An indicator-based approach for cross-realm coastal biodiversity assessments

Ecosystem status assessments are generally separated into realm-specific analyses (terrestrial, freshwater, estuarine or marine), but without integrating these into a coherent assessment of coastal biodiversity across the land–sea interface. Trends in assessment indicators in coastal versus non-coastal areas have also rarely been considered. In this study we aimed to compile the first cross-realm national biodiversity assessment for the South African coast using three key indicators. The ecological condition, ecosystem threat status, and ecosystem protection level of coastal ecosystem types (n = 186) were determined and compared with those of non-coastal ecosystem types (n = 444). Nearly half (46.9%) of the South African coastal habitat has been degraded compared with 20% of non-coastal areas. Proportionately, there are three-times (60%) as many threatened coastal ecosystem types (or 55% by area) as there are threatened non-coastal ecosystem types (19%, 6% by area). Despite the impacted state of coastal biodiversity, protection levels are generally higher in the coastal zone (87% of ecosystem types have some protection) compared with non-coastal areas (75%), although fewer coastal ecosystem types have met their biodiversity targets (24%, vs 28% for non-coastal ecosystem types). These results illustrate the importance of using a cross-realm approach for status assessments, management and conservation of coastal biodiversity. The assessment methods described are flexible and widely applicable to other regions.

) and 2018 (Skowno et al. 2019a). The analyses are undertaken per realm: terrestrial, inland aquatic (rivers and wetlands), estuarine and marine. There are two headline indicators reported for the ecosystem types: ecosystem threat status and ecosystem protection level. The former reports the proportions of ecosystem types that are Critically Endangered, Endangered, Vulnerable or of Least Concern; the latter indicates the proportions of each ecosystem type that are Well Protected, Moderately Protected, Poorly Protected or Not Protected. Although not a headline indicator, ecological condition is also assessed, which broadly ranges from natural to permanently lost, and which underpins the assessment of ecosystem threat status, and, in some cases, informs the assessment of ecosystem protection level.
The NBA is based on the best available information and is iteratively improved over time. Assessing the status of biodiversity in the coastal zone is a case in point. Although there were regular cross-realm meetings in the first two assessments, and the headline indicators were the same in all cases, teams assessing the state of biodiversity in each realm largely worked in isolation. This is exemplified by the fact that the maps representing the various realms in 2004 and 2011 did not align at the land-estuary-sea interface (Harris et al. 2019a). This spatial misalignment precluded a cross-realm assessment for the coast and presented significant challenges for a range of other coastal spatial work (e.g. systematic conservation planning). While preparing for the third assessment (i.e. NBA 2018), the need for a more integrated approach for the coastal zone became clear. Not only would this resolve many challenges for coastal biodiversity assessment, prioritisation and management in South Africa, but it would also contribute a methodology that could be applied in other countries to support more-informed coastal management and conservation.
Our aim was to compile a status assessment of South Africa's coastal biodiversity. Specifically, we: (i) define and delineate an ecologically determined coastal zone that is seamless across realms; (ii) determine the ecological condition, ecosystem threat status, and ecosystem protection level of coastal ecosystem types in South Africa; and (iii) compare these indicators between coastal and non-coastal areas. We also identify challenges that arose during this first coastal assessment and highlight key areas of improvement for future analyses that can also guide others undertaking a similar assessment for the first time. Given that coasts are generally areas of intensive use, we predict that the ecological condition and ecosystem threat status will be proportionately worse on the coast compared with non-coastal areas. However, South Africa has many land-based and marine protected areas along the coast, and therefore we predict that most coastal ecosystem types will meet or partly meet their biodiversity targets.

Study area
The study area is the South African mainland (landmass to the seaward edge of the 200 nautical mile exclusive economic zone) but particularly the coast (Figure 1). Given that the purpose was a biodiversity assessment, we used an ecologically determined coastal zone rather than an administrative boundary. This required creating a conceptual framework and generating a seamless, fine-scale map of ecosystem types at the land-estuarysea interface, from which an ecologically determined coastal zone could be identified. Details of this process and the 186 coastal ecosystem types are given in Harris et al. (2019a). Briefly, terrestrial and marine ecosystem types that respectively had an oceanic and land-based influence were considered coastal, and all estuaries were included. The inland aquatic realm was excluded because it was not considered part of the ecologically determined coastal zone based on the available evidence at the time (Harris et al. 2019a). The basic ecological unit in the status assessment is an ecosystem type.

Coastal biodiversity assessment framework
There were five main steps in the process of assessing the status of coastal biodiversity in South Africa ( Figure 2). Once the project management was set up, we decided on the indicators to be included in the assessment: ecological condition, ecosystem threat status, and ecosystem protection level. There are strong differences in the nature of the three realms, the types of data that are available, and the typical realm-specific approach to analyses. Therefore, methods for assessing the indicators had to be similar enough to be comparable for an integrated assessment of coastal biodiversity, but sufficiently flexible to account for these differences. Therefore, realm teams all: (i) worked towards the same categories within each indicator, even if there were slight variations in the finer details (e.g. in ecological condition); (ii) agreed on a set of principles or an approach (e.g. to use the IUCN Red List of Ecosystems [RLE] criteria for assessing ecosystem threat status) but then used realm-specific methods to reach the same end point (e.g. assessed different RLE criteria depending on data availability); and (iii) shared data (e.g. all realms used a single map of protected areas for the protection-level assessments). In this way, there was standardisation in what each of the categories of each indicator represents, but flexibility in how the realm teams performed the calculations to arrive at equivalent answers (see the methods below for details of the analyses) ( Figure 2). Finally, the realm-specific results were compiled specifically for the coastal zone.

Ecological condition
Ecological condition was estimated using different, realm-relevant approaches, with the results aligned to the same set of categories of ecological condition (Figure 3). In the terrestrial assessment, land-cover change was used to measure habitat loss. The dataset was derived from the 1990 and 2013/14 national land-cover products, which were reclassified and combined into a simple land-cover-change map and refined further (see Skowno et al. [2019bSkowno et al. [ , 2021 for details). Ideally, terrestrial ecological condition should also include disruption of biotic processes, and environmental degradation, especially where invasive species and/or overutilisation or overgrazing are prevalent, but there were no nationwide data for these components, except for the Albany Thicket biome and Little Karoo region (see Skowno et al. [2021] for details). Ultimately, these land-cover data allow a binary classification of terrestrial ecological condition: natural and not natural ( Figure 3). Ecological condition in the estuarine and marine realms is assessed on a relative, continuous scale because of the cumulative nature of pressures in these realms that, in turn, cause levels of degradation rather than (necessarily) outright loss. In the estuarine assessment, a multi-criterion estuarine health index (EHI) was used (DWAF 1999;Turpie et al. 2012;van Niekerk et al. 2013). The EHI comprises both abiotic (hydrology, hydrodynamics and mouth condition, water chemistry, sediment processes) and biotic components (microalgae, macrophytes, invertebrates, fish, birds). A multidisciplinary group of estuarine scientists assessed the EHI based on measured data and their collective understanding of likely impacts affecting each estuary. The EHI scores indicate how much of the natural ecological patterns and processes have been lost as a percentage of the pristine condition. These scores were then classified into six categories of ecological condition, with the bottom two categories grouped together (van Niekerk et al. 2013(van Niekerk et al. , 2019 (Figure 3).
Ecological condition in the marine realm ) was based on an assessment of cumulative pressure (e.g. Halpern et al. 2007;Teck et al. 2010). The assessment followed a simplified version of the method by Teck et al. (2010), whereby a matrix of pressure-impact weightings was calculated based on scores of functional impact and recovery time associated with each pressure on each ecosystem group. Scores were based on expert knowledge, supported by evidence in the scientific literature when possible (reviewed in Chapter 4 of Sink et al. [2019]). The scores were also moderated against the concept of complete loss owing to severe and very severe impacts to ensure a robust link to the degradation categories of the IUCN RLE criterion C3 ( Marine Estuaries Set up project management: appoint a project manager, define realms, identify a lead scientist per realm, determine deadlines and budgets Identify realm teams and reference committees Agree on indicators to be assessed, and principles or an approach for assessing an indicator. Standardise indicator categories (including terminology) and what each category means (e.g. distinction between a moderately modified and severely modified area) Align boundaries of ecosystem types at the land-estuary-sea interface. Create a cross-realm coastal zone within which biodiversity will be assessed. Share pressure data (where relevant). Compile standardised datasets (e.g. one dataset of protected areas) Generate maps of ecosystem types. Identify which ecosystem types are coastal or non-coastal. Map pressures. Use standardised datasets for analyses where they exist Cross-check methods among realms, co-learning and aligning where possible. Also cross-check outputs to ensure that the indicator categories all mean the same thing across realms Perform assessment calculations, refining analyses in a learning-by-doing approach. Adjust and refine analysis thresholds as needed so that the results reflect indicator categories with the same meaning across realms Geographic integration of realm outputs to the coastal zone Identify a realm-specific approach to assess biodiversity under each indicator based on available data and suitable methods Figure 2: Conceptual framework of the coastal biodiversity assessment, showing the close links between the coastal integration and realm-specific work in all five stages of the assessment process. This allowed standardisation of the process and outputs, but with a sufficient level of flexibility to accommodate realm-specific requirements values of cumulative impact into four categories to align with the IUCN RLE thresholds (Figure 3) based on natural breaks in the distribution of the cumulative-impact values, supplemented by expert judgement .
Consideration was given to the influence of pressures and ecological condition in adjacent realms, where possible and relevant. For example, coastal development on the dunes (seashore vegetation types; terrestrial assessment) was considered a pressure to adjacent sandy beaches (marine assessment) because of the strong linkages between beaches and dunes. Ecological condition classes were aligned as much as possible to allow for cross-realm comparisons ( Figure 3). Note, though, that realm teams treated the heavily, severely, and critically modified classes differently; see Figure 3 for nuances. A map of coastal ecological condition was compiled by extracting the data for the coastal terrestrial, estuarine, and coastal marine ecosystem types from each realm. The total proportion of natural/near-natural, moderately modified, heavily modified and critically modified habitat was calculated for the coast, compared between coastal and non-coastal areas, and compared among realms.

Ecosystem threat status
Ecosystem threat status in all realms was calculated using the IUCN RLE framework (Keith et al. 2013;Bland et al. 2017). There was a 'core' assessment in each realm that evaluated ecosystem types against the main criteria that were most appropriate for that realm (Table 1; Supplementary Table S2). To complement this, additional data were compiled and evaluated under several 'supplementary' assessments to ensure that the ecosystem risk assessments were based on the best available data (Table 1; Supplementary Table S2) .) A map of coastal ecosystem threat status was compiled by extracting the data for the coastal terrestrial, estuarine, and coastal marine ecosystem types from each realm assessment. The total proportions of Critically Endangered, Endangered, Vulnerable and Least Concern ecosystem types were calculated for the coast, compared between coastal and non-coastal areas, and compared among realms.

Ecosystem protection level
Ecosystem protection level indicates the extent to which ecosystems are adequately protected or under-protected relative to their biodiversity target (Reyers et al. 2007). These biodiversity targets represent the minimum proportion of each ecosystem type that needs to be kept natural to near-natural to maintain a viable, representative sample of that biodiversity in the long term (Desmet and Cowling 2004;Reyers et al. 2007  List of Ecosystems (RLE) approach, which uses percentage degradation of ecosystems or percentage disruption of biotic processes, and to South Africa's original Department of Water and Sanitation (DWS) framework (Present Ecological State categories using the letters A to F), which is applied to rivers, inland wetlands and estuaries. For the coast, there are four categories of ecological condition: (i) natural/near-natural; (ii) moderately modified; (iii) heavily modified, which includes the severely modified habitat in the terrestrial and marine realms; and (iv) critically modified, which includes severely modified habitat in the estuarine realm relationships and varies between 16% and 36% of the original extent of each ecosystem type (Desmet and Cowling 2004). For estuarine and marine ecosystem types, a 20% target was used. Historically, the 20% target was established at the World Parks Congress in 2003 and has been used in the NBA since then. Its use in these analyses was also informed by an international literature review (Porter et al. 2011) and a dedicated NBA workshop held in 2016. Ecosystem types meeting their biodiversity targets are considered Well Protected (100% of the target met). Ecosystem types considered Moderately Protected (>50% of the target met), Poorly Protected (5-50% of the target met) or Not Protected (<5% of the target met) are collectively referred to as under-protected ecosystem types (Figure 4).
The South African Protected Areas Database (DFFE 2021) was updated to include new protected areas, and the data were cleaned and prepared for the assessment (Figure 1e). The proportion of each ecosystem type in protected areas was calculated ( Figure 4). Importantly, only areas in a natural/near-natural state contributed towards meeting biodiversity targets. Therefore, in the terrestrial assessment, non-natural areas within protected areas do ≤2 000 km 2 ≤20 000 km 2 ≤50 000 km 2

Sub-criterion B2 (i) -The number of 10×10 km grid cells occupied (AOO) and an observed or inferred continuing decline in spatial extent.
For the NBA 2018, the absolute rate of habitat loss was used to identify ecosystems with ongoing declines (>0.4% year −1 ). Supplementary assessments used expert input and the threatened species database to identify restricted-distribution ecosystems with very high levels of impacts from overgrazing and invasive species and poor fire management. Assessments: Terrestrial Core; Terrestrial Supplementary; Marine Supplementary (see Sink et al. [2019] for the variation from the IUCN approach) ≤2 ≤20 ≤50

Sub-criterion B3
Assessments: Estuarine Supplementary Criterion C: Environmental degradation Sub-criterion C3 Assessments: Estuarine Core; Marine Core Criterion D: Disruption of biotic processes or interactions Sub-criterion D3 -Disruption of biotic processes, since 1750, based on change in a biotic variable affecting a fraction of the extent of the ecosystem and with relative severity, as indicated by the table on the right. For the NBA 2018, ecosystem degradation data from the Albany Thicket biome and Little Karoo region were used. The severely degraded class in these datasets was considered to be ≥90% severity, the extent of severe degradation was expressed as a percentage of the remaining habitat circa 2014.

High-risk ecosystem types
The results from the ecosystem threat status and ecosystem protection level analyses were cross-tabulated to determine which ecosystem types are both threatened and under-protected. In particular, high-risk ecosystem types were identified; these are Critically Endangered or Endangered ecosystem types that are Not Protected.

Ecological condition
There is more than twice the proportion of modified habitat in the coastal zone compared with that in the rest of the country, with almost half (47%) of the coastal extent in a modified state compared with only a fifth (20%) of non-coastal areas (Table 2; Figure 5). This confirms the disproportionate amount of pressure on the coastal zone when compared with non-coastal terrestrial (22% modified) and non-coastal marine areas (18% modified).
Although the proportion of natural/near-natural habitat is similar in non-coastal terrestrial (78%) and marine (82%) areas, 13% of the non-coastal marine extent is moderately modified, compared with 21% of non-coastal terrestrial habitat that is critically modified. Among realms within the coastal zone, estuaries have the largest extent of modified habitat (77%), but the least amount of critically modified habitat (5%). There is more natural/near-natural coastal terrestrial habitat (60%) compared with that for coastal marine habitat (49%), but also more critically modified habitat on the landward side of the coast (38%). This is because terrestrial pressures tend to be binary, usually resulting in intense modification of natural habitat as compared with marine (and estuarine) pressures that overlap spatially and accumulate over time (recall, there is also no assessment of the moderately modified category in the terrestrial assessment) (Figure 3). Broadly speaking, the places with the most heavily to critically modified ecological condition are associated with coastal areas with the highest human population density ( Figure 5). These include the areas around Cape Town, Mossel Bay, Cape St Francis, Gqeberha (Port Elizabeth), Durban and Richards Bay.
Notably, these areas are all associated with ports and harbours, and their related burgeoning development in the surrounding area.

Ecosystem threat status
Given the concentration and overlap of pressures at the land-sea interface, and thus a more-degraded ecological condition of the constituent ecosystem types, the South African coast is proportionately more threatened than the rest of the country (entire non-coastal territory combined). In fact, the coast (60% threatened ecosystem types) has proportionately three-times more threatened    Figure 3 for nuances in ecological condition classes among realms ecosystem types than the rest of the country (19%), which is proportionately nine-times more by extent (coastal: 55%, non-coastal: 6%) (Table 3; Figure 6). Of the 186 ecosystem types in the coastal zone, 112 are threatened, comprising 21 Critically Endangered, 37 Endangered, and 54 Vulnerable ecosystem types (Table 3; Figure 6). Most of the threatened ecosystems in the coastal zone are in the aquatic realms, with estuaries proportionately having the most threatened ecosystem types: 86% of types and 99% by area are threatened. The marine realm follows with the second-highest proportion of threatened ecosystems in the coastal zone: 67% of types and 69% by area are threatened.
Within that, 83% of the shore extent is threatened. Although the terrestrial coastal ecosystem types are less threatened (43% of types, 26% of extent threatened), most of the Critically Endangered ecosystem types are within this zone, especially among the semi-coastal vegetation types (26% of types, 16% of extent). This means that pressures on land act more locally and intensively than they do in the aquatic realms. Coastal development and mining, for example, have a distinct footprint, causing complete habitat loss within that footprint. In contrast, pressures in the estuarine and marine realms tend to have wide-reaching, chronic impacts that accumulate over time and that escalate as the intensity increases (e.g. pollution and fishing). The result is many more threatened aquatic ecosystem types over a much broader area in the Endangered and Vulnerable categories, compared with fewer threatened ecosystem types in the highest risk category (Critically Endangered) on land.

Ecosystem protection level
Although 87% of coastal ecosystem types have some level of protection (162 of 186 types), only a quarter are Well Protected (24%), which accounts for 9% of the coastal extent (Table 4; Figure 7). In terms of the extent of protection, each realm has its own strength: coastal terrestrial ecosystem types are the most Well Protected (18% of extent); estuarine ecosystem types have the highest extent of protection across the Well, Moderately, and Poorly Protected categories (89% of extent); and coastal marine ecosystem types have the largest extent of protection in the top-two categories, namely Well Protected and Moderately Protected (61% of extent). However, it is important to note that the bulk of estuarine ecosystem types are Poorly Protected, and estuaries are the least Well Protected ecosystem types (by both the proportional types and extent) in the coastal zone. There is good progress towards protecting coastal ecosystem types, but protected area networks need to be strengthened to afford protection to about half of the coastal zone that is currently Poorly Protected or Not Protected (39% of types, 49% of extent), especially for coastal terrestrial and estuarine ecosystem types.

High-risk ecosystem types
There are 102 threatened coastal ecosystem types that are under-protected (Moderately Protected, Poorly Protected, and Not Protected) in South Africa (Table 5). This means that all but 10 of the 112 threatened coastal ecosystem types (i.e. 91%) are currently under-protected. Of these, 13 are at the greatest risk, being Critically Endangered or Endangered and Not Protected: seven terrestrial, three estuarine, and three marine ecosystem types (Table 5). There are two hotspots of these high-risk ecosystem types (Figure 8): one around the Orange River mouth (five high-risk types), and one around Durban (three high-risk types).

The status of South Africa's coastal biodiversity
For the first time, we were able to quantitatively compare the status of biodiversity in coastal and non-coastal areas because of the improved alignment among realm assessment methods, and the seamless map of ecosystem types (Harris et al. 2019a). The South African coast is indeed a hotspot of cumulative pressure, which drives poor ecological condition. Therefore, as predicted, the coast is disproportionately impacted and more threatened than non-coastal land and sea areas. This calls for careful, improved management of key pressures-particularly living resource use, freshwater resource use, coastal development, mining, and pollution-and consideration of cumulative and cross-realm impacts. For example, reduced freshwater flow in estuaries affects the estuaries themselves, as well as many other coastal ecosystem types (e.g. river-influenced marine ecosystem types that rely on supplies of terrigenous muds, and beaches and dunes that rely on terrigenous sand). Furthermore, ports and harbours were key drivers of cumulative degradation in the surrounding area. As South Africa looks to strengthen  Table 3: Number of ecosystem types per threat category, given per across-shore zone (inland to offshore). Percentage of ecosystem types per threat category is given in round brackets, and the percentage extent is given in square brackets (extent of terrestrial ecosystem types is given as the percentage of remaining natural habitat) its ocean economy by including new ports and harbours through Operation Phakisa (RSA 2014; Findlay 2018), decisions regarding their placement must be made with the utmost care, particularly given that almost half of the South African coast is already in a modified state. A positive finding was that the coast is proportionately more protected than the rest of the country and, as predicted, most ecosystem types at least partly meet their biodiversity targets. In fact, nearly two-thirds (61%) of the ecosystem types have met at least half of their target. This is especially important given the rich diversity the coast supports (Griffiths et al. 2010;Harris et al. 2014), the myriad of benefits it provides (Harris et al. 2019b) and its threatened status (this study). Furthermore, there are notable opportunities to improve the status of coastal biodiversity in South Africa (see Supplementary Box S1 for priority actions, and Harris    Table 4: Number of ecosystem types per protection level category, given per across-shore zone (inland to offshore). Percentage of ecosystem types per protection-level category is given in round brackets, and the percentage extent is given in square brackets (note the extent of terrestrial ecosystem types is given as the percentage of remaining natural habitat) et al. [2019b] for details). These include: (i) to protect the 13 high-risk ecosystem types, especially in the two identified hotspots; (ii) to improve the ecological condition of degraded estuarine and marine habitat within existing protected areas; (iii) to look for synergies in land-sea protection and improve protection of estuaries; and (iv) to secure coastal biodiversity and benefits using the new National Coastal and Marine Spatial Biodiversity Plan (Harris et al. 2022a(Harris et al. , 2022b) through its intended implementation in the emerging national Marine Spatial Planning process.

Global application of the assessment methods
The methodology applied in this biodiversity assessment can be easily implemented in other countries. By undertaking such assessments, countries can contribute to global initiatives (e.g. IUCN RLE) and their own international reporting, and the results can add value to countries' National Biodiversity Strategy and Action Plans (SANBI and UNEP-WCMC 2016). Moreover, the outputs can be used to guide national priorities for protected area expansion by identifying highly threatened ecosystem types, under-protected ecosystem types, and perhaps most importantly the high-risk ecosystem types. Although some countries may perceive a paucity of data as a limitation to undertaking these assessments, there are many options available to compile national maps from global datasets and surrogates, even if this means starting with relatively simplified foundational data and analyses (e.g. SANBI and UNEP-WCMC 2016; Kirkman et al. 2019). Some ecosystem types can be mapped from Google Earth or similar available imagery (Harris et al. 2011(Harris et al. , 2019avan Niekerk et al. 2013) or using remote sensing (e.g. Dunga 2020). The IUCN Global Ecosystem Typology (GET) (Keith et al. 2020; https://global-ecosystems.org) is recommended as a starting point for a national ecosystem classification system. However, to assess the indicators correctly, the original ecosystem types need to be mapped in areas that have been permanently modified so that the loss of natural habitat can be reported. For example, the original shore type (from the IUCN GET MT1 and MT2 biomes) needs to be mapped rather than the IUCN GET category, MT3.1 Artificial Shorelines. Development and artificial structures should instead be mapped as coastal development as part of the pressures included in the assessment of ecological condition.
Even in our relatively mature assessments (third NBA), we still acknowledge data gaps and shortcomings. However, we encourage countries not to allow data limitations to preclude undertaking an assessment. It is preferable to work with the best available data and iteratively improve the results over time than to wait for comprehensive data from the outset.

Future work
Nineteen knowledge gaps and corresponding research priorities to improve future coastal biodiversity assessments in South Africa were identified and grouped into four themes: foundational data and knowledge (n = 3); pressures, pressure impacts and monitoring (n = 8); assessment (n = 6); and benefits and messaging (n = 2). Full details are given in Supplementary Table S3. Those of global relevance or that require international research and discussion are highlighted here.
Although the current approach to assessing the status of coastal biodiversity is a substantial advance, more-explicit attention is required for a robust cross-realm integration and for cross-checks of the analysis results among realms. For example, if the dunes are in poor ecological condition (heavily to critically modified), the adjacent beaches cannot be in good ecological condition (natural/ near-natural) because the two are a single geomorphic unit. In cases where the separate realm assessments produced inconsistent results, such as in this hypothetical example, additional filters would need to be applied in the condition assessment. We did try to account for cross-realm effects where possible, but more-explicit attention needs to be paid to this aspect in future assessments. Furthermore, although excellent for the rest of the marine realm, the resolution of 30-m pixels for the assessment of shores is not high enough for this narrow interface. Perhaps one way to best improve cross-realm consistency in results-at least between the terrestrial and marine realms-is to assess the seashore (comprising the foredunes: terrestrial [IUCN GET MT2 biome], and shores: marine [IUCN GET MT1 biome]) separately from the rest of the land and sea at a resolution that is appropriate for this narrow ecotone.
The approach of performing core and supplementary assessments for the first use of the IUCN RLE worked well, and it will allow for reassessments of selected ecosystem types as new and improved data are found or collected. However, because South Africa is an ecologically megadiverse country, we needed to adjust some of the criteria thresholds to yield realistic results. This also highlighted that improving the application of the IUCN RLE criteria is an important area of research, which could be advanced by testing the scale and thresholds of the assessment, and creating models of ecosystem collapse, including refining thresholds for classifying ecological condition from habitat degradation. Developing conceptual models of ecosystemtype functioning, the effects of pressures, and the pathways of degradation and collapse will assist. In terms of scale, systems such as mangroves (IUCN GET: MFT1.2 Intertidal forests and shrub lands) are being red listed elsewhere (Sievers et al. 2020  . Similarly, the southern Benguela was red listed (Bland et al. 2018), whereas in the NBA of 2018 this same area was found to be closer in extent to an ecoregion comprising many ecosystem types . We believe that it is important to engage more deeply across ecological realms and in the global context on ecosystem classification philosophies and scales, red-listing approaches, criteria, thresholds, results and implications. We note that ecosystem protection level is purely a representation-based indicator and does not consider management effectiveness within protected areas. Efforts are underway to develop a parallel indicator of management effectiveness (e.g. Kirkman et al. 2021), but this approach is also problematic in that effective management can be difficult to define and can be different for various components of biodiversity. Reserve effectiveness is partly accounted for in the estuarine assessment by scoring fishing pressure within estuaries, precluding those systems with high to very high fishing pressure from contributing to the Well Protected category. However, it is similarly noted that the effectiveness of estuarine protection can be truly judged only   by additionally considering environmental flow allocations, fisheries control measures, and land-use and infrastructuredevelopment measures (van Niekerk et al. 2019). Similarly, in the marine environment, protection of marine biodiversity within protected areas depends strongly on management effectiveness, which in turn depends on adequate resources for effective monitoring, enforcement and compliance . We encourage discussion and research on indicators of management effectiveness to complement the ecosystem protection level indicator. Another important metric that we are working towards quantifying in the future is the rate of change. The foundational data and methods are still being refined in the marine and estuarine assessments, such that the results from each NBA cannot always be compared directly, precluding analyses of trends at this time. However, tracking rates of change is more advanced in the terrestrial assessment (Skowno et al. 2021). Comparing land-cover maps through time shows that habitat loss in coastal and semi-coastal vegetation types is between two-times (1990-2014) and three and a half-times (2014)(2015)(2016)(2017)(2018) faster than in non-coastal vegetation types, and is still accelerating (Harris et al. 2019a;Skowno et al. 2021). When combined with the other indicators, rate of change can help to prioritise ecosystem types that most urgently need management actions, such as highly threatened, under-protected ecosystem types with high rates of habitat loss.

Conclusions
Coasts globally are intensively used spaces that require more deliberate attention to cross-realm and cumulative impacts in management decision-making. Certainly, an important starting point is to perform a robust assessment of the status of coastal biodiversity, which can be tracked over time. The method used in this study-comprising the indicators of ecological condition, and especially ecosystem threat status and ecosystem protection level-can achieve that. The standard indicators are applicable and sufficiently flexible to be quantified in all realms; although improved integration of data and results is still required for the landestuary-sea ecotone, the standardised approach allowed implementation of the methods used separately by each realm team, so that the findings could later be aggregated for the ecologically determined coastal zone. The assessment also brought to light key areas of research for the global scientific community, notably around application of the IUCN GET and RLE. We encourage discussion on, and further testing and refinement of, coastal biodiversity assessments in support of securing coastal biodiversity and ecosystem services, especially in the face of accelerating global change that is so strongly felt in this intensively used zone.