Introduction

To counter the biodiversity crisis, there is an urgent need to protect important habitats to ensure the stability of global ecosystems. In response, the International Union for Conservation of Nature (IUCN) has called for 30% of the Earth’s overall land and sea area to be protected by 2030. However, as of April 2023 only 2.9% of the ocean was highly protected (www.mpatlas.org; Morgan et al. 2018). There is also concern that protected areas could be situated in locations where increased protection offers marginal benefit for biodiversity (Devillers et al. 2015, Woodley et al. 2021). One barrier to protected area implementation is that limited knowledge of animal distributions and abundance hampers the identification of the most critical locations. In the marine realm, data collection is logistically and financially challenging; however, there are numerous, pressing threats to biodiversity that include fishing, climate change, pollution, shipping and energy generation (Halpern et al. 2008).
Seabirds are a highly threatened animal group (Croxall et al. 2012) with a marine foraging niche (they depend upon the sea for food). Their relative detectability and accessibility (i.e. above water, colonial, terrestrial breeding) compared to other marine species has provided sufficient data to merit global analyses of conservation priority (Dias et al. 2019a) and designate marine protected areas (MPAs) (e.g. North Atlantic Current and Evlanov Sea-basin MPA; Davies et al. 2021). During the breeding season, seabird distribution is focussed around colonies, as birds regularly return to perform parental duties such as incubating eggs and feeding chicks, and area-based conservation measures, such as MPAs, are more feasible than during more dispersive migratory and non-breeding seasons (Oppel et al. 2018). To identify MPAs for seabirds, a spatial representation of their marine foraging niche is required. This can come directly from bird-borne tracking devices or at-sea surveys, which locate hotspots of occurrence or abundance to delineate candidate MPAs (Lascelles et al. 2016). Alternatively, such data can be entered into Ecological Niche Models (ENMs), also known as species distribution models, to make predictions beyond surveyed areas. ENMs build statistical relationships between species space use and remotely-sensed environmental variables, and then predict these relationships over broad areas sampled by remote sensors (e.g. satellite imagery). For seabirds, these environmental variables describe important marine habitat and bio-physical processes such as seamounts, frontal systems and productivity blooms, which can characterise their foraging niches, for example the foraging niche of a seabird population could be described as specialising on upwellings at pelagic seamounts. The ability of ENMs to predict seabird foraging niches beyond surveyed areas, has seen them regularly used for marine spatial planning and identification of priority conservation areas (Nur et al. 2011, Žydelis et al. 2011, Lavers et al. 2014, Dias et al. 2019b).
However, a key hurdle for ENMs is to extrapolate predictions beyond the geographical and temporal range of training data, known as model ‘transferability’ (Randin et al. 2006, Yates et al. 2018). A growing body of literature suggests limited ENM transferability between different regions (Redfern et al. 2017, Mannocci et al. 2020). In seabirds, Warwick-Evans et al. (2018) found that the foraging niche of chinstrap penguinsPygoscelis antarcticus was transferable between colonies at <100 km, while Péron et al. (2018) found that foraging niche of Scopoli’s shearwaterCalonectris diomedea was transferable locally but not regionally (>200 km). Transferability at large scales (>1000’s of km) does not apply in grey petrelProcellaria cinerea wintering distributions (Torres et al. 2015), nor red-billed tropicbird Phaethon aethereusforaging niches (Diop et al. 2018). These studies support contrasting regional marine habitats and differing associations with habitat by different populations (local adaptation) as major barriers to transferability. Local adaptation is driven by strong philopatry in seabirds, many of which have evolved population-specific foraging behaviours suited to local biotic and oceanographic conditions (Peck & Congdon 2005, Mendez et al. 2017, Gilmour et al. 2018). To improve transferability, ENMs can be trained with data from different regions to extract commonalities in local adaptation and better generalise a species’ niche (Matthiopoulos et al. 2011). With the growing availability of multi-colony seabird tracking datasets (e.g. Ropert-Coudert et al. 2020; www.seabirdtracking.org), it is now feasible to train such ENMs and conduct a comprehensive evaluation of the transferability of breeding seabird foraging niches. This assessment is particularly warranted in tropical regions, which have received significantly less effort from seabird tracking studies relative to higher latitudes (Bernard et al. 2021), and ENMs could be particularly important for filling gaps in knowledge on seabird distributions.
A simple, pragmatic and generally effective alternative to ENMs for defining important areas of seascape for breeding seabirds is the foraging radius approach (Birdlife International 2010, Thaxter et al. 2012, Soanes et al. 2016, Critchley et al. 2020). At-sea observations and/or tracking data are used to calculate the distance breeding seabirds forage from their colony (hereafter a “foraging radius”), this is mapped with a circle centred on the colony to represent the sea area within which the breeding seabird population feeds (hereafter a “foraging circle”). Averaging foraging radii from multiple colonies can generalise a species’ foraging radius which can then be applied to colonies lacking information on the at-sea distribution of breeding residents (Thaxter et al. 2012). With a representative sample of colonies, the process can be applied globally (e.g. https://maps.birdlife.org/marineibas). However, the foraging radius approach overestimates foraging habitat, as not all areas within the foraging circle will be used. A foraging circle can be refined with information relating to habitat preference, such as bathymetry (Birdlife International 2010, Soanes et al. 2016) or prey availability (Grecian et al. 2012), where unsuitable habitat is ‘cut out’ from the circle. Extending this approach by using a range of remotely-sensed variables commonly used in ENMs allows a more holistic approximation of unsuitable habitat, which could improve the accuracy of foraging circle refinement. As such, this new approach better meets a key goal of systematic conservation planning, minimising the cost (in terms of size) of protected areas while maximising the confidence they contain core habitat (Margules & Pressey 2000).
The Great Barrier Reef (GBR) Marine Park, Australia is arguably the most thoroughly managed large MPA in the tropics (Fernandes et al. 2005). In addition to the mega-diverse coral reef community, for which the park was created, the GBR’s cays support numerous globally and regionally significant breeding populations of tropical seabirds (King 1993). However, numerous breeding GBR seabird populations are in decline (Heatwole et al. 1996, Batianoff & Cornelius 2005, Hemson 2015, Woodworth et al. 2020). Despite this knowledge of seabird population change, the ability of managers to understand the causes of these changes and respond to them is limited by the lack of information on where they forage. Limited tracking studies underpin much of what we know (Fig. 1), with no standardised boat or aerial-based seabird survey data existing (but see CSIRO (2020) for a survey dataset from the adjacent Coral Sea). Consequently, seabird foraging resources have not been considered in the designation of marine park zoning that dictates the location of permitted activities within the GBR Marine Park. Overcoming these data gaps could inform future amendments to marine park zoning and guide other management interventions, and potentially mitigate seabird population declines.