As marine top predators, seabirds respond to changes in the oceanography, ocean productivity, and their lower trophic-level prey (i.e., zooplankton, squid, and fish), shifting their distributions over multiple temporal scales (i.e., seasonally, inter-annually). Thus, groups of species with similar requirements may exhibit similar distribution patterns, forming assemblages that co-exist in space. To date, these biogeographic patterns have been characterized in some regions, like the South Pacific (Ainley and Boekelheide, 1983), the Eastern Tropical Pacific (Ballance et al., 1997), the North Pacific/Pacific Arctic Ocean (Piatt and Springer, 2003; Santora et al., 2017; Sigler et al., 2017), the Southern Indian Ocean (Hyrenbach et al., 2007), and the South Atlantic (Veit, 1995). However, despite the progress describing seabird distributions globally, we still lack a mechanistic understanding of how these patterns are generated (Ainley et al., 2012). Therefore, documenting these patterns and investigating their environmental drivers are critical steps toward understanding the macro-ecology of seabirds and their role in marine ecosystems. Moreover, this information has important conservation and ecosystem management implications globally, since the forecasted environmental changes are expected to affect the physical structure and productivity of the marine environment, and disrupt biotic communities (Polovina et al., 2008; Proud et al., 2017). However, to date, most studies have focused on habitat relationships of individual species, with few aiming to characterize how entire seabird assemblages are distributed over biogeographical scales and respond to environmental gradients.

Over macro-scales (i.e., 1000 km), seabirds respond to changing hydrographic properties, suggesting that specific assemblages are associated with the specific physical and biological characteristics of certain water masses (Ashmole, 1971; Hunt and Schneider, 1987). For example, seabird assemblages vary in response to changes in sea surface water temperature and salinity (Pocklington, 1979; Ainley and Boekelheide, 1983; Wahl et al., 1989; Force et al., 2015). In contrast, studies conducted in high latitude systems (e.g., the Antarctic and the Artic) highlight the important role of sea ice cover structuring the marine avifauna, with the presence of unique “ice-associated” assemblages (Ainley et al., 1993; Commins et al., 2014; Renner et al., 2016).

Other environmental drivers influence seabird assemblages over meso-scales (i.e., 100 km), with bathymetry, and seasonal changes in hydrography playing a key role, especially in broad continental shelves (Veit, 1995; Hunt et al., 2014; Santora et al., 2017). In addition to bathymetric gradients (i.e., the shelf-break), the boundaries between water masses with different productivity, temperature, density, or velocity (i.e., frontal zones) greatly influence species distributions at these spatial scales. In particular, these hydrographic features can act as biogeographic boundaries, affecting the overall seabird abundance and species richness (Scales et al., 2014). For instance, the tropical and subtropical transition front conspicuously delineates seabird distributions in the South Pacific (Ainley and Boekelheide, 1983) and the South Atlantic (Commins et al., 2014). Moreover, in the Southern Ocean, the Polar Front delineates the distribution of distinct latitudinal seabird assemblages and enhances the abundance of certain species that forage at the front (Bost et al., 2009; Force et al., 2015). While this evidence underscores the influence of hydrography on the distribution and abundance of seabirds, major areas of the global ocean have not been surveyed and characterized.

Macro-scale studies of seabird assemblages in the Eastern South Pacific Ocean (ESPO) have been restricted to the southern tip of South America and the Humboldt Current (Jehl, 1973; Brown et al., 1975; Weichler et al., 2004; Spear and Ainley, 2008). Thus, there is a lack of knowledge regarding seabird distributions farther west, between the west coast of South America and the eastern limit of the South Pacific Gyre (Figure 1) (but see Miranda-Urbina et al., 2015). This area provides a great natural experiment to examine the environmental factors driving seabird distributions, because it encompasses a strong gradient in primary productivity, and water mass properties (temperature and salinity). This gradient extends from the low temperature and low salinity water of the Humboldt Current to the East, which constitutes one of the most productive systems of the world, to the high temperature and high salinity waters of the South Pacific Gyre to the West, which is the largest low-productivity domain in the world (Longhurst, 2007; Morel et al., 2010).

Our study aims to quantify the geographical patterns of seabird abundance, species richness, and assemblage distribution in the ESPO, and to assess how environmental conditions influence these patterns. We hypothesize that seabird assemblages are spatially structured by the underlying patterns in the oceanic environment. To test our hypothesis, we employ a spatially explicit approach by applying multivariate techniques to investigate the biogeographic structure of seabird assemblages, in relation to oceanographic properties.

Our analyses revealed a clear spatial pattern of seabird species composition across the ESPO, with four distinct assemblages spreading from the Humboldt Current (east) to Easter Island (west), and with evidence of some latitudinal separation, associated with the presence of the Desventuradas (north) and Juan Fernández (south) archipelagos (see Supplementary Table S4 for a list of species breeding in the different islands). Three inter-related environmental variables had an important influence on the distribution of these assemblages: SSS, SST, and sea surface chlorophyll α. These dynamic oceanographic characteristics captured a major longitudinal gradient spanning the study area: cold, nutrient-rich, and low salinity waters in the east (Humboldt Current) and warm, nutrient-poor, and high salinity waters in the west (South Pacific Gyre). Moreover, this co-variability was evident in the VIF analysis, which identified cross-correlations among these three variables. Two more variables showed a high importance in the models regarding assemblage composition: Julian day and depth. Julian day captures the seasonality of seabird distributions, which is obvious for species that breed in the study area during the boreal summer (November–April) and disperse across the Pacific Ocean thereafter. Thus, the species-specific breeding and migration cycles influence the overall composition and species richness (see below) of the seabird assemblage in the study area. The influence of depth may be related to onshore–offshore gradients in seabird distributions, as has been previously documented in the broad continental shelves in the Bering Sea (Hunt et al., 2014; Santora et al., 2017) and the South Atlantic (Veit, 1995). While the Chilean continental shelve is relatively narrow, abrupt bathymetric changes (e.g., shelf-breaks and slopes) also structure seabird assemblages, likely due to changes in water flow and prey distributions (Schneider, 1997). In particular, coastal species from the Humboldt Current assemblage may be linked to the shallow waters of the continental shelf, unlike the far-ranging oceanic species.

Previous works have described SST and SSS as important environmental variables influencing seabird assemblages (Hyrenbach et al., 2007; Commins et al., 2014; Force et al., 2015). Moreover, these variables are regularly used to describe water masses, with seabird species distributions often mirroring their changes at macro-scales (Pocklington, 1979; Wahl et al., 1989). However, there is little understanding about the mechanisms underlying these seabird–water mass relationships, because seabirds are not directly consuming the nutrients tightly associated with these physical tracers. Yet, SST and SSS are indicators of oceanographic processes and features that influence the abundance and composition of marine biota, both planktonic and nektonic (Longhurst, 2007). Thus, seabird assemblages likely reflect the distribution of their lower trophic level prey (Ashmole, 1971; Pocklington, 1979). For instance, several characteristic seabird species from the Humboldt Current [e.g., Humboldt Penguin (Spheniscus humboldti), Peruvian Booby (Sula variegata)] forage on schooling fish (e.g., Engraulis ringens, Strangomera bentincki) (Jahncke and Goya, 1998; Herling et al., 2005). Because the distribution of these schooling fishes in the ESPO is restricted to the Humboldt Current, those seabird species highly dependent on this resource would be similarly constrained, thus geographically delineating a distinct seabird assemblage. On the other hand, tropical and subtropical seabirds [e.g., Masked Booby (Sula dactylatra), Red-tailed Tropicbird (Phaethon rubricauda)] are highly dependent on flying fishes (Exocoetidae) and squids (Ommastrephidae) (Carboneras et al., 2019; Orta et al., 2019) that inhabit warm surface waters and are not present in the colder, nutrient-rich waters of the Humboldt Current. Therefore, the distribution of tropical and subtropical seabirds in the ESPO could be driven by the availability of these epipelagic prey and the subsurface predators (predatory fishes and tunas) that drive them to the surface and into the air (Ballance and Pitman, 1999; Spear et al., 2007). To date, different authors have proposed that patterns of prey distribution may drive the biogeographic structure of seabirds over macro and mega-scales (Ainley and Boekelheide, 1983; Abrams, 1985; Sydeman et al., 2010; Sigler et al., 2017). While our findings point in that direction, further research is needed to fully understand how these predator–prey relationships influence species distributions. Many factors can mediate between the abundance and availability of prey, including the degree of prey aggregation, their vertical distribution, and their habitat associations (Benoit-Bird et al., 2011, 2013; Suryan et al., 2016). In particular, our results underscore the importance of productivity (chlorophyll α concentration) as a driver of seabird assemblages. Previous studies documented this same influence in the Pacific and Indian oceans (Pocklington, 1979; Ballance et al., 1997; Ribic et al., 1997). Altogether, these results highlight the need to understand how the structure of marine food webs influences the identity and the abundance of the various seabird prey, especially when many oceanic seabirds have broad diets (Ballance and Pitman, 1999; Spear et al., 2007).

In the same way the spatial pattern in seabird assemblages was related to environmental factors, species richness was also influenced by these factors. Particularly, our study shows that species richness was highly related to Julian day and SST (CV). We used SST (CV) as a proxy for the presence of water mass boundaries and fronts. These areas have been described as important features in the distribution of seabirds via two mechanisms. They can act as areas of primary productivity enhancement and prey concentration, which often leads to higher seabird abundance (Bost et al., 2009; Scales et al., 2014), and they can act as biogeographic boundaries, separating different seabird assemblages (Commins et al., 2014; Force et al., 2015). Daily transects crossing these boundaries would be expected to capture species present in both water masses and along the front, thus yielding high species richness. In particular, areas of higher SST (CV) likely correspond to the western limit of the Humboldt Current, where sharp changes in oceanographic conditions occur (Thiel et al., 2007). Interestingly, because the daily transects belonging to the Humboldt Current group are located close to shore, the area influenced by this current system seems to be very restricted to the coast (≈100 km).

The influence of Julian day on species richness reflects the strong seasonality in the distribution of several species that breed in the area [e.g., Juan Fernández Petrel, Stejneger’s Petrel (Pterodroma longirostris), pink-footed shearwater]. For instance, the Juan Fernández Petrel breeding population during the austral summer, which has been estimated at 1,000,000 pairs (BirdLife International, 2018), migrates to northern latitudes and spends the rest of the year in the North Pacific. Thus, the presence of these breeding species in the study area is highly dependent on the time of the year, which in turn influences the overall species richness in the study area, especially in the vicinity of their breeding islands.

The abundance of seabirds in the study area is related to several environmental factors indicative of a mixture of oceanographic processes. Distance to the colonies and depth (SD) are indicative of onshore–offshore gradients in seabird abundance, as has been suggested as an explanation for the at-sea distribution of seabirds (Ballance et al., 1997; Hyrenbach et al., 2006; Mannocci et al., 2014). The spatial restriction imposed on seabirds during the breeding season has been postulated as one reason behind this relationship, resulting in increased abundances near colonies. Usually, distance to the coast of the breeding colonies (i.e., islands or mainland) is linked to the depth. However, changes in seafloor topography because of ridges or seamounts play an essential role in aggregating top predators, such as tunas and seabirds (Morato et al., 2008, 2010). Two important bathymetric features are present in our study area: the Nazca Ridge and the Easter Seamount Chain (Ray et al., 2012). The importance of depth (SD) in our models could be related to these topographic features in addition to onshore–offshore gradients away from breeding colonies. Further surveys are necessary to clarify the extent of seabird halos around these breeding colonies, and the influence of ridges and seamounts on seabird distribution in the ESPO. The other two variables also important in explaining changes in seabird abundance (mixed layer depth and SSS) are related to the influence of coastal upwelling (Ainley et al., 2005). The thickness of the mixed layer depth is greatly influenced by upwelling of nutrient-rich and cold waters into the mixed layer, affecting localized primary productivity and seabird prey availability (Ballance et al., 1997). In our study area, both variables follow a longitudinal gradient increasing toward the west, as they move away from the Humboldt Current. This gradient is also reflected in the habitat characteristics of the four seabird species clusters, which correspond to four distinct biogeographic domains (Table 4).

The geographical distribution of these four groups, identified by the cluster analysis, revealed a longitudinal and latitudinal spatial pattern of seabird assemblages. Previous biogeographical studies from the area, focusing on coastal ecosystems, have identified similar patterns, assigning four different ecoregions in the area: Humboldt, Easter Island, Juan Fernández, and Desventuradas (Rovira and Herreros, 2016). At both extremes of this longitudinal spatial structure, we could identify two well-defined, spatially divided groups (Humboldt and Easter Island). The NMDS analysis supports this pattern by placing these groups in the opposite extremes of the ordination (related to axis 1). The species composition of these two groups shows virtually no overlap, since no species are shared between them (Supplementary Table S2). However, the other two groups do not show such a clear spatial differentiation, with a more diffuse (Desventuradas) and aggregated (Juan Fernández) geographical distribution around the two archipelagos. In the NMDS ordination space both groups are not differentiated in axis 1, but show a segregation in relation to axis 2, meaning that species composition does differ between both groups. The assemblages identified for both groups using at-sea distributions reflect those assemblages of nesting seabirds in Juan Fernández and Desventuradas, respectively (Figure 2). Both archipelagos are different at the level of their islands’ physiography; the three islands of Desventuradas are relatively small, in the range of 0.25–2.2 km², with maximum heights of 166–479 m, while the three islands of Juan Fernández are in the range of 5–93 km², and maximum heights of 375–1650 m. Despite their proximity, it seems that differences in the physiographic and climatic factors (Bahamonde, 1987; Hajek and Espinoza, 1987) affect the existence of different vegetation in these archipelagos (Hoffmann and Marticorena, 1987), which in turn provides nesting habitat for substantially different breeding seabird assemblages (see Supplementary Table S4). However, at-sea distributional differences become diluted away from the breeding islands, with no clear spatial segregation. This result has important management implications, since conservation actions in the surrounding waters of one archipelago could have implications for seabird communities in both archipelagos.

This study shows that seabird assemblages in ESPO are spatially structured through a large-scale longitudinal gradient that extends from the Humboldt Current System to Rapa Nui. The observed pattern is related to Julian day, ocean depth, and surface levels of chlorophyll, temperature, and salinity. How seabirds respond to environmental factors and how they partition the ocean habitat is crucial to understand their ecology, life-history traits, and conservation threats (Croxall et al., 2012; Lewison et al., 2012). Knowledge of the distribution patterns and assemblage composition during the time they are at sea would be beneficial for declaring and delimiting marine protected areas in the ocean. For example, four large marine protected areas (Nazca-Desventuradas, Mar de Juan Fernández, Crusoe and Selkirk Submarine Mountains, and Motu Motiva Hiva) were recently established in the ESPO, based on criteria related to the existence of a great diversity of invertebrates and fish, some with a high degree of endemism (e.g., Friedlander et al., 2013). However, seabirds were only partially considered in these spatial conservation actions, although they play an essential role supporting the other components of marine biodiversity (e.g., Graham et al., 2018). The results of this study will be useful for decision-makers and political authorities to assess the degree of compliance with the conservation objectives established in the decree creating marine protected areas and recommend actions that allow the integral protection of the biodiversity they contain.

The datasets generated for this study are available on request to the corresponding author.

GL-J and JS designed and conceived the study. JS, MP-T, NL, and DM-U collected the data. JS and KH analyzed the data and wrote first versions of the manuscript. JS, KH, and GL-J corrected and wrote the final version of the manuscript. All authors reviewed and contributed to the writing of the final version of the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.