Progress towards a representative network of Southern Ocean protected areas.

Global threats to ocean biodiversity have generated a worldwide movement to take actions to improve conservation and management. Several international initiatives have recommended the adoption of marine protected areas (MPAs) in national and international waters. National governments and the Commission for the Conservation of Antarctic Marine Living Resources have successfully adopted multiple MPAs in the Southern Ocean despite the challenging nature of establishing MPAs in international waters. But are these MPAs representative of Southern Ocean biodiversity? Here we answer this question for both existing and proposed Antarctic MPAs, using benthic and pelagic regionalizations as a proxy for biodiversity. Currently about 11.98% of the Southern Ocean is protected in MPAs, with 4.61% being encompassed by no-take areas. While this is a relatively large proportion of protection when compared to other international waters, current Antarctic MPAs are not representative of the full range of benthic and pelagic ecoregions. Implementing additional protected areas, including those currently under negotiation, would encompass almost 22% of the Southern Ocean. It would also substantially improve representation with 17 benthic and pelagic ecoregions (out of 23 and 19, respectively) achieving at least 10% representation.


Introduction
Global threats to ocean biodiversity have generated worldwide momentum to improve its conservation and management. Marine protected areas (MPAs), areas of ocean where human activities are limited or prohibited, have been increasingly promoted by policy-makers, scientists and conservationists as a tool for mitigating ocean threats, conserving biodiversity, and managing fisheries [1][2][3]. Numerous studies demonstrate that MPAs, especially no-take MPAs (also known as marine reserves), lead to increases in biomass, density, and diversity of life in the MPA [4][5][6]. Notably, these MPA benefits can extend to fisheries. MPAs have been shown to facilitate the recovery of depleted fisheries, provide spillover effects, and lead to larger fish [7][8][9]. Furthermore, because they maintain all trophic levels of the ecosystem and increase both species and genetic diversity, MPAs can enhance resilience to environmental impacts, including those related to climate change [10][11][12]. Several international targets have recommended the adoption of representative networks of MPAs in national waters and in areas beyond national jurisdiction. At the 2002 World Summit on Sustainable Development, participating States agreed to designate a representative global network of MPAs by 2012 [13]. This call was further reiterated at the 2003 International Union for the Conservation of Nature (IUCN) World Parks Congress, which called for protected areas encompassing 20-30% of all marine habitats also by 2012 [14]. The 2010 Aichi Biodiversity Targets, adopted by the Convention on Biological Diversity as part of its Strategic Plan for Biodiversity 2011-2020, offered a new deadline of 2020 to designate 10% of the global oceans in ecologically representative MPAs [15]. Then, in 2014 the IUCN World Parks Congress recommended that 30% of the ocean be protected in an ecologically representative network [16]. Finally, in 2015 the United Nations adopted the Sustainable Development Goals, including goal 14 which aims to conserve 10% of coastal and marine areas by 2020 [17]. Evidence-based conservation science research often suggests protection targets of at least 30% and often higher are required to effectively conserve biodiversity and ecosystems [18,19].
A major criteria for conservation, including Aichi target 11, is that protected areas be ecologically representative since their efficacy is substantially enhanced when they are representative of the biodiversity of a region [20]. A widely used method for examining representivity is by determining coverage of ecoregions by the protected area network [21,22]. Ecoregions are spatial regions, typically within a large spatial domain, defined in such a way that each ecoregion defines a characteristic set of species communities and habitats that are distinct from those of other ecoregions within the domain [23,24]. Direct sampling of biodiversity at large spatial scales is generally impractical, necessitating the use of proxies or modelling approaches to achieve broad spatial coverage. Species distribution and related modelling methods can be used to infer broad-scale biodiversity patterns based on spatially-limited sampling (e.g. [25,26]). However, such approaches present difficulties for our purposes of assessing Southern Ocean MPA representativeness. Predictions of species distributions would need to be available at circum-Antarctic scale, and from a sufficiently diverse suite of species in order to be suitably representative of broader Southern Ocean biodiversity To date, Southern Ocean applications of such models have tended to be regional in scope (but see e.g. [25,[27][28][29][30] for circum-Antarctic applications), and focused on a relatively restricted number of species. Here we therefore use heterogeneity of habitats and geomorphic features as proxies for biodiversity. This approach is well established in the terrestrial and marine realms (see e.g., [31][32][33][34][35]).
In line with global MPA goals, roughly 18.45% of national waters, globally, have been protected to date. Meeting these targets in areas beyond national jurisdiction has proven a more difficult challenge, with only 1.18% of the high seas protected thus far [36]. Further, MPAs have generally been found to not be ecologically representative, especially within waters under national jurisdiction [37,38].
National governments and the Commission for the Conservation of Antarctic Marine Living Resources (CCAMLR) have successfully adopted multiple MPAs in the Southern Ocean despite the challenging nature of establishing MPAs in international waters. The Southern Ocean encompasses roughly 10% of the global oceans (Fig 1), most of which is considered high seas. This area is primarily governed by a multi-lateral Convention on the Conservation of Antarctic Marine Living Resources (CAMLR Convention). This Convention is carried forward by CCAMLR, a Commission of 25-Member States plus the European Union. Within CCAMLR's waters are five sets of sub-Antarctic islands that fall under national jurisdictions (Fig 1), which are managed in accordance with Convention rules [39]. CCAMLR has the explicit objective to conserve marine living resources and employs a science-based precautionary and ecosystem-based management approach [39]. In doing so CCAMLR is arguably the world's most successful international management body for marine living resources [40-43].  . Additionally, CCAMLR has been negotiating a large MPA network in the East Antarctic, the Weddell Sea, and the Antarctic Peninsula (Fig 1). Adjacent to the CCAMLR Area other sub-Antarctic MPAs exist (e.g., Macquarie Island Marine Park,162,000 km 2 declared in 1999) however these are outside of CCAMLR jurisdiction.
Here we assess progress towards establishing a representative network of MPAs, including its level of protection, in the Southern Ocean. We examine CCAMLR and nationally governed protected areas that have been adopted as well as those currently under negotiation. We map the location of existing and proposed MPAs and calculated no-take areas. We then assess whether these proposed and existing MPAs are representative of Southern Ocean biodiversity and ecosystems using existing benthic and pelagic regionalizations as a proxy for biodiversity [45,46].

Materials and methods
Here, we use the CAMLR Convention Area as our study region (Fig 1). This region is circumpolar, with its northern boundary between 60 and 45˚S aligning approximately with the Polar Front and its southern boundary aligning with the coast of Antarctica and ice shelf boundaries. Due to the lack of data under ice shelves, our study area does not include areas under ice shelves (thus it omits the area under the Ross Ice Shelf, which is technically part of the MPA) nor does it include some of the sub-Antarctic region situated above the Polar Front which falls outside the bounds of the CCAMLR Convention Area.

Pelagic regionalization
Pelagic cluster data representing 20 different categories [46] was downloaded from [70]. Following the methods of two previous pelagic Southern Ocean regionalizations [24,71], the pelagic regionalization data we used was based on a non-hierarchical clustering algorithm to reduce the number of grid cells, followed by further refinement using a hierarchical clustering algorithm [46]. During the latter, clusters comprised of only one datum were merged into parent clusters (which occurred in five instances in cluster groups 2, 3, 8 and 13). The regionalization used summer climatological sea surface temperature [63], depth [72] and the proportion of time covered by sea ice as input variables [64]. These data were originally calculated south of 40˚S, but were bound to the CCAMLR Convention area for this analysis (Table 3). For further details on methods underpinning the generation of the pelagic regionalization data used in this analysis see [46].   [73]. Sub-Antarctic MPAs with boundaries extending outside the CCAMLR Area (Kerguelen, Crozet and Prince Edward Islands) were constrained to the CCAMLR Area. We then calculated the total area encompassed by each existing and proposed MPA in ArcGIS. The total proportion of protected area is given by:

Total proportion of protected area
where A pi = area of each MPA, indexed by i, located within the CCAMLR Convention area; n is the number of MPAs; and A is the total CCAMLR Area. This metric was calculated for each of the nine CCAMLR MPA planning domains as well as for the entire CCAMLR Area. To report the proportion as a percentage, we multiplied the total proportion by 100. This metric was also calculated for the no-take areas in the CCAMLR Area. To calculate the total proportion of no-take area, A pi = the no-take area of each MPA, indexed by i, located within the target CCAMLR Convention area. As with the total MPA area, we calculated the no-take metric for each CCAMLR MPA planning domain as well as the entire CCAMLR Convention area.

Fraction of ecoregion protected
We calculated the area and proportion of each benthic and pelagic ecoregion that falls within the boundaries of existing and proposed MPAs, including ecoregions encompassed by no-take zones. Benthic ecoregion and pelagic cluster files (see above) were downloaded and projected into the ESRI:102020 projection, South Pole Lambert Azimuthal Equal Area [73]. The pelagic regionalization was originally projected out to 40˚S, thus we constrained the data to the CCAMLR Area. Pelagic cluster 18 fell outside the area of analysis as it only occurs north of the CCAMLR Area. The benthic and pelagic ecoregion data files were intersected with the MPA shapefiles. We then calculated ecoregion areas included in each existing and proposed MPA,

PLOS ONE
Towards a representative network of Southern Ocean protected areas including areas encompassed by no-take zones. Mean fraction of each ecoregion protected = Where m j is the number of MPAs in ecoregion or pelagic cluster j and A pij is the area of each MPA, i, overlapping areas of ecoregion or pelagic cluster j. A j is the total area of ecoregion or pelagic cluster y. We calculated this for both existing and existing + proposed MPAs. This metric was also calculated for the no-take areas in each existing MPA. Note: n = 1-23 for the

PLOS ONE
Towards a representative network of Southern Ocean protected areas benthic analysis and 1-20 for the pelagic analysis (representing 23 benthic ecoregions and 20 pelagic clusters). We also calculated the number of benthic ecoregions and pelagic clusters that have at least 10% of their total area protected (per Aichi Target 11 [15]) and at least 30% of their total area protected (per IUCN guidelines [16]). This = the number of times that ð P m j 1 A pij =A j Þ is � 0.1 and � 0.3, respectively. We calculated this percentage for existing MPAs, existing + proposed MPAs, and no-take zones within existing MPAs.

Protection equality
Finally, we calculated the protection equality of the existing and proposed MPA system using parallel methods to [22]. These metrics were developed by [78] and are measures of how equitably the different benthic ecoregions and pelagic clusters are represented in the MPA system (i.e., a Gini coefficient). We used the "ProtectEqual" package in R (version 3.5.1) [79], developed by [80], to calculate protection equality values based on the proportion of each ecoregion and pelagic cluster protected. Protection equality values can range from 0-1 with high numbers indicating a higher protection equality.

Total proportion of protected area
Seven MPAs currently exist in the Southern Ocean resulting in 11.98% of the CCAMLR Area falling under general protection and 4.61% falling under strict no-take protection ( Table 1; Fig  1). Of the 11.98% area protected, nationally managed MPAs account for more than half of this (7.21%) and CCAMLR-governed MPAs account for the latter (4.6%). MPAs implemented in the CCAMLR Area which fall under national jurisdiction are: the Heard Island and McDonald Islands (HIMI) marine reserve (~71,000 km 2 ; adopted in 2002 and expanded in 2014; governed by Australia), the South Georgia and South Sandwich Islands MPA (~1.24 million km 2 ; adopted in 2012 and expanded in 2019; governed by the United Kingdom), the Prince Edward Islands MPA (~180,000 km 2 ; adopted in 2013; governed by South Africa), and the Crozet and Kerguelen Islands MPAs (~1.14 million km 2 ; adopted in 2017; governed by France). Note that the northern boundaries of the Prince Edward Islands, Kerguelen and Crozet MPAs extend beyond the CCAMLR Convention Area boundary. CCAMLR has also collectively adopted two MPAs: the South Orkney Islands Southern Shelf MPA (~94,000 km 2 ; adopted in 2009) and the Ross Sea region MPA (~1.55 million km 2 ; adopted in 2016). Three large MPA proposals also remain under negotiation at CCAMLR in the East Antarctic (proposed at~1 million km 2 ), the Weddell Sea (~2 million km 2 ) and in Domain 1 (~466,000 km 2 ) (Fig 1).
Of the nine planning domains established by CCAMLR (Fig 1),

Fraction of ecoregion protected
In the CCAMLR Area, 23 benthic ecoregions have been identified [45]. Of these, 12 benthic ecoregions are at least partially protected in no-take zones of existing Southern Ocean MPAs (0.21-64.0%; median = 7.58, mean = 14.33; Table 2). However, only six of these benthic ecoregions have 10% or more no-take protection; only two benthic ecoregions have 30% notake protection (Table 4). Across all zones of existing Southern Ocean MPAs, 13 benthic ecoregions are at least partially represented in existing MPAs (0.21-99.95%; median = 32.75; mean = 34.22; Fig 2; Table 2). Nine of these benthic ecoregions are at least 10% represented in existing MPAs; seven benthic ecoregions are at least 30% represented in the existing MPAs (Table 4). Ten benthic ecoregions are not represented in the current Southern Ocean MPA network (Fig 2; Table 2). Table 4. Number of benthic ecoregions and pelagic clusters that have � 10% and � 30% represented (out of 23 benthic ecoregions and 19 pelagic clusters considered in this analysis). See Table 2 and Table 4 for names and descriptions of benthic ecoregions and pelagic clusters.

PLOS ONE
Towards a representative network of Southern Ocean protected areas proposed MPAs in the Southern Ocean MPA network would result in 17 benthic ecoregions being at least 10% protected; 12 of these benthic ecoregions would achieve being at least 30% protected (Table 4).
In the Southern Ocean, 20 pelagic clusters have been identified [46], however only 19 of these fall within the CCAMLR Area (cluster 18 only occurs outside the CCAMLR Area). Of

PLOS ONE
Towards a representative network of Southern Ocean protected areas these, 18 pelagic clusters are at least partially protected in no-take regions of existing Southern Ocean MPAs (1.19-35.29%; median = 6.11; mean = 11.11; Table 3). However, only seven of these pelagic clusters are at least 10% protected in no-take zones; only two are at least 30% protected in no-take zones (  Table 3). Eleven of these pelagic clusters are at least 10% represented in existing MPAs; six are at least 30% represented in existing MPAs (Table 4).
Proposed MPAs in the East Antarctic, Weddell Sea and Domain 1 (Antarctic Peninsula) currently being negotiated by CCAMLR would increase representation of almost all pelagic cluster types (6.98-99.78%; median = 50.65; mean = 47.17; Fig 3; Table 3). Including these additional proposed MPAs in the Southern Ocean MPA network would result in 17 of the pelagic clusters being at least 10% protected; and 13 of the clusters being at least 30% protected (Table 4).

Protection equality
The protection equality of the no-take zones of existing MPAs were 0.18 and 0.41 for benthic ecoregions and pelagic clusters, respectively (Table 5). For all zones of existing MPAs, the protection equality values increased to 0.26 and 0.44 for benthic and pelagic regions, respectively. Including the existing and the proposed MPAs increased the protection equality values to 0.52 and 0.67 for benthic and pelagic regions, respectively (Table 5).

Discussion
CCAMLR has successfully adopted two MPAs in the Southern Ocean, with the Ross Sea being the world's largest international MPA at~1.55 million km 2 . CCAMLR jurisdiction MPAs encompass 4.6% of the CCAMLR Area, mostly comprised by the large Ross Sea region MPA. Nationally implemented MPAs encompass 7.21% of the CCAMLR Area. Collectively almost 12% of the Southern Ocean is encompassed in MPAs, thus the region meets the 10% area targets of the Convention on Biological Diversity [15] and the United Nations Sustainable Development Goals [17], and in surpassing the proportion ice-free areas protected on the Antarctic continent [22]. No other high seas management body has achieved this level of protection. It exceeds the global average of 7.91% [36]. Many national waters have not reached the 10% target (e.g., Norway at 0.83%), however, others have far surpassed it (e.g., United States, France, and Australia all have greater than 40% of their national waters protected) [36]. Indeed, among the 66 large marine ecosystems in the world, the Antarctic has the 2 nd largest area encompassed by MPAs and the Ross Sea MPA is considered to contain a high level of ecological representativeness for Antarctic biodiversity [37].
Despite having more than 10% of the Southern Ocean protected, only 4.61% is encompassed in no-take areas, largely comprised of the South Orkney Islands Southern Shelf MPA, HIMI marine reserve, and a large proportion (~70%) of the Ross Sea region MPA. Multiple studies point to the importance of MPAs having no-take areas to be effective at conserving biodiversity, including fish populations [4,5,[81][82][83]. Furthermore, the Ross Sea region MPA has a limited 35-year duration, meaning that this proportion might not receive protection after this time if the MPA is not renewed. Moreover, while some targets call for 10% protection, many studies suggest that less than 30% is insufficient to protect biodiversity, conserve ecosystem services-including sustaining commercial fisheries-and to achieve socioeconomic priorities set by these targets [9,84,85]. Others have argued that protection targets closer to 50% protection are required to curb biodiversity loss [18,86,87].
Beyond percentage targets, current protected areas do not provide a representative sample of the Southern Ocean's biodiversity. Global targets call for protected areas to be ecologically representative [14][15][16], meaning that protection should encompass the full range of biodiversity present in a region [88]. Overall, current MPA distribution is largely biased towards sub-Antarctic regions and the Ross Sea. Thus, within currently established no-take areas in the Southern Ocean, only two benthic ecoregions have 30% protection. The Ross Sea ecoregion meets this threshold, due to the large-scale MPA in that region and the South Sandwich Islands ecoregions also has this level of protection due to recent (2019) expansions in no-take areas [55]. For pelagic clusters, only 13 (shallow parts of sub-Antarctic plateaus near Kerguelen and South Georgia) and 20 (Crozet shelves) have 30% protected in sub-Antarctic MPAs. At the 10% threshold, still only six benthic ecoregions and seven pelagic clusters are protected in notake areas. Factoring in all existing MPAs, including no-take and multi-use zones, the Southern Ocean MPA network is still not representative of all benthic ecoregions and pelagic clusters, thus it is not representative of Southern Ocean biodiversity. This is in line with global MPA trends where, while there has been an overall increase in representation, overall 61% of the benthic ecoregions in national waters remain unprotected [38] and globally, most largemarine ecosystems do not have greater than 10% representation [37].
The adoption of additional proposed MPAs in the Weddell Sea, East Antarctic and Antarctic Peninsula would increase representation in the Southern Ocean MPA network. All of these regions encompass parts of CCAMLR's MPA planning domains (Fig 1) and original priority areas [89]. With the addition of these proposed MPAs, roughly three-quarters of the benthic ecoregions and almost all pelagic clusters would be 10% represented. However, as noted above, to conserve biodiversity these additional MPAs should have no-take zones and further should have long duration (e.g., [81]). Even with the addition of pending MPA proposals, some areas remaining poorly represented. These include the benthic ecoregions of the South Atlantic (mostly in northern Domain 4), Amundsen (mostly in Domain 9), Central Indian-Wilkes subregion (in Domain 7), East Indian Abyssal (mostly in Domain 7), Ob and Lena (mostly in Domain 5), and Pacific Basin (largely in Domain 9). Additional MPAs (to those existing or currently being proposed) would allow for complete representation.
The Southern Ocean MPAs also do not have equitable protection in terms of proportionality protected across benthic ecoregions. For existing MPAs, the benthic ecoregions fall within the lowest quartile and the pelagic clusters are in the second lowest quartile for equality protection [78],. However these numbers are comparable to protection equality values for national MPAs globally [38]. While these values are much higher for the network of MPAs achieved by currently existing and proposed MPAs, protection equality still measures at less than 50% for benthic ecoregions and 60% for pelagic clusters which puts them in the second highest quartile [78]. While not completely equitable, these values are much higher than the values for MPAs globally inside national waters [38].
This assessment of representativeness was undertaken on the basis of large-scale benthic and pelagic regionalizations. The large-scale regions provide a helpful broad measure of progress but do not go far enough to plan for capturing biodiversity patterns, internal heterogeneity, genetic diversity and cryptic species [45]. Although biological and ecosystemlevel data are more difficult to work with, and typically do not have consistent circumpolar coverage, consideration of such data might also provide a more nuanced assessment of the strengths and gaps in current and proposed MPAs (e.g., [90][91][92]). Furthermore, at smaller scales, MPAs may be designed to protect vulnerable or critical habitats that are missed in broad-scale regionalizations. Ensuring protection of all ecoregions and replicated protection of particular ecoregions across different ocean basins is one possible means of addressing this.
The urgency of the threats to the Southern Ocean and the need for protection is critical now more than ever before. The Southern Ocean supports international commercial fisheries for Patagonian and Antarctic toothfish (Dissostichus eleginoides and D. mawsoni; sold as Chilean sea bass) and Antarctic krill (Euphausia superba) [44]. Pressure on these fisheries has increased in recent years [93] and is likely to continue, and at the same time climate change pressures on Southern Ocean ecosystems are also increasing [94][95][96][97][98]. The cumulative impacts of fishing and climate change are likely to have greater effect than either impact alone [99][100][101]. Increasing numbers of studies show that MPAs, especially no-take marine reserves, can be a proactive and precautionary tool to enhance resilience to environmental change, including climate change and warming [10-12, 102, 103]. Importantly, the MPAs need to be well designed, with representation being one of many elements. Key biodiversity areas, vulnerable and rare species should be considered in MPA design, as well as connectivity (e.g., [6,104]). Further, the MPAs need to be well managed and enforced [105,106], a significant challenge for large-scale MPAs in a place as large and remote as the Antarctic [107].
Nonetheless, protected areas alone will not suffice to conserve Antarctic marine biodiversity [108]. CCAMLR may need to enact other precautionary management measures targeted at reducing or even eliminating fish catch in some areas [109]. Given the international nature of climate change and threats to Antarctic biodiversity, successful deployment of such measures by CCAMLR will require collaboration with other appropriate international organizations and initiatives, including those of the United Nations [110][111][112][113]. Integration across these management bodies will broaden CCAMLR's toolbox [114] for taking action on conserving the globally significant biodiversity and living resources of the Southern Ocean [115][116][117][118].