Hydro-meteorological dynamics of rainfall erosivity risk in the Amazon River Delta-Estuary Open Access

Since 1988, the Intergovernmental Panel on Climate Change (IPCC) has been synthesizing technical-scientific studies based on scenarios related to climate change and its implications in social, economic, and environmental spheres (Silveira et al. 2016; IPCC 2023).

In the Amazon context, the watershed plays a crucial role as a source of atmospheric water vapor, influencing regional and global climate patterns (Werth & Avissar 2004; Wang & Dickinson 2012; Tang et al. 2024). Additionally, the forest acts as a significant carbon sink, absorbing CO₂ from anthropogenic activities (Malhi et al. 1998, 2009; Brienen et al. 2015). These ecosystem services are directly associated with landscape variability and meteorological-climatic factors, which influence total precipitation over the basin.

Specifically, the Amazon River Delta-Estuary is a dynamic environment where interactions between water, soil, vegetation, and the atmosphere regulate essential processes such as carbon balance, evapotranspiration, and soil stabilization (Abril et al. 2014). This macroenvironment, influenced by hydrometeorological aspects, provides fundamental support for research on climatology and the impacts of climate change.

Soil erosion, a natural process increasingly affected by climate variability and land use changes, has been widely studied (Pacheco et al. 2019; Borrelli et al. 2021; Delgado et al. 2022; Riquetti et al. 2023). However, conventional methods require ongoing enhancements to reflect environmental transformations. The link between soil erosion and climatic variability is crucial for multiple fields, including water resource management, land-use planning, biodiversity conservation, and climate change adaptation.

The erosive potential of rainfall is related to the R-factor, concerning the Universal Soil Loss Equation (USLE), developed in a series of classic works, notably by Wischmeier (1959), Wischmeier & Smith (1965, 1978), Foster et al. (1981), Lombardi Neto & Moldenhauer (1992), and Vieira & Lombardi Neto (1995). These authors also highlight the differences regarding the Erodibility component (K) of the USLE, which characterizes the natural potential of the profile composed of the rock-soil sequence to residual soil, constituting the material covering the earth's surface, where geological and pedogenesis characteristics prevail, in contrast to the LS factor, which reflects morphogenesis (slope length and steepness).

Rainfall erosivity is strongly linked to large-scale and meso-scale hydrometeorological events, such as the El Niño-Southern Oscillation (ENSO), Atlantic Meridional Mode (AMM), and the Intertropical Convergence Zone (ITCZ). These phenomena control the Amazon's precipitation variability, directly influencing droughts and floods (Rudari et al. 2005; Marengo & Espinoza 2016; Michot et al. 2024). Studies indicate that topsoil erosion in the Amazon River Basin increased by over 600% between 1960 and 2019, particularly in the Madeira, Solimões, Xingu, and Tapajós sub-basins (Riquetti et al. 2023). In these regions, precipitation was classified as moderately to highly aggressive, with topography being the primary erosion factor in the Andean part and precipitation playing the dominant role in the eastern basin.

The relationship between extreme events and landscape dynamics can amplify erosive processes in watersheds with high runoff capacity and inadequate vegetation cover, resulting in increased sediment transport to water bodies (Gomes et al. 2021a). This phenomenon poses a direct risk to Amazonian riverbank communities, which often inhabit vulnerable areas. Additionally, erosion shapes specific fluvial topographic features along the Amazon River and its tributaries, including meander bars, floodplains, and oxbow lakes (Mertes et al. 1995; Silva et al. 2007; Guerrero et al. 2024). An example is the ‘collapsed lands’ phenomenon, a seasonal geomorphological process influenced by interactions between river systems, rainfall, and the physical landscape (Magalhães & Vieira 2018).

Given this scenario, the R-factor application to evaluate rainfall erosivity in different regions has many successful results (Da Silva 2004; Montebeller et al. 2007; Oliveira et al. 2009; De Assis Silva et al. 2010; Almeida et al. 2011; Eltz et al. 2013; Valvassori & Back 2014; Costa & Blanco 2018; Da Silva Barbosa et al. 2018; Rosa & Sousa 2018; Casanova-Ruiz et al. 2024; Das et al. 2024). Considering its sensitivity to climate change, the R-factor is a key metric for predicting future impacts on soil erosion and supporting mitigation and adaptation strategies.

Therefore, this study aims to analyze the risk of rainfall erosivity in the Amazon River Delta-Estuary, considering different hydrometeorological and land cover perspectives. The research hypothesis acknowledges that hydrometeorological variability events influence erosion potential. However, they also reflect land use and land cover processes.

The Amazon Sedimentary basin is part of the South American Plate, bounded by the Andes and the Brazilian Shield, and covers an area of 500,000 km², where the maximum sedimentary thickness reaches 5,000 m. Two geological structures mark the central area, the Gurupá Arch in the west and the Marajó Rift in the east. They extend over 53,000 km², with a maximum sedimentary thickness of over 16,000 m (Ferreira et al. 2021).

The sedimentary environment of the Amazon sedimentary basin dates back to the Paleozoic era (Ordovician). During the Cenozoic era, environmental and filling changes occurred, reflecting the influence of Andean movements, which transformed the entire Amazon landscape (Campbell et al. 2006; Soares et al. 2021). The evolutionary context highlights sedimentation during the late Miocene (∼5.3 Ma) of the Solimões geological formation, characterized by fluvial sequences of sandstones and fine clayey sediments. This scenario depicted a basin extending to the east, associated with a plain dominated by rivers, lakes, and swamps, with its tectonic control related to the central Andes. This sedimentary response led to the deposition of a continental sedimentary sequence linked to a low global sea level (Latrubesse & Franzinelli 2002; Latrubesse et al. 2010).

Tertiary deposits (mixed siliciclastic carbonates of the Marajó geological formation, Paleocene-Eocene) indicate the transition to shallow marine deposition environments (Rossetti & Góes 2008). Morphologically, the context that led to the recent hydrographic structure stands out. The structures to the southwest of Marajó Island were active until the Late Pleistocene, where Quaternary tectonics influenced sediment deposition east of Marajó Island. It is likely that the definitive detachment of Marajó Island from the continent occurred in the Holocene. Therefore, a new drainage system was established, formed by main channels in the S-SE direction (Rossetti & Valeriano 2007; Rossetti et al. 2008).

The origin of Marajó Island is thus initially linked to the presence of tectonic control associated with a ‘paleo’ Tocantins River that flowed into the N-NE, into the Atlantic. This gradually repositioned, resulting in the widening of the paleovalley, and the gradual establishment of a southward channel, called the ‘Pará River’, where segments predominantly oriented along a main E-W zone attributed to the Holocene (Rossetti & Valeriano 2007).

The final geometry results in a tropical delta/estuary dominated by tides, represented by the Northern Channel, Marajó Bay, and Southern Channel. Marajó Bay is part of the estuary's most characteristic region; the Northern Channel is influenced by the dynamics of the Amazon River mouth; and the Southern Channel, by tides and the dynamics of the Tocantins River mouth (Mansur et al. 2016). The hydrodynamic characteristics of the estuarine system indicate that the current direction varies considerably, with diverse currents emerging along the water column during tidal changes (Menezes et al. 2013). This hydrodynamics has the consequence that, with the exception of the northeast region, the coast of Marajó Island is influenced by freshwater from the rivers, where a water salinity of 10% is recorded throughout the island, at a distance of approximately 60 km from the coast (Cohen et al. 2008).

This region is called the Delta-Estuary due to being an estuary that flows into the Atlantic Ocean, with drainage passing through several channels constituted by numerous insular environments.

This area is close to the equator, with large water units, solar energy, and vegetation resources (Pereira Lira et al. 2020). The study region encompasses the Af (Tropical zone without dry season) and Am (Tropical zone monsoon) climate types, according to the climatic zoning by Alvares et al. (2013) based on Köppen's criteria (1936).

The Amazon River Delta-Estuary presents a socio-ecological system (SES) consisting of a biogeophysical unit and its associated social actors and institutions, where socio-ecological systems are recognized as complex and adaptive, bounded by spatial and functional ecosystem boundaries (Brondizio et al. 2016). In this context, it includes 50 cities (41 in Pará and 9 in Amapá), where small towns (<20,000 inhabitants) predominate (68%) (Costa et al. 2019).

The delta as a coupled SES, as defined by Brondizio et al. (2016), differs from the approach proposed in this work, as it considers municipal boundaries as the internal reference unit in the region, while the adopted configuration is limited to the extent of the island strip derived from the sedimentation of the mouth of the Amazon River. The proposed limitations are not exclusive and can also derive from a political-administrative criterion if necessary. The different data sources are presented in Table 1. The data period is the same for all sources, starting in January 1985 and ending in December 2022.

Since precipitation is the basis for erosivity data, the final scale was the CHIRPS reference (∼5 km). The lack of a time series of rainfall and other meteorological parameters in the island of Marajó justifies the chosen database. The National Water and Sanitation Agency (ANA) presents in its portal (HIDROWEB 2024) the last fluviometric stations after the tidal influence in the estuary and the rain gauge stations were discontinued over time, and became the satellite information essential for studies developments.

Precipitation data were obtained from the virtual platform of the Climate Hazards Center (CHIRPS: Rainfall Estimates from Rain Gauge and Satellite Observations), while Land Use and Land Cover data were acquired from the virtual platform of the Annual Land Use and Land Cover Mapping Project in Brazil (MapBiomas).

Rainfall in the Amazon River basin needs satellite instruments for regional evaluation and pattern definitions. The spatial resolution of rainfall datasets and conventional rain gauge networks represent the primary information for understanding rainfall distribution and its relationship with environmental factors (Mu et al. 2021; López-Bermeo et al. 2022). The CHIRPS product is combined with gauge data (Funk et al. 2015). It was validated for the Brazilian Amazon and showed the best results for monthly rainfall obtained using data from the rain gauge stations (Cavalcante et al. 2020).

The CHIRPS dataset was selected for its ability to provide spatially distributed rainfall measurements rather than relying on data from a single weather station, reducing interference from local hydrometeorological phenomena, ensuring that rainfall erosivity estimates are based on a broader spatial scale, mitigating potential errors and measurement biases associated with limited station coverage.

Annual Land Use and Land Cover Mapping Project in Brazil (MapBiomas) since 1985 applies the random forest algorithmic to Landsat archive using Google Earth Engine and classified five main classes: forest, non-forest natural formation, farming, non-vegetated areas, and water; it was broken into sub-classification levels at a 30 m pixel resolution.

Souza Jr et al. (2020) describe the land cover data classification and validation process. For the Amazon biome, a set of pre-existing land cover maps was randomly selected to train the Random Forest Classifier to extend the analysis to the all-time series. The average overall accuracy is 96.8% based on a stratified random sample for collection number 8. Overall, the MapBiomas dataset was chosen due to its continuous annual monitoring, allowing the assessment of erosivity trends in relation to land cover dynamics in the region.

The precipitation raster data were stored in a specific directory. Subsequently, a script was developed (using the ‘raster’ and ‘sf’ packages), and through the RStudio software version 2023.12.1+402, all the rasters were clipped in a single processing step (loop), using the study area mask (in shapefile format) as the basis for this procedure. Later, with a new script, the basic statistical parameters (mean, standard deviation, maximum, and minimum) of each raster were automatically calculated (using the ‘xlsx’ and ‘openxlsx’ packages).

The Land Use and Land Cover (LULC) data, acquired in raster format, were imported into QGIS software version 3.34.4, where they were individually processed according to the study area. Subsequently, the areas (using the ‘r.report’ tool from GRASS/QGIS) of the classes of each land use and land cover corresponding to the years 1985 to 2022 were calculated. Additionally, Fragstat software version 4.2 was used to identify metrics related to the patch size of the defined land use and land cover classes: forest, non-forest natural formations, wetland ecosystems, agriculture, sandy environments, non-vegetated areas, and water bodies.

As pointed out by Da Silva (2004), Table 2 presents the classification of erosive potential as weak, moderate, moderate to strong, strong, and very strong, according to Carvalho (1994), adapted to S.I. metric units by Foster et al. (1981).

The chosen methodological approach is supported by studies that have used the same application, such as De Sousa Teixeira et al. (2022), Neves & Lollo (2022), Back & Poleto (2018), Hernández et al. (2018) and Pontes et al. (2017a).

First, descriptive statistics of the precipitation and EI_m data were calculated. Specifically, the mean, maximum, and minimum values, as well as the standard deviations, were determined. Subsequently, the normality of the data was assessed using the Shapiro-Wilk (S–W) test, considering a statistical significance level of 5%. Upon finding non-normality (p-value<0.05), Spearman correlation was applied to the monthly mean erosivity index in relation to the AMM.

Spearman rank correlation was selected due to its nonparametric nature, making it suitable for analyzing nonlinear relationships between precipitation erosivity, climate variability, and land cover changes. Unlike Pearson's correlation, which assumes a linear relationship, Spearman's method effectively captures monotonic trends, allowing for a more robust assessment of associations in complex environmental data sets. This allows the evaluation of data sets that do not follow a linear distribution, especially when considering meteorological data, which undergoes variation.

Applying the S–W test made it possible to assess the homogeneity of variance and normality for rainfall and erosivity. The spatial aggregation applied aimed at avoiding spatial inconsistencies. The validation was performed at pixel level with previous data sources during the same period. Studies show a high correlation in the spatiotemporal variability between rainfall and erosivity. They support the methodology used to evaluate the relationship between erosivity in the Amazonian context. Almagro et al. (2017) assessed the impact of climate change on rainfall erosivity across Brazil; Riquetti et al. (2020) assessed the impact of climate change on long-term mean annual rainfall erosivity and mean annual precipitation in South America; and Dos Santos Silva et al. (2020) determined the spatial and temporal distribution of rainfall erosivity in the Amazon from 334 rain gauge stations over a 20-year time series.

In the process, cluster analysis (k-means) and principal component analysis (PCA) were applied to a group of specific periods of rainfall erosivity to aid the interpretation of the datasets. PCA was performed between R and land use and land cover classes for 2022 to investigate possible relationships between these two information.

The main objective is to relate variables by grouping behavioral similarities, being decisive for the statistical treatment. This choice is appropriate for the dataset under consideration and reflects the spatial scale adopted. Precipitation has a strong regional component of response to climate variability in the Delta-Estuary region, which makes correlation a detailed assessment process involving aspects of seasonality, spatial distribution and climate response to extreme events at different scales. Rizzo et al. (2020) describe variations in precipitation behavior with variations in land use and land cover (Southeastern Amazon), highlighting changes in the amount of rainfall over time but not in the total duration of the rainy and dry seasons, neither a significant change (p>0.05) in seasonality behavior.

In terms of meteorological systems, the pronounced seasonality of precipitation is mainly due to the latitudinal migration of the ITCZ, which is climatologically located further south in the months of March and April (De Souza et al. 2005; Berry & Reeder 2014), leading to the occurrence of extreme precipitation events in the Delta-Estuary region of the Amazon River (Santos et al. 2017).

Another factor responsible for the high volume of precipitation in the eastern region, as well as in the western portion of the Delta-Estuary, are the Squall Lines (LI), characterized by large clusters of cumulonimbus clouds, generated by the circulation of sea breeze, penetrating into the continental portion (Cohen et al. 1989, 1995; Fisch et al. 1996, 1998; Sousa et al. 2021).

In the less rainy quarter, SON, and in August, the low precipitation values are more noticeable in the northwest, north, northeast, and east portions. In June and July, precipitation was more pronounced in the west region, with significant dryness in the south, east, and north regions. The months of December and January are transitional, with precipitation peaks accentuating in the eastern portion. During this period, the summer circulation in South America, with the formation of the Bolivian High (at high levels), induces the occurrence of deep convection in most of the Amazon River basin, with sporadic intense rains, as evidenced by the outliers in Figure 4 (Santos et al. 2015, 2017).

Outliers are values that deviate from average patterns; depending on the time series, they can be sequence, point or pattern anomalies (Ma et al. 2017). Figures 3 and 5 show the outliers as point anomalies derived from the CHIRPS evaluation. The values associated with each pixel vary due to climatic variability and the heterogeneity of rainfall distribution along the island. The result is the dispersion effect, which is greater during the rainiest than during the least rainy season.

The Atlantic Dipole, which refers to an atypical change in sea surface temperature (SST) in the Tropical Atlantic Ocean, can alter the meridional circulation of the atmosphere (Hadley) and increase precipitation in the northern and northeastern regions of Brazil during its negative phase (AMM), in which the waters of the North Tropical Atlantic are colder and those of the South Tropical Atlantic are warmer (Aragão 1998), favoring the southern position of the ITCZ. The positive phase of the AMM is responsible for precipitation reduction in the region, and the negative phase contributes to rainfall formation in the region, given the north-south modulation of the ITCZ over the Atlantic Ocean, characterizing the proportional dynamics of the Atlantic Dipole (Assis et al. 2023).

In this context, during the years 1986 and 1994, the region of the Delta-Estuary region of the Amazon River recorded the highest monthly averages of AMM, with values of −3.81 and −3.50, respectively. This indicated a significant increase in EI_m due to increased precipitation. On the other hand, in the years 2005 and 2010, the monthly averages of AMM were 3.67 and 4.97, respectively, resulting in a spatially lower EI_m due to reduced precipitation.

Considering the classification by Carvalho (1994), adapted to S.I. metric units by Foster et al. (1981) (Table 2), for all 38 years of the historical series under analysis, R was classified as very high erosivity (R> 9,810 MJ·mm⁻¹·ha⁻¹·h⁻¹·year⁻¹). This is associated with factors contributing to intense precipitation in the region, such as squall lines, Easterly waves, moisture sources associated with vegetation cover evapotranspiration, meso-scale effects, and primarily the ITCZ (Bastos et al. 2002).

The comparison highlights the impact of ENSO variability on rainfall erosivity, with La Niña years (2017, 2018) showing higher erosivity values, especially in the northern and central regions, whereas the El Niño year (1987) exhibits lower erosivity across the study area. The ENSO-neutral year (1993) presents an intermediate erosivity pattern, suggesting a strong dependence of erosivity dynamics on climate oscillations.

The reason for this difference in the hydrometeorological context is related to ENSO, which presents, as discussed by Minuzzi (2010), the warm phase, El Niño, characterized by the warming of surface waters in the Pacific Ocean in the equatorial region; and La Niña, which is the opposite, consisting of the cooling of these waters. In this context, specifically, the Southern Oscillation refers to the anomalous dynamics of tropical atmospheric pressure, as an atmospheric consequence of El Niño, tending to reduced pressure in the Pacific and elevated in other areas of the tropical region, whose abnormal descending movements inhibit cloud formation and reduce rainfall in the northern region of Brazil, while in La Niña, the process is reversed (Aragão 1998; Marengo et al. 2011; Matias et al. 2024). As a result, rainfall erosivity is enhanced during La Niña periods and reduced during El Niño periods.

El Niño and La Niña conditions mainly affect the northern Amazon basin (Delta-Estuary region) and central parts. Tropical North Atlantic SST anomalies are concentrated in the dry season (September to November), while ENSO mainly affects the rainy season (February to April). As a result, increasing rainfall in the wet season and decreasing rainfall in the dry season may influence the frequency of erosivity, extreme floods and drier conditions in the northern Amazon Basin during the same period (Gloor et al. 2015).

Extreme ENSO-induced drought events, such as those observed between 2015 and 2016, resulted in soil and atmospheric water stress in the eastern Amazon from October 2015 (mid-dry season) to April 2016 (mid-rainy season, wet season), both observed in forest cover layers (Brum et al. 2018; Fancourt et al. 2022). This perspective shows the relevance of land use and cover evolution in the Amazon Delta-Estuary, where forest fragmentation affects soil loss and forest regeneration capacity.

Other studies have found similar results for tropical regions. For example, in tropical Australia, rainfall erosivity ranged from 1,080 to 33,500 MJ·mm⁻¹·ha⁻¹·h⁻¹·year⁻¹, with 90% of erosivity concentrated in the summer months, emphasizing the strong seasonality of erosive rainfall (Yu 1998). On a global scale, the average rainfall erosivity was estimated at 2,190 MJ·mm⁻¹·ha⁻¹·h⁻¹·year⁻¹, with the highest values recorded in South America, the Caribbean, Central East Africa, and Southeast Asia (Panagos et al. 2017). In the Dosquebradas Basin, Colombia, rainfall erosivity exhibited a marked seasonal pattern (Hoyos et al. 2005). In Eastern Africa, Fenta et al. (2017) found high climate variability, calculating a mean erosivity of 3,246 ± 1,895 MJ·mm⁻¹·ha⁻¹·h⁻¹·year⁻¹, and determining that approximately 55% of the studied area had high to very high erosivity, indicating significant potential for erosion processes in a landscape characterized by seasonal rainfall and strong latitude influence.

Future projections indicate a potential global increase in rainfall erosivity of 26.2–34.3% by 2070, driven by climate change and intensified precipitation regime leading to a 30–66% increase in global soil erosion rates (Panagos et al. 2022). Additionally, more than 76% of the global land area is expected to experience an increase in rainfall erosivity, with regions facing more pronounced changes from mid-century onwards (Chen et al. 2024).

The relationship between erosivity, non-forest natural formations, and agriculture is related to soil water erosion. In these areas, given the low density of vegetation and, consequently, the lack of interception of raindrops by canopies, such as forest formations, there is a greater potential for soil water erosion caused by rainfall.

In this context, evaluating the granulometric composition under a temporal analysis in a region of the Amazon River Delta-Estuary, De Lima et al. (2015) concluded that the granulometric composition of the area is influenced by dynamic processes such as erosion, transportation, and deposition in the Marajó Bay region. Dos Santos Silva et al. (2020), aiming to determine the spatial and temporal distribution of rainfall erosivity in the Amazon, found that the region was classified as having very strong erosivity. Gomes et al. (2021b), analyzing the susceptibility to soil water erosion in the Capim River basin (PA/MA), pointed out that accelerated urbanization processes can increase soil erosion, especially during the rainy season.

De Souza Negrão et al. (2022), evaluating vulnerability to coastal erosion on Amazonian beaches, concluded that areas with higher anthropogenic occupation configured environments at greater risk of coastal erosion, characterized as critical areas. Riquetti et al. (2023), after studying soil erosion in the Amazon River basin over the past 60 years of deforestation, found a significant influence of deforestation resulting from the expansion of agricultural and livestock activities on soil erosion rates in the Amazon basin, especially in the sub-basins of the Madeira, Solimões, Xingu, and Tapajós rivers. Rizzo et al. (2023), in a qualitative analysis of soil erosive susceptibility in the Pequiá Stream Watershed in the state of Maranhão (Brazil), concluded that the severity of erosion can occur in response to the formation and renewal of pastures on slopes.

The Amazon River Delta-Estuary thus appears as a region of intense hydrodynamics marked by processes of erosive potential, in addition to the intense natural sedimentation of the area. Its natural characteristics contrast with the land use process that is underway, which does not prioritize conservation practices for soil loss control. Future scenarios of sea-level rise and intensification of extreme weather events (especially those with higher rainfall volumes) have the potential to alter the delta's configuration, leading to intensive losses through erosive processes and consequent increase in the region's socio-environmental vulnerability.

The applicability of CHIRPS for erosivity assessment is enhanced by the fact that the output does not include the complex microscale hydrologic behavior of the estuarine system, such as tidal loading, river discharge variability and sediment transport processes, which are critical for a comprehensive assessment of erosivity. Furthermore, the output does not include the complex microscale hydrological loads of the estuarine system, such as tidal influence, river discharge variability and sediment transport processes, which are critical for a comprehensive assessment of erosivity.

The limit stays the grid of precipitation estimates. This may not represent all local variations due to spatial resolution limitations, especially in areas with high hydroclimatic heterogeneity. However, CHIRPS remains a viable dataset for this study due to its pixel-based representation. It allows for a more comprehensive spatial analysis than traditional point-source precipitation data from a single weather station.

Unlike single-location interpolation-dependent datasets, CHIRPS provides multiple spatial sampling targets, reducing potential biases introduced by localized precipitation anomalies and enabling representative calculations of erosivity potential across the study area. It is essential to recognize that CHIRPS-derived precipitation values must be validated against station-based observations, where possible, to verify erosivity estimates and address any residual uncertainties in rainfall quantification.

The region of the Amazon River Delta-Estuary exhibited very high rainfall erosive potential in all years of analysis, influenced by hydrometeorological events such as the AMM and the phases of ENOS, and spatially by different types of land use and land cover, confirming the hypothesis raised. The hydrometeorological dynamics of the rainfall erosion risk in this area strengthens the argument for its protection, especially its wetland status in island areas.

Ecosystems and urban areas may suffer from soil loss. The main human settlements are along rivers and coastlines, where problems such as sea-level changes need to be discussed in more detail through specific studies. The Amazon River basin is connected to a large estuary bordering all coastlines north of Brazil; similar research in comparable ecosystems suggests a range of potential research applications.

Methodological development can be improved, especially by reactivating rain gauge stations. Climate variables such as precipitation, evapotranspiration, and temperature are necessary to monitor climate variability and understand the impact of extreme precipitation events (increases and decreases) and sea-level changes.

Land use types, rainfall erosivity, and sea-level changes are interlinked and represent an emerging problem. Although the western part of Marajó and the islands that form the archipelago behave as wetlands, the eastern part has less influence from flooded areas and natural fields that favor the productive sector (livestock and agriculture). Thus, the loss of topsoil and the increase in eroded areas contribute to the degradation of productive zones, affecting the urban economy and increasing the social vulnerability of the population living along the riverbanks. Soil erosion is related to surface runoff and rainfall intensity. The surface runoff and soil erosion higher during the high rainfall season means a splash raindrop strength could be amortized by forest cover or produce a removal of soil particles in exposed areas (without vegetation cover), which demonstrates the importance of land management in agricultural and livestock production areas, which are dominant in the region with the greatest potential for erosion identified on Marajó Island.

The dynamics of the ITCZ and Squall Lines, associated with the warm and cold phases of ENOS, as well as the phases of the Atlantic Dipole, shape the erosive potential due to their direct influence on the region's rainfall regime. Despite the small variation in land use and cover areas in the study area over the years, rainfall erosivity is directly related to the type of land cover and how it can enhance or diminish rainfall erosive potential depending on the type of interception over the soil. The climate dynamics discussed by the Intergovernmental Panel on Climate Change (IPCC) explain that humid tropical forests are a relevant group of ecosystems for water cycling and precipitation. Thus, the assessment of the ENSO and the AMM, integrated to the erosion potential, provides a positive initiative to understand how this system could interact and evolve towards a resilient scenario, allowing a conservationist trend or a scenario of losses (lower resilience) with the intensification of land use and increased losses of damaged soils and wetlands.

The trends in climate variability and land use development processes suggest that the region should have public policies aimed at encouraging conservation practices, preserving the most sensitive areas, especially those where the region's productive potential is concentrated, such as agriculture and animal husbandry, which are main sources of the area's income.

Prospective research includes the need for a monitoring system structure for urban environments and sensitive areas (such as riparian forests). They represent supporting strategies to mitigate the effects of hydrological and meteorological changes, especially in a flooded area. The Marajó Archipelago Environmental Protection Area (APA) is a sustainable use unit created in 1989, and the Marajó Archipelago Sustainable Territorial Development Plan (PDTS-Marajó), launched in 2007, are territorial planning actions that represent public policies that contribute to mitigating climate events but depend on political and social articulation that strengthens outlining specific management measures.

To enhance research on rainfall erosivity in the Amazon River Delta-Estuary, new studies should, whenever possible, evaluate the interactions between soil physical properties and rainfall intensity. Furthermore, investigating the role of vegetation cover and structure in soil stabilization under the influence of seasonal precipitation patterns in the region with greater precision may provide better natural mitigation strategies under different climate scenarios. Additionally, regional climate models should incorporate detailed hydrological simulations to assess how projected climate change scenarios may amplify erosion risks.

From a political perspective, municipal governments within the region must monitor coastal erosion processes, develop risk and disaster action plans, and implement climate change adaptation strategies. Strengthening environmental enforcement in areas with high erosive potential is crucial to prevent anthropogenic interference that could trigger intense erosion processes. Moreover, these governments can reinforce community-based land management initiatives, increasing local resilience and ensuring conservation measures align with the region's socioeconomic needs.

Y.R., A.L. and C.S. conceptualized the study, wrote and reviewed. V.F., L.R., I.O. and M.D. wrote, reviewed, and edited the article. P.B. reviewed and edited the article.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.