Mapping the spatial distribution of wetlands in Argentina (South America) from a fusion of national databases

Keywords: Argentina, global wetland datasets, local spatial datasets, national databases, spatial dataset integration, spatial distribution of wetlands, wetland mapping, wetland types diversity.

There is worldwide recognition of the social, economic, and ecological values of wetlands, although their global extent has significantly declined in the 20th century (Convention on Wetlands 2021). Economic growth and population density are identified as the main underlying forces leading to systematic degradation and loss of wetlands (Finlayson and D’Cruz 2005; van Asselen et al. 2013; Davidson 2014). In addition, the close relationship of the ecological functions of wetlands with the hydrological regime highlights their sensitivity to climate-change processes. Changes in the water balance owing to increases or decreases in evapotranspiration, the increase in extreme events, and increase in floods as a result of sea-level rise, pose serious risks to biodiversity and ecosystem functions (Day et al. 2008; Erwin 2009; Sandi et al. 2018, 2020; Were et al. 2019; Taillardat et al. 2020; Xi et al. 2021), and can trigger synergistic processes with local and regional land uses (Faleiro et al. 2013; Finlayson et al. 2019; Ponzio et al. 2019). In this rapidly changing world, there is a race against time to prevent wetland loss and degradation with the development of effective conservation policies. Wetland policy and management require reliable information about wetland occurrence, extent, and diversity, which can be fulfilled by developing wetland maps and inventories as stated by Ramsar Convention on Wetlands (Ramsar Convention 1990; Davidson et al. 2018).

Wetland mapping is not a trivial matter. Reports on the extent of wetlands at the global scale show large differences from 8% (9.167 × 10⁶ km²) reported by Lehner and Döll (2004) to 21.1–21.6% (27.5 × 10⁶–29 × 10⁶ km²) given by Tootchi et al. (2019). Factors responsible for the disparity in the estimates in global databases can be attributed to wetland definitions, the diversity of mapping criteria, and the intrinsic ecological characteristics of wetlands (Hu et al. 2017). In addition, despite the wide availability of open remote-sensing scenes and products, most wetlands are not easy to delineate only considering these data sources (Brisco 2015; Gallant 2015; Guo et al. 2017). On regional and local scales, wetlands present large spatio-temporal changes owing to their geomorphological settings and hydrological regimes. Wetlands can be small patches embedded in terrestrial landscapes (meadows, small shallow lakes, etc.) or they can occupy large areas assembled in complex wetland mosaics (Neiff 1999; Neiff and Malvárez 2004; Junk and Cunha 2005; Kandus et al. 2010; Minotti et al. 2013; Benzaquén et al. 2017). This complexity enhances the perception that accurate wetland mapping is an unattainable task for many regions.

South America–Caribbean has the third largest regional extent of wetlands (Davidson et al. 2018) and is the main contributor to tropical wetlands, the majority belonging to the Amazon region (Gumbricht et al. 2017a). Nevertheless, there is a large information gap on the spatial distribution and diversity of wetland types in South America. Although substantial progress has been made, national inventory programs are still incipient (Olmsted 1993; Maltchik 2003; Junk et al. 2013; Cortés-Duque and Rodríguez-Ortíz 2014; Estupinan-Suarez et al. 2015; Kandus and Minotti 2018; Maltchik et al. 2018; Benzaquén et al. 2020). In addition, there is a paucity of researchers working on remote sensing of wetlands in South America (Kandus et al. 2018), and wetland classification scheme agreement or typologies are generally lacking.

Wetland mapping in Argentina does not differ from this South American situation. Argentina signed the Ramsar Convention (1990) and its amendments in 2000. Many Ramsar sites have been established and management plans were put forward since then. But Argentina has been slow in adopting definitions, delineations, and classifications of its wetlands and building an inventory according to their extent and ecological characteristics, a situation shared with most countries in South America (Wittmann et al. 2015). During the 1990s, the first attempt to assess Argentina’s wetland distribution was based on wetland regions defined for the whole of South America (Canevari et al. 1999). In this assessment, an aquatic or bird ecology expert from each region was asked to point out the main wetlands, their characteristics, and threats. The maps provided a general location of the known main wetland areas but did not give an idea of their actual coverage. Kandus et al. (2008) provided the first wetland map attempting to close this gap in Argentina, by interpreting the wetland character on a soil digital database published some years before. Nevertheless, during the past two decades, numerous sources of cartographic information were generated by governmental institutions. Although they were developed for different purposes and without explicitly labelling wetlands, they show wetland occurrence at a more detailed geographic scale than do global wetland maps.

Integrating these seemingly disparate local sources may provide a quick way to know the distribution of wetlands at the country scale with an acceptable spatial resolution as well as a way to speed up national wetland inventory processes. By this reasoning, the objectives of this work are to map the spatial distribution of wetlands and their broad types in Argentina’s mainland through the integration of local open data sources and to compare this overall distribution to published sources using other approaches.

Argentinas’s variety in climate, physiography and lithological complexity provides rich hydrogeomorphological conditions for the development and maintenance of wetlands. Argentina spans between latitudes 21°46′52″ and 55°03′21″S. It crosses subtropical, temperate, and cold zones. Longitudinally, it extends ∼1500 km in the north to less than 300 km in the southern tip of the continent. The Andes mountain range in the west separates the country from Chile and the Pacific Ocean, generating a sharp west–east altitudinal contrast (Pereyra et al. 2004; Veblen et al. 2007; Morello et al. 2012). On the basis of this environmental heterogeneity at the country scale, wetland regions were delineated in the framework of the National Wetland Inventory of Argentina (Benzaquén et al. 2017; Fig. 1). Each region presents a particular variety of wetland type (Fig. 2). The Patagonian Andes present a great number of streams and deep glacial lakes (Fig. 2a). Bog peatlands are located in the very cold and southern tip of the country at the bottom of glacial valleys (Fig. 2b). Palustrine wetlands appear as isolated oases in arid environments, such as fens in steppes of the Patagonia region (Fig. 2c) or wet meadows in the Pune–High Andes region (Fig. 2d). Salt marshes and beaches are typical along the Atlantic Ocean and De la Plata Estuary in the Coast region (Fig. 2e, f). Extensive marshes with hydrophytic prairies and riparian forests are characteristic of the large river floodplains of the Paraná–Paraguay–Uruguay region (Fig. 2g, h). Salt flats and lakes are noticeable in arid and semiarid central and north-west of the country, in Pune–High Andes and Hills–Valley regions (Fig. 2i). Palm savannas and flooded grasslands occupy the interfluve lowlands of the warm north of the Chaco and Paraná–Paraguay–Uruguay regions (Fig. 2j). Mosaics of waterlogged grasslands, shrublands, and forests are common views in the extensive lowlands of the Chaco region, under warm weather with seasonal summer rains (Fig. 2k). Alkaline and freshwater shallow lakes and marshes dot everywhere in the flatlands of the Pampa region (Fig. 2l).

Fig. 1.  Argentina Wetland regions (adapted from Kandus et al. 2017). Boxes A–E are the locations of sites mentioned in the Discussion section.

Fig. 2.  Wetland diversity of Argentina. Pictures show wetland types in the regions of Fig. 1. (a) Lacustrine wetland at ‘Los Glaciares’ National Park. (b) Peatlands in Tierra del Fuego. (c) Palustrine wetland of Patagonian stepe at ‘Perito Moreno’ National Park (fens). (d) Palustrine wetlands in the Puna of Jujuy (wet meadows). (e) Coastal wetland with mussel’s coverage at Puerto Madryn City in Patagonia. (f) Smooth cordgrass (Spartina alterniflora) salt marsh at Samborombón Bay. (g) Palustrine wetland with herbaceous prairies at the Paraná River floodplain. (h) Levee forest of Albizia inundata at the Paraná River floodplain. (i) Salinas Grandes salt lake. (j) Palustrine interfluve at the Pilcomayo River National Park. (k) Mosaic of Palustrine wetlands in the Bajos Submeridionales of Santa Fé Province. (l) Palustrine wetlands (marshes and shallow lake) in the wet Pampa. Photographs: M. Perez Safontas (a), G. Mataloni (b), B. Vilá (d), P. Kandus (c, e–g, i, l), M. Romano (h), P. Minotti (j), A. Carbajo (k).

We searched for spatial data sets developed by Argentine Government Agencies covering the whole country and which included geographic objects that could be interpreted as wetlands. Instituto Geográfico Nacional (2019) develops and maintains digital vector layers for topographic mapping, originally mapped from aerial flights and updated afterwords with satellite imagery. Vector land-cover layers have been developed by Instituto Nacional de Tecnología (1993) from MODIS images using supervised classification using FAO’s LULC scheme (Volante et al. 2009), whereas Ministerio de Ambiente y Desarrollo Sostenible (MAyDS) developed its dataset for the National Native Forest Inventory from Landsat 5 TM+ images by visual interpretation (Secretaría de Ambiente y Desarrollo Sustentable de la Nación 2007). Servicio de Hidrografia Naval (2016a, 2016b) is responsible for bathymetric information, which is delivered as electronic nautical charts or as raster maps. More information and current access to these databases are provided in Table 1. Antarctica and Argentine territories in the South Atlantic islands were not included in this analysis.

Table 1.  National cartographic sources to assess wetland type distribution in Argentina.

The national database providers do not take into account information on wetland hydrogeomorphic attributes or functions, because they developed their GIS layers with a focus on cartographic visualisation. Nevertheless, to provide a common base to integrate their wetland distribution, we grouped all cartographic wetland objects considering the wetland typologies of the Ramsar Convention (1990) and hydrogeomorphic frameworks (Brinson 1993, 2011; Semeniuk and Semeniuk 2016). We considered landform, dominant water source and its transport, geomorphic setting, and whether they were of natural or man-made origin.

We considered the following broad classes: Riparian, Lacustrine, Palustrine, Peatland, Coastal, Underground and Artificial. Each class includes a variety of wetland types that may differ in terms of hydrology, shape, size, and vegetation cover.

This class includes permanent and temporary rivers and streams, together with their associated banks, islands, and natural levees. Woody land-cover types bordering streams and rivers such as riparian and gallery forests are also considered in this group. Most watercourses are expressed as line features, whereas lower reaches of large rivers and their associated forests are usually represented as polygons.

The Lacustrine class gathers all Ramsar lake types. We included both deep and shallow freshwater and alkaline lakes with areas over 8 ha, permanent or temporary, without differentiating between deep waters and lacustrine fringe. Lakes are represented as polygon features in the data sources, sometimes with a direct reference to a lake or salt lake (salar) entity, others just as open water cover.

The class includes peat bogs and mires where the peat layer is being produced and accumulated progressively, with at least 30-cm-thick organic matter on its surface (Joosten and Clarke 2002). These wetlands were interpreted from polygon land-cover databases where the vegetation description refers to Sphagnum moss or peat.

This class gathers cartographic representations of marine and estuarine intertidal environments, including sand or rocky beaches, and coastal dunes land-cover types as well. Some coastal features such as the landward extent of estuaries, gulfs, and bays appear as polygons with named extents. Most intertidal and shallow marine areas can also be interpreted from bathymetry, using the Ramsar criteria that consider less than 6 m deep at low tide (Ramsar Convention 1990).

We grouped here a miscellaneous set of inland wetlands in terms of hydric regime, water chemistry and vegetation cover. They appear as polygon features representing lakes and shallow lakes smaller than 8 ha, pools and wet meadows, swamps and marshes, playas, salt flats, and oases, ‘flooded vegetation’ or ‘permanently flooded’ land covers, or polygons with local type denominations such as Esteros, Bañados, and Vegas. Esteros are wetlands located in paleo-valleys and low concave areas, generally covered by hydrophytic vegetation, permanently or regularly flooded with freshwater, and generally having inorganic soils. Bañados are present in relatively flat lowlands with herbaceous or shrub vegetation, seasonally flooded or waterlogged intermittently, having fresh or brackish water and inorganic soils. Vegas corresponds to temporarily or permanently saturated wetlands dominated by an herbaceous cover of grasses and sedges such as wet meadows, with mineral soils with abundant organic matter (Movia 1984). Under certain environmental conditions, their dense vegetation can give rise to a peat layer (Mazzoni and Vázquez 2004). They receive different local designations such as the Mapuche term mallín in Patagonia or bofedales in the northern high Andes. Depending on the geographic area, the term vegas may also include various types of wetlands with different eco-hydrogeomorphic features, such as peat bogs (Izquierdo et al. 2015) or mountain wet prairies (Irisarri et al. 2012; Mazzoni and Rabassa 2018).

The class considers inland subterranean hydrological systems, geothermal sources, and karst environments. Because our cartographic databases represent surface features, their presence could also be reflected in attributes with toponyms indicating seeps, and cold or hot springs, in point or line features.

All man-made water features were considered in this class, such as reservoirs, agricultural or aquaculture ditches and channels, and other types of artificial ponds.

We developed a spatial data fusion approach that would allow us to assess the known coverage of wetland types in mainland Argentina, and at the same time provide a basis for comparison with already published wetland maps or datasets covering the country.

To build this wetland layer, named herein WetCarto_AR, we designed a two-tier process, by creating a different layer for each wetland category and then populating a raster template. First, we selected the features belonging to the same wetland class from each cartographic source and merged them as a separate polygon layer. The main great rivers of Argentina, (i.e. Paraná or Uruguay rivers) were already available as polygons. For linear features, as vector databases did not provide information on stream, river, or channel width, we used a linear buffer of 20 m to represent their areal extent as polygon features. As shallow marine intertidal areas could be interpreted from the Servicio de Hidrografía Naval (SHN) raster nautical charts portraying depth contours, we developed a polygon dataset by digitising the existing 5-m bathymetric contour and closing it to the continental boundary of Argentina from Instituto Geográfico Nacional (IGN). We attributed each polygon layer with a numeric code that represented the wetland class and the hierarchy to be used when populating the raster template. We also included the provider source and the corresponding Ramsar type whenever possible, to evaluate the contribution of each source.

In a second step, we built a 25-m cell grid template covering our study area, using an equal-area projection (WKID: 102033 Authority: Esri). This cell size matches the finest resolution of existing global datasets and is small enough to capture most spatial coincidences and differences among local or global databases. Using the template, we rasterised each wetland layer assigning its class code. Finally, we ensembled all raster layers into a single raster on the basis of the following hierarchy: underground > peatland > artificial > lacustrine > coastal > palustrine.

Table 3 provides the broad wetland classes associated with their Ramsar type and their dataset source. Riparian wetlands come from IGN, which provides temporary and permanent watercourses, whereas riparian forests come from MAyDS and INTA. Lacustrine wetlands come from IGN and Instituto Nacional de Tecnología (INTA). IGN contributes to every wetland type in the Palustrine class, with INTA contributing mainly with bañados. Coastal wetlands are derived mainly from the SHN, with some additions from INTA and IGN layers. The Artificial class included only elements derived from IGN. Using our search criteria, Underground wetlands could not be identified in any of the data sources.

Table 3.  Wetland classes and types adapted from Ramsar classification and national data sources used for mapping wetlands in Argentina.

Fig. 3 presents the occurrence of wetlands in WetCarto_Ar discriminated by broad class. Wetlands cover 13.5% of mainland Argentina. They are distributed throughout the country, with a greater concentration in the north-east, mainly in the Paraná–Paraguay–Uruguay region. In the western and southern portions of the country, wetlands occur as patches or as thin strips along watercourses, whereas the central portion stands out for its apparent absence of wetlands. Palustrine wetlands are the dominant class followed by Riparian and Lacustrine.

Fig. 3.  Wetland distribution in Argentina according to the WetCarto_AR. The bar chart represents the proportion of each wetland class considering total wetland extent.

The north-east is dominated by palustrine and riparian wetlands characteristics of the Paraná–Paraguay–Uruguay region. Through the Patagonia region, a variety of palustrine, lacustrine and riparian wetlands appear as patches of different sizes. Coastal wetlands extend from Del Plata Estuary coasts along the entire Atlantic Ocean. Peatlands are barely represented, occurring only at the southern tip of the continent in the Patagonia region. Artificial wetlands, represented by channels and hydroelectric and water reservoir dams, are distributed everywhere, particularly associated with the fluvial network. In the Lacustrine class, some lakes appear more visible than others due to their large sizes, such as the deep-water lakes of glacier origin along the western side of the Patagonia region, the large alkaline shallow lakes of Pampa, Hills–Valleys, and Puna–High Andes regions.

Fig. 4 shows the overlay between WetCarto_Ar and the selected layers. Most of our analysis area is non-wetland. WetSoil_AR (Fig. 4d) presents the highest percentage of wetland coverage (17.8%). The global layers (Fig. 4a–c) have a significantly lower coverage (GLW-3 5.8%, CIFOR-SWAMP 7.9%, and SCG 6.6.1 3.5%).

Fig. 4.  Comparison between WetCarto_AR and the other wetland layers. (a) GLWD; (b) CIFOR-SWAMP (c) SDG 6.6.1. (d) WetSoil_AR. Wetland area coincidence (Wc, blue); wetlands only present in WetCarto_AR, Commission (Co, gray); wetlands not present in HumeAR_2021, Omission (O, purple).

The legend in Fig. 4 presents the resulting overlay categories as percentages of our analysis area. WetSoil_AR and CIFOR-SWAMP have the highest wetland area coincidences, but also the largest amount of wetland not identified in WetCarto_AR (commissions). On the contrary, GLW-3 and SCG 6.6.1 omit most of the WetCarto_AR wetland areas.

All datasets show coincidences with the most extensive wetlands in the country. The GLWD-3 layer (Fig. 4a) shares most lakes and identifies the large salt lakes in the centre and north-west of the country (Hills–Valleys and Pampa regions) and the open water bodies and floodplains of the main rivers in the Patagonia region. The SDG 6.6.1 (Fig. 4c) dataset also shows a bias towards open water bodies but includes many small palustrine wetlands, particularly in Pampa region, not contemplated by WetCarto_AR. The CIFOR-SWAMP layer (Fig. 4b) identifies wetland areas in the Chaco and Monte regions that are absent both in WetCarto_AR and the other maps; however, it omits many of the small shallow lakes present in all the other models. SDG 6.6.1 and WetCarto_AR are coincident in mapping coastal wetlands. Palustrine is the most omitted wetland class.

The multi-source fusion layer WetCarto_AR shows the integrated view of the wetland distribution acknowledged by Argentinas’s national mapping agencies. IGN layers provide wetland objects for nearly all the wetland classes considered. Nevertheless, the remaining sources not only made non-overlapping contributions but also included wetland types not represented in IGN.

Our dataset ensemble approach tried to overcome some of these wetland representation problems carried over from using cartographic data sources made with other objectives. Wetlands represented as linear features are a clear example of under-representation (Davidson et al. 2018). In the case of our Riparian class, although buffering the lines may have over- or underestimated the surface of the wetlands, it provided an attempt to incorporate these environments into the national estimates. The resulting width of 20 m for streams and low-order rivers is not far away from the 32 ± 7 cm found by Allen et al. (2018) for many head streams. Coastal wetlands were better represented in WetCarto_AR by considering estuarine and marine bathymetric information. Spatial resolution is a clear limiting factor in landscapes where wetlands are tiny single patches embedded within terrestrial landscapes (Guo et al. 2017; Pande-Chhetri et al. 2017). Considering this, an additional care was taken in the wetland layer compilation, using a predefined hierarchy to assure that the least represented classes were not lost or overlapped by more extensive ones.

Wetland occurrence is markedly uneven in Argentina and WetCarto_AR shows that. Continuous mosaics of palustrine and riparian wetlands are common landscapes in the Paraná–Paraguay floodplains (Neiff 1999; Iriondo 2004; Neiff and Malvárez 2004) whereas the rest of the country appears as an extensive upland matrix where wetlands are isolated patches of a different size (Benzaquén et al. 2017). Palustrine, Lacustrine and Riparian classes are well portrayed as they represent the main and well-known wetland types (lakes, rivers, marshes). The Palustrine class is the most abundant and gathers very different wetland types. Although bañados and esteros are typical of large river floodplains (Aceñolaza et al. 2008; Neiff et al. 2009; Marchetti et al. 2013, 2016), vegas is distributed as small patches throughout Patagonia (Mazzoni and Rabassa 2013, 2018) and in the Puna–High Andes regions (Izquierdo et al. 2015). Lacustrine class includes the emblematic glacial lakes in the Andean portion of the Patagonia region but also salt lakes and flats with enormous economic and social value in regions such as Puna–High Andes, and Hills–Valleys (Karlin et al. 2010a, 2010b; López Steinmetz et al. 2020). Coastal wetlands include extensive marshes almost exclusively dominated by species of Sporobolus and Sarcocornia (Bortolus 2006; Isacch et al. 2006; Gonzalez et al. 2019), but to the south, in the Patagonian sector, coastal wetlands occur as a very thin strip linked to beaches and shoals that widen only at the mouths of the rivers. The central area of the country has very low slopes, few watercourses of endorheic basins, and a marked deficit in water balance (Pereyra et al. 2004), which may have contributed to the absence of recognisable wetlands by the different agencies. Springs, most peat bogs, fens, or small wet prairies, and even antropic wetlands such as rice paddies, are under-represented or absent despite their ecological, cultural, and socioeconomic values. These wetlands may be missing owing to their small size compared with the mapping scale, or because they appear only in local databases but are not present in national datasets, which frequently are also outdated.

All datasets analysed show coincidences with the most extensive and well-known wetlands of the country, such as the large floodplains and delta of the Paraná River, or the great lakes of the Patagonian Andes. However, the layers show a great disparity in the way they identify the rest of the wetland extent. To highlight some of these inconsistencies, Fig. 5 compares the different datasets with greater detail in five selected places. Sites A–C correspond to landscapes made up of complex mosaics of upland and wetland patches of the Chaco and Pampa regions. These mosaics are completely identified as wetlands by WetSoil_AR. Site A, in the semi-arid portion of the Chaco region, corresponds to alluvial deposits covered by riparian forests, shrublands and grasslands, seasonally flooded by the Bermejo River, are culturally valued by local communities, and have high biodiversity heritage (Firpo Lacoste 2018). CIFOR-SWAMP identifies nearly the whole area as wetlands, but they barely appear in WetCarto_AR, and are almost missing in the two others. Site B is in the Bajos Submeridionales, Chaco region, one of the largest, least known, and most threatened wetlands in Argentina (Fundación Vida Silvestre Argentina and Fundación para el Desarrollo en Justicia y Paz 2007). It has seasonal wetlands, flooded during summer rains and covered by a mosaic of freshwater and saline plant communities. Most layers show only the fluvial channel, except for CIFOR-SWAMP and WetSoil_AR. Site C corresponds to a grassland and agriculture matrix in the Pampa region, with Palustrine wetlands being scattered everywhere. This area has rains throughout the year and moderate flood probability (Fluet-Chouinard et al. 2015), but shows a strong interannual variability (Tanco and Kruse 2001), which translates into a huge variation in the occurrence and extent of water bodies between wet and dry periods (Fig. 6a, b). Nearly all layers fail to show wetlands, but the presence of soil complexes and series dominated by hydric soils (Fig. 6c) are strongly indicative of the actual occurrence of wetlands (Vepraskas and Craft 2016; Tiner 2017). Site D shows the Santa Cruz River valley and its surrounding plateaux in Patagonia region. This cold and arid environment presents marshes, fens, and small shallow lakes (Lancelotti et al. 2009). Although all datasets underestimate the presence of wetlands, and CIFOR-SWAMP barely depicts the wetlands associated with the main channel, high-resolution images portray an abundance of tiny wetlands (Fig. 7). Site E shows fluvial and paleo glacier valleys of Tierra del Fuego, home of its peat bogs and mires (Iturraspe et al. 2012). Although 95% of the Argentinian peatlands are located in the southernmost portion of the Patagonia region (Rabassa et al. 1996), they are almost absent in all the layers.

Fig. 5.  Detail comparison of wetland occurrence and extent (light blue) between layers focusing on the boxes of Fig. 1. (a) Alluvial deposits of the Bermejo River; (b) Bajos Submeridionales; (c) Wet Pampa; (d) Santa Cruz River floodplain and surrounding plateaux; (e) Tierra del Fuego mires and peatbogs. Background: Sentinel-2 near-infrared band.

Fig. 6.  Detail of palustrine wetland landscape in Pampa region (see Fig. 5c). Sentinel-2 images colour composition of spectral bands near infrared, short-wave infrared and visible red: (a) 24 December 2020, dry period; (b) 23 February 2017, water-excess period. Vegetation coverage (red and green); low vegetation coverage and nude soil (cyan); water (blue). (c) Soil Chart Scale 1:50 000. Natraquolls (1), Natraqualfs, as main taxons of the soil series (2); series with Natraquolls and Natraqualfs as secondary soil taxons (3); series where Natraquolls and Natraqualfs are present as taxons of third order (4); series without presence of hydric soils (light pink, 5). Water bodies and streams in blue. Soil chart source: Instituto Nacional de Tecnología Agropecuaria (1993, see http://anterior.inta.gov.ar/suelos/cartas/).

Fig. 7.  Detail of Santa Cruz River floodplain and neighbourhood in Patagonia region (see Fig. 5e). Colour composite of Planet Team image 20210102_143540_27_2401 (Planet Team, San Francisco, CA, USA, see https://api.planet.com), spectral bands: near infrared, visible red and green. Vegetation coverage of vegas in deep red; river water in blue or cyan, flood plain in greenish colours; herbaceous and shrubby steppe in reddish-brown colours.

In Argentina, small wetlands are scattered over the entire surface of the country, but their identification is most critical in the arid and semiarid regions because of their provision of water for local communities, livestock production, and the maintenance of biodiversity (Blanco and de la Balze 2004; Perotti et al. 2005; Mazzoni and Rabassa 2013; Epele et al. 2021). The low spatial resolution of the WetSoil_AR map presents little sensitivity to identify small wetlands in terrestrial landscape matrices, but they are also poorly represented in global datasets despite their more detailed spatial resolution.

The use of multitemporal remote-sensing approaches combined with machine learning is considered as an efficient approach for wetland landscape mosaics (Allen et al. 2012; Farda 2017). The representativeness of the diversity of wetland types, then, should be a critical factor in the selection of training areas to feed classification algorithms. Although using this approach, the SDG 6.6.1 dataset fails to record extensive complex wetland landscapes in our area of analysis, highlighting the importance of sample design for training areas (Amani et al. 2019; Tamiminia et al. 2020), but also taking into account knowledge of the types of wetland present.

The integration of wetland data from several local sources, developed with different objectives and methods, can provide valuable bottom-up information to feed and validate global models when wetland maps or inventories are lacking or incipient. Nevertheless, the wetland datasets presented in this work, local or global, have value as general models to assess the spatial distribution of wetlands and to compare different regions with fairly similar criteria. As expected, none of these datasets seems to be useful tools for local management, because they may not guarantee the accuracy and completeness required.

Our approach stresses that the contribution of non-wetland national databases, even when generated for other purposes, should not be underestimated, because they can be easily used anywhere as a starting point for countries or regions where wetland inventories are still missing.

This work presents a wetland dataset built with a bottom-up approach, that makes visible the known spatial extent of wetlands in Argentina. WetCarto_AR identified 13.5% of mainland Argentina as wetlands, which is higher than the estimates of global datasets but lower than a previous value based on the interpretation of soil information. It highlights the dominance of palustrine, lacustrine and riparian wetland types, and also shows that certain wetland types, such as springs, mountain wet meadows or peat bogs, will appear only when the mapping objectives are focused on their identification.

Websites of open datasets used are listed in Tables 1 and 3.

The authors declare that they have no conflicts of interest.

This work was supported by the National Agency for Scientific and Technical Promotion of Argentina under Grants FONCyT MinCyT PICT-2014-0824, PICT-2014-2860 and PICT-2019-0036.

P. Kandus set up the objectives and the initial version of the manuscript. P. Minotti proposed initial ideas about local and regional database integration and contributed with technical considerations of database management. I. Fabricante analysed and integrated the databases and developed tables and figures. All authors edited and reviewed the different manuscript versions. All the authors read and approved the final paper.