Diets and trophic niches of the main commercial fish species from the Celtic Sea
Abstract
The characterization and quantification of diets of nine commercially important Celtic Sea fish species (black-bellied angler Lophius budegassa, blue whiting Micromesistius poutassou, Atlantic cod Gadus morhua, haddock Melanogrammus aeglefinus, European hake Merluccius merluccius, megrim Lepidorhombus whiffiagonis, European plaice Pleuronectes platessa, common sole Solea solea and whiting Merlangius merlangus) was undertaken November 2014 and November 2015 to gain a better understanding of fish feeding strategies, prey preferences, competition for resources and, more generally, increases knowledge of marine ecosystem functioning. Prey were classified into 39 taxonomic groups. A feeding overlap index and multivariate analyses were used to classify the fishes into four main trophic groups where interspecific competition for resources may be important: piscivorous species, omnivorous species, planktivorous species and invertebrate benthic feeders. Ontogenetic changes in diet were also considered for L. budegassa, G. morhua, M. aeglefinus, M. merluccius and M. merlangus through partitioning into size classes. This revealed an important shift in the diet of M. merluccius from omnivory to piscivory, whereas M. aeglefinus exhibited no significant ontogenetic change in diet, remaining an invertebrate benthic feeder. Feeding strategies of these species were also investigated using the Shannon niche-breadth index and other descriptors, such as the total number of taxonomic groups of prey and the mean number of prey in gut contents.
Introduction
The Celtic Sea comprises the shelf area between the south of Ireland, the south-west of the U.K. and the north-west of France. Most of the area is shallower than 200 m and is bounded to the west by a steep and rocky slope with reefs of deep-water corals (ICES, 2016a). The variety of habitats (including substrata, oceanographic conditions, etc.) in the Celtic Sea accommodates a diverse range of fish, crustacean and cephalopod species and supports economically important fisheries for pelagic fish, whitefish, flatfish and invertebrate species exploited mainly by Belgium, France, Ireland, Spain and the United Kingdom. Within the European common fisheries policy, annual trends in abundance and stock assessments of commercial species are used to provide scientific advice (ICES, 2014), in which management options are based on monospecific approaches. Options are proposed for a given stock regardless of the interactions with other commercial and non-commercial species. Previous analyses have shown however, that fishing also affects the size structure of fish populations and, therefore, the mean trophic level of the community, by decreasing the relative abundance of larger piscivorous fish, which is followed by an increase in the abundance of smaller pelagic fish species (Trenkel & Rochet, 2003; Pinnegar et al., 2003; Blanchard et al., 2005). With a view to ecosystem-based fisheries management, e.g. shifting from the monospecific stock-by-stock approach to provide more integrative management advice, it will be crucial to understand and quantify better the biological interactions between species and more specifically fish predation.
Multi-species models, such as multi-species virtual population analysis (MSVPA) (Pope, 1979; Helgason & Gislason, 1979), ecopath-with-ecosim (Christensen & Walters, 2004) or the more recent stochastic multi-species model (SMS) (Lewy & Vinther, 2004; Kinzey & Punt, 2009), have been developed to promote ecosystem-based fisheries management and to investigate ecosystem functioning. In such modelling approaches, predation between species is explicitly modelled since it represents an important cause of mortality for fish. The abundance of a prey may be negatively correlated with the abundance of its predators (Link et al., 2009) and, more generally, fishing pressure on some species may influence the whole food web through indirect effects caused by underlying changes in predation intensities. Multi-species models are now used in various European and world seas [see ICES (2016b) for an overview of multi-species models currently used in European seas, Jurado-Molina et al. (2005) for their use in the Bering Sea, Kinzey & Punt (2009) on the Aleutian Shelf and Tsou & Collie (2001) on the Georges Bank]. The main drawback of these models is that they rely on robust data for local fish diets in order to estimate accurate predation parameters. Indeed, fish diet may depend on several factors, such as ontogeny (especially the development stage or size), habitat, sex, season, geographical position, depth and, more generally, on various abiotic and biotic environmental factors, in particular the local availability of prey (Pinnegar et al., 2003).
Although data on the diet of predators may be abundant for some European seas, diet data for the Celtic Sea are old, sparse or even non-existent for some of its main commercially exploited fish species. In the North Sea, important stomach sampling campaigns were conducted in 1981 and 1991, in which more than a hundred thousand stomachs were collected and analysed, distributed across the five main exploited predators, the Atlantic cod Gadus morhua L. 1758, the haddock Melanogrammus aeglefinus (L. 1758), the whiting Merlangius merlangus (L. 1758), the Atlantic mackerel Scomber scombrus L. 1758 and the saithe Pollachius virens (L. 1758) and 29 other fishes (Daan, 1989; ICES, 1997). These data have contributed to a deep understanding of the trophic ecology and functioning of the North Sea ecosystem and have supported the development of various modelling approaches. In contrast, existing ecosystem models in the Celtic Sea mostly rely on data collected in adjacent areas, such as the Irish Sea and the Bay of Biscay (Guénette & Gascuel, 2009; Moullec, 2015), based on the assumption that, at least for some predators, the environmental conditions (especially the availability of preferred prey) are not very different from the Celtic Sea, precluding relevant diagnostics on past and future states of the ecosystem.
It is therefore essential to characterize and quantify the local feeding habits of predator fish in the Celtic Sea. Among previous diet studies in the Celtic Sea, Du Buit (1982) focused on the diet of some teleosts, including European hake Merluccius merluccius (L. 1758), M. aeglefinus, G. morhua, M. merlangus, megrim Lepidorhombus whiffiagonis (Walbaum 1792) and pollack Pollachius pollachius (L. 1758). This study was followed by a series of papers analysing the feeding habits of M. merlangus (Du Buit & Merlinat, 1987), L. whiffiagonis (Du Buit, 1992), G. morhua (Du Buit, 1995) and M. merluccius (Du Buit, 1996) in more detail, combining stomach content data with a digestion rate model to estimate daily rations. More recently, Pinnegar et al. (2003) focused on the diet of five major predators in the Celtic Sea (M. merluccius, M. merlangus, L. whiffiagonis, G. morhua and P. virens), linking stomach-content data to the local availability of prey using the Chesson index (Chesson, 1978). Mahe et al. (2007) compared the diet of M. merluccius in the Celtic Sea and the Bay of Biscay.
The underlying aim of this work is to provide local data that could provide a basis for reliable and coherent multi-specific modelling approaches. To achieve this goal, fish were collected and the contents of their guts were analysed to update and complete the trophic knowledge of the nine main commercial fish species in the Celtic Sea. For five species, two size classes were considered in order to evaluate a potential ontogenetic diet shift. Furthermore, to study the ecosystem trophic structure, competitive relationships between species were analysed using the concept of trophic niches and feeding-overlap indices. The results are described and discussed in the light of observed trends and functioning in other adjacent ecoregions and past observations in the Celtic Sea.
Materials and methods
Sampling
Fish were sampled in the Celtic Sea (Fig. 1) during the EVHOE survey (evaluation halieutique de l'ouest part of the international bottom trawl survey, IBTS) in November 2014 and November 2015, using a GOV (grande ouverture verticale) demersal trawl with a codend of 20 mm stretched mesh, towed for 30 min at a speed of c. 6·5 km hour–1 (c. 3·5 knots) by R.V. Thalassa. All fishing operations were carried out during daytime. On board, fish were measured and weighed. Their guts were dissected, weighed and frozen for full analysis in the laboratory. A total of 1002 guts were collected across nine main commercial fish species, including six demersal species: common sole Solea solea (L. 1758), European plaice Pleuronectes platessa L. 1758, L. whiffiagonis, M. aeglefinus, black-bellied angler Lophius budegassa Spinola 1807 and M. merluccius, two benthopelagic species: M. merlangus and G. morhua and one bathypelagic species: blue whiting Micromesistius poutassou (Risso 1827) (Fishbase, Froese & Pauly, 2016). Lophius budegassa, G. morhua, M. aeglefinus, M. merluccius and M. merlangus were subdivided into size classes and sampling was performed to ensure that at least 30 non-empty guts were obtained for each size class of each fish species. Size classes were chosen before the sampling, according to the main modes observed in the size distribution of each species obtained during previous EVHOE surveys in the same period of the year. The size classes for each predator species and the number of non-empty guts are summarized in Table I.
) for nine commercial fish species in the Celtic Sea.| Number of fish guts | ||
|---|---|---|
| Species | Total | Non empty guts |
| Lophius budegassa | 94 | 65 |
| <16 cm | 44 | 31 |
| >16 cm | 50 | 34 |
| Micromesistius poutassou | 122 | 99 |
| Gadus morhua | 100 | 97 |
| <60 cm | 66 | 63 |
| >60 cm | 34 | 34 |
| Melanogrammus aeglefinus | 122 | 119 |
| <22 cm | 55 | 52 |
| >22 cm | 67 | 67 |
| Merluccius merluccius | 192 | 106 |
| <21 cm | 64 | 45 |
| >21 cm | 128 | 61 |
| Lepidorhombus whiffiagonis | 110 | 67 |
| Pleuronected platessa | 64 | 45 |
| Solea solea | 40 | 37 |
| Merlangius merlangus | 104 | 62 |
| <23 cm | 34 | 21 |
| >23 cm | 70 | 41 |
Fish gut contents
In the laboratory, each fish gut was thawed and opened in order to identify prey to the lowest taxonomic level possible: macroscopically for the larger fish and with a stereomicroscope for the smaller ones. Of the 4931 prey items recorded, 43% were identified to the species level. The analysis of digestive contents was used to evaluate the frequency of occurrence of each prey taxon, as well as their relative abundance in the diet of each predator and size class. Although some prey items were identified at a low taxonomic level (such as species), results are presented and further analyses performed by grouping them into higher taxonomic groups. The classification was chosen in order to provide more robust indicators given the number of gut contents analysed, while retaining sufficient information on the trophic position of prey, necessary for the analysis of predator trophic niches. The frequencies of occurrence take into account only individuals that have at least one prey item in their gut contents.
Trophic niche breadths and diet-overlaps
The normalized Shannon's index was used to evaluate the trophic niche breadth of each fish [see Shannon & Weaver (1949) for the original definition in information theory and Colwell & Futuyma (1971) or Hurlbert (1978) for its use as a niche-breadth index]. The normalized index is equal to 0 when the considered predator is fully specialist and preys only upon one prey type. It is equal to 1 when the predator is generalist and consumes all possible prey types equally. This index was calculated based on the relative abundances of prey. The niche breadth of predator species i is given by: hi = − (lnN)−1
, where N is the total number of prey taxon groups and pi , kis the relative abundance of prey taxon k in the diet of predator species i.
Numerous feeding overlap indices exist (Colwell & Futuyma, 1971; Hurlbert, 1978; Wallace, 1981) that aim to measure how species share a trophic niche and thus to evaluate the intensity of exploitative competition, i.e. indirect competition through the exploitation of resources (MacArthur & Levins, 1964; Tilman, 1982). Feeding overlap indices were used to characterize intraspecific changes in the diet with ontogeny, as well as interspecific competition for resources.
The classical Renkonen similarity index (Renkonen, 1938), also known as the Schoener index (Schoener, 1970), was used to evaluate the niche overlap by constructing the similarity matrix S including all predators, some of which were subdivided into size classes. Coefficients si,j of matrix S, evaluating the similarity between feeding niches of species (with eventual size class) i and j, are given by: si,j = 
, where pi,k and pj,k are relative abundances of prey species k for predator i and predator j, respectively. The Renkonen similarity index ranges from 0 when there is no feeding overlap, i.e. when there is no common prey in the diet of the two predators, to 1 when the relative distribution of prey is identical for both predators. The corresponding dissimilarity matrix D = 1 – S was then computed and, based on this dissimilarity matrix, hierarchical clustering performed using the unweighted arithmetic average method [see for example, Legendre & Legendre (2012) for a detailed description of this clustering method]. The resulting dendrogram gives a visual representation of similarity–dissimilarity between each predator species or size class and is used to group predators sharing a similar trophic niche.
In order to visualize predators and prey in the same representation and at the same time, taking into account their dissimilarities, a non-metric multidimensional scaling (NMDS) was carried out (Shepard, 1962; Kruskal, 1964).
Results
Diet description
General descriptors and indices concerning the diets of the nine predators are shown in Table II. Occurrences and relative abundances of each prey taxonomic group in the diet of these predators are shown in Tables III and IV, respectively.
| Vacuity rate (%) | Number of non-empty guts | Number of prey taxonomic groups | Number of prey taxa per fish (mean ± s.d.) | Prey abundance per fish (mean ± s.d.) | Shannon index | |
|---|---|---|---|---|---|---|
| Lophius budegassa | 30·9 | 65 | 12 | 1·35 ± 0·57 | 1·82 ± 1·42 | 0·48 |
| <16 cm | 29·5 | 31 | 7 | 1·32 ± 0·54 | 1·94 ± 1·82 | 0·35 |
| >16 cm | 32·0 | 34 | 9 | 1·38 ± 0·6 | 1·71 ± 0·94 | 0·5 |
| Micromesistius poutassou | 18·9 | 99 | 12 | 1·54 ± 0·69 | 6·12 ± 12·59 | 0·4 |
| Gadus morhua | 3·0 | 97 | 22 | 3·53 ± 1·87 | 13·62 ± 16·68 | 0·56 |
| <60 cm | 4·5 | 63 | 21 | 3·3 ± 1·91 | 14 ± 19·4 | 0·51 |
| >60 cm | 0·0 | 34 | 18 | 3·94 ± 1·74 | 12·91 ± 10·09 | 0·6 |
| Melanogrammus aeglefinus | 2·5 | 119 | 23 | 3·97 ± 2·03 | 15·13 ± 17·51 | 0·46 |
| <22 cm | 5·5 | 52 | 18 | 3·48 ± 1·75 | 11·23 ± 11·99 | 0·46 |
| >22 cm | 0·0 | 67 | 20 | 4·34 ± 2·16 | 18·16 ± 20·38 | 0·42 |
| Merluccius merluccius | > 44·8* | 106 | 16 | 1·22 ± 0·5 | 1·72 ± 1·71 | 0·55 |
| <21 cm | > 29·7* | 45 | 8 | 1·24 ± 0·57 | 2·16 ± 2·42 | 0·43 |
| >21 cm | > 52·3* | 61 | 13 | 1·2 ± 0·44 | 1·39 ± 0·76 | 0·54 |
| Lepidorhombus whiffiagonis | 39·1 | 67 | 16 | 1·36 ± 0·62 | 2·63 ± 3·75 | 0·55 |
| Pleuronectes platessa | 29·7 | 45 | 19 | 2·82 ± 1·53 | 7·02 ± 7·14 | 0·59 |
| Solea solea | 7·5 | 37 | 20 | 3·38 ± 1·59 | 6·7 ± 4·48 | 0·58 |
| Merlangius merlangus | 40·4 | 62 | 16 | 1·35 ± 0·6 | 1·89 ± 1·36 | 0·58 |
| <23 cm | 38·2 | 21 | 10 | 1·43 ± 0·6 | 2·1 ± 1·41 | 0·5 |
| >23 cm | 41·4 | 41 | 14 | 1·32 ± 0·61 | 1·78 ± 1·33 | 0·56 |
- * Vacuity rates are greater than indicated values since many empty guts had been excluded on board while not counted.
Lophius budegassa
Fish represent 83·3 and 63·8%, respectively, of the diet of the first and second size classes of this species (Table IV). Small L. budegassa feed particularly on Gobiidae, whereas large individuals prey on more diverse fishes including Perciformes (Gobiidae) and Gadiformes [Norway pout Trisopterus esmarkii (Nilsson 1855), poor cod Trisopterus minutus (L. 1758), M. poutassou and M. merluccius]. The remaining part of the diet is composed of carideans, but also cephalopod molluscs for the large L. budegassa (Tables III and IV).
Micromesistius poutassou
Micromesistius poutassou is the only species that preys on calanoid copepods (Tables III and IV). Copepods constitute about 30% of the diet by abundance, the rest being composed of amphipods (mainly hyperids) and carideans (mainly Pasiphaeidae). Fishes were only part of the diet in some individuals larger than 18 cm and cannibalism was observed for one individual.
Cod Gadus morhua
Gadus morhua mainly feed on malacostracans (brachyurans, anomurans and carideans) and to a lesser extent fishes and molluscs (Table III and IV). Fishes are rare in the diet of smaller G. morhua (2·7% abundance for 30·2% occurrence) and become a more important food source for larger ones (16·6% abundance for 73·5% occurrence), particularly Gadiformes [especially T. esmarkii, M. merluccius and M. poutassou) and Perciformes (Gobiidae and horse mackerel Trachurus trachurus (L. 1758)] species. Among crustacean prey, Munida rugosa, Crangon allmanni, Liocarcinus vernalis and Goneplax rhomboides dominate the diet. Norway lobster Nephrops norvegicus was also observed, especially in the second size class (an overall 1·44% abundance for 7·22% occurrence).
Haddock Melanogrammus aeglefinus
Considering the relative abundance data (Table IV), M. aeglefinus mainly consume sea urchins (more than 50% of the total prey items are Echinocyamus pusillus), bivalves (essentially Abra spp.) and amphipods. Although their relative abundance is not significant, ophiuroids are found in more than 50% of the non-empty guts and polychaetes in more than 40% (Table III). The main difference between the two size classes is that large individuals prey more on sea urchins, whereas small ones feed more on bivalves.
Hake Merluccius merluccius
There is a pronounced shift in M. merluccius diet between size classes. Small individuals feed essentially on malacostracans (80% abundance); the remaining part being composed of perciform fishes (Tables III and IV). Among malacostracan prey items, M. merluccius feed more particularly on carideans (Pasiphaea sivado, Crangon allmanni) and to a lesser extent on amphipods. The largest M. merluccius mainly feed on fish (67% abundance) and on malacostracans (26% abundance). Among fish prey, Gadiformes (M. poutassou, but also T. minutus and T. esmarkii) and Perciformes (Gobiidae and T. trachurus) are most frequent. Nevertheless, Clupeiformes (Atlantic herring Clupea harengus L. 1758), Osmeriformes [silver smelt Argentina silus (Ascanius 1775)] and Stomiiformes are also observed in gut contents.
Megrim Lepidorhombus whiffiagonis
Lepidorhombus whiffiagonis feed essentially on various malacostracans, which represent 58·5% of the abundance (Table IV), with a particular affinity for mysids (8·33% abundance), C. allmanni (4·69% abundance) and M. rugosa (4·17% abundance), fish (34·2% abundance) and to a lesser extent cephalopods. Among fish prey, T. trachurus constitutes the main source of food (16·7% abundance), but other fishes such as Gobiidae, Argentinidae, M. merluccius and T. minutus were also observed in their gut contents.
Plaice Pleuronectes platessa and sole Solea solea
Pleuronectes platessa and S. solea have a relatively similar diet, composed of benthic invertebrate species. Considering the abundance data, about 30% of their diet is composed of bivalves (Table IV), mainly Abra alba and Abra prismatica. For the remaining part of their diet, P. platessa feed mainly on polychaetes (36·7% abundance) and S. solea on amphipods (28·6% abundance), in particular Ampeliscidae. When analysing the occurrence data (Table III), malacostracans, molluscs, polychaetes and echinoderms were found in more than 40% of the non-empty guts of both species.
Whiting Merlangius merlangus
Merlangius merlangus mainly consume malacostracans and especially carideans (C. allmanni), amphipods and mysids (Tables III and IV). Smaller individuals also include polychaetes (15·9% abundance) in their diet, with a particular affinity for Lagis koreni, whereas the larger ones are more piscivorous, feeding on Gadiformes: M. poutassou represents 19·5% of prey in the larger size class for 13·7% occurrence. Trachurus trachurus and T. esmarkii are also observed in the diet of larger M. merlangus.
Feeding strategy and niche breadth
When the size classes are not considered, the most selective species are L. budegassa and M. poutassou, both of which prey only upon species from 12 of the selected taxonomic groups (mainly fish and malacostracans for L. budegassa and copepods and malacostracans for M. poutassou). One major difference between their feeding strategies is the mean number of prey per predator: 1·82 for L. budegassa and 6·12 for M. poutassou (Table II). This observation is confirmed by the small values of Shannon's index (Table II), corresponding to small niche breadths and to specialist feeding strategies. Melanogrammus aeglefinus also have a small niche breadth according to Shannon's index, but have the highest number of different prey (23 taxonomic groups), indicating a strong preference for a few prey types that constitute their main source of food (sea urchins) together with a generalist–opportunist feeding strategy. Melanogrammus aeglefinus also have the highest mean number (15·13) and most diverse prey types (3·97) in their guts (Table II). Pleuronectes platessa and S. solea could be classified as benthic invertebrate feeders, sharing their feeding niche with M. aeglefinus, but in contrast to the latter, they have a high Shannon's index (Table II), indicating that they probably focus less on a particular prey type. Despite the fact that the feeding niche of G. morhua is different from that of P. platessa and S. solea, their feeding strategies seem to be similar: G. morhua can prey on a large number of different prey without any preference for any particular type. Finally, L. whiffiagonis, M. merlangus and M. merluccius have a relatively similar feeding strategy, preying on an intermediate number of prey types, without a pronounced preference for any of them and generally having less prey in their guts.
Diet overlap
Within species interclass diet overlap
Comparing within species interclass similarity indices (Table V), M. merluccius and L. budegassa have the lowest values (0·46 and 0·51, respectively), suggesting a pronounced shift in their diet. Merluccius merluccius feed mainly on crustaceans in the first class, whereas fishes largely dominate the diet in the second class. Concerning the L. budegassa, both classes are piscivorous, but preference in fish prey varies between the size classes: small individuals mainly consume Gobiidae, whereas large individuals feed more on gadoids. Melanogrammus aeglefinus and G. morhua species exhibit the smallest difference between size classes (intra-class similarity is 0·68 and 0·66, respectively).
| Lophius budegassa (< 16 cm) | L. budegassa (> 16 cm) | Micromesistius poutassou | Gadus morhua (<60 cm) | G. morhua (>60 cm) | Melanogrammus aeglefinus (<22 cm) | M. aeglefinus (>22 cm) | Merluccius merluccius (<21 cm) | M. merluccius (>21 cm) | Lepidorhombus whiffiagonis | Pleuronectes platessa | Solea solea | Merlangius merlangus (<23 cm) | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| L. budegassa (> 16 cm) | 0·51 | ||||||||||||
| M. poutassou | 0·14 | 0·29 | |||||||||||
| G. morhua (< 60 cm) | 0·18 | 0·23 | 0·22 | ||||||||||
| G. morhua (> 60 cm) | 0·26 | 0·44 | 0·28 | 0·66 | |||||||||
| M. aeglefinus (< 22 cm) | 0·05 | 0·08 | 0·18 | 0·10 | 0·11 | ||||||||
| M. aeglefinus (> 22 cm) | 0·06 | 0·09 | 0·17 | 0·15 | 0·15 | 0·68 | |||||||
| M. merluccius (< 21 cm) | 0·33 | 0·46 | 0·35 | 0·55 | 0·41 | 0·12 | 0·14 | ||||||
| M. merluccius (> 21 cm) | 0·44 | 0·78 | 0·25 | 0·25 | 0·44 | 0·09 | 0·11 | 0·46 | |||||
| L. whiffiagonis | 0·47 | 0·54 | 0·22 | 0·59 | 0·51 | 0·08 | 0·13 | 0·65 | 0·45 | ||||
| P. platessa | 0·10 | 0·14 | 0·13 | 0·18 | 0·15 | 0·55 | 0·28 | 0·13 | 0·14 | 0·14 | |||
| S. solea | 0·10 | 0·16 | 0·43 | 0·18 | 0·19 | 0·63 | 0·40 | 0·20 | 0·16 | 0·29 | 0·62 | ||
| M. merlangus (< 23 cm) | 0·18 | 0·34 | 0·33 | 0·55 | 0·40 | 0·12 | 0·13 | 0·65 | 0·33 | 0·59 | 0·28 | 0·21 | |
| M. merlangus (> 23 cm) | 0·23 | 0·48 | 0·43 | 0·47 | 0·45 | 0·19 | 0·22 | 0·59 | 0·50 | 0·60 | 0·17 | 0·37 | 0·57 |
Interspecific diet overlap
The highest feeding overlap is found between the first and second largest classes, corresponding to M. merluccius and L. budegassa (similarity is 0·78; Table V), both species being piscivorous, consuming Gadiformes, Perciformes and completing their diet with carideans (Table III). The species that are the most distinctive in their feeding habits are L. budegassa and M. aeglefinus, with a similarity varying between 0·05 and 0·09 depending on the size classes considered. The diet of M. aeglefinus is especially different from the other Gadiformes species in this study, even for those with comparable size ranges like G. morhua, M. merlangus and M. merluccius (the similarity between M. aeglefinus and these species is never higher than 0·22). The similarity between the latter predators varies between 0·25 (G. morhua size class 1 and M. merluccius size class 2) and 0·65 (M. merlangus size class 1 and M. merluccius size class 1) and most often lies between 0·4 and 0·6. The feeding overlap for these three Gadiformes is always higher between the first classes than between the second classes, suggesting a divergence in their diet with increasing size.
Trophic niches
Four main feeding niches were identified (Fig. 2). The first group of species is composed of benthic invertebrate feeders, i.e. S. solea, P. platessa and M. aeglefinus. The second group is composed of M. poutassou, which is the only species that preys upon calanoid copepods. The third group includes piscivorous species, L. budegassa and M. merluccius in the second size class. The last group contains omnivorous species (G. morhua and M. merlangus regardless of their size class, L. whiffiagonis and M. merluccius of the first class) that prey mainly upon malacostracans and to a lesser extent on fishes.
Predators belonging to the trophic groups obtained through hierarchical clustering are also well grouped in the two-dimensional representation created by non-metric dimensional scaling (Fig. 3). Predator species superimposed with polygons of prey species occupy the same area, with piscivorous species on the top left, invertebrate benthic feeders are located on the right, planktivorous species at the bottom and omnivorous species in the centre.
, benthos and crustaceans; 
, echinoderms; 
, fishes; 
, molluscs; 
, others. (See Discussion
In this study, the diets of nine commercial fish species of the Celtic Sea are described and compared through gut-content analysis. These species belong to various taxonomic groups, including Gadiformes, Pleuronectiformes and Lophiiformes and are considered as benthic, demersal, benthopelagic or pelagic according to their links with the bottom. Multivariate analyses highlighted that four main trophic niches within interspecific competition for resources may be important. A first group of omnivorous species, composed of M. merlangus, G. morhua, L. whiffiagonis and the first size class of M. merluccius (<21 cm), feed mainly on crustaceans and fish (especially the largest individuals), sometimes completing their diet with molluscs and polychaetes, as in the case of small M. merlangus. The second group is composed of piscivorous species, L. budegassa and the large size class of M. merluccius (>21 cm). The third group, composed of S. solea, P. platessa and M. aeglefinus, are benthic feeders that prey upon crustaceans, molluscs, echinoderms and polychaetes. Finally, M. poutassou was the only fish for which zooplankton constitutes the main prey. The resulting trophic classification of these predators does not follow their taxonomic classification: M. aeglefinus has a diet closer to the flatfishes S. solea and P. platessa than to other Gadiformes, whereas L. whiffiagonis was grouped with M. merlangus, G. morhua and small M. merluccius.
Diets are known to depend on several factors such as the time of day, season, area, year and, more generally, on all factors that affect the current local availability of prey. While the contributions of the major taxonomic groups of prey (fishes, crustaceans, molluscs, echinoderms and polychaetes) observed here are generally consistent with many other studies, it appears that at the species level, or even at intermediate taxonomic group levels, differences could exist. These are probably due, in many cases, to the local availability of prey rather than to different feeding preferences from one region to another.
Diets in the Celtic Sea and other European Seas
Benthivorous species
Even though the feeding ecology of the two flatfishes S. solea and P. platessa differs, S. solea feeding essentially during the night and P. platessa during the day (Amara et al., 2001), this study shows that competition between these two species is high (feeding overlap of 0·62; Table V). In the North Sea, P. platessa and S. solea have a strong affinity for polychaetes and bivalves and they prey to a lesser extent on crustaceans and echinoderms (Braber & de Groot, 1973; Rijnsdorp & Vingerhoed, 2001). In these previous studies, it appeared that S. solea preyed more on crustaceans than P. platessa, as observed in the present study, with a prevalence of amphipods such as Ampelisca species. For P. platessa, the same pattern has been observed in the Irish Sea (Wyche & Shackley, 1986; Carter et al., 1991), where bivalves (especially A. alba) and polychaetes (L. koreni) dominate the diet, but echinoderms (Echinocardium cordatum and ophiuroids) also contribute. Present results are also consistent with these latter studies, suggesting no significant difference between the diet of P. platessa in the Celtic and Irish Seas.
Melanogrammus aeglefinus has a large feeding overlap with S. solea and P. platessa, especially when considering its first size class. Many sea urchins (E. pusillus) were found in the M. aeglefinus diet, although this was not observed in the Celtic Sea during the period 1975–1978 (Du Buit, 1982). While echinoderms are generally an important resource for M. aeglefinus, the literature reported many more Ophiurea than Echinodea species (Du Buit, 1982; Mattson, 1992; Schückel et al., 2010). Tam et al. (2016) have shown that the relative proportions of echinoderms and fishes in the diet of M. aeglefinus depend on the area considered: while fishes are almost absent in its diet in the Celtic Sea, they dominate in the North Sea or in the North Atlantic Ocean (Rockall Bank). Such differences are probably due to the local availability of prey (change of the benthic community composition and/or sea urchin recruitment dynamics), reflecting the feeding plasticity of this fish.
Omnivorous species
Small M. merluccius, as well as G. morhua, L. whiffiagonis and M. merlangus, are omnivorous species. The diets of M. merlangus and G. morhua have already been studied in the Celtic Sea (Du Buit, 1982; Du Buit & Merlinat, 1987; Du Buit, 1995; Pinnegar et al., 2003), where it appears that M. poutassou, Trisopterus spp. and T. trachurus are their preferred fish prey. Similarly, in the present analysis, M. poutassou is present in the diet of M. merlangus, whereas G. morhua feeds on more diverse fish prey, including L. whiffiagonis and M. merluccius. In the North Sea, various Gadidae species including G. morhua, M. aeglefinus and M. merlangus were reported in non-trivial proportions in the diet of large M. merlangus (Hislop et al., 1991), but these prey were not observed in the Celtic Sea. It can be noted that among the numerous crustacean prey of G. morhua, the commercially important species N. norvegicus is often observed. This affinity has already been observed in the Celtic Sea (Du Buit, 1995) and in the Irish Sea (Armstrong, 1982; Link et al., 2009) where its consumption may be particularly important, inducing a high mortality rate for N. norvegicus (Brander & Bennett, 1989).
The diet of L. whiffiagonis appeared to be closer to that of G. morhua and M. merlangus than to the other Pleuronectiformes, P. platessa and S. solea, as already observed in previous studies. The general trend is that L. whiffiagonis feeds on crustaceans, fish and to a lesser extent on cephalopods., feeding mainly on Trisopterus spp., silvery pout Gadiculus argenteus Guichenot 1850, M. poutassou and sprat Sprattus sprattus (L. 1758) in waters surrounding Scotland (Du Buit, 1984) or Trisopterus spp., M. poutassou and T. trachurus in the south Celtic Sea (Du Buit, 1992). The present observation of a preference for crustacean prey is similar to that made by Du Buit (1984) in the Celtic Sea, although a particularly high consumption of T. trachurus and Gobiidae was noted among the fish prey and has not previously been reported (Du Buit, 1984; Du Buit, 1992).
Piscivorous species
While small M. merluccius are omnivorous, large M. merluccius are piscivores, with T. trachurus, M. poutassou and Trisopterus spp. as preferred prey in the Celtic Sea (Du Buit, 1996; Mahe et al., 2007), as observed in this study, with nevertheless a particular affinity for M. poutassou. In Mediterranean waters, European pilchard Sardina pilchardus (Walbaum 1792) and European anchovy Engraulis encrasicolus (L. 1758) dominate the M. merluccius diet (Carpentieri et al., 2005). Despite the fact that interstage cannibalism was not observed in this present study, it has also been reported, with variable intensity (Du Buit, 1996; Preciado et al., 2015; Mahe et al., 2007). Cannibalism may occur in M. merluccius, even at early life stages: Guichet (1995) reported cannibalism starting at a length of 15 cm. Although cannibalism occurs in a wide range of marine fish species, among those studied here, M. merluccius and G. morhua are probably the species for which this mechanism is the most common and intense, indicating a strong capacity of intraspecific regulation of their population.
In the Celtic Sea, L. budegassa prey mainly on Gadiformes and Gobiidae and, to a lesser extent, on other fish species, crustaceans and cephalopods. Their feeding habits have mainly been studied in southern European seas, such as the Cantabrian Sea (Preciado et al., 2006) and some parts of the Mediterranean Sea (Negzaoui-Garali et al., 2008; Stagioni et al., 2013). Studies on other Lophius spp., especially monkfish Lophius piscatorius L. 1758, are more common and cover a larger area of the European seas (Crozier, 1985; Thangstad et al., 2002; Laurenson & Priede, 2005; Fariña et al., 2008). The preferred fish prey of L. budegassa in the Celtic Sea are more similar to those of L. piscatorius in northern European seas, which feed particularly on T. esmarkii and T. minutus (Crozier, 1985; Laurenson & Priede, 2005), than the prey of L. budegassa in the southern European seas, where M. merluccius seems to be an important fish prey (Preciado et al., 2006; Negzaoui-Garali et al., 2008; Stagioni et al., 2013). The present study reveals that, in the Celtic Sea, L. budegassa consume numerous Gobiidae, almost the only fish prey of small L. budegassa (<16 cm). This was also observed in the Mediterranean Sea (Lopez et al., 2016), but not in the Irish Sea (Crozier, 1985) for L. piscatorius, indicating possible differences in the feeding habits of these species.
Planktivorous species
Finally, the diet of M. poutassou, which is the only bathypelagic species in this study, is very different from those of the other Gadiformes studied. Micromeistius poutassou is a planktivore, eating crustaceans, particularly euphausiids (krill), copepods, amphipods (especially hyperiids) and carideans (e.g. Pasiphaeidae species) (Sorbe, 1980; Cabral & Murta, 2002; Prokopchuk & Sentyabov, 2006; Andrey et al., 2010; Bachiller et al., 2016). A high consumption of krill has been observed in almost all of these previous studies. The fact that this was not observed in the Celtic Sea is probably due to its absence at such depths and during the season in question. Large individuals are also able to consume fish and several studies have also reported cannibalism (Dolgov et al., 2010). Even if M. poutassou does not compete for resources with other species of this study, at least when considering non-larval stages, competition may nevertheless be important with other size-comparable pelagic fishes highly abundant in the Celtic Sea, such as T. trachurus, Trisopterus spp. and S. scomber.
Ontogenetic shifts
Diet of fishes may depend on their size, most of them exhibiting ontogenetic shifts. The shift observed for M. merluccius was more pronounced than for M. merlangus or G. morhua. Merluccius merluccius exhibit a pronounced ontogenetic diet shift, occurring between lengths of 15 and 30 cm, toward almost exclusive piscivory, putting it at the top of the food web. This shift has also been reported in several studies, either in the Celtic Sea (Pinnegar et al., 2003; Mahe et al., 2007) or other European seas (Guichet, 1995; Carpentieri et al., 2005). Conversely, M. aeglefinus showed no difference in its diet between the two size groups, as already observed in numerous European seas (Mattson, 1992; Tam et al., 2016). The three other Gadiformes increased their consumption of fishes with ontogeny. In their second size class, the proportion of fishes in their diet nevertheless differs substantially, resulting in a decrease of the feeding overlap compared with the ones observed between the smallest size classes. These ontogenetic niche differentiations lead to a decrease of interspecific competition for resources between these species. Owing to its particular feeding niche, feeding competition between M. aeglefinus and other Gadiformes is almost non-existent in the Celtic Sea. Rowlands et al. (2008) showed that the niche differentiation between M. aeglefinus and G. morhua–M. merlangus begins at earlier life stages, whereas feeding overlap between larvae and juvenile G. morhua and M. merlangus remains high. The niche differentiation between these two latter species is weaker, occurring more progressively and later in their life stages. Ontogenetic shifts were not investigated for M. poutassou and the three flatfishes in this present study due to low accessibility of small size classes during the survey. Although fishes generally represent a minor contribution in the diet of M. poutassou, it appears that their consumption may significantly increase with ontogeny in some particular seas such as in the Barents Sea (Dolgov et al., 2010). Finally, like other omnivorous species, L. whiffiagonis also seems to increase its piscivory progressively, with a shift occurring between lengths of 15 and 20 cm (Rae, 1963).
Feeding strategies
Feeding strategies differ between predators regardless of their prey preferences. They are characterized by complementary indices, such as the total number of prey groups, the mean number of prey in the gut and various other indices evaluating feeding niche breath (Colwell & Futuyma, 1971), such as the Shannon index used in this study.
When considering the mean number of prey per individual, the abundance was high for G. morhua and M. aeglefinus, intermediate for M. poutassou, S. solea and P. platessa and low for L. budegassa, L. whiffiagonis, M. merlangus and M. merluccius. A high mean number of prey may suggest smaller prey relative to predator size or (without exclusivity) prey with a lower energy content with respect to individual predator energy requirements. Pinnegar et al. (2003) estimated mean predator to prey size ratio for some of these fishes, showing that G. morhua feeds on much smaller prey relative to its size (size ratio of 4·07) than M. merlangus (3·18) or M. merluccius (2·60). When considering the marine biomass size spectrum, small prey are more numerous than larger, so it follows unsurprisingly that prey encounter rates and potential consumption rates are higher. Gut vacuity rates may depend on the time of day, but also on the mean number of prey observed in non-empty guts. The particularly high vacuity rates observed for M. merluccius and L. budegassa were also reported in numerous other studies (Crozier, 1985; Preciado et al., 2006; Mahe et al., 2007) confirming that these species feed less often, but on larger prey, to compensate energetic requirements, in contrast to G. morhua and M. aeglefinus (Du Buit, 1982; Mattson, 1992). It should be noted that comparisons of vacuity rates and mean number of prey in gut contents between species imply that digestion time is similar for all prey and predators, whatever their species and size, which is obviously not the case.
The number of prey taxonomic groups, as well as the evaluation of niche breadth using Shannon's index, highlights the degree of generalism of a species. Lophius budegassa and M. poutassou are specialist species with a small niche breadth, feeding upon a small number of taxonomic groups. Melanogrammus aeglefinus also has a small niche breadth, but can prey secondarily on many taxonomic groups. Hence, it shows a strong preference for a few prey types (E. pusillus), but has the ability to prey upon many others, possibly when its preferred prey are scarce. Other species are clearly generalists with a relatively large niche breadth and either a large number of prey taxonomic groups, as for G. morhua, S. solea and P. platessa, or an intermediate number of prey taxonomic groups, as for M. merlangus, M. merluccius and L. whiffiagonis. The observations made on feeding strategies are consistent with previous ones. Lophius budegassa is a non-generalist opportunist feeder: although L. budegassa may eat almost all the prey it can catch, it has a particular feeding behaviour, based on a sit and wait strategy (Fariña et al., 2008), drastically limiting the number of prey species available, so it follows that only a restricted part of all the potential prey is accessible. In contrast, G. morhua is a generalist and opportunistic predator, eating almost all adequate sized prey present in its environment, making its stomach content a good indicator of the local species assemblage (Link et al., 2009).
It is nevertheless hard to reliably classify feeding strategies without any information on the local and current availability of prey (including at least their size distribution) since previous interpretations implicitly assume that all prey taxonomic groups are equally present in the environment, which is obviously not the case. Estimation of preferences, or prey-specific attack rates, would require their inclusion through the use of a predation model. One classical and simple way to achieve this is to use the Chesson index to evaluate preferences (Chesson, 1978). It is however, difficult to provide maps on prey abundance even using scientific survey data, especially since the sampling gear (GOV trawl) was not designed to catch benthic macrofauna, plankton or even pelagic fishes. Using the Chesson index, Mahe et al. (2007) concluded that M. merluccius longer than 32 cm positively select only M. merluccius in the Celtic Sea. Preciado et al. (2015) have shown that M. merluccius cannibalism particularly occurs near nursery areas and its intensity depends on the availability of appropriately-sized juvenile M. merluccius and other fish prey. This may explain why cannibalism was not observed in the present study, while underlining the difficulties in using such preference indexes without including the prey-size distribution.
Comparisons of the observed diets with previous studies in the Celtic Sea and other European seas have revealed similarities, but also specificities, of the current diets of fish, emphasizing the importance of collecting local data on a regular basis. Monitoring of local inter-annual changes in diets will improve understanding of prey preferences, feeding adaptation abilities, biomass flow among species and of the possible evolution of the Celtic Sea food web in the future. In addition, it would also be interesting to consider seasonal variations in diets, since relative abundances of preferred prey vary over the year, as well as other environmental factors such as water temperature, which may have consequences for fish diets, condition and fitness. Finally, it would be necessary to consider smaller size classes to investigate shifts in diet with ontogeny in greater depth.
Diet analyses, more than just improving our knowledge on the ecology of the considered species and trophic interactions among them, also contribute to the calibration of multi-specific or ecosystem models. The dataset analysed in this study will eventually support the operational implementation of the ecosystem approach to fisheries management in the Celtic Sea. Indeed, the ecopath-with-ecosim model could now be balanced using local and more reliable diet matrices for the main commercial fish species, contributing to a better understanding of the current state of the Celtic Sea ecosystem and providing a relevant framework to test scenarios. Indeed, it is well known that fishing activity modifies the food-web structure and stability by removing species from their environment and by discarding dead individuals back to the sea, which may support part of the scavenger community. Bottom trawling might alter the benthic habitat, which has direct consequences for the benthic invertebrate community, indirect consequences for their predators by bottom-up control and possibly, at a larger scale, for the functioning of the whole ecosystem. Additionally, environmental change and, more precisely, climate change may, among other effects, induce changes in the living area of some species, modifying the local species assembly.
The work was part of the EATME project supported by IFREMER, France Filière Pêche and Région Bretagne. The authors thank G. Allanic for his help onboard the R. V. Thalassa and in the laboratory.
