Evaluating the effectiveness of a large multi‐use MPA in protecting Key Biodiversity Areas for marine predators

Marine protected areas can serve to regulate harvesting and conserve biodiversity. Within large multi‐use MPAs, it is often unclear to what degree critical sites of biodiversity are afforded protection against commercial activities. Addressing this issue is a prerequisite if we are to appropriately assess sites against conservation targets. We evaluated whether the management regime of a large MPA conserved sites (Key Biodiversity Areas, KBAs) supporting the global persistence of top marine predators.


| INTRODUC TI ON
For the conservation of marine species and ecosystems, particularly for those where harvesting of living resources is pursued, it is imperative to ensure both sustainable production and adequate long-term protection are appropriately maintained (Margules & Pressey, 2000).
In the marine realm, both the UN Convention on Biological Diversity (CBD) Decision X/2 and the Sustainable Development Goals (SDGs) highlight the need for sustainable use of marine resources (Aichi Target 6 and SDG 14) (CBD, 2010;UN General Assembly, 2015).
Additionally, the CBD has set a global conservation target of 10% of coastal and marine areas to be protected through effective management (encompassed in Aichi Target 11) (CBD, 2010). The generally accepted route to achieving these outcomes is through robust fisheries management and, more recently, in combination with the designation of marine protected areas (MPAs) (Edgar et al., 2014).
However, while there has been increased emphasis on the establishment of MPAs (Lubchenco & Grorud-Colvert, 2015), their location and effectiveness have been questioned (Gill et al., 2017;UNEP-WCMC & IUCN, 2016;Zupan et al., 2018). Robust fisheries management will always be a necessity, and capacity shortfall often limits protected area efficacy of both small and large MPAs (Gill et al., 2017;O'Leary et al., 2018). There are also concerns regarding whether MPAs can ensure the long-term persistence of species and whether they cover the most important sites for biodiversity (Barr & Possingham, 2013;Klein et al., 2015).
A suite of criteria now exist for identifying important sites for marine biodiversity (Dunn et al., 2014;Lyons, 2019); a critical requirement for both MPA delineation and assessment (Ehler & Douvere, 2009;Smith et al., 2019). Of these, the overarching framework for identifying critical sites for species is that of Key Biodiversity Areas (KBAs) (Eken et al., 2004). KBAs are sites important for the global persistence of biodiversity, identified as containing a significant proportion of a species' global population or ecosystem extent; the criteria include, but are not limited to, thresholds for threatened and geographically restricted species or ecosystems, and congregations of species during key life stages (IUCN, 2016). These global criteria are applicable to all macro-organisms, and KBA identification follows a standardized and quantitative set of guidelines that have recently been released (IUCN, 2016; KBA Standards & Appeals Committee, 2019). Coupled with these recently established guidelines, the proliferation and enhanced resolution of animal tracking data have now made it feasible to identify marine biodiversity hotspots, along with associated abundance estimates of species within these hotspots Lascelles et al., 2016;Soanes et al., 2016).
The marine conservation community can therefore apply these standardized protocols to identify sites of high biodiversity value and assess whether they are sufficiently represented within the boundaries of established MPAs or whether further areas need to be afforded legislative protection.
We applied the new KBA criteria ( KBA Standards & Appeals Committee, 2019) to identify sites critical to the persistence of biodiversity for a suite of marine top predators (12 seabird species, two mammal species) which breed at South Georgia and the South Sandwich Islands (SGSSI). Because KBAs can only be designated based on the current known presence of biodiversity, we selected taxa which are widely regarded as indicators of the broader biodiversity and state of the ecosystem (Boersma, 2008;Furness & Camphuysen, 1997;Moore, 2008), and for which sufficient data exist. We then addressed the extent to which there is spatiotemporal overlap between these sites and key fisheries, to assess whether the key objectives of the MPA could be potentially met. We focused particularly on two objectives of the MPA, namely to "better protect important biodiversity" and to "protect foraging areas used by spatially constrained krill-eating predators." Assessing whether these objectives are being met will enhance the ability of the Government of SGSSI to implement adaptive management regimes within the MPA and will facilitate understanding about the broader factors that may play a role in driving species population trends.
Our approach focuses on evaluating whether established conservation investment is delivering desired protection, and is the first utilizing the new KBA guidelines (KBA Standards & Appeals Committee, 2019) for identifying critical biodiversity sites at sea for marine top predators. This approach is readily applicable for use by practitioners requiring the identification of KBA sites elsewhere.

| Study area
South Georgia and the South Sandwich Islands lie within the Antarctic Circumpolar Current, south of the polar front (Figure 1), and are a hotspot of marine biodiversity (Rogers, Yesson, & Gravestock, 2015;Trathan et al., 2014). Their position means they are subject to strong seasonal variations in light, temperature and sea ice, which leads to strong seasonality in primary production and the abundance of Antarctic krill (Euphausia superba), a key prey item for many predators that breed at the islands (Barlow et al., 2002;Croxall, Prince, & Reid, 1997). Specifically, the islands support key populations of seabirds and pinnipeds (Hart & Convey, 2018;Lynch et al., 2016;Trathan, Daunt, & Murphy, 1996). At South Georgia, top predators breed primarily along the north coast and in the north-west of the  et al., 1996). Furthermore, the seascape north-west of the islands hosts a region of high primary productivity which supports local krill stocks (Rogers et al., 2015).
The islands are surrounded by a large sustainable-use MPA (SGSSI MPA, 1.07 million km 2 ) which was designated in 2012, and encompasses the entire exclusive economic zone (EEZ) (Trathan et al., 2014). The primary objective of the MPA (IUCN Category VI) is to ensure the protection and conservation of the region's rich biodiversity, through a number of measures that reduce the risk of biodiversity loss, while also allowing for sustainable fisheries operations and ecotourism (GSGSSI, 2013;Rogers et al., 2015;Trathan et al., 2014). Impetus for the MPA came from the desire to conserve species and habitats in the face of climate variability and change, and previously high levels of illegal, unregulated and unreported fishing (IUU), and incidental mortality (bycatch) (GSGSSI, 2017;Trathan et al., 2014).

| Data considered for KBA identification
We collated data on IUCN Red List threat status (IUCN, 2018), breeding site locations, population sizes and at-sea locations (derived from tracking data) sampled using global positioning systems (GPS) and platform terminal transmitters (PTT), for 14 species of higher predators (Table 1) which breed at approximately 815 sites across SGSSI (Appendix S1, Sheet: "Pops_data_sources" & Appendix S2, section "Species overview"). We used the most recent population estimates available for each breeding site, based on published information (Table 1)  seabi rdtra cking.org) and other stakeholders, representing 12 different species (Table 2). Tracking data came from species-rich sites with high abundance of biodiversity. Sites included a primary site in the north-west of South Georgia, Bird Island and six other regions on the northern coastline of South Georgia. Two key sites to the west and south of South Georgia would benefit from future sampling efforts, but the southwest coast of South Georgia provides generally poor breeding habitat for many of the species considered in this study (Appendix S2, Figure S1). Tracking data spanned the 1990/91 austral summer to the 2015 austral winter (Appendix S1, Sheet: Tracking_data_sources). No species have been tracked from the South Sandwich Islands. Data were analysed at the level of homogeneous units, which account for specific stages in the annual cycle of an organism where distribution may be more or less constrained during a given stage, location or age of the organism. Specifically, we refer to these homogeneous units as "data-groups," where each data-group represents a species, from a particular breeding site and during a specific breeding stage, and accounts for age or sex differences where necessary (e.g. Figure 2). Initially, 69 data-groups were distinguished from the collated tracking data (Appendix S2, Figure S2). Each of these data-groups was assessed in a stepwise fashion to determine whether at-sea distribution data were suitable for attempts to identify representative core-area use sites (step 1, e.g. sufficient sample size), then whether these sites were representative of the sampled population (step 2) and finally whether these sites would be regarded as manageable units as per the KBA guidelines (step 3). The final at-sea sites, identified through the approaches outlined below, were assessed to determine whether they met the global KBA criteria (See Appendix S2 for further details and Table S1 for details pertaining to each data-group).  ?
Note: Breeding site types, * = breeding colonies; † = Bird Island breeding colony only; ‡ = breeding zones/regions; § = main haul-out regions; ¶ = specific haul-out beaches. ** = trend based on Bird Island colony only. Breeding site-specific population estimates available in Appendix S1 (Sheet: "Pops_data_sources"). References (alphabetic characters) listed in Appendix S3. Population sizes refer to the number of breeding pairs and adult females for seabirds and seals, respectively.

| Delineating KBAs
Delineating KBAs for marine predators at sea requires the identi- We used recently established methods which have been derived for seabird species to identify representative at-sea areas used by a threshold number of individuals based on (a) tracking data (Lascelles et al., 2016, Heraah et al., 2019 and (b) foraging radii (Soanes et al., 2016) and species distribution models .
The above methods were originally developed for the identification of marine Important Bird and Biodiversity Areas (mIBAs) (Lascelles et al., 2016). However, the key outcome from these methods is a spatial polygon representative of the sampled population

| Tracking data
The primary method used to identify marine KBA sites utilized the raw tracking data Lascelles et al., 2016) (Figure 2). Tracks from GPS and PTT devices were filtered following standard protocols to remove outlying locations based on speed thresholds and ARGOS location classes (Freitas, 2012;Sumner, 2016), remove points on land, and regularize sampling intervals across all data-groups (Calenge, 2006). Additionally, TA B L E 2 Tracking data considered for the identification of Key Biodiversity Areas within the South Georgia and the South Sandwich Islands MPA. For a more detailed breakdown of tracking data, see Appendix S1 (Sheet: "Tracking_data_sources") and Appendix S2 ( Figure  S2)

F I G U R E 2
Overview of approach using tracking data to identify Key Biodiversity Areas at sea (method adopted from the protocol to identify marine Important Bird and Biodiversity Areas (Lascelles et al., 2016)) shown for the example of the data-group, adult Grey-headed Albatrosses during post-guard from Bird Island, South Georgia (n individuals = 37, n tracks = 193): (i) interpolated tracks, (ii) core foraging areas of each individual bird, (iii) assessment for representativeness of tracking for sampled population, where data are simulated across sample sizes from 1 to n individuals -1, (iv) polygons where core foraging areas of at least 10% of tracked individuals overlap (selected according to representativeness value), (v) [green] core-area polygons with abundance estimates that meet KBA criteria, (vi) refined [blue] polygons with minimized boundary-to-area ratio suitable for management. Black boundary indicates South Georgia MPA data-groups were removed when the sample size was insufficient or sampling intervals were too sparse (Appendix S2: Further details). supplementary material). Where representativeness was < 70% for a data-group, the tracking data were deemed inadequate to describe the space use of the population (Appendix S2, Table S1).
To enhance practicability of management zones, the identified spatial polygons were aggregated to minimize the boundary-to-area ratio via a custom R script utilizing the smoothr package (Strimas-Mackey, 2018). Specifically, any isolated polygon or hole within a larger polygon, which was smaller than 5% of the total area identified, was removed or filled, respectively. For the remaining polygons, the great circle distance between centroids was calculated. Using this distance matrix, a hierarchical cluster analysis was implemented to identify which polygons could be grouped specified by a threshold distance of 5% of the maximum distance between polygons. The final boundaries of sites identified for each data-group were delimited by a minimum convex polygon (Figure 2, vi).

| Foraging radii and species distribution models
Where tracking data were unavailable for species, the alternate methods of using foraging radii and previously developed species distribution models for near-shore foraging species were used (see Appendix S2, "KBA sites per data-group," for further details of methods specific to sites identified for each data-group). The foraging radius approach was applied to breeding sites holding > 1% of the global population of a species. This method consists of defining a radius around the colony based on the mean-maximum foraging distance achieved by the species (derived from tracking data collected elsewhere) (Soanes et al., 2016). This approach was applied for Gentoo Penguins (17 km radius (Ratcliffe, Adlard, Stowasser, & Mcgill, 2018;Tanton, Reid, Croxall, & Trathan, 2004)) at South Georgia and for Chinstrap Penguins (60 km radius (Ratcliffe & Trathan, 2011)) at the South Sandwich Islands. We also used the foraging radius approach for Antarctic Fur Seals (150 km radius (Boyd, 1999)), but bound this site by an established species distribution model (Boyd, Staniland, & Martin, 2002)) and further understanding of the species foraging ecology (Lunn, Boyd, Barton, & Croxall, 1993).
A recent Chinstrap Penguin tracking study, a near-shore foraging species during the breeding period, at the South Orkney Islands showed a high degree of overlap between the areas identified as important using tracking data and those predicted from species distribution models . Therefore, for Macaroni Penguins which also forage in the near-shore environment during the brood-guard and crèche periods, we used the final boundaries of South Georgia island-wide predicted distribution models for Macaroni Penguins during these different periods (Scheffer, Ratcliffe, Dias, Bost, & Trathan, 2015). As these projected distributions reflect the likely distribution for this species across the whole of South Georgia as oppose to the site-based tracking data approach, we based abundance estimates for these data-groups on the island-wide population estimates for species during the breeding period.

| Fisheries in identified KBAs
Three fisheries (details outlined in Table 3)  We evaluated the role of the MPA in conserving globally important sites of biodiversity by assessing the overlap between the identified KBAs and the operational areas of the main fisheries within the MPA. This analysis was carried out for each data-group separately to account for variation in foraging distributions. Due to differences in diets and foraging behaviours of predators, overlap with the krill fishery was only assessed for krill-dependent species and overlap with demersal longline and pelagic trawl fisheries only for species which have historically been recorded as bycatch (Table 3). We first assessed temporal overlap for data-groups which met the KBA criteria during the operating periods of the fisheries. Then, when temporal overlap was possible, we assessed spatial overlap as the proportion of the KBA layer which intersected with potential fishing grounds within the SGSSI MPA.

| KBA identification
Representative core areas at sea which met the global KBA criteria were identified for 19 data-groups, featuring nine species (one seal, eight seabird species) of the 69 data-groups assessed (Table 4 and Appendix S2 ("KBA sites per data-group")). After accounting for  Figure S1).

| Potential interactions with fisheries in KBAs
For four species, comprising five of the 19 data-groups, there was temporal overlap with the KBA site and the respective fisheries operating period (Table 4). For four of these data-groups, there was also the opportunity for direct interaction with a fishery through spatial overlap (Figure 4). For the remaining five species and 14 of 19 data-groups, the relevant fishery would be closed during the period for which we identified KBAs (Appendix S2, "Temporal overlap of KBA sites with fisheries operating periods").

| Krill fishery
Six species were recognized as krill-dependent predators (Table 3). For the four species which had representative sites at sea that met the KBA criteria, all except one were delimited entirely within the MPA (Table 4).
Two of nine data-groups had the potential for temporal interaction with the krill fishery, as for all other data-groups their use of the MPA was during the fishery closure period (Appendix S2). These data-groups the potential for spatial overlap with the krill fishery as 88.8% of this KBA is beyond the 30-km no-take zone (Table 4). By contrast, the South Georgia island-wide KBA identified for Gentoo Penguins, based on a 17 km foraging radius, lies entirely within the pelagic no-take zone.

| Demersal longline fishery
Of the six species recognized to be at risk of bycatch in the demersal longline fishery (Table 3), four had representative sites at sea that met the KBA criteria. For these four species, two of nine data-groups, both for the Wandering Albatrosses, had KBAs where a potential for interaction with the demersal longline fishery would be possible (Table 4). These data-groups were adult Wandering Albatrosses during the (a) brood-guard period (April) and (b) post-guard period (April-August), where 82.0% and 98.1% of the KBAs fell within the MPA, respectively (Figure 4 iii, iv). During both periods, KBAs were situated in the region where the longline fishery is legally allowed to operate (waters between 700 and 2,250 m deep). However, areas of these KBA sites are also off limits to demersal longline fisheries because they fall within the 30-km no-take zone, the no bottom fishing zones (0-700 m depth) and two of the main benthic closed areas ( Figure 1). Therefore, for both Wandering Albatross data-groups, the proportion of the KBAs for which there is potential for interaction with fisheries within the MPA is 23.5% and 27.8% for the broodguard and post-guard period, respectively (Table 4).

| Pelagic trawl fishery
For two species, Black-browed Albatrosses and White-chinned Petrels, we recognized the potential for negative interactions with the pelagic trawl fishery ( for the 30-km no-take zone (Figure 1, Figure 4v).

| D ISCUSS I ON
Using a collation of contemporary tracking data and knowledge of species breeding populations, we identified the first marine KBAsfollowing the new standards and guidelines -both within and beyond the borders of the South Georgia and South Sandwich Islands large MPA. This distribution of KBAs reflects the contrasting foraging strategies of top predators assessed in this study (Appendix S2, Figure  S2). Critically, the primary objective of the MPA is to protect marine biodiversity, habitats and critical ecosystem function (Trathan et al., 2014). Therefore, considering that for only five data-groups there was the possibility of spatiotemporal overlap with a unique KBA site and relevant fishery within the MPA, the current conservation measures (Table 3) in the context of interaction with fisheries appear to be achieving the desired goals for the 14 top predators considered in this study. Coupled with the seasonal closures of the krill and demersal longline fisheries throughout the entire MPA, protection of these marine predators at sea is also promoted by regulations on gear used and fishing practices (Table 3). These mitigation measures facilitate the achievement of objective I of the MPA, protection for all species considered in this study (Croxall, Prince, & Reid, 2004;GSGSSI, 2013GSGSSI, , 2017. For krill-eating Macaroni Penguins, there is potential for spatial overlap with the krill fishery during the post-moult period (May-August); however, the estimated krill stock taken by both this species and the krill fishery is negligible. As such, direct competition during this period is likely to be low under the current krill harvesting levels . It seems likely, therefore, that the foraging areas of the six krill-eating predators (Table 3)  Heard Island (Patterson et al., 2016)   EEZs. Conservation measures in many of these areas, beyond seasonal closures, require seabird bycatch mitigation measures to be used within the fisheries which are similar to those in the SGSSI MPA, all of which have greatly reduced seabird bycatch rates. Concerns which still remain for many of these species, however, are the effects of distant-water pelagic longline fisheries and IUU fishing, mostly in waters beyond the jurisdiction of MPAs (Clay et al., 2019;Michael et al., 2017;Österblom & Bodin, 2012). This threat is believed to be a key driver in continued declines of some albatross and petrel populations, including those at SGSSI (Table 1) (Krüger et al., 2018;Pardo et al., 2017;Poncet et al., 2017). Therefore, efforts must still be made across fisheries management organizations to implement and enforce best-practice bycatch mitigation both within areas beyond national jurisdiction and the EEZs of other coastal states (Carneiro et al., in press;Clay et al., 2019;Melanie, White, Smith, Crain, & Beck, 2010).

Grey-headed
While the links between both local and distant-water fisheries and marine top predator population declines have been well-established, of growing concern is the impact of climate change on predator populations and their prey (Atkinson et al., 2019;Krüger et al., 2018;Pardo et al., 2017). Of particular importance for SGSSI is the impact of climate change on the distribution of Antarctic krill, an important prey item for numerous top predators which breed at the islands (Boyd, 1999;Croxall et al., 1997;Forcada & Hoffman, 2014).
Recent evidence suggests that over a 90-year period, krill distribution has shifted southward by approximately 440 km, likely as a result of warming seas and a reduction in sea-ice cover (Atkinson et al., 2019). The shifting distribution of krill may in turn influence the breeding success of top predators as these species are constrained in foraging duration and distance when rearing offspring (Lunn et al., 1993;Weimerskirch, 2007). Therefore, just as for predators which breed in high northern latitudes (Divoky, Douglas, & Stenhouse, 2016;Macias-Fauria & Post, 2018), there is a critical need for continued monitoring efforts to assess the effects of shifting prey distributions (due to climate change) on predator populations.
Spatially explicit analyses of krill consumption by predators would be particularly informative, especially in understanding if recovery of marine mammals or changes in other predator species distributions have occurred in particular areas as a result of changing krill distributions. Identifying KBAs at sea may serve as a key baseline with which to compare spatial distribution of sites identified in future.
Because South Georgia is a comparatively well-studied archipelago (Hart & Convey, 2018;Lynch et al., 2016;Rogers et al., 2015;Trathan et al., 2014Trathan et al., , 1996, prior conservation successes for some albatross and petrel species have been possible (John P Croxall, 2008;Hays et al., 2019). However, to enhance the identification of at-sea KBAs for marine top predators that inhabit remote sites in future, several limitations of this study which apply to marine predator datasets globally (e.g. incomplete population estimates and representativeness of tracking data) will need to be overcome. Although the population counts and tracking data were not available for the same time periods, it is unlikely that the KBAs would have changed substantially if we used more contemporaneous population estimates.
This is because many of the sites identified for each data-group were from declining populations of globally threatened species (  (Borboroglu & Boersma, 2013;Boyd et al., 2002;Lynch et al., 2016).
Improved knowledge of the spatiotemporal distribution of top predators during all major life history stages is crucial for a holistic understanding of population-level habitat use and overlap with threats (Carneiro et al., in press;Clay et al., 2019;Reisinger et al., 2018 (Oppel et al., 2018). Additionally, for some species, the at-sea distribution of major colonies at South Georgia and all colonies at the South Sandwich Islands remains to be investigated (Appendix S2, Figure S1, and "Future research"). Despite these knowledge gaps, the network of KBA sites is probably well-justified for the species considered in this study, particularly near-shore foraging species-penguins and Antarctic Fur Seals-as they account for their most plausible island-wide breeding ranges. During the non-breeding period when all species considered in this study (excl. Gentoo Penguins) are wide-ranging (Appendix S2, Figure S2) and site-based conservation approaches such as protection or management of KBAs are less effective, likely conservation solutions will be the mitigation of the broad threats marine predators face across the oceans (Clay et al., 2019;Halpern et al., 2015). Future effort should also be directed towards recovering populations of previously over-exploited cetaceans (Zerbini et al., 2019).
In a more localized context, environmental management plans should also consider the fact that sites meeting the global KBA criteria are those sites which "contribute significantly to the global persistence of biodiversity" (IUCN, 2016). This presents caveats to the KBA approach that may either promote or mask the conservation requirements of species at a regional scale. For example, if a species is locally abundant but globally rare (such as the Antarctic fur seal), higher priority might be given to the conservation of a species in systematic conservation planning procedures (Smith et al., 2019). In contrast, species which are globally abundant but experiencing local population declines may not yield sites which meet global KBA criteria (such as the South Georgia Black-browed Albatross population which is considered locally vulnerable (Poncet et al., 2017)). Thus, context-specific decisions must be made as to how and when the utility of KBAs can be used to achieve local, national and global goals, and when additional data sources or approaches will be required to achieve conservation goals at varying spatiotemporal scales (Smith et al., 2019).
In recognition of the globally threatened species and species with significant proportions of their respective global populations that breed at SGSSI, our study informed policy and management processes at a local level through the utility of the new global KBA initiative.
Ensuring both the conservation of species and sustainable harvesting of biological resources is a critical factor for the continued success of the MPA, as revenue generated from fisheries is often key to supporting the ongoing monitoring and management of MPAs (Melanie et al., 2010). Furthermore, the objectively defined sites identified in this study play a critical role towards meeting the 2020 Aichi Biodiversity  (Hays et al., 2019;Phillips et al., 2016). Precedence to address the effects of competition for resources, particularly with a growing interest in mesopelagic fisheries (St John et al., 2016), and the future resilience of systems to climate change will also be critical to consider.
Ultimately, recognizing sites through the new KBA framework now provides a harmonized approach to identify sites critical to biodiversity across all taxa (IUCN, 2016; KBA Standards & Appeals Committee, 2019). Therefore, we encourage practitioners to adopt this framework both for the development of future projects investigating species distributions and for the retrospective analysis of animal tracking data.

This research was funded by the Pew Bertarelli Ocean Legacy
Project of the Pew Charitable Trusts and Bertarelli Foundation. We would also like to thank the British Antarctic Survey GIS team and the many individuals who each contributed to the successful development of these large tracking and population estimate datasets.

DATA AVA I L A B I L I T Y S TAT E M E N T
Tracking data used in this study are detailed in Appendix S1 (Sheet: Tracking_data_sources), where users may view the tracking dataset IDs and make appropriate requests for the data via the BirdLife International Seabird Tracking Database (www.seabi rdtra cking.