Inter-specific niche partitioning and overlap in albatrosses and petrels: dietary divergence and the role of fishing discards

Abstract

Although fisheries discards are recognized as a key food source for many seabirds, there have been few thorough assessments of their importance relative to natural prey, and of their influence on the trophic structure of pelagic seabird communities during the non-breeding period. Competition for resources in Procellariiformes appears to be reduced mainly by avoiding spatial overlap, which is supposed to influence diet composition. However, artificial food sources provided by fisheries might relax niche partitioning, increasing trophic niche overlap. Using bycaught birds from pelagic longline fisheries, we combined the conventional diet and stable isotope analyses to assess the importance of fishing discards in the diet of eight species of Procellariiformes. Both methods revealed the high contribution of trawl discards to the non-breeding diet of three neritic species and a moderate contribution in several other species; discards from pelagic and demersal longline fisheries were considerably less important. There was a clear contrast in diets of neritic vs. oceanic species, which are closely related taxonomically, but segregate at sea. Niche partitioning was less clear among neritic species. They showed an unexpectedly high level of diet overlap, presumably related to the large volume of trawl discards available. This is the first study combining the conventional diet and stable isotope analyses to quantify the importance of fishery discards for a community of non-breeding seabirds, and demonstrates how the super-abundance of supplementary food generates high levels of overlap in diets and allows the coexistence of species.

Introduction

Seabirds are important consumers of global marine resources with an estimated annual consumption comparable to total fisheries landings (Brooke 2004; Karpouzi et al. 2007). Fisheries discards, which include non-target and undersized fish, squid, and other invertebrates, used bait, and offal from onboard processing, are a key food source for many scavenging seabirds, including gulls, skuas, gannets, albatrosses, and petrels (Garthe et al. 1996; Oro et al. 1996; Montevecchi 2002). The negative aspects of interactions with fisheries are particularly obvious for albatrosses and large petrels (Procellariiformes), which are recorded as bycatch in substantial numbers (Anderson et al. 2011; Jiménez et al. 2014). However, in the provision of discards as a potentially major dietary component, fisheries may also have a major role in the structuring of seabird communities, with implications for marine food webs, in general.

During the breeding season, seabirds are central-place foragers and competition among sympatric species is thought to be intense (Phillips et al. 2008). As many seabirds disperse during the non-breeding period, competition among species from the same breeding location is much reduced (Bodey et al. 2014); however, there is increased potential for overlap among closely related species from different populations (Phillips et al. 2008; Quillfeldt et al. 2013). Niche partitioning has been proposed as the main mechanism explaining the coexistence of closely related species of seabirds, as competition for resources is reduced by the avoidance of spatial overlap (Nicholls et al. 2002; Frere et al. 2008; Navarro et al. 2009a; Quillfeldt et al. 2013), use of different habitats (i.e., oceanographic features; Waugh and Weimerskirch 2003; Bugoni et al. 2009), and consumption of different prey (Cherel et al. 2002; Connan et al. 2014). Very productive areas with high levels of environmental heterogeneity can facilitate coexistence in closely related species that exhibit some degree of niche divergence (Waugh and Weimerskirch 2003). Fisheries also tend to concentrate in highly productive areas and are known to affect seabird distributions (Ryan and Moloney 1988; Garthe 1997). Because the discards provided by vessels supplement natural food sources, they could be a cause, rather than a consequence, of spatio-temporal overlap in species that might otherwise segregate at sea. Hence, this abundant and predicable source of supplementary food could be the major determinant of trophic structure.

Trawl fisheries account for over 50% of total estimated global discards (Kelleher 2005). Therefore, in continental shelf regions with both high trawl fishing activity and high abundance of scavenging seabirds, we could anticipate very high levels of consumption (Oro et al. 1997). Other fishing techniques that provide fewer discards (e.g., pelagic longlining) might also attract birds, particularly highly pelagic species foraging over extensive, but less productive, ocean regions where prey patches are less predictable. Communities of scavenging seabirds tend to be most speciose in the Southern Hemisphere, and for example, there are globally important populations of many albatrosses and petrels in the southwest Atlantic Ocean which aggregate around different fisheries to feed on discards (Bugoni et al. 2008; Favero et al. 2011; Jiménez et al. 2011). Over shelf waters, 1000 s of birds may feed behind demersal trawl vessels targeting Argentine hake (Merluccius hubbsi) off Argentina and Uruguay, which produce large amounts of discards (Favero et al. 2011; Dirección Nacional de Recursos Acuáticos—DINARA—unpublished data). The oceanic region from the shelf break to high seas influenced by the Brazil-Falkland Confluence is fished extensively by several pelagic longline fleets targeting swordfish (Xiphias gladius), tuna (Thunnus spp.), and sharks (Huang 2011; Jiménez et al. 2016), which produce fewer discards (pieces of large fish, including viscera, and used baits) and vessels attract at most a few 100 seabirds (Bugoni et al. 2008; Jiménez et al. 2011). Over the shelf break and slope, seabirds also feed on discards from the demersal longline fishery for Patagonian toothfish (Dissostichus eleginoides) (Favero et al. 2013; DINARA unpublished data).

Many species, including great albatrosses (Diomedea spp.), mollymawks (Thalassarche spp.), and medium-sized petrels (e.g., Procellaria spp.), some of which are closely related, overlap in the southwest Atlantic. Tracking data indicate some segregation in at-sea distribution and habitat use; northern royal albatrosses (D. sanfordi) tend to forage on continental shelves, and wandering albatrosses (D. exulans) on the shelf slope and deeper waters, whereas Tristan albatrosses (D. dabbenena) mainly forage in warm subtropical areas, and so overlap only with the northern part of the distribution of the wandering albatross (Nicholls et al. 2002; Xavier et al. 2004; Cuthbert et al. 2005; Reid et al. 2013). Ring recoveries and data from fishing vessels suggest that southern royal albatrosses (D. epomophora), like northern royal, also prefer shelf waters (Moore and Bettany 2005; Jiménez et al. 2014). White-chinned petrels (P. aequinoctialis) are distributed mainly in shallow continental shelf and highly productive regions (Phillips et al. 2006), whereas spectacled petrels (P. conspicillata) prefer warm tropical and subtropical oligotrophic and mesotrophic waters (Bugoni et al. 2009). Despite this evidence of niche partitioning, these species overlap extensively in the Brazil–Falkland Confluence region. They are commonly observed feeding on fisheries discards (Jiménez et al. 2011), yet its importance relative to natural prey, and the potential influence on niche overlap, is poorly understood.

Separating naturally captured prey from those obtained through fisheries discards is not always possible, as the same species or groups may be obtained by either means. The importance of fishing discards for seabird species has been addressed using several methods, including the conventional and isotopic studies (Hudson and Furness 1988; Colabuono and Vooren 2007; Navarro et al. 2009b; Bugoni et al. 2010; González-Zevallos and Yorio 2011; Mariano-Jelicich et al. 2014). Each approach has advantages and disadvantages: the conventional assessment allows the identification of many prey species but provides only a brief snapshot of diet and tends to underestimate the importance of easily digested items. The use of stable isotope analysis (SIA) avoids the biases of conventional diet assessments and reflects prey assimilated over different timescales depending on the tissue, but provides limited taxonomic resolution (Barrett et al. 2007; Inger and Bearhop 2008; Karnovsky et al. 2012). In this study, we combined the conventional diet and stable isotope analyses to determine the degree of dietary segregation in albatrosses and petrels in a major resource and biodiversity hotspot, the southwest Atlantic, and in particular to assess the relative importance of fishing discards during the non-breeding period. It was hypothesized that fishery discards allowed overlap in the diet of the different seabirds. Based on the reported spatial and habitat segregation for albatrosses and petrels and the quantity of food supplied by longline vs. demersal trawl fisheries, we also predicted higher trophic niche overlap among species feeding mainly on discards from the latter fishery.

Material and methods

Conventional diet assessment

Samples were obtained from albatrosses and petrels caught incidentally during pelagic longline fishing by Uruguayan commercial and research vessels in 2005–2012 and 2009–2013, respectively, and Japanese commercial vessels in 2009–2011 operating off Uruguay under an experimental fishing license (Jiménez et al. 2010, 2014, 2015a). All vessels fished in shelf break, slope, and deeper waters off Uruguay, and Uruguayan commercial vessels also operated in international waters (Jiménez et al. 2014).

The digestive tracts (esophagus, proventriculus, and ventriculus) of 126 and 27 specimens of six species of albatrosses and two of petrels, respectively, were examined. These included four great albatrosses (Diomedea spp.): northern royal (n = 36), southern royal (n = 23), wandering (n = 12), and Tristan albatrosses (n = 6); two species of mollymawks: black-browed (Thalassarche melanophris, n = 32) and white-capped albatrosses (T. steadi, n = 17); and white-chinned (n = 19) and spectacled petrels (n = 8). Species of great albatrosses were identified in the laboratory; northern royal and southern royal albatrosses were distinguished by their plumage, and wandering albatrosses were separated from Tristan albatrosses by a morphometric discriminant function (Cuthbert et al. 2003). White-capped albatrosses were identified by molecular analysis (Jiménez et al. 2015b). Birds were captured on the shelf break and deeper waters off Uruguay and in adjacent international waters, and a few in international waters off southern Brazil (details in Appendix S1, Fig. S1.1). All the Thalassarche and Procellaria spp. were bycaught during the non-breeding season (April–November). Because the breeding season of great albatrosses lasts almost a year and adults take a sabbatical year before breeding again, both breeding and non-breeding seasons overlap temporally. Only two Tristan albatrosses (with unfeathered brood patches captured in January) and two wandering albatrosses (ringed birds caught in late July and early September; late chick-rearing) were breeding birds. Based on gonadal examination, a few other birds (a wandering albatross and two spectacled petrels) were about to breed.

Frozen bycatch specimens were thawed and the digestive tract removed. Food items found in the esophagus, proventriculus, and ventriculus were examined separately. Few undigested or partially digested prey were found, so most of the data on diet composition are based on cephalopod beaks and fish otoliths. The cephalopod beaks were identified using Xavier and Cherel (2009) and checked against reference collections held at the University of Coimbra or the Centre d’Etudes Biologiques de Chizé. As the ratio of upper:lower beaks frequently differs from unity (Xavier et al. 2011), the number and species identification of individual squid was determined from the 709 lower beaks (611 upper beaks were also collected). The lower rostral length (LRL) and the lower hood length (LHL) of the squid and octopod beaks (except those highly eroded), respectively, were measured to ±0.01 mm with a digital vernier calliper, and allometric equations used to predict mantle length and mass (Xavier and Cherel 2009). Wet mass was reconstructed separately from all measurable lower beaks (fresh and old) and then only for fresh beaks. The former was used to compare the cephalopod diet between species, and the latter to estimate the proportion by fresh mass represented by each cephalopod species and the cephalopod component in the overall diet, including fish. Fresh beaks were those with wings and lateral walls in near perfect condition, often with buccal muscle attached (Xavier and Cherel 2009). Because squid and octopod beaks can accumulate in the stomach (mainly the ventriculus) for long periods (Furness et al. 1984), the abundance and reconstructed mass of these accumulated beaks represent cephalopods ingested in the previous weeks or months. In contrast, fresh beaks represent cephalopods ingested in the last few days (Spear et al. 2007).

Fish prey were identified in almost all cases from otoliths using the available literature (Reid 1996; Volpedo and Echeverría 2000; Rossi-Wongtschowski et al. 2014, DINARA unpublished), online catalogues (http://www.cmima.csic.es/aforo/) and fish samples collected from different fisheries by DINARA. The number of fish present in each sample was estimated from the number of paired otoliths of similar sizes, and often from unpaired otoliths. Otolith length and width were measured to ±0.01 mm and fish standard length and mass estimated using regression equations; where no relationships were available, regression equations were used from closely related species or subspecies (Wöhler 1997; Waessle et al. 2003; Morley and Belchier 2002; González-Zevallos et al. 2010; Crec’Hriou et al. 2015). Otoliths are digested faster than squid beaks by seabirds and the level of erosion depends on their shape, size, thickness, and time since ingestion. Experiments in captivity (Furness et al. 1984; Casaux et al. 1995) and dedicated stomach content analyses (Spear et al. 2007) show that otoliths can be completely digested from within one to several days. Thus, otoliths were classified into four categories of erosion: (1) no sign of wear or digestion, (2) slight, with some signs of smoothing of margins (3) moderate, and (4) heavily eroded with margins rounded or broken, or with a considerable part of the otolith missing after fracture. Otoliths in categories 1–3 were measured to reconstruct the fish mass. A correction factor was applied to compensate for erosion: 10% for otoliths from category 2 and 20% for those in category 3 (Reid 1995; Arata et al. 2004). In some stomachs, remains of fresh fish (skeletal material with attached flesh) were found, but otoliths or other distinctive structures to identify the species were absent, in which cases, the occurrence of one unknown fish was assumed.

The analyses of conventional diet took account of the time scales represented by different types of sample. The overall species composition of the cephalopod and fish components was determined separately. The frequency of occurrence, number, and reconstructed mass of all accumulated beaks (fresh and old) were considered to represent the cephalopod diet during the non-breeding period (few breeding birds were sampled). In contrast, fish species identified from otoliths (including eroded) were considered to represent the fish component of the diet in the last week. Data just from the fresh cephalopod beaks and fish remains (including whole prey and otoliths in categories 1–3) were assumed to represent the diet in recent days, and used to estimate the relative contribution of each component and species, reducing the biases caused by different digestion rates. Reconstructions of fresh mass were performed for all but one species with at least 12 stomachs sampled. This exception was the black-browed albatross, because the stomachs held a considerable number of unidentified fresh fish remains (see below), and two crustaceans for which the original wet mass could not be assessed, and hence, any reconstruction of the overall diet would have led to an overestimation of the importance of cephalopods.

We employed hierarchical cluster and non-metric multidimensional scaling (MDS) analyses in the software Primer (version 6) to analyse similarities in the diet between seabird species (Clarke and Gorley 2006). These were based on the Bray–Curtis similarity index and used the default parameters for the software. Results of the cluster analyses were integrated in the MDS output to produce ellipses of similarities. We arbitrary set two similarity thresholds: >50% and >75%, which would represent a moderate or high overlap in diet among species. This procedure was conducted three times, first with a matrix of presence–absence of all the identified prey by species to compare the non-breeding diet using as many prey as possible. Second and third, we conducted the analyses with the percentage of wet mass reconstructed from fresh prey (see above) both by species and by component (cephalopod, demersal fish, and pelagic fish), respectively. Only one fresh beak belonging to a deep-sea squid (fam. Mastigoteuthidae) was recorded, so the cephalopod component was not classified according to depth preference.

Stable isotope analysis

As tissue samples were not available for all individuals for which stomach contents were examined, these were also taken from birds captured incidentally during the same trips and in the same fishing area (see Appendix S1, Fig. S1.2; Jiménez et al. 2014, 2015b). Stable isotope ratios were analysed in eight birds per species (except for Tristan Albatross; see below). Samples from southern royal, northern royal, white-capped, and black-browed albatrosses, and most white-chinned petrels were selected at random from birds incidentally captured in the same year (April–November 2009). Spectacled petrels and wandering albatrosses are captured incidentally in low numbers, and so, samples for SIA were from four and five birds from the conventional diet analysis and additional four samples from 2011 and three from 2008 to 2009, respectively. The only four samples available for Tristan albatross were from the non-breeding birds included in the conventional diet assessment. Overall, 33% of the birds from the conventional diet assessment were represented in the isotopic analysis.

Seabirds usually replace feathers during the non-breeding season. Not all body plumage is replaced annually and so includes feathers that may have been grown in multiple non-breeding seasons (Ginn and Melville 1983). To focus on the diet in the southwest Atlantic, we sampled growing feathers from the head (new feathers less than two-thirds grown, with remains of waxy sheath at its base; scores 2–4 of Ginn and Melville 1983). An individual body feather is presumed to take at least 2–3 weeks to grow 6–10 cm long (Jaeger et al. 2010); therefore, the isotope ratios of nitrogen (15N/14N; δ15N) and carbon (13C/12C; δ13C) of our sample of growing feathers from the head (from several mm in petrels to few cm in some albatrosses) might represent the prey consumed over a shorter period. About 2–4 (albatrosses) and 4–8 (petrels) growing feathers were pooled for each bird. Feathers were cleaned of surface contaminants using 2:1 chloroform:methanol followed by a methanol rinse, air dried and then homogenized by cutting into small fragments with stainless steel scissors (Cherel et al. 2013).

SIA was also carried out on fish and squid tissues obtained from stomach contents or from fishing vessels (see Appendix S2, Table S2.1). This included samples of muscle from large pelagic fishes, including tuna (Thunnus alalunga = 6, T. albacares = 4), swordfish (Xiphias gladius = 5) and blue shark (Prionace glauca = 4), small pelagic fishes (Scomber sp. = 1, Trachurus sp. = 2), demersal fishes (Merluccius hubbsi = 3, Bassanago albescens = 5, Macrourus carinatus = 2), and from squid baits (Illex argentinus = 3), and of squid beaks (fresh or slightly eroded) from Histioteuthis eltaninae (n = 5), H. atlantica (n = 5), H. macrohista (n = 3), and Illex argentinus (n = 3). All prey, except large pelagic fishes, were represented in the conventional diet samples. Identifiable remains from large pelagic fishes are unlikely to be represented in stomach contents, because scavenging birds feed mainly on soft body parts and viscera, and are unable to swallow the head (hence the absence of otoliths). Squid beaks and muscle samples were oven-dried over 24 h at 60 °C and ground to a fine powder. For analysis of δ¹³C, lipids were extracted from subsamples of muscle (~50 mg) placed in filter paper envelopes using 2:1 chloroform:methanol and ultrasound for 1 h. This procedure was repeated twice and then subsamples were oven-dried at 40 °C for 24 h (Colabuono et al. 2014). Muscle subsamples of prey with and without the lipids extracted were analysed separately. Carbon and nitrogen isotope ratios for samples of seabird feathers and prey (i.e., beaks and muscle) (~0.7 mg weighed within tin capsules) were measured by isotope ratio mass spectrometry using a Flash EA 112 elemental analyzer coupled with a Delta Plus isotope ratio mass spectrometer (Finnigan MAT) at the Stable Isotope Laboratory of Facultad de Agronomía (Universidad de la República, Uruguay). Stable isotope values are presented in delta (δ) notation in units of parts per thousand (‰), δX = [(R _sample/R _standard) − 1], where R is the ratio of the heavy to the light isotope of element X. Standards for this equation were the stable isotope values of Vienna Pee Dee Belemnite (PDB) and atmospheric nitrogen for δ¹³C and δ¹⁵N, respectively. Analytical error based on repeated measurements of internal standards was of 0.2‰ (SD) for both nitrogen and carbon.

A Bayesian multisource stable isotope mixing model (SIAR: Stable Isotope Analyses in R; Parnell et al. 2010) was used to estimate the ranges of probable contributions of each prey group to the diet of each seabird species. Diet-tissue discrimination factors for albatross or petrel feathers have not been published (see Bond and Jones 2009). However, Caut et al. (2009) reviewed trophic discrimination factors (TDFs) for various taxa, including birds, and calculated mean values in bird feathers of 2.16 ± 1.52‰ (SD) for ¹³C and 2.37 ± 1.13‰ for ¹⁵N, which were, therefore, the values we assumed in all our models. Soft parts, which represent the bulk of the squid consumed by birds (mainly mantle and arms), are consistently highly enriched in ¹⁵N over beaks, but slightly depleted in ¹³C (Hobson and Cherel 2006). Thus, δ¹⁵N values from our samples of lower beaks were converted into those expected for mantle tissue. For Illex argentinus, the values of δ¹⁵N and δ¹³C in muscle were assumed to be 5.2‰ higher and 0.8‰ lower than lower beaks, respectively, based in data from Illex coindetii (Hobson and Cherel 2006). For Histioteuthis spp., we assumed that soft body parts were enriched by 4.8‰ in ¹⁵N and depleted by 0.8‰ in ¹³C over lower beaks using an average estimate for several cephalopod species (Hobson and Cherel 2006). After this correction, lower beaks of I. argentinus showed similar isotope ratios to those in the mantle of longline bait of this species; therefore, they were pooled as a single group for further analyses. Bayesian mixing models were employed using a non-informative Dirichlet prior distribution, with zero concentration dependencies, and 5 × 10⁵ iterations, thinned by 15, with an initial discard of 5 × 10⁴ interactions in the MCMC estimation. A total of six potential prey sources were included in the model: Histioteuthis spp., Illex, small pelagic fish, pelagic longline fish (including all large pelagic fishes), demersal trawl fish (M. hubbsi and B. albescens), and demersal longline fish (M. carinatus). The species from the last three groups are known to be the discards from the respective fisheries. The trawl fishery produces the largest amount of discards, comprising undersized hake, non-target demersal fish (in Uruguay and northern Argentina, this includes B. albescens and other species that can be fully discarded; and Helicolenus dactylopterus, Cheilodactylus bergi, and other species that may be retained), squid (mainly I. argentinus, which is largely retained), and other invertebrates, and offal from onboard processing (DINARA unpublished data). Both longline fisheries provide substantially fewer discards. In the pelagic longline fishery, the discards comprise pieces of swordfish, tuna, and sharks, including their viscera, and parts or whole individuals of other large fishes, and the used baits. Discards from the demersal longline fishery are mainly non-target demersal fish, including Macrouridae, and offal. None of these demersal and large pelagic fishes are expected to be obtained naturally in substantial amounts. The cephalopods, including Histioteuthis spp., are thought to be taken directly by the seabirds (either caught close to the surface, or scavenged postmortem) (Croxall and Prince 1996; Cherel and Klages 1998), but I. argentinus may also be made available to surface predators either as a discard from the trawl fishery (see above and Discussion) or as used baits discarded from pelagic longline vessels. Squid beaks from stomachs of fish, including tuna, can become available to seabirds as offal discarded from fishing vessels (Vaske Júnior 2011); however, the limited information available suggests that this is unlikely to be a major source of beaks found in the stomachs of scavenging seabirds (Xavier et al. 2013). Some of the small pelagic fishes could also be used baits (see “Results” section), although they are also naturally available. Similar cluster hierarchical and MDS analyses to those conducted for the conventional diet analyses were performed using the mean values of raw δ¹³C and δ¹⁵N values for seabirds, and the median values for the contribution of each prey for the different albatross and petrel species.

Bayesian ellipses (SIBER: Stable Isotope Bayesian Ellipses in R, Jackson et al. 2011) were used to describe the isotopic niche space occupied by the different seabird species during the non-breeding season. The Standard Ellipse Area (SEA) encompassing 40% of the data after small sample size correction (SEAc) was used to estimate the percentage overlap (regarding the smaller SEAc of each species pair) among isotopic niches. The posterior estimates of the Bayesian Standard Ellipse Area (SEAB) were used to compare isotopic niche widths between species.

Results

Dietary analysis

A total of 40 prey species were identified in 153 stomachs of the eight species of albatrosses and petrels: at least 28 species of cephalopods (squid and Octopoda), 12 of fishes, and 1 of crustaceans. Crustacea were only found in a black-browed albatross (remains of two Chaceon notialis) and two northern royal albatrosses (fragments of unidentified exoskeleton).

Cephalopod component of the diet

At least nine species of squid were identified in the stomachs of Tristan albatrosses, representing those ingested in recent weeks or months of the non-breeding period. The most important family by number and mass was Histioteuthidae (mainly Histioteuthis bonnellii corpuscula), followed by Chiroteuthidae (Table 1). Thirteen species of squid were determined in the stomachs of wandering albatrosses, of which the most important by number was Histioteuthidae and by mass was Onychoteuthidae, represented by just two individual Kondakovia longimana, and Histioteuthidae, of which H. eltaninae and H. atlantica were the most important (Table 1). A total of 10 and 12 species of squid were found in the diet of southern and northern royal albatrosses, respectively, of which I. argentinus (Ommastrephidae) was by far the most important species (Table 2). The second most important by mass was Onychoteuthidae, represented by a few individual Moroteuthis spp. in southern royal albatross, and Histioteuthidae in northern royal albatross (Table 2). Octopoda were only recorded in the diet of southern and northern royal albatrosses (Table 2).

Table 1 Frequency of occurrence, numbers, and reconstructed wet mass for all the cephalopods ingested (fresh and old material combined) by Tristan and wandering albatrosses

Table 2 Frequency of occurrence, numbers, and reconstructed wet mass for all the cephalopods ingested (fresh and old material combined) by southern and northern royal albatrosses

Six and three species were identified in the stomachs of black-browed and white-capped albatrosses, respectively (Table 3). Similar to the royal albatrosses, the most important squid for white-capped albatross was I. argentinus, both by numbers and mass, whereas Histioteuthidae were the most common squid (mainly H. atlantica) consumed by black-browed albatrosses (Table 3). Very large numbers of lower squid beaks were found in the stomachs analysed of white-chinned and spectacled petrels, of which 69.3 and 48.9% were highly eroded, respectively (Table 4). A total of eight species were identified for white-chinned petrels, of which the most important were Histioteuthidae. Nine squid species were present in spectacled petrel stomachs, with Lycoteuthis lorigera (Lycoteuthidae) the most abundant, but representing just 4.3% of wet mass, partly because almost all beaks were eroded, preventing mass reconstruction. In contrast, five of six lower beaks of I. argentinus were measurable, and hence, this squid was the most important by reconstructed mass (Table 4).

Table 3 Frequency of occurrence, numbers, and reconstructed wet mass for all the cephalopods ingested (fresh and old material combined) by black-browed and white-capped albatrosses

Table 4 Frequency of occurrence, numbers, and reconstructed wet mass for all the cephalopods ingested (fresh and old material combined) by white-chinned and spectacled petrels

Fish component of the diet

No fish was recorded in Tristan albatross stomachs, and just one unidentified fish was found in a spectacled petrel. Most of the identified fish species in the other albatrosses and petrels were demersal, and a few were pelagic and small; some of the latter were confirmed to be longline baits (fresh, and located at the esophagus or attached to a hook) (Table 5). Only two Macrouridae and a bait fish, Sardinops sagax, were found in wandering albatrosses. The main fish species eaten by the two royal albatross species were B. albescens and M. hubbsi (Table 5). Macrouridae and a few Urophycis cirratus were also found in both albatrosses. Only three (one Macrouridae, one M. hubbsi, and one Scomber sp. bait) of the 17 fish found in black-browed albatross were identified. The fish diet of white-capped albatross was dominated by B. albescens. Finally, M. hubbsi and B. albescens were the most important fish by number for white-chinned petrel, which had also consumed the demersal fish Cynoscion guatucupa and H. dactylopterus, and some pelagic fishes that were longline baits (Table 5).

Table 5 Frequency of occurrence and percentage of the total numbers of fishes found in the stomach content analysis of wandering (WA), southern royal (SRA), northern royal (NRA), black-browed (BBA) and white-capped (WCA) albatrosses, and white-chinned petrel (WCP)

Dietary overlap and segregation

Using only fresh material, we were able to reconstruct the fresh mass of each component in the diet of five seabird species. The diets of all species, except wandering albatross (reconstructed mass: cephalopods 80.2% and fish 19.8%), were dominated in mass by fish (Table 6). The relative importance of fish (~80%) and cephalopods (20%) was similar in the closely related southern and northern royal albatrosses; in the former, these were largely the demersal fishes B. albescens, M. hubbsi, and Macrouridae species. Fish were even more important for white-capped albatross and white-chinned petrel (91.2 and 89.6%, respectively), of which the main species by mass were B. albescens, or both M. hubbsi and B. albescens, respectively (Table 6). Illex argentinus was usually the most important fresh cephalopod in diets (Table 6).

Table 6 Fresh cephalopod and fish contribution in the dietary analysis of wandering (WA), southern royal (SRA), northern royal (NRA), and white-capped (WCA) albatrosses and white-chinned petrel (WCP)

The cluster and MSD results underlined the similarity in diet between some species, particularly the royal albatrosses (Fig. 1). Based on the presence–absence of each prey species, northern and southern royal albatrosses showed the greatest, and Tristan albatross and spectacled petrel showed the least overlap among species pairs (Fig. 1a). Based on reconstructed wet mass (Fig. 1b), there was extensive dietary overlap between the two royal albatross species, and a moderate overlap between these species and white-capped albatross. However, the diet of both royal albatrosses showed greater overlap with white-capped albatross and white-chinned petrel if comparing the reconstructed wet mass by component (fish vs. squid; Fig. 1c). The most distinct diet using these various metrics was that of the wandering albatross because of the predominance of squid (Fig. 1c).

Fig. 1

Left non-metric multidimensional scaling (MDS) with superimposed clusters at similarity levels of 50% (grey) and 75% (black). Right hierarchical clustering (Bray–Curtis similarity). a Presence–absence data of all squid and fish species identified in the stomach content analysis of eight species of albatross and petrels. b Percentage of reconstructed wet mass of squid and fish species found in the stomach content analysis of the five main species. c Percentage of reconstructed wet mass for the three main components (squid, demersal fish, and pelagic fish) found in the stomach content analysis of the five main species. SRA southern royal albatross, NRA northern royal albatross, BBA black-browed albatross, WCA white-capped albatross, WCP white-chinned petrel, SP spectacled petrel, WA wandering albatross, TA Tristan albatross

Stable isotope analyses

Isotopic ratios of growing feathers ranged from the mean (±SD) values of both δ¹³C and δ¹⁵N of −16.67 ± 0.67‰ and 15.14 ± 0.67‰, respectively, in spectacled petrel, to −15.55 ± 0.98‰ and 18.47 ± 1.09‰ in northern royal albatross (see Appendix 2, Table S2.1). There were significant differences among species in feather δ¹³C and δ¹⁵N values using a multivariate ANOVA (Wilks’ λ, F _14,102 = 5.29, P < 0.05), but in univariate tests, the differences were significant in δ¹⁵N (ANOVA, F _7,52 = 10.99, P < 0.05), but not in δ¹³C (ANOVA, F _7,52 = 1.862, P > 0.05). Post-hoc tests (Tukey tests < 0.05) indicated several pairwise differences in δ¹⁵N values: northern royal albatross > all species, except southern royal and black-browed albatrosses; southern royal and black-browed albatrosses > spectacled petrel and wandering albatross; black-browed albatross > Tristan albatross; white-capped albatross > spectacled petrel.

Niche overlap and segregation

Levels of niche overlap also varied between the seabird species (Table 7; Fig. 2a). Excluding Tristan albatross because of the small sample size, the corrected Standard Ellipse Area (SEAc) ranged between 0.914‰² in wandering albatross and 3.380‰² in white-capped albatross (Table 7). The overlap between isotopic niches of the great albatross species (from SEAc estimates) was low (southern royal overlapped with 10% of the SEAc of northern royal) or null (wandering vs. both northern and southern royal albatross). In addition, wandering albatross showed a significantly narrower niche width (SEAB, P < 0.05) than either royal albatross species, whereas these two species showed similar niche widths (Fig. 2b). White-capped and black-browed albatrosses showed considerable niche overlap (52%, Table 7) and similar niche widths according to SEAB (p > 0.05; Figs. 2a, b). The isotope niches of the Procellaria petrels differed substantially; there was a little overlap (15%, Table 7), and the niche width of white-chinned petrel was significantly wider than in spectacled petrel (Fig. 2b). Relative differences between species appeared to reflect habitat preferences; the two most oceanic species (wandering albatross and spectacled petrel) showed the narrowest niche widths (Fig. 2b) and least overlap (Fig. 2a; Table 7) with most of the neritic species. No differences were detected between the niche width of black-browed albatross and any other species (Fig. 2b). Furthermore, the isotopic niches of all neritic species overlapped extensively with the SEAc of black-browed albatross (35–78%, Table 7; Fig. 2a).

Table 7 Corrected standard ellipse area (SEAc; ‰²) and overlap of SEAc between species pairs in ‰² and percentage of the smaller SEAc of the species pair

Fig. 2

a Stable carbon and nitrogen values of re-growing feathers and Standard Ellipse Area (SEA) estimated using Stable Isotope Bayesian Ellipses in R (SIBER, Jackson et al. 2011) for seven species of albatross and petrels (eight sample per species). b Density plots of the Bayesian Standard Ellipse Area [SEAB, mean (black dots) and their credibility intervals 50, 75, and 95%]. SRA southern royal albatross, NRA northern royal albatross, BBA black-browed albatross, WCA white-capped albatross, WCP white-chinned petrel, SP spectacled petrel, WA wandering albatross

Mixing model

The results of the Bayesian stable isotope mixing model showed some differences in the contribution of each of the six prey groups to the diet, although credibility intervals (CI) overlapped in most cases (Fig. 3; Appendix 2, Table S2.2). Prey groups overlapped to some extent in their δ¹³C and δ¹⁵N values, particularly demersal trawl fish and Illex (Appendix 2, Table S2.1). The Bayesian mixing model indicated a scenario with a greater contribution of demersal trawl fish and Illex to the diet of southern royal (mean and 95% CI = 24%, 6–42% and 23%, 5–42%, respectively), northern royal (23%, 5–41% and 23%, 5–43%, respectively), and black-browed (27%, 8–48% and 26%, 7–43%, respectively) albatrosses than the other four potential prey sources. In decreasing order of importance, the relative importance of the latter to the diet of the three albatrosses was as follows: demersal longline fish, Histioteuthis, and pelagic longline fish, followed by small pelagic fish (Fig. 3, Appendix 2, Table S2.2). Results for white-capped albatross were similar in some respects, with a greater contribution in the diet of both demersal fish and Illex (20%, 4–35% and 19%, 4–35%, respectively) but with somewhat higher proportions of the remaining groups, particularly pelagic and demersal longline fishes (Fig. 3, see Appendix 2, Table S2.2). In contrast, the model indicated a greater contribution of Histioteuthis (30%, 17–42% and 31%, 16–45%, respectively) and small pelagic fish (26%, 8–44% and 29%, 8–51%, respectively), followed by demersal longline fish and pelagic longline fish, to the diet of wandering albatross and spectacled petrel (Fig. 3, see Appendix 2, Table S2.2). Demersal trawl fish and Illex were of little relevance, as expected for these two predominately oceanic seabirds. Histioteuthis (23%, 7–37%) was important for white-chinned petrels. However, in clear contrast to the spectacled petrel, white-chinned petrels consumed more demersal trawl fish and Illex (17%, 2–32% and 17%, 2–33%, respectively) (Fig. 3, see Appendix 2, Table S2.2). Finally, the diet of Tristan Albatross was based on very few birds sampled, and thus, the mixing model output should be viewed with caution; nevertheless, it suggested a similar order of importance of the six prey groups as in wandering albatross (Histioteuthis and small pelagic fish, followed by demersal and pelagic longline fishes, and Illex and demersal trawl fish), but with smaller differences between groups due to the high overlap among credibility intervals (Fig. 3, see Appendix 2, Table S2.2).

Fig. 3

Range of possible proportions of six prey groups in the diet of eight species of albatrosses and petrels. Histioteuthis spp. are likely to be obtained naturally (mostly), Illex argentinus as trawl discards, natural prey or longline baits, small pelagic fishes as natural prey or longline baits. Decreasing bar widths represent 5, 50, 75, and 95% Bayesian credibility intervals computed by Stable Isotope Analysis in R. SRA southern royal albatross, NRA northern royal albatross, BBA black-browed albatross, WCA white-capped albatross, WCP white-chinned petrel, SP spectacled petrel, WA wandering albatross, TA Tristan albatross

The results of the cluster and MSD analyses from both stable isotope outputs (i.e., mean δ¹³C and δ¹⁵N values, and posterior estimates of the Bayesian model) indicated similarities in diet among some species pairs (Fig. 4a, b). The highest overlap was in the diets of both royal albatrosses and black-browed albatross, which clearly contrasted with the diets of wandering albatross and spectacled petrel (Fig. 4a, b). The diets of white-capped albatross and white-chinned petrel were similar to each other, and most like the diets of the royal and black-browed albatrosses. The diet of Tristan albatross was closest to white-capped albatross and white-chinned petrel (Fig. 4a, b).

Fig. 4

Left non-metric multidimensional scaling (MDS). Right hierarchical clustering (Bray–Curtis similarity). a Mean values of δ¹³C and δ¹⁵N. b Mean posterior estimates of the Bayesian stable isotope mixing model for the proportion of the six prey groups (Histioteuthis spp., Illex argentinus, pelagic longline fish, small pelagic fish, demersal trawl fish, and demersal longline fish) in the diet of eight species of albatross and petrels. Superimposed clusters in MDS are at similarity levels of 98% (grey) and 99% (black) in (a) and 75% (grey) and 95% (black) in (b). SRA southern royal albatross, NRA northern royal albatross, BBA black-browed albatross, WCA white-capped albatross, WCP white-chinned petrel, SP spectacled petrel, WA wandering albatross, TA Tristan albatross

Discussion

To our knowledge, this is the first study to use a combination of conventional diet and stable isotope analyses to examine the role of fishing discards in structuring trophic relationships within a seabird community. The conventional diet assessment provided a robust taxonomic identification of the prey consumed by several albatross and petrel species in the southwest Atlantic, including, for the first time, migrants from colonies in the New Zealand region (white-capped, northern royal, and southern royal albatrosses). The conventional and isotopic approaches provided complementary information, overcoming some of the biases or limitations of the individual methods, to reveal the major importance of trawl discards in the non-breeding diet of three species (southern and northern royal, and black-browed albatrosses), moderate importance in two species (white-capped albatross and white-chinned petrel), and minor relevance in two others (wandering albatross and spectacled petrel). Although the study was of birds killed on pelagic longliners, discards from this type of fishery, or from demersal longlining, contributed substantially less than other sources to their diets, according to the stable isotope mixing model. This indicates that our sample was not biased to birds specialised in pelagic longline discards. Moreover, the mean and range of stable isotope ratios measured in feathers were the same as in random samples of these same seabird species taken at colonies (see below), and therefore, there is no reason to consider that the birds that we sampled were atypical of these populations. Segregation in the diet and isotopic niche between neritic and oceanic species was confirmed. The availability of abundant resources in shelf waters, partly supplied by trawl fishing, appears to reduce the degree of segregation among neritic species pairs; this was especially evident for both royal albatrosses, which had similar diets based on stomach contents, although their isotopic niches were more distinctive (see below). Our results applied to the non-breeding season, which are of particular relevance as most natural mortality in seabirds occurs during the winter, when the diet of pelagic species, such as the Procellariiformes, is much more difficult to study than when they attend colonies during breeding.

Relevance of fishing discards

In this study, the conventional analysis of stomach contents proved to be effective for accurately identifying prey species, but not for quantifying the importance of large pelagic fishes obtained as discards. However, use of the Bayesian stable isotope mixing model with these fish included as a potential food source revealed their importance relative to other prey groups. Bayesian stable mixing models are sensitive to the initial inputs, especially to the trophic discrimination factors (TDFs) (Bond and Diamond 2011). Due to the lack of TDF values for Procellariiformes, we used TDFs for bird feathers that are an average across many different taxa and a wide SD to allow for the uncertainty (Caut et al. 2009). Therefore, we acknowledge that caution is required when interpreting our estimates of the contribution of prey groups. In addition, this method may not provide reliable estimates of the relative contribution of groups with similar isotopic signatures (Illex vs. demersal fish, see below), for which stomach contents provided better resolution. Furthermore, the proportion of each prey varied according to the number of sources. We acknowledge the small sample size for stable isotope ratios of some prey groups, and overlap in the credibility intervals between groups in the mixing model. Given these biases, similar results were not expected from the two methods of quantifying the number of prey groups consumed and their proportion in the diet. Indeed, the estimates for the proportions of each prey group varied between methods (Fig. 5, see below). However, the combined results from the conventional and isotopic approaches provide a compelling indication of the main prey in the diet of the study species, with the possible exception of Tristan albatross for which sample sizes were the smallest. The diet of wandering albatross and potentially that of spectacled petrel (see below) included the highest contributions of prey that are likely to be taken naturally (Histioteuthis spp. and probably many of the small pelagic fish). However, the diet of northern royal, southern royal, black-browed and white-capped albatrosses, and white-chinned petrel were dominated by prey obtained artificially, i.e., from fisheries discards.

Fig. 5

Comparison of the conventional analysis of stomach contents (% of the reconstructed fresh mass, see Table 6) and the Bayesian stable isotope mixing model for the five species in which we could reconstruct the wet mass of the prey. Mixing models outputs are the range of possible proportions of four and five prey groups in the diet of each study species. Blue shark was used as pelagic longline fish (marked with an asterisk) and demersal trawl fish and Illex were pooled as a single source in the models with four and five prey groups, respectively (see “Discussion” section for specifications). Decreasing bar widths which represent 5, 50, 75, and 95% Bayesian credibility intervals computed by Stable Isotope Analysis in R. Histioteuthis spp. are likely to be obtained naturally (mostly), Illex argentinus as trawl discards, natural prey or longline baits, small pelagic fishes as natural prey or longline baits. WA wandering albatross, SRA southern royal albatross, NRA northern royal albatross, WCA white-capped albatross, WCP white-chinned petrel

Based on reconstructed wet mass from the conventional diet, fishing discards from trawl fisheries are a major food source for four of our study species in the southwest Atlantic during their non-breeding season (southern and northern royal, and white-capped albatrosses, and white-chinned petrel). There is a major trawl fishery for Argentinean hake operating over the Argentinean and Uruguayan shelf (mainly from 50 to 200 m isobaths), where these seabirds are relatively abundant (Favero et al. 2011; Jiménez et al. 2015c; DINARA unpublished data). Undersized hake (landing of hake <35 cm is not allowed) are the most likely source of food, which is also supported by the length of the hake estimated from otoliths found in stomach contents of the bycaught birds (mean ± SD of reconstructed hake length = 25 ± 8 cm, n = 32; this study). Bassanago albescens is highly abundant in deep shelf waters from 35° to 42°S (Figueroa and Ehrlich 2006; García et al. 2010), is discarded only in trawl fisheries, and is more buoyant than other discarded demersal fish so will remain closer to the surface and more accessible to birds behind vessels (Seco Pon 2014). Another prey likely to be obtained as discards is I. argentinus; this is the most important squid captured by fisheries in the shelf waters off southern Brazil, Uruguay, and Argentina, which is abundant in our study area mainly during winter (Sacau et al. 2005), and discarded in large amounts from the trawl fishery for hake (Seco Pon 2014). However, it can also be obtained naturally, and so the proportions that are taken directly by the seabirds (either caught close to the surface, or scavenged postmortem), obtained as pelagic longline baits, or as trawl discards, are unknown, although this last source is likely to be major.

Another potentially relevant demersal trawl fishery in the study region targets white mouth croaker (Micropogonias furnieri), in particular, and operates largely in coastal areas of the Rio de la Plata within the Argentinean–Uruguayan Common Fishing Zone (ZCPAU). However, we did not find evidence of this fish in the conventional diet assessment of any seabird species and it was not, therefore, included as a potential prey source in the mixing models. Indeed, there was no evidence of any coastal prey in the diet of the sampled seabirds (see Tables 1, 2, 3, 4, 5), except for a single Cynoscion guatucupa in the stomach of a white-chinned petrel.

Stable isotope analyses supported that both demersal trawl fish (M. hubbsi and B. albescens) and I. argentinus were major items in the diet, particularly of southern and northern royal albatrosses, but also white-capped albatross and white-chinned petrel (see below). Furthermore, although we could not reconstruct the diet of black-browed albatross by wet mass because of the unidentified fish, the stable isotope results suggested that its diet was close to that of both royal albatross species and that discards from the trawl fishery were a main food source. This is supported by the high abundance of black-browed albatrosses, which in winter can reach >1000 birds behind a single trawler (Favero et al. 2011; Jiménez et al. 2015c). It should be noted that two sources might have the same contribution in a stable isotope mixing model if they occupy the same isotope space. Hence, the similarity in isotopic signatures of demersal trawl fish and Illex (see Appendix 2, Table S2.1.) could explain the differences in relative importance between the conventional diet and Bayesian mixing models. Nevertheless, demersal trawl fish and Illex are clearly of major importance to both species of royal albatrosses and black-browed albatross, as supported by the Bayesian mixing model including demersal trawl fish and Illex combined as a source (Fig. 5; Appendix 2, Fig. S2.1), and by a previous study of black-browed albatrosses (Mariano-Jelicich et al. 2014). Note that this mixed source—demersal fish and Illex—encompasses the bulk of the discards of this particular fishery. A Bayesian mixing model with our original six sources, but including isotopic data from Illex and M. hubbsi sampled in our study area (Saporiti et al. 2015), further supported our conclusion that these two prey, particularly M. hubbsi, were important in the diet of both species of royal albatrosses, and black-browed albatross (Fig. S2.2).

The Bayesian mixing model output indicated that discards from pelagic and demersal longline fisheries did not predominate in the diet of any species. Across the study area, discards from the demersal longline fishery for toothfish consist of offal from onboard processing, and non-target demersal fish, including Macrourus spp. In terms of reconstructed fresh mass, Macrourus spp. were eaten by wandering albatross, southern and northern royal albatrosses, and white-capped albatross. However, consistent with the mixing model output, these prey did not predominate in the diet of any species.

The Bayesian stable isotope mixing models suggested that discards from pelagic longline fishing were of low-to-moderate importance in all species. However, some Illex and small pelagic fish used as pelagic longline baits are undoubtedly ingested by scavenging seabirds (see Table 6). All the birds in our study were bycaught on pelagic longliners, so the relatively low importance of this fishery might indicate that vessels are not attended by birds for prolonged periods, as suggested from vessel-based observations of those that are banded or have diagnostic external marks (Bugoni et al. 2010). Spectacled petrel and particularly wandering albatross are the two most oceanic of our study species known to overlap extensively with pelagic longline fishing (Bugoni et al. 2009, Jiménez et al. 2016). Histioteuthis spp. and small pelagic fish dominated their diet according to the mixing models. The predominance of Histioteuthis spp. in the diet of wandering albatross was confirmed by the conventional diet analysis of fresh prey only (Table 6), or fresh and old material combined (Table 1). These species are also important in the diet during the breeding season (Xavier et al. 2003) along with other squid [e.g., Kondakovia longimana and Taonius sp. B (Voss)], and fish discarded from the demersal longline fishery for toothfish (Xavier et al. 2003; Ceia et al. 2012). Although we analysed few stomachs of spectacled petrel, almost half of the squid beaks were highly eroded and so remained undetermined. The most important squid identified was Lycoteuthis lorigera, but the lower beaks were eroded and thus excluded from isotopic analyses as they were potentially consumed outside the study area. Although Histioteuthis was recorded, its importance remained unclear. However, stomach analysis of seven bycaught birds in southern Brazil found that Histioteuthidae was the main squid family targeted by spectacled petrel (Colabuono and Vooren 2007), suggesting that may be important prey. The isotopic signatures of small pelagic fish used in the mixing model were from Scomber spp. and Trachurus spp., which are often used as baits by pelagic longliner and are consumed by both seabird species (Jiménez et al. 2011, 2012). Therefore, the provenance of this dietary component could be as discards; however, no evidence for either prey was found in the stomach contents of wandering albatross or spectacled petrel.

A study in southern Brazil found that discards from the pelagic longline fishery were the main dietary source for the community of Procellariiformes (Bugoni et al. 2010). Note that this was based on SI ratios in seabird blood, and thus other TDF values, modelled in relation to three sources in Isosource (Phillips and Gregg 2003), which is not Bayesian and had no variation in TDF. Therefore, the comparison with our mixing models is not straightforward. This study used stable isotope ratios of muscle from blue shark as a proxy of pelagic longline fishing discards, as birds frequently feed on shark liver. We generated an additional Bayesian mixing model for wandering albatross and spectacled petrel that included four potential sources: Histioteuthis, small pelagic fish, blue shark (see Appendix 2, Table S2.1 for isotopic values), and demersal longline fish. Illex and demersal fish from the trawl fishery were excluded as they were of minor relevance for these seabirds (see above). The model again supported the higher importance of Histioteuthis and small pelagic fish than either blue shark or demersal longline fish for spectacled petrel (Appendix 2, Fig. S2.3). For wandering albatross, Histioteuthis was the most important prey, followed closely by both small pelagic fish and demersal longline fish; however, blue shark was least important (Fig. 5, Appendix 2, Fig. S2.3). This latter scenario is closest to our stomach content analysis and the previous studies of diet of wandering albatross at the breeding site (Xavier et al. 2004; Ceia et al. 2012). Hence, our results do not support the suggestion that discards from pelagic longline fishing are a major diet component for wandering albatross, spectacled petrel, or indeed any of the species sampled in our study.

Niche overlap and segregation

The stable isotope ratios measured in growing head feathers in our study matched those in body feathers of the same albatross species sampled at colonies (Cherel et al. 2013); of our species, northern royal albatross exhibited the highest, and wandering albatross the lowest values of both δ¹⁵N and δ¹³C. The δ¹³C values in these feathers were as expected for a region influenced by the Brazil–Falkland confluence, i.e., typical of subtropical, or a mixture of subtropical, subantarctic, and continental shelf waters (Phillips et al. 2009; Cherel et al. 2013). The exception was the value for a single southern royal albatross (δ¹³C = −18.65‰), which may have been a recent arrival from a more southerly or oceanic region. Therefore, our results indicate that SI values in small, growing feathers provide a good proxy of the foraging ecology of highly mobile seabirds while using a relatively well-defined region.

Although direct evidence is lacking, it is usually assumed that population sizes of large marine predators are limited by resources (Nicholls et al. 2002; Lewis et al. 2006; Conners et al. 2015). Foraging competition in Procellariiformes appears to be reduced by avoiding spatial overlap, which is presumed to lead to the targeting of different prey (Cherel et al. 2002; Nicholls et al. 2002; Connan et al. 2014). Our comparison between closely related species supported this hypothesis. The diets of the neritic southern and northern royal albatrosses were well differentiated from the more oceanic wandering albatross, as of the neritic white-chinned petrel from the predominantly oceanic spectacled petrel. Over shelf waters, the super-abundant and artificial food source provided by trawl fishing seems to relax niche partitioning, allowing coexistence, and thus high trophic niche overlap among closely related species. Indeed, niche partitioning was less apparent between the neritic congeners black-browed and white-capped albatrosses, which fed mainly on demersal trawl fish and Illex (Fig. 3). Isotopic niches of both albatrosses overlapped and did not differ statistically in isotopic niche width, although the latter tended to be wider in white-capped albatross (Fig. 2a), suggesting some level of niche partitioning. Both northern and southern royal albatrosses showed surprising similarity in diets, dominated by discards from the trawl fishery (Table 6; Figs. 1, 4), and exhibited similar niche widths. However, they showed low isotopic niche overlap (Fig. 2a). This is unlikely to be caused either by dietary divergence (Table 6; Figs. 1, 4) or small sample size, as Cherel et al. (2013) also found higher values of δ¹⁵N in feathers of northern royal than southern royal albatross. A plausible explanation is that both species coexist in the area and feed on the same abundant resources provided by fishing, but on a finer scale, they overlap little on space or time when feeding at trawlers. The slightly lower δ¹⁵N and δ¹³C in southern royal albatross (albeit not significant possibly due to the constrained study area) might suggest feeding on average in shelf waters slightly to the south or further offshore than northern royal albatross. Nevertheless, the unprecedented high levels of dietary overlap among these closely related species of albatrosses highlight the key influence of fisheries discards.

In summary, this study confirmed the expectation that ecologically similar species should avoid spatial overlap (neritic vs. oceanic) or show divergent diets, suggesting that niche partitioning reduces competition. However, there was also convincing evidence that closely related species of neritic albatross can show high levels of trophic overlap when feeding on a super-abundant resource. Although supported by all the evidence gathered in this paper, the lack of a control study of a similar community without fisheries discards (e.g., Oro et al. 1996) does not allow us to conclude explicitly that the availability of trawl fishery discards increases niche overlap. Niche overlap as a consequence of super-abundant resources will also occur if natural resources are super-abundant. However, in that case, the diet of neritic species should mainly consist of natural prey, including squid (e.g., Histioteuthis spp.), epipelagic fish, and carrion (Cherel and Klages 1998), for which there was a little evidence (with the exception of some of the I. argentinus). It, therefore, appears that the availability of trawl fishery discards on the continental shelf can produce high levels of dietary overlap and permit the coexistence of species within communities of non-breeding seabirds.