum (Figure 4b), attributable to fundamental differences in observational sensitivity: GNSS measurements are preferentially sensitive to local/regional hydrological signals, whereas GRACE resolves basin-scale integrated TWS variations [51].

3.3. Interannual Fluctuations of Hydrological Variability in the Amazon Basin (2008–2021) and Their Response to Climate Change

Among all PCs derived from PCA, the first PC typically represents the dominant, spatially coherent response, while the remaining PCs primarily capture local features and lack regional consistency [53,54,55]. As this study focuses on the overall hydrological dynamics of the Amazon region, we do not analyze the other PCs. Wavelet time–frequency analysis of the averaged GRACE-derived TWS time series and the first PC of GNSS data reveals a prominent non-stationary signal within the 2–6 years band, characteristic of interannual hydrological variability in the Amazon Basin. Xavier et al. (2010) [24] demonstrated that ENSO-driven extreme climate events modulate TWS fluctuations primarily through anomalous precipitation and temperature regimes. Accordingly, we analyzed interannual precipitation and temperature anomalies and incorporated the ENSO index to investigate linkages between hydrological variability and extreme climate events (Figure 9). Key findings are summarized as follows:

• As shown in Figure 9a, the interannual fluctuations derived from GRACE observations for the Amazon basin are strongly anticorrelated with those of the first PC of the CMCs calculated from GNSS data, demonstrating a high degree of correspondence. This indicates that GNSS observations effectively capture surface load signals induced by hydrological mass changes in the Amazon basin.
• Comparison of Figure 9a–c reveals abrupt changes in both precipitation and temperature across the Amazon basin during 2010–2011, 2011–2012, 2013–2014, 2015–2016, and 2020–2021, closely aligned with corresponding anomalies in hydrological mass and minimal time lag. This suggests that precipitation is the primary driver of hydrological mass variability in the basin.
• Comparison with the ENSO index variation in Figure 9d reveals a strong correlation between hydrological mass changes in the Amazon basin and ENSO activity. ENSO-related extreme climate events, particularly anomalous precipitation and temperature variations, significantly influence the dynamic processes of regional hydrological mass.

In conclusion, ENSO-driven extreme climate events, primarily through precipitation anomalies, are the dominant mechanism regulating interannual hydrological mass fluctuations in the Amazon Basin. These findings are essential for advancing understanding of regional hydrological cycles, assessing anthropogenic impacts, and improving forecasts of future water storage trends under climate change.

4. Discussion

The low correlation between GNSS vertical displacement time series and HYDL deformation reveals limitations in current hydrological models for representing groundwater dynamics, highlighting the complexity of groundwater variability in the Amazon Basin and the need for improved high-resolution hydrological models, particularly in quantifying groundwater exchange processes. Although GRACE provides large-scale observations that complement the high-resolution, point-based monitoring of GNSS, signal attenuation in GRACE data remains a challenge that can be mitigated through advanced data assimilation techniques. Integrating next-generation gravity satellites (e.g., GRACE-II) with distributed hydrological models holds promise for improving the resolution of hydrological processes in complex terrains.

From 2008 to 2021, the Amazon Basin experienced multiple extreme flood and drought events across parts or all of the region. The 2011–2012 flood, initiated by heavy rainfall in the northwestern Amazon around November 2011, was documented by Espinoza et al. (2014) [56]. The 2013–2014 flood, primarily affecting the southwestern Amazon, began with increased precipitation from September 2013, intensified in January 2014, and culminated in unprecedented flooding in February 2014 [57]. A severe El Niño event in 2015–2016 triggered widespread drought, while La Niña events occurred in 2010–2011 and 2020–2021. As shown in Figure 9, interannual fluctuations in the first PC of GRACE data and GNSS vertical displacement time series correspond closely with these extreme climate events. Notably, Knowles et al. (2020) [58] reported similar GNSS station displacement anomalies during these periods, but their analysis was limited to individual stations and lacked comparison with GRACE data. In contrast, this study employs a basin-wide analysis, integrating GNSS and GRACE datasets to robustly characterize the hydrological anomalies associated with these events.

Examining the seasonal and interannual variability of water storage in the Amazon Basin is crucial for advancing our understanding of regional hydrological processes and climate change impacts. This study analyzes interannual oscillations in GNSS and GRACE observations to elucidate key patterns of hydrological mass variations across the basin. The strong correlation between GNSS- and GRACE-derived elastic load displacements highlights both the seasonal and interannual dynamics of regional hydrology. Furthermore, interannual fluctuations in GRACE-derived TWS and elastic load displacements can be attributed to the dynamic hydrological processes operating at interannual timescales in the Amazon Basin.

However, it is worth mentioning that potential uncertainties in GNSS-derived displacements include the following: (1) residual signals from imperfectly modeled NTAL and NTOL, particularly during convective storms [38]; (2) unaccounted inelastic crustal responses to long-term loading; and (3) contamination by ~6-year geophysical signals potentially associated with core–mantle coupling [59]. Although preprocessing with ESMGFZ products removed major non-hydrological loads, sub-decadal variability may still reflect non-hydrological contributions. Future studies should apply multivariate decomposition techniques [53] to isolate these components.

5. Conclusions

This study integrates GNSS and GRACE observations to investigate seasonal and interannual hydrological dynamics in the Amazon Basin. Firstly, (1) GRACE data reveal significant spatial heterogeneity in TWS variations, with annual amplitudes exceeding 65 cm in the central floodplains and decreasing to less than 25 cm in peripheral mountainous regions. This pattern likely reflects the combined effects of precipitation gradients and topographic influences on runoff accumulation. Additionally, (2) vertical crustal displacements derived from GNSS exhibit strong consistency with GRACE-inferred hydrological loading deformations, as indicated by a high mean correlation (mean PCC = 0.72) and a substantial RMS reduction (35%). In contrast, the weaker correlation with HYDL model outputs (mean PCC = 0.64, RMS reduction = 24%) underscores the model’s limited capacity to capture groundwater dynamics, which account for ~35% of total TWS changes in the Amazon. Furthermore, (3) PCA of GNSS vertical displacements reveals abrupt interannual fluctuations during 2010–2011, 2011–2012, 2013–2014, 2015–2016, and 2020–2021, coinciding with extreme precipitation anomalies driven by ENSO. These events correspond to 20–30% deviations from mean TWS levels in GRACE observations, confirming that ENSO primarily modulates basin-scale hydrological mass variability through precipitation anomalies. The anti-correlation between GNSS-derived CMCs and GRACE TWS fluctuations further corroborates this mechanistic link, demonstrating that dense geodetic networks effectively capture climate-driven hydrologic signals.

In conclusion, this study establishes a synergistic framework that integrates geodetic and satellite gravimetry to quantify climate-hydrology interactions in the Amazon. By leveraging GNSS networks to validate GRACE-derived hydrology and identify model limitations, this approach is easily transferable to other critical basins (e.g., Congo, Mekong) where in situ monitoring is sparse. As climate warming exacerbates ENSO extremes, the operational integration of these techniques could strengthen early warning systems for hydrological extremes, thereby supporting adaptive water resource management in vulnerable ecosystems.