Comprehensive bathymetry and intertidal topography of the Amazon estuary

The characterization of estuarine hydrodynamics primarily depends on knowledge of the bathymetry and topography. Here, we present the first comprehensive, high-resolution dataset of the topography and bathymetry of the Amazon River estuary, the world’s largest estuary. Our product is based on an innovative approach combining spaceborne remote sensing data, an extensive and processed river depth dataset, and auxiliary data. Our goal with this mapping is to promote the database usage in studies that require this information, such as hydrodynamic modeling or geomorphological assessments. Our twofold approach considered 500 000 sounding points digitized from 19 nautical charts for bathymetry estimation, in conjunction with a state-of-the-art topographic dataset based on remote sensing, encompassing intertidal flats, riverbanks, and adjacent floodplains. Finally, our estimate can be accessed in a unified 30 m resolution regular grid referenced to the Earth Gravitational Model 2008 (EGM08), complemented both landward and seaward by land (Multi-Error-Removed ImprovedTerrain digital elevation model, MERIT DEM) and ocean (General Bathymetric Chart of the Oceans version 2020, GEBCO_2020) topographic data. Extensive validation against independent and spatially distributed data, from an airborne lidar survey, from ICESat-2 altimetric satellite data, and from various in situ surveys, shows a typical vertical accuracy of 7.2 m (riverbed) and 1.2 m (non-vegetated intertidal floodplains). The dataset is available at https://doi.org/10.17632/3g6b5ynrdb.2 (Fassoni-Andrade et al., 2021).

semi-diurnal tide propagating upstream from the Atlantic Ocean to the interannual hydro-meteorological climatic events frequently occurring over the upstream catchment. These exchanges between the river and the ocean result in sporadic flooding events, which profoundly impact the riparian communities' socio-economic conditions (Andrade and Szlafsztein, 2018;Mansur et al., 2016). The morphology of the river bed is known to primarily condition the estuary's hydrodynamics, particularly the propagation of the tidal wave (Gallo and Vinzon, 2015), which is expected to affect the dynamics of the 35 riverine floods and the extent of the associated flooding (Kosuth et al., 2009). This dynamic environment with high ecological diversity is essential for nutrient cycling and carbon fluxes (Sawakuchi et al., 2017;Ward et al., 2015), for navigation (Fernandes et al., 2007), and the transport and accumulation of sediment (Nittrouer et al., 2021).
The Amazon estuary extends from the continental shelf up to Óbidos city, corresponding to the longest tidally-influenced reach in the world with 800-910 km (Kosuth et al., 2009;Nittrouer et al., 2021;Fig. 1). This river flow is drained 40 downstream towards the ocean through the main channel until the confluence with the Xingu River, around 300 km upstream of the mouth, where it is divided into two long channels, hereafter called South Channel (locally named Gurupá Channel) and North Channel (Fig. 1). Downstream of this branching, the estuary appears as a complex network of dendritic tidal channels and islands (Fricke et al., 2019). The estuary is classified as macrotidal (Dyer, 1997;Gallo and Vinzon, 2005) and semi-diurnal (Kosuth et al., 2009) with a tidal range between 4 and 6 m at the mouth. The 2 (lunar semi-diurnal) and 45 2 (solar semi-diurnal) tidal constituents are the dominant components at the ocean boundary, with amplitudes of 1.5 and 0.4 m there (Gallo and Vinzon, 2005). At the upstream limit of the estuary in Óbidos, the range of the drought-flood annual cycle of the river height typically amounts to 6 m, and the tidal effects remain sensible only during the drought season (Kosuth et al., 2009). So far, the quantitative investigation of the estuary's hydrodynamics and the interaction mechanisms between the tide and the river flow has been limited by the lack of sufficiently-resolved bathymetric databases (e.g., Gabioux 50 et al., 2005). Past hydrodynamical studies of the Amazon estuary thus relied on approaches based on box models (Prestes et al., 2020) and/or on coarse hydrodynamical models (e.g. Gallo and Vinzon, 2015). Still, these past studies revealed rich hydrodynamics of the estuary, comprising contrasted patterns of bottom friction (Gabioux et al., 2005), active non-linear deformation of the tidal waves (Gallo and Vinzon, 2005), a distinct structure of the salinity front (Molinas et al., 2014(Molinas et al., , 2020 and a prominent role of the intertidal flats in the flow variability (Gallo and Vinzon, 2005). The interplay between the fluvial 55 variability of the water level and its tidal variability is particularly known to exert a central control on the estuary's sedimentation pattern (Fricke et al., 2019). While the geometry of the Amazon estuary is known to have been little influenced by anthropic effects up to now, it appears essential to document it in its current state, at a time when the human influence is rising and is expected to induce profound, long-lasting impacts on the continental sediment supply to this estuary (Latrubesse et al., 2017). 60 The present paper aims to present a novel topography and bathymetry dataset of the whole Amazon River estuary, from its upstream limit 1000 km inland, to its terminal estuary at its oceanic outlet, covering the riverbed as well as the intermittently-flooded river banks and adjoining floodplains. These various domains being inherently different in nature, our https://doi.org/10.5194/essd-2021-32 methodological approach is twofold. Over the always-wet part of the riverbed, we rely on a traditional methodology to construct the bathymetry based on comprehensive, systematic digitization of existing nautical charts. In contrast over the 65 intermittently-dry intertidal zones and floodplains, our mapping is achieved through an original, state-of-the-art approach based on space-borne remote sensing. Our dataset is regularly gridded at 30 m resolution and elevations are referenced to EGM08 (Pavlis et al., 2012). It covers the river streams, riverbanks, and floodplains and extends downstream of the estuary mouths over the near-shore ocean shelf and open-ocean coastline, covering the domain shown in Fig. 1.
Section 2 presents the data sources and the methods used to build the dataset. Sections 3 presents the validation against 70 independent databases. Section 4 shows the topographic mapping and cross-section along the river and floodplain. Section 5 discusses the significance and caveats of the dataset, and Section 6 explains the access to the various forms of our dataset.

Bathymetry of the river bed
Over the always-wet part of the various streams of the Amazon estuary, the approach relies on systematic digitization of sounding points of bed elevation harvested from a comprehensive ensemble of nautical charts published by the Brazilian 80 Navy (available at www.marinha.mil.br/chm/dados-do-segnav/cartas-raster). Although technically straightforward, this task was by far the most tedious part of the procedure, on account of the large geographical extent of the domain (21'500 km 2 , a grey polygon in Fig. 1). We digitized more than 500'000 individual points in a total of 19 charts, which identification and limits are shown in Fig. 1. The primary bathymetric surveys utilized in these charts were carried out by the Brazilian Navy on different dates varying between 1953 and 2019, with a reasonably large fraction of them done after 2000 (see Table A1  85 for further details). Figure 2a displays an example of a digitized nautical chart. One issue with the maps we could access was the vertical referencing of the digitized elevations. Depending on the map considered, the bed elevation values were provided with respect to two different reference water surface elevations (WSE): either the level of the 90 th percentile of water surface elevation (hereafter noted WS90) or the average level of the low tide during the spring tide (termed as syzygy, noted SYZ). 90 We inferred the vertical elevation of each of these two references from the available records of the tide gauge stations scattered along the river down to the river mouth. The tidal and limnigraphic records from the seven stations we could access, listed in Table A2 and Table A3 (locations in Fig. 1), were provided by the Agência Nacional de Águas (ANA), by the Brazilian Navy, and by Instituto Brasileiro de Geografia e Estatística (IBGE). The vertical water level of both references (WS90 and SYZ) was deducted explicitly from the temporal records by computing the level of the 90 th percentile to infer 95 WS90 and by computing the average level of the low tide during spring tides for SYZ. These references were computed with respect to the geoid considering the absolute leveling published in Calmant et al. (2012) and Callède et al. (2013), complemented by a dedicated geodetic survey field that we conducted in January-February 2020 (Text A1). At the river mouth, downstream of the downstream-most tidal stations (blue points in Fig. 1), the SYZ level was calculated using the combination of the mean sea surface height provided by the ocean general circulation model of Ruault et al. (2020) and of a 100 proxy of the syzygy level estimated by FES2014 tidal atlas (Carrère et al., 2016; available at www.aviso.altimetry.fr/en/data/products/auxiliary-products/global-tide-fes.html). This proxy of SYZ was defined classically according to equation 1, from the sum of amplitudes of 2 and 2 tidal constituents (Pugh and Woodworth, 2014).  Fig. 2c. The WSE for each of the 110 500'000 digitized points was then inferred from the values along the two blue lines and blues points in Fig. 1 via a nearestneighbor interpolation method. After, the WSE was subtracted from the water depths resulting in bed elevation values referenced to EGM08. The bed elevation points were then interpolated using the topo-to-raster method (Hutchinson, 1989), which is essentially an interpolation method suited to hydrological objects to create a regular elevation grid with 30 m spatial resolution. 115 In the interpolation, a river boundary was considered, as shown in Fig. 1 (grey polygon) and exemplified in Fig. 2b (bathymetry boundary). This boundary is a polygon defined considering a flood frequency comprised between 96% and 100%. The flood frequency map was calculated from the Global Surface Water (GSW) Monthly Water History v1.1 data (Pekel et al., 2016; available at https://global-surface-water.appspot.com), which represents the space-borne Landsat-based monthly record of water presence on a global scale with a spatial resolution of 30 m. A Google-Earth engine code (Gorelick 120 et al., 2017), described in Fassoni- Andrade et al. (2020b), was used to create it considering all GSW monthly images from the period from January 2015 to December 2018, hereby totalizing 48 months.

Topography of periodically flooded areas
Intertidal banks and floodplains are areas periodically flooded by tides and riverine floods, respectively. We define them as 130 the areas comprised between 0% and 96% of the above-described flood frequency map (Fig. 3a). In the past studies devoted to coastal mapping, intertidal topography has been mapped through remote sensing data, the waterline method being one of the most widely adopted techniques (see Salameh Salameh et al. (2020). This method has proven tractable with moderate-resolution space-borne imagery; however, it requires simultaneous water level knowledge at the exact time of each acquisition, along the remotelysensed water lines. It also relies on spatial interpolation of the isobaths between successive waterlines, which can be problematic if the waterlines are sparse. Recently, alternative methods have been developed using a flood frequency map to 140 estimate the coastal topography pixel-by-pixel (Dai et al., 2019;Tseng et al., 2017). In these approaches, reference water levels (for instance, minimum, average, and maximum) were assigned to reference flood frequencies (respectively 100%, 50%, and 0%). Contrasting these approaches that do not explicitly require the knowledge of the water level's temporal variation, Fassoni-Andrade et al. (2020b) related the function of water level exceedance probability and a flood frequency map to estimate the topography of the water bodies. The authors showed that the terrain elevation at a given pixel is defined 145 as the water level, which probability of exceedance is equal to the flood frequency there.
This straightforward approach requires a flood-frequency map, and the water level exceedance probability functions to generate the terrain elevation map. It has been applied in situations where the temporal dynamics of water filling and draining is slow. Here, the same method is applied to estimate the floodplain topography and coastal topography of the Amazon estuary, where the water level variability is spread over a broader spectrum, from intra-daily timescales to seasonal 150 or interannual timescales (Kosuth et al., 2009). As we did not have access to vertically-leveled tide gauge archives in the Amazon estuary coastal area, two domains were considered for topographic estimation. The first domain considered the riverbanks, intertidal zone, and floodplains along the channels (described in Section 2.2.1), where the approach described in Ocean, the method was adapted to estimate tidal variation considering the water level exceedance function from a tidal 155 station (Section 2.2.2).

Riverbanks, intertidal zone, and floodplains
In the intertidal zone and floodplains along the Amazon River, WSE records from limnigraphic and tidal stations were considered over the period 2015-2018 (Table A2 and Table A3)  station). Like WSE estimation along the river (Section 2.1), the water level duration curve was inferred separately along the North Channel and the South Channel by interpolating curvilinearly the curves obtained at each station. Then, the water level duration curves were extrapolated through a nearest-neighbor interpolation over the whole of the estuary inter-tidal areas and floodplains, i.e., everywhere upstream estuary mouths. Therefore, the terrain elevation at any pixel was estimated considering the water level, in which the probability of exceedance is equal to the flood frequency at the same pixel 165 (Fassoni-Andrade et al., 2020b). In permanently flooded areas, i.e., where the flood frequency is 100%, the method considers the topography equal to the lowest WSE observed, as in the river.

Coastal ocean
The challenge to estimate the topography in the coastal area using the methodology of Fassoni- Andrade et al. (2020b), where vertically-leveled tide gauges are lacking, is to infer spatially-distributed water level exceedance functions. We considered 175 the downstream-most Amazon estuary station of Ponta do Céu (Table A2; location in Fig. 1) and computed the water level exceedance function there. This curve's shape was then assumed to be the same throughout the coastal region, but with variable amplitude, proportional to the local tidal amplitude. In this semi-diurnal macro-tidal region, a reasonable proxy of the tidal amplitude over the region can be thought of as the sum of the amplitudes of 2 and 2 tidal constituents, as these two constituents are the dominant ones downstream of the river mouth (Gallo and Vinzon, 2005 (2) 185 in which the tidal amplitude at a point is given by 2 × ( 2 + 2), and the PC subscript refers to Ponta do Céu values.
As a verification, Fig. 3c shows the observed and estimated exceedance probability functions at Colares station, located about 150 km to the east of the Amazon estuary ( Fig. 1). Both curves represent the anomaly with respect to the average. The two curves look very similar, with a root-mean-square deviation (RMSD) of 12.7 cm. 190 After estimating the water level exceedance probability at each point, the vertical reference was adjusted by matching the mean level with the height of the mean sea surface estimated by the ocean circulation model of Ruault et al. (2020). Similar to the water level exceedance probability functions estimated along the river, the coastal area's water level duration functions were inferred for all pixels of the flood frequency map considering the nearest neighbor interpolation method. Therefore, the terrain elevation at a pixel was estimated considering the water level, in which the probability of exceedance is equal to the 195 flood frequency at the same pixel (Fassoni-Andrade et al., 2020b).  (Vieira, 2015) in 205 cooperation with Governo do Estado do Amapá and Exército Brasileiro (GEA/EB).

ICESat-2 space-borne data
The topography was further validated against Ice, Cloud, and land Elevation Satellite-2 (ICESat-2) data. Launched in September 2018, the satellite provides measurements of the surface level from the transmission of laser pulses in the green wavelength (532 nm) by the Advanced Topographic Laser Altimeter System (ATLAS) instrument. ATLAS beams provide 210 six tracks, divided into three pairs, on the Earth surface along the ICESat-2 orbit. The beam pairs are separated by ~3.3 km in the across-track direction, and each spot on the surface is ~13 m in diameter of footprint (Neuenschwander et al., 2020). The accuracy expected from ATLAS is approximately 25 cm for flat surfaces and 119 cm in the case of a 10-degree surface slope (Neuenschwander et al., 2020).
ATL08 v.3 dataset, derived from ATLAS measurements, provides every 100 m along-track heights above the WGS84 215 ellipsoid for the land and vegetation (available at https://nsidc.org/data/ATL08). ATL08 data can also represent water surface elevation, and therefore a criterion has been used to separate these from the measurements over the land surface.
Some studies have also shown that the ATLAS instrument can penetrate water and provide information on the bottom (Ma et al., 2020;Parrish et al., 2019). Since the Amazon River has a high concentration of sediments (Martinez et al., 2009), we assume that target information from the water represents only the water surface elevation. 220 Two regions were selected for the topography validation: upstream of the Xingu River, where the tidal amplitude is small (~ 40 cm; Kosuth et al., 2009), and along the oceanic coastal area (see Fig. 1 inset map for the locations of both these regions: "Tidal Riverine area" and "Coastal Area", respectively). In both regions, cloudy conditions and measurements with uncertainty above 50 cm (as indicated by the dataset flags) were discarded. Criteria to remove the points derived from the water elevation were considered in each case. In the tidal riverine region, only ATL08 points from the October-December 225 seasons of 2018 and 2019 were considered as this period represents the low-water season of the Amazon River. Besides, since the flooded areas should show a very low variability of the space-borne measurements along-track, it is easy to detect them in the individual along-track data. Thus, each track's water level was evaluated, and points below 50 cm above the water elevation were discarded. For the coastal area, tracks with a visually markedly different elevation between ocean and continent were selected. Each selected track was evaluated, and points below 1 m above the water elevation were discarded.  Fig. 1). The water depth was obtained by acoustic Doppler current profiler (ADCP) instrument, and a WSE 235 was considered here for estimating the bed elevation concerning EGM08. In section A, we considered the WSE at Óbidos station on the survey day (28 November 2019). In sections, B, C, and D, the WSE at Porto de Moz station on the day of the survey (6 February 2007) was used considering the water surface declivity obtained by the WSE estimated along the river (Fig. 2c). Finally, the WSE measured every 15 minutes at Porto de Santana station on 5 June of 2008 was considered in sections E and F and is described in Callède et al. (2010). Besides, metrics were evaluated considering all points (excluding 240 outliers) in the round-trip survey and repetitions. In section A, four cross-sections were acquired. In sections B, C, and D, only one cross-section was obtained. In section E, six cross-sections, and in section F, nine cross-sections were obtained.

Ancillary databases
The GEBCO is a global terrain model referred to mean sea level with a spatial resolution of 15 arc seconds (approximately 460 m in the Amazon estuary; available at https://www.gebco.net/data_and_products/gridded_bathymetry_data/gebco_2020/).
Since GEBCO data has integer values at intervals of 1 meter, the top-to-raster interpolation was used considering the 1 m isolines to generate data consistent with float values wiping out staircases artefacts. Besides, a low-pass filter with a 9 × 9 (19 × 19) window moving average, i.e., 4.5 km × 4.5km (9.5 × 9.5 km, respectively), was used in the region above 255 (below) the -200 m isobath to reduce the noise caused by in situ multibeam sounding swaths edges.
These combined databases allowed a unified mapping of the topography and bathymetry of the Amazon estuary. However, since MERIT DEM represents the topography of 2010 and some areas in the coastal region may have been eroded or accreted between 2010 and the 2015-2017 period considered in the flood frequency mapping, a procedure was implemented to correct this issue considering three types of regions: 1) erosion areas; 2) accretion areas, i.e., regions where MERIT 260 product does not have topographic information; and 3) GEBCO regions that represent the continent due to sparse spatial resolution, whereas it should represent transition areas or the ocean. The procedure was performed as follows: 1) MERIT DEM areas with topographic information in eroded areas were selected and replaced by GEBCO data, which covers both the continent and the ocean. In the case of substitution to continent GEBCO data, the region is corrected again in step three. ocean, including regions not previously corrected in step one, were removed, and new values were estimated by topo-toraster interpolation considering the neighboring pixels. These areas were manually selected considering the polygons generated from the elevation reclassification into three classes: less than -8 m, between -8 m and 1 m, and greater than 1 mvisually-defined criteria. Figure A1 shows an example of these steps and a corrected area. Finally, to ensure a smooth 270 transition between the nautical charts and GEBCO, an area was selected and replaced considering topo-to-raster interpolation from the neighboring pixels. This area was defined by a buffer of ~2 km around the transition limit, i.e., considering 4 km width. Considering the ICESat-2 data, the estimated topography was also well represented in the riverbanks/floodplain and coastal area with of 0.8 and 0.8, RMSE of 1.5 m and 1.8, and a bias of 0.9 m and -1.5, respectively. However, Fig. 4f shows a bias related to the flood frequency in which overestimations are observed at low flood frequencies (e.g., ~3 m for flood-frequency of 0%). As shown in Fassoni- Andrade et al. (2020a), this bias may be related to the Landsat images used in the flood frequency map that do not represent the flood extent in flood-prone vegetated areas. Thus, the flood frequency in these areas 285 is considered only from situations when the water level exceeds the vegetation height, and, therefore, the flood frequency is underestimated, mechanically overestimating the terrain elevation.   Figure 5 shows the comparison between the in situ cross-sections of the Amazon River and our product. The Amazon river's 295 bed elevation was well represented, with an average RMSE of 8.4 m and an average bias of 4.3 m for the six sections considered together. Keeping aside the small-scale (typically: sub-kilometric) features not resolved by the coarser bathymetric digitized charts, the shape of the cross-sections appear appropriately captured by our product, along the steep banks as well as in the median part. The smallest errors are observed in Óbidos (Section A; RMSE 3.5 m) and North Channel 2 (Section E; RMSE 2.6 m), possibly due to the shorter time lag between the dates of the surveys (< 5 years), keeping in 300 mind the possible variability of the bed bathymetry (Vital et al., 1998). Furthermore, the WSE considered in these sections are from nearby stations, reducing the vertical reference uncertainties. low tidal influence (~40 cm or less), has high levees that limit the overbank flow and sediment accumulation in the 320 floodplain. In the central reach, the stronger tidal range (~1-2 m) and the associated tidal flow suppress the levees' heights, inducing strong overbank transport with high sedimentation rates in the floodplain. In the lower reach with an even stronger tidal range (~4 m), the river canalized transport predominates, and there is little space for sediment accumulation in the floodplain. dots represent the MERIT DEM, and the horizontal lines represent the average (blue) and the maximum (red) river WSE (2015)(2016)(2017)(2018). Note the difference between the average WSE and the floodplain elevation tends to reduce from section 1 to section 5 (h values in grey); that is, the space for sediment accommodation in the floodplain reduces due to sediment accumulation. It can also be observed that the height of the levees is similar to the maximum WSE in sections 1 to 5 (upper 335 and central reach), but from section 6 and further dowstream, the topography, represented by the black points, is higher than the maximum elevation and the river has no flooded banks. This observation in the estuary's lower reach has more uncertainty because it considers the MERIT DEM data (Yamazaki et al., 2019), which were not validated here. Fricke et al.

Summary and significance of the dataset
Our dataset provides the first-ever consistent, high-resolution, and vertically-referenced topography of the Amazon estuary.
Our product's accuracy typically amounts to 8.4 m over the river bed and 1.2 m over non-vegetated inter-tidal floodplains (2015-2018 period). These values appear in line with similar remote, poorly-surveyed tropical or deltaic shorelines (Khan et 350 al., 2019;Salameh et al., 2019). Our mapping is based on an innovative approach using remote sensing data, an extensive and novel dataset of river depth, and auxiliary data over the adjoining areas. We believe that this new approach opens unprecedented opportunities for a straightforward estimation of coastal topography worldwide. The validation approach uses an independent and spatially-distributed dataset of various origins (in situ and remotely-sensed), which provides vital support to our findings' quality. 355 Our overarching goal in assembling this dataset is to characterize the topography and bathymetry of the world's largest estuary. This dataset has many potential applications, such as hydrodynamic modeling, flooding hazard assessment, sedimentology, ecology, physical or human geography, among others. For hydrodynamic modeling, for instance, where the knowledge of topography is instrumental for the accuracy of the results, as well as for geomorphological assessments, which are usually performed with satellite images extracting horizontal information (e.g., width, length) but most often lacking the 360 vertical information, we believe this dataset offers a substantial potential of scientific progress. The dataset can also support ecological studies such as vegetation distribution and carbon balance.
The availability of high-resolution space-borne imagery promised by ongoing operational initiatives (such as the European program Sentinel or the upcoming Constellation Optique 3D -CO3D -mission) opens excellent prospects for frequent revisit updates and improvement of the intertidal part of our product. Keeping in mind the Amazon estuary's energetic 365 morphodynamics, such updates will ensure a perennial quality of our dataset.

Caveats
Our product's main limitation lies in the long timespan of our raw bathymetry data collection (encompassing five decades, Table A1). This limitation is probably sensible regarding the supposed characteristic timescale of the variability of river bed through erosion and accretion processes, as revealed from our validation. Repeated shipborne bathymetric surveys are 370 needed, although the geographical extent of the domain makes it hardly tractable at this mega-delta scale. In particular, it will be opportune to consider the future releases of bathymetric charts by the Brazilian Navy along the Amazon estuary, as they become available in the future years, in case they are based on updated primary bathymetric surveys. The issue is less severe for the intertidal topography, as the timespan of our primary data period is inferior by one order of magnitude (4 years only). Another limitation of our dataset over the intertidal flats and floodplains results from the approach based on remotely-375 sensed imagery of GSW product to estimate the flood frequency. Indeed, it is not expected to work well over the Amazon estuary areas that are densely vegetated. In addition, topographic mapping bias due to flooded vegetation could be avoided using satellite radar data to map the water extent even in flooded vegetation, such as ALOS/PALSAR (Arnesen et al., 2013).
Another issue of using Landsat images for coastal topography estimation is the flood extent representation only every 16 days (Landsat has a sun-synchronous orbit). The tide's temporal variability occurs on an hourly scale, and the amplitude of 380 2 tidal constituent would be observed in the same phase, introducing a bias in the mapping. More investigations are needed using images with more significant temporal variability. The upcoming SWOT satellite mission that will provide for the first time a frequent mapping of the water surface elevation and water extent over continental and riverine areas offers a bright prospect to curb these limitations.

Data availability 385
The dataset generated from this work (listed below) is available at http://dx.doi.org/10.17632/3g6b5ynrdb.1 (Fassoni- Andrade et al., 2021). All other datasets used in our work are open-source data cited within. 1) Bathymetry of the Amazon estuary (Bathymetry.tif and Bathymetry.nc); Elevation in meters relative to the EGM2008 geoid.
2) Topography of the non-forested portion of the lower Amazon floodplain (Topography.tif and Topography.nc); Elevation 390 in meters relative to the EGM2008 geoid.

410
Series of stage values are relative to a so-called "gauge zero", which simply corresponds to the lowest mark on the graduated staff and is referred to an arbitrary datum that is different in each gauge station. Therefore, stages from one gauge cannot be compared in an absolute way to stages at other gauges. It is not possible to obtain the corresponding water surface elevation to derive, for example, the slope of the water surface or relate the water level to a Digital Elevation Models-DEM of the surrounding watershed. Yet, the slope information is a key-parameter for the hydrodynamic modeling of the flow in the 415 basin. The "zero values" of the Almeirim and Porto de Moz gauge stations were surveyed using GNSS (Global Navigation Satellite System) geodetic receivers installed over gauges benchmarks. The data surveyed was computed in Precise Point Positioning (PPP) technique (Héroux and Kouba, 1995), using the GINS software (Marty et al., 2011)