National Wetland Map 5 – An improved spatial extent and representation of inland aquatic and estuarine ecosystems in South Africa

The improved representation of inland aquatic (freshwater) and estuarine ecosystems and associated data was a key component of the 2018 National Biodiversity Assessment, and is an essential step in enhancing defensible land use planning and decision making. This paper reports on this enhancement of the National Wetland Map version 5 (NWM5) for South Africa and other data layers associated with the South African Inventory of Inland Aquatic Ecosystems. Detail is provided on (i) the extent of wetlands mapped in NWM5, compared to previous versions of the NWMs; (ii) the improved extent of inland wetlands mapped in focus areas in NWM5 relative to NWM4; (iii) the type of cover associated with the wetlands (inundated, vegetated or arid); (iv) the ecotone between rivers or inland wetlands and estuaries mapped as freshwater—estuarine transition zones; and (v) level of confidence for the inland wetlands in terms of how well the extent and hydrogeomorphic units were captured for each sub-quaternary catchment of South Africa. A total of 4 698 824 ha (3.9% of South Africa) of inland aquatic ecosystems and artificial wetlands have now been mapped, with NWM5 delineating 123% more inland wetlands (2 635 809 ha or 2.2% of SA) compared with NWM4. The estuarine functional zone, which encapsulates all estuarine processes, associated habitats and biota, was refined for 292 systems totalling 188 944 ha, with the addition of 42 micro-estuaries totalling 246 ha. Nearly 600 000 ha (0.5% of SA) of artificial wetlands were mapped in SA. Inland wetlands are predominantly palustrine (55%), some arid (34%) and few inundated systems (11%). Transition zones between freshwater ecosystems and estuaries formed a small fraction (<1.5%) of river total extent (164 018 km), indicating an ecotone where biota and processes continuously vary from freshwater to estuarine. The majority of inland wetlands (~70%) had a low confidence ranking for designation of extent and typing. Future improvements of the map should be focussed on catchment-based improvements particularly in strategic water-source areas, areas of high development pressure and those with low confidence designation of wetland type.

headline indicators of the National Biodiversity Assessments (NBA), as well as the listing of threatened ecosystem types under the National Environmental Management: Biodiversity Act (NEMBA), Act 10 of 2004 (RSA, 2004). Owing to the poor representation of inland wetlands in previous NWMs (Mbona et al., 2015;Schael et al., 2015;Melly et al., 2016), as well as the need to improve the representation of South Africa's estuaries, a significant effort was made to improve the National Wetland Map version 5 (NWM5) in preparation for the NBA 2018.
In addition to national obligations in the improvement of the NWM, South Africa also has international obligations in reporting the extent, biodiversity and integrity of its wetlands (inland wetlands and estuaries). The results of the NBAs are used by the Department of Environment, Forestry & Fisheries (DEFF) to inform the global Convention on Biological Diversity (CBD), whereas the extent and quality of wetland types are important for monitoring and reporting on the Sustainable Development Goals (SDGs) to the United Nations Environment Programme (UNEP) (UN, 2015). The extent of vegetated and inundated wetlands is, for example, required under indicators related to SDG6 ('Ensure availability and sustainable management of water and sanitation for all'), whereas ecosystem diversity, protection and restoration is reported under SDG15 ('Protect, restore and promote sustainable use of terrestrial ecosystems, sustainably manage forests, combat desertification, and halt and reverse land degradation and halt biodiversity loss'). The improvements to NWM5, therefore, had to include both ecosystem diversity information, ecological condition, as well as cover type, including types that are inundated (lacustrine), vegetated (palustrine) or arid.
The National Freshwater Ecosystems Priority Areas (NFEPA) atlas compiled in 2011 ) used a number of sub-national datasets to improve the spatial extent of inland wetlands mapped in NWM3, and modelled some of the wetland ecosystem types to Level 4A of the Classification System (Ollis et al., 2013). The NFEPA wetlands dataset was subsequently used in the NBA 2011 for the assessment of the headline indicators of inland wetlands (Driver et al., 2012) and adopted by the South African National Biodiversity Institute (SANBI) as NWM4. Several studies indicated that NWM4 showed up to 46% omission errors compared to wetlands mapped at finer scales and that the hydrogeomorphic (HGM) units were incongruent between the modelled and fine-scale mapped wetlands (Mbona et al., 2015;Schael et al., 2015;Melly et al., 2016). The shortcomings of the NWM4 could be attributed, on the one hand, to the modelling of the extent of the wetlands using space-borne Satellite Pour l'Observation de la Terre-5 (SPOT) and Landsat multispectral imagery with inappropriate spatial resolution for detecting some of the small-scale inland wetlands in the semi-arid to arid South Africa (Thompson et al., 2002). Apart from omission errors, an evaluation of the NWM4 for Gauteng also revealed a commission error (a false presence of inland wetlands) of 32% (Van Deventer et al., 2018a). Essentially, the spectral bands of these multispectral images are unable to distinguish vegetated wetlands from adjacent terrestrial vegetation, resulting in poor representation of these systems in NWM4. Extrapolating these statistics to the rest of the country suggested that about half of the South African inland wetlands might not be represented in the NWM, of which two-thirds are truly wetlands, resulting in only a third of South Africa's wetlands being represented in NWM4. Thus, the improved representation and typing of inland wetlands was crucial for improved assessment of the Ecosystem Threat Status (ETS) and Ecosystem Protection Levels (EPL) for the NBA 2018.
Since the NFEPA wetland maps of 2011, several national and sub-national wetland datasets have been published, which have enabled the improvement of the extent and wetland types across the country (Van Deventer et al., 2018a;Van Deventer et al., 2018a,b). The NFEPA wetlands had a tremendous impact at several levels . In addition, several funding sources, made available during the onset of the NBA 2018, facilitated the improvement of the extent of the wetland map in a number of district municipalities. Other examples are the National Land Cover data (GTI, 2015;, which mapped the open water bodies of natural and artificial wetlands, and the Leaf Area Index (LAI), as a proxy for vegetation biomass, which was generated for South Africa through inversion modelling from the Moderate Resolution Imaging Spectroradiometer (MODIS) imagery (Cho et al., 2017). In addition to the vegetation bioregions included in the National Vegetation Map update for the NBA 2018 , cover types for wetlands could be estimated. Thus several opportunities were aligned to improve the extent of the wetlands in the NWM5, the typing of inland wetland ecosystems and the determination of the cover type.
Estuaries require a more accurate delineation of the spatial extent as the estuarine functional zone represents a 'development setback line' and feeds into supporting estuarine and coastal management processes and legislation (Van Niekerk et al., 2018;2019a;Harris et al., 2019). Since there are less than 300 functional estuaries in South Africa, it was achievable to place extensive effort into addressing previous deficiencies in the updated estuary delineation of 2018 (Van Niekerk et al., 2019a, in press). As in the case of the inland wetlands, several projects enabled improvement in assessment of coverage of estuarine functional zones (EFZs) and the demarcation of micro-estuaries for the first time in the NBA process (Van Niekerk et al., in press). An estuary is a partially enclosed permanent water body, either continuously or periodically open to the sea on decadal time scales, extending as far as the upper limit of tidal action, salinity penetration or back-flooding under closed-mouth conditions (Van Niekerk et al., 2013, modified from Cameron andPrichard, 1963). Extremes to this generic definition are during floods, when an estuary can become a river mouth with no seawater entering the formerly estuarine area, or, when there is little or no fluvial input, an estuary can be isolated from the sea by a sandbar and become fresh or hypersaline (modified from Van Niekerk and Turpie, 2012;Van Niekerk et al., 2013). Veldkornet et al. (2015) highlighted that critical estuarine habitats, such as salt marshes and swamp forest, were excluded from the NBA 2011, thereby under-representing the functional zones of these systems. Additional supporting information, available since the conclusion of the previous NBA 2018, is the significant effort that has gone into updating the National Estuarine Botanical Database with field observations and more detailed mapping (Adams et al., 2016), the Light Detection and Radar (LiDAR) data collection for parts of the coast, promising a spatial accuracy in mapping between 5 and 10 cm in the x, y and z spatial components, the use of supporting datasets (such as the 5 and 10 m interval above mean sea level contours [DRDLR:NGI, 2017]) and the Shuttle Radar Topography Mission (STRM) 30 m data which is readily available in digital format (USGS, 2004). Estuaries are dynamic ecosystems and without a proper understanding of changes over time, these ecosystems cannot be assessed and managed appropriately. Google Earth provided such a time-series dataset, allowing the mapping of the ever-changing processes that characterise South African estuaries, e.g., changes in mouth configuration and inundation patterns.
Micro-estuaries (i.e. estuaries < 500 m in length and/or < 2 ha in size), river outlets, coastal seeps, ephemeral systems, and waterfalls, were only represented as x,y point data in 2011, but in the current NWM5 iteration effort was made to map some of these smaller features (Van Niekerk et al., 2019a;. It was also deemed important that transition zones above estuaries and into coastal freshwaters be identified and demarcated. These are essentially river-estuary ecotones immediately upstream of estuaries but not including the riverestuarine interface (REI). The REI is defined as the area in an estuary within which salinity ranges from 10 to zero under the influence of the upstream limits of back-flooding or tidal intrusion.
An ecotone is a 'zone of transition between adjacent ecological systems, having a set of characteristics uniquely defined by time and space scales, and by the strength of the interactions between adjacent ecological systems' (Di Castri et al., 1988 p.19). Previous research dealt with processes within relative homogenous landscapes (Holland et al., 2012), but a growing appreciation is that abiotic and biotic components move through heterogeneous landscapes and the boundaries (ecotones) between such units are dynamic. In aquatic ecosystems, physicochemical change for animals is far more pronounced than in the terrestrial environment; an example is the freshwater and estuary ecotone with a substantial change in chemical condition (e.g. salinity) (Thorp, 2015). This ecotone is critical for satisfying the life-cycle needs of many organisms and can be characterised by high productivity and ecological diversity (due to edge effects). River-estuary ecotones are areas which, during any time of their hydroperiod, would host estuarine and riverine biota but only have abiotic characteristics of freshwater ecosystems. These ecotones have previously been poorly defined, owing to their dynamic nature over long hydrological cycles and research being largely deficient with respect to their nature and characteristics (see also Rundle et al., 1998). They are essential supporting habitats to estuarine systems, and require proper mapping for the purpose of policy formulation and their protection. In addition, coastal aquatic ecosystems are also poorly understood and mapped for South Africa. A number of coastal depressions were identified according to the criterion '…organisms of estuarine origin (algae, crustaceans, and fish, which are relicts in the case of lakes cut off from the sea since the last Ice Age) but are normally uninfluenced by the sea' (Noble and Hemens, 1978 p. 37).
SANBI, in collaboration with the Council for Scientific and Industrial Research (CSIR), coordinated the improvement of the NWM5, supported by several other institutions. The aim was to improve the representation of the spatial extent and type of inland wetland and estuarine ecosystem types of South Africa in NWM5. We report the total spatial extent of inland wetlands and estuaries mapped in NWM5 as follows: • The extent of wetlands mapped in NWM5 in comparison to previous versions of the NWMs • The improvement in the representation of inland wetlands mapped in focus areas in NWM5 relative to NWM4 • The type of cover associated with the wetlands (inundated, vegetated or arid) • Ecotones between rivers or inland wetlands and estuaries mapped as river-estuary ecotones • The level of confidence for the inland wetlands in terms of how well the extent and hydrogeomorphic units were captured for each sub-quaternary catchment of South Africa Our intention is to inform users of the improvements and shortcomings of NWM5 so that it is appropriately used in planning and decision making, whilst enabling better planning for the wetland inventory of South Africa. During the integration and mapping phase, the attributes of inland wetland types had to be completed for Levels 1, 3 and 4A of the Classification System. Back-drop imagery included the freely available 50 cm colour orthophotography available through the ArcGIS online web map service from DRDLR:NGI dated from 2012 to 2013. SPOT imagery was also used in some instances, dated to similar years. Unfortunately, most of these images were largely taken during the dry season, possibly to avoid cloud cover, and therefore were less suitable for the purpose of wetland mapping. Where the data capturer found it difficult to judge the extent or HGM type of the wetland, historical images available through Google Earth were accessed to support the mapping of inland wetlands.

Improving representation of inland wetlands
Decisions regarding the extent and ecosystem type for features were guided primarily by three principles: to always map the maximum extent of a wetland, if possible, to map the original extent (historic maximum), and to retain the extent and typing done in fine-scale datasets. The focus areas were then reviewed by national wetland experts (Freshwater Consultancy Group Pty Ltd (FCG) and Wetland Consultancy Services (Pty) Ltd (WCS)) and corrections implemented by data capturers. Subsequently, the available data for the remainder of the provinces were integrated with limited mapping of large floodplains, wetlands in Ramsar sites, and nine limnetic depressions. Limnetic depressions were considered unique wetland types where the maximum depth exceeds 2 m at the average annual low-water level of an open waterbody (Ollis et al., 2013). The nine limnetic depressions mapped in NWM5 included Barberspan (North West), De Hoop (Western Cape), Groenvlei (Western Cape), Lake Banagher, Lake Chrissie and Tevredenpan (Mpumalanga), Lake Fundudzi (Limpopo), Lake kuZilonde (KwaZulu-Natal), and Lake Sibaya (KwaZulu-Natal) (compiled from Hill, 1969;Miller, 1998;Noble and Hemens, 1978; https://www.lakepedia. com/). Following the integration of the provincial datasets, these were reviewed and edits implemented.

Mapping artificial wetlands as a separate layer
To better represent the original wetland extent for the NBA 2018, artificial wetlands were compiled as a separate feature class layer in the ArcGIS geodatabase. This was to assess disturbance effects on wetland ecosystem types in the NBA Where artificial wetlands were situated within an inland wetland ecosystem type, the artificial wetland polygon had to be merged with the adjacent polygon to represent the original extent prior to the modification. Isolated artificial wetlands were completely deleted from the NWM5.

Improving the representation of estuaries
The mapping convention for determining the 'estuary functional zone' (EFZ) was based on a precautionary approach with the departure point being the first inland 5 m contour above mean sea level, to capture all estuarine processes and biotic responses. This dataset was then adjusted to address the shortcomings identified by Veldkornet et al. (2015), e.g., exclusions of swamp forest (freshwater mangrove) and salt marsh areas contiguous to estuaries. In addition, all habitat features excluded from ground-truthed vegetation maps were also incorporated in the new delineation (Adams et al., 2016). The EFZ boundary, where possible, mapped the maximum extent of the historical and present geographical boundary (although resources were only available to do this accurately in KZN). The location of estuary outlets was determined by the maximum extent of migration of the estuary mouth or outlet (furthest north and south) as identified from any historical image, i.e., Google Earth or historic aerial photographs. Areas identified by coastal LiDAR datasets (corrected to mean sea level and verified by experts) or the 1:100 year flood line delineation (only available for the Groot Berg or Breede estuaries) were also incorporated. All 'island' type features created by high elevation areas surrounded by estuarine floodplains were incorporated into the EFZ. In addition, in the case of small, incised estuaries (i.e. where the 0 m, 5 m and 10 m contour are close to one another) with relatively high river inflow, delineation was based on the 10 m contour above mean sea level to accommodate mapping uncertainty and lateral movement. The EFZ was also extended to incorporate environments that are predominantly surrounded by estuarine habitats or processes (i.e. more than 75% of feature is surrounded by S-and U-bends). In addition, habitat features that support estuarine functioning were also included to ensure future health, i.e., upstream inland wetlands that influence estuarine water quality by filtering nutrients. This included the incorporation of small areas of inland aquatic ecosystems contiguous to estuaries, e.g., seeps and springs. Overall, the 2018 revised EFZs strived to incorporate all vegetation ecotones that have elements of estuarine habitat, e.g., mosaics of swamp and dune forest. The EFZ was broadened to include novel ecosystems such as marinas and harbours adjacent to estuaries as they directly influence estuarine functionality and biodiversity. Where possible, names were changed to provincial standards, e.g., the KwaZulu-Natal (KZN) provincial gazettes and Eastern Cape conservation plan.
Estuarine and provincial inland aquatic datasets, now including inland wetlands and the extent of some large rivers, were merged into a national dataset. The draft NWM5 was checked for overlapping polygon (topology) errors, features aligned along provincial boundaries as well as between inland aquatic ecosystems and estuaries. The total extent (hectares) of all the NWM versions 1 to 5 was updated in ArcGIS and calculated as a percentage of the total of South Africa's extent. Similarly, the surface area of each estuarine and inland aquatic ecosystem type is summarised and the percentage calculated for the total extent of South Africa. These statistics of NWM4 and NWM5 are then reported for the 10 focus areas to highlight the improvements of the digitising efforts.

Determining inundated, arid and palustrine wetland extent for SDg reporting
To determine the extent of inundated wetlands, the seasonal and permanent water and wetland classes from the 30 m National Land Cover data of 2013/14 (GTI, 2015) were extracted. The extent of the inundation was then calculated as a percentage of the extent of all inland wetlands. Subsequently, the remaining extent of the inland wetlands was combined with the LAI predicted using the 463 m spatial resolution MODIS image of 13 March 2013 (Julian Day 73) to indicate potential ranges of vegetation biomass (Cho et al., 2017). A LAI range of 0-1 was used to distinguish low to no vegetation and therefore likely to be arid, whereas a LAI range from 1 to 8 was considered to indicate dense grass to tree cover, and therefore more likely to be palustrine wetlands (Cho, 2018). The total amount of pixels for both processes was extracted using the zonal statistics tool in ArcGIS 10.3 (ESRI, 1999(ESRI, -2014.

Mapping river-estuary ecotones
In addition to the EFZs, the transition zones between the rivers or inland wetlands and estuaries were also mapped as river reaches where riverine abiotic processes dominate, and where riverine and estuarine biota occur (during any time of their hydroperiod). These areas are not subjected to tidal action or back flooding and at no stage experience an increase in salinity as a result of tidal penetration. These river-estuary ecotones are poorly understood and have previously not been delineated, owing to their dynamic nature over long hydrological cycles. River lengths, varying between 0.5 km and 30 km, were defined based on sampling data, topography and expert opinion. The steeper the gradient, the smaller the ecotone. Thus rivers with extensive lowland reaches such as the Breede have longer ecotones. Ephemeral rivers upstream of estuaries were excluded from this delineation. The extent of the river-estuary ecotone was mapped and included in the national rivers database of the NBA 2018 (Smith-Adao et al., 2018).
River-estuary ecotones were identified based on expert knowledge of the occurrences of estuarine-associated fish and invertebrate assemblages. These assemblages within river-estuary ecotones are a mixture of typically estuarine species (e.g. moony Monodactylus falciformis), species adept at completing their entire lifecycle in both habitats (e.g. estuarine roundherring Gilchristella aestuaria), freshwater species that may have an estuarine phase of their life-history (e.g. multi-specific freshwater prawns of the genus Macrobrachium) and catadromous Anguillidae eels and swimming crabs Varuna litterata that are either resident, or migrate through coastal freshwaters whilst recruiting to the upper catchment or migrating back as adults to spawn far offshore in the abyssal depths. Typically estuarine species include, but are not limited to, 'facultative catadromous' fish such as freshwater mullet Pseudomyxus capensis that may opportunistically spend most of their life in freshwater but return to spawn in estuaries or the sea. Supplementary Material I lists systems where these species assemblages have been observed, and hence where river-estuary ecotones exist.
River-estuary ecotones were not incorporated into the EFZ, as they are not subject to estuarine abiotic processes, but they should be highlighted as estuarine supporting areas in planning legislation and approaches to ensure that future developments, or discharges or abstractions, do not disrupt or degrade estuarine connectivity and ultimately condition.

Rating Description 1 -Low
Desktop mapping of the extent of inland wetlands was done by non-wetland specialists for a part of or to the full extent of the sub-quaternary catchment (SQ4) (Mostly DRDLR:NGI for non-wetland typing purposes).

-Low to Medium
Desktop mapping of inland wetland extent was captured by interns for the purpose of NWM5 to the full extent or a part of the SQ4. This may also include areas where data from wetland specialists had been incorporated, but the dataset was either not typed to the HGM unit, or complete for all HGM units, or was based on old imagery.

-Medium
Desktop mapping of the extent of inland wetlands and HGM typing was done by wetland specialists for the full extent of the SQ4.

-Medium to High
Desktop mapping of the extent of inland wetlands and HGM typing, as well as field verification and revision by experts was completed for the full extent of the SQ4.

-High
Inland wetlands have been mapped and verified for a period of > 10 years over multiple hydrological cycles/ hydroperiod for the full extent of the SQ4. Verification may include field observations as well as soil and/or vegetation surveys.
Coastal depressions were refined from Noble and Hemens (1978) using available grey and scientific literature, as well as expert opinion (Van Deventer et al., 2018a). The limnetic status of these systems has been added to the NWM5 database under fields related to the hydroperiod. Other inland wetlands located on sandy coastal plains near the coast or estuarine systems were also attributed as 'coastal' in the NWM5 dataset.

Confidence ranks of the inland wetlands
The final step was a confidence map, generated for the inland wetlands based on the extent of a sub-quaternary catchment (SQ4) that was mapped in full, as well as the degree of expertise involved in the wetland mapping, the completeness of the HGM unit and extent to which the hydroperiod is known for a wetland (Table 2). Statistics are reported for the number of SQ4s relative to the total number of SQ4s in South Africa for each of the five ranks, as well as the percentage of surface area of South Africa which is likely to be mapped according to the ranks.

Improvement of the National Wetland Map 5 compared to previous versions
A total of 4 596 509 ha of inland aquatic ecosystems and artificial wetlands has been mapped in South Africa, constituting about 3.8% of the surface area of the country (Table 3; Fig. 1; Supplementary Material II). The extent of South Africa has been calculated as 121 973 563.7 ha using the provincial boundaries of the Municipal Demarcation Board (MDB) of 2011, however the marine reserves have been excluded from the shapefile by the CSIR. Aquatic ecosystems, including estuaries, inland wetlands, and some river channels, totalled 3 998 120 ha (3.3%), while wetlands (inland wetlands and estuaries) totalled 2 851 888 ha (2.3%). The extent of the ecosystems and attributes represented in NWM5 has increased compared with the previous versions of the NWMs (Fig. 3). In addition, the artificial layer has been separated from the inland aquatic data, and now forms part  In the NWM5, inland wetlands constituted more than 157 000 polygons, totalling more than 2,6 million ha or 2.2% of South Africa (Table 3). The inland wetlands have increased in extent from NWM4 (NFEPA wetlands of 2011) to NWM5 by 23%; now comprising 2.2% of the surface area of South Africa (Fig. 2). The extent of the 290 EFZs in NWM5, which falls within the boundary of South Africa, increased by 21% from the previous version (from 0.14% to 0.16% of the country's surface area) (  Fig. 3). Wetland flats modelled in NWM4 were corrected to depressions in NWM5 resulting in a marked decrease of 90%. Hillslope seeps have  been amalgamated into the seeps category, resulting in a 153% increase in the spatial extent, when comparing the combination of hillslope seeps and seeps of NWM4, to seeps of NWM5 (Fig. 3). When the percentage of HGM units are compared between NWM4 and NWM5, relative to the total extent of inland wetlands mapped, NWM5 mapped more valley-bottom and seep systems (3-9% more in the extent) compared to NWM4. The extent of depressions, wetland flats and floodplains were greater in NWM4 compared to NWM5 (0-6% more in extent). It is interesting to note that floodplain wetlands achieved a similar percentage (21%) of the total spatial extent of inland wetlands.

Improvement of the representation of inland wetlands in focus areas
The majority of the 10 focus areas, where wetlands data were integrated and additional wetlands mapped in NWM5 for the NBA 2018, showed an increase in the extent of inland wetlands compared to NWM4 (Table 5; Fig. 4). Five of the ten focus areas (Amathole, Buffalo City, Cape Winelands, Lejweleputshwa, and the Frances Baard municipalities), however, showed a reduction  in the extent of inland wetlands. For these areas, commission errors from remote sensing and probability mapping, which had been included in previous versions of the NWM, were removed in NWM5. The four focus areas (Eden, Ehlanzeni, Vhembe, and the West Rand District municipalities) showed increases in the extent of inland wetlands of between 54% and 127%. The West Rand District Municipality achieved the largest increase in extent of nearly 600% compared to NWM4. The extent of inland wetlands for the Hogsback and Tevredenpan study areas increased from the NFEPA (NWM4) wetlands to NWM5 by 107% and 290%, respectively, following in-field visits and corrections (Table 5). The total extent of inland wetlands ranged from <1% to 43% of the surface area of the respective study areas.

Cover types of inland wetlands (inundated, palustrine or arid) for SDg reporting
Inundated wetlands made up an estimated 278 719 ha or 11% of the extent of South African inland wetlands (Fig. 5) and are distributed across the country. Inland wetlands which are more likely to be vegetated or palustrine (LAI > 1) are found in the Fynbos, Grassland and Savanna Biomes of South Africa, totalling an estimated 1 447 932 ha or 55%. Arid systems are in the central Karoo and Northern Cape Provinces primarily, and a total of 909 157 ha or 34% have been estimated where the LAI < 1.

Estuarine support areas (river-estuary ecotones) and coastal depressions
A total of 1 931 km of transition rivers (Fig. 6) have been identified (Supplementary Material I) which constitutes about 1% of the total length (164 018 km) of rivers as identified by . Almost 30 000 ha of inland wetlands were found to coincide with the coastal regions of South Africa ( Table 6). The majority of these are coastal depressions, of which Groenvlei (357 ha) and Lake Sibaya (8 233 ha) were the only limnetic depressions.

Confidence ranking of the inland wetlands of NWM5
Much of the country has been mapped by non-wetland-experts with limited understanding of wetlands with a low confidence overall (76% of country's surface area, Fig. 7, Table 2). Almost 17% of the surface area of the country has been attended to at a desktop level through the integration of existing data and/or the mapping of wetlands by interns trained during the update of NWM5 (representing low-medium confidence). Only 7% of the country has been mapped and typed to HGM units by wetland experts (i.e. medium confidence), and a further 0.04% of the country including site visits and subsequent improvements to the representation (extent and ecosystem typing) of the map. No area has been mapped and refined following long-term research (Category 5 = 0%) (i.e. high confidence).

DISCuSSION
In the past four years, a significant effort has been made to improve the representation of various datasets in the South Figure 5. Distribution of cover type of inland wetlands across South Africa  African Inventory of Inland Aquatic Ecosystems (SAIIAE), particularly the National Wetland Map version 5 (NWM5) in preparation for the National Biodiversity Assessment 2018, and for reporting to international conventions on behalf of South Africa. The effort involved more than 30 data capturers from more than 10 organisations, at a total estimated cost of R7 million (Van Deventer et al., 2018a). The results paid off, in that a total amount of 4 596 509 ha (3.8% of South Africa) of inland aquatic ecosystems and artificial wetlands have been mapped for South Africa. Despite this tremendous effort, the majority of sub-quaternary catchment extents were at a low confidence that the extent and hydrogeomorphic unit have been well represented at a desktop level. Wetlands in arid to semi-arid regions are poorly detected through remote sensing indices, which often use open water indices for extracting wetland extent. As a result, few palustrine and arid systems are well represented in regional to global wetland maps. The extent of wetlands for Africa, for example, has been estimated at 22 440 000 ha or 0.7% of the surface extent of Africa in the Global Lakes and Wetlands Database (GLWD) (Lehner and Döll, 2004). The more recent Global Inundation Extent from Multi-Satellites (GIEMS) dataset (Fluet-Chouinard et al., 2015), at a spatial resolution of 420 x 460 m (total about 19.3 ha per pixel), estimates the extent of wetlands in South Africa at 4.2% of the surface area. It does not, however, distinguish between natural or artificial wetlands, or amongst estuaries, rivers or inland wetlands. Previous wetland maps for South Africa, done using remote sensing and prediction modelling, have proved to underrepresent the full extent of arid and palustrine wetlands. In semi-arid to arid countries, such as South Africa, heads-up digitising and in-field verification are essential to improve our national maps.
The investment made in the NWM5 showed that a significant improvement in the representation of inland wetlands can be achieved. Commission errors associated with the extent of wetlands were removed for some focus areas, resulting in a decrease of the incorrect representation of inland wetlands. Several focus areas, however, showed an increase in the extent of wetlands. During the update of the NWM5, the base data of wetlands for the West Rand District Municipality was combined with the wetlands probability map of Collins (2018), resulting in an increase in extent by nearly 600%; though without in-field verification, the confidence ranges between low and moderate.
In-field verification of wetlands by wetland ecologists (Mr Anton Linström and Ms Nancy Job) for the Tevredenpan and Hogsback study areas, resulted in an increase of more than 100% and a moderately high confidence rank (Van Deventer et al., 2019). Investment in similar efforts is therefore crucial for improving the representation of wetland extent to a moderate confidence level. Further improvements should be done for selected priority areas, at catchment levels, particularly in the strategic water source areas (Nel et al., 2017;Le Maitre et al., 2018), areas of high development pressures, and others identified through the NBA 2018 assessment of threatened ecosystems.
Mapped estuarine area increased by almost 5% from the previous version. Some smaller systems moved to the microestuaries category and four new small estuaries were mapped and added. However, significant improvements were made in the incorporation of all estuarine and estuarine-associated habitats in the updated delineation. In addition, the dataset now includes 42 micro-estuaries. The use of time-series data vastly improved the mapping of the estuarine extent and allowed for the incorporation of dynamic features such as the river or estuary mouth position. LiDAR data showed significant promise, but unfortunately the datasets were not post-processed adequately, e.g., included the return signal from treetops instead off ground level, to be used without expert judgment. In the future, this type of dataset would benefit from extensive post-processing to increase reliability and assist with increasing mapping accuracies. The incorporation of 1:100 year flood lines allowed for the incorporation of all relevant sediment processes but was unfortunately only limited to two large systems. However, they supported the use of the 5 m contour above mean sea level as a proxy for sedimentary and inundation processes. While the 5 and 10 m above mean sea level contour dataset (DRDLR:NGI, 2017) proved to be useful in supporting the delineation of estuaries in the lower reaches where the floodplain opens up on the coastal plain, it was less useful for delineating the incised, upper reaches correctly, e.g., the Palmiet was delineated a third shorter than the measured extent. The use of expert judgement in combination with a precautionary approach is advisable in delineating the upper reaches of most systems with limited extent. Estuary delineation in South Africa is still largely based on spatial and habitat features, as detailed information on soil moisture, sediment particle size, redox potential, and total organic matter are not available on a national scale. The latter have been used in international approaches on regionalscale delineations and therefore highlight future research requirements (Adam, 1992;Caeiro et al., 2003;Junk et al., 2013).
Although the index on SDG6, 'Ensure availability and sustainable management of water and sanitation for all', requires only the reporting of lakes (the open water) and vegetated wetlands at a national scale, we propose the inclusion of arid systems which are not permanently or seasonally inundated, nor vegetated, as an additional category. The results showed that 11% of South Africa's inland wetlands are inundated and, of the remaining wetlands, 55% could be vegetated and 34% arid. Both the land cover (GTI, 2015) and LAI (Cho et al., 2017) products, at 30 m and 463 m spatial resolution, are unfortunately at a very coarse spatial resolution and considered inadequate to accurately determine the true extent of inundation and vegetation cover of inland wetlands. Closer inspection of sites showed an underrepresentation of the inundated waterbodies and an over-estimation of the vegetated and arid systems. Finer-scale data or in-field verification is therefore required to verify the extent of inundation and vegetation of wetlands, whereas in-field verification of the limnicity of systems is crucial. Time-series analysis should be incorporated to determine the full hydroperiod and phenology of wetlands. These results are therefore only potentially broadscale indicators of inundation, vegetation, and arid systems for the purpose of reporting at a national scale.
Demarcating the river-estuary ecotones highlights the importance of river reaches just above estuaries and the need to consider their role in maintaining estuary condition. In the past five years, there has been a trend to plan new wastewater discharge points out of estuaries, but in freshwater reaches just above them (i.e. in river-estuary ecotones) to allow for the application of less stringent treatment standards and license agreements applicable to river discharges rather than the more onerous requirements applicable to estuaries as the receiving environments. Benefits of these transition zones include refuge offered to estuarine biota from adverse conditions such as hypersalinity, eutrophication, hypoxia and temperature extremes sometimes experienced in estuarine environments. They are also areas used by estuarine and marine biota to clean themselves of pathogens and parasites that cannot withstand osmotic stress in these transitional freshwater reaches. Fish and invertebrates can also benefit from greater diversity and abundance of grazing and prey that may be unavailable or limited in the adjacent estuarine habitat. Euryhaline species benefit from lower predation levels from stenohaline predators. These river-estuary ecotones also provide extended habitat for euryhaline estuarine and marine species, a function particularly important during low-flow and drought periods when the REI breaks down and/or the estuary may be cut off from the sea. Weirs, impoundments and other instream obstacles in the transition zone may greatly reduce the availability and benefits of this habitat to estuarine-associated species. This includes a reduction in the production and transport of detritus from the ecotone to the estuary downstream. Recruitment of estuarine-associated species will be limited to animals that are able to complete their entire lifecycle in this habitat. Alien and extralimital fish may have a similar impact. Many of these fish are predatory, out-compete their indigenous counterparts and thrive in transitional waters where they can provide an effective barrier to any larval or juvenile fish and invertebrates trying to recruit from downstream. Identifying these ecotones requires that the relevant lead agents, i.e., DWS and DEFF, collaborate more closely on issues within these support areas that can potentially impact on estuaries.

CONCLuSION
A total of 4 596 509 ha (3.8% of South Africa) of inland aquatic ecosystems, including inland wetlands, estuaries and some river channels in the National Wetland Map version 5, and an artificial wetlands data layer, have been mapped as part of the South African Inventory of Inland Aquatic Ecosystems (SAIIAE). The datasets and associated attributes have informed the National Biodiversity Assessment for 2018, as well as the Sustainable Development Goal reporting for Indicator 6 through the Department of Water and Sanitation to the United Nations Environment Programme. Significant effort is required to improve the confidence of the representation of the inland wetlands in the future updates of the National Wetland Map.