© 2006 International Council for the Exploration of the Sea
Feeding habits of demersal fish in Icelandic waters: a multivariate approach
Marine Research Institute Skúlagata 4, PO Box 1390, 121 Reykjavík, Iceland
*Correspondence to A. Jaworski: FRS Marine Laboratory, PO Box 101, 375 Victoria Road, Aberdeen AB11 9DB, UK; tel: +44 1224 876544; fax: +44 1224 295511. e-mail: jaworskia{at}marlab.ac.uk.
Stomach data were examined to assess the key factors that determine diet composition in some of the most important demersal fish species in Icelandic waters and to identify major feeding guilds. The data were collected during the groundfish surveys conducted by the Marine Research Institute in 1992. The factors examined included geographic position, depth, season, and fish size. Data were analysed using multivariate methods: canonical correspondence analysis (CCA), non-metric multi-dimensional scaling (NMDS), and hierarchical clustering. For the CCA, important explanatory variables for the observed feeding patterns were found using forward stepwise selection. Fish size was the most important explanatory variable for most species, reflecting distinct ontogenetic shifts in diets. A large variation in diet composition was observed, and the CCA model explained 6–16% of the total variation. The spatial and seasonal variability in diets reflected, in general, patterns of prey availability. Among the main predators, the two major feeding guilds were (i) species preying mainly on echinoderms, supplemented with fish and other benthic invertebrates, and (ii) species preying mainly on crustaceans and fish.
Keywords: Amblyraja radiata, Anarhichas lupus, Anarhichas minor, feeding habits, Gadus morhua, Hippoglossoides platessoides, Iceland, Melanogrammus aeglefinus, Pollachius virens, Sebastes marinus
Received 6 April 2006; accepted 12 July 2006.
 |
Introduction |
|---|
 Top Introduction Material and methodsResultsDiscussionReferences |
|---|
Feeding ecology of exploited fish species has seldom been considered in evaluating their population dynamics in fisheries management. In general, little is known about trophic interactions between exploited fish species and other organisms in the ecosystem, or of the factors determining strengths of predator–prey and competitive interactions (Garrison and Link, 2000; Link and Garrison, 2002; Link et al., 2002). Such information is needed as input to ecosystem-based models. It has been recognized that this information is often ecosystem-specific (Hanson and Chouinard, 2002).
Iceland is located at the boundary between warm Atlantic water and cold water from the Arctic. Consequently, there are marked fluctuations in salinity and temperature, particularly over the north Icelandic continental shelf (Malmberg and Kristmannsson, 1992). The current system and distribution of water masses create favourable conditions for many marine species. The Icelandic continental shelf is considered to be moderately productive. Productivity is higher in the southwest than in the northeast, and higher on the continental shelf than in the oceanic regions (Gudmundsson, 1998). The fish that are economically and ecologically important components of the Icelandic marine ecosystem include cod Gadus morhua, haddock Melanogrammus aeglefinus, saithe Pollachius virens, redfish Sebastes marinus, Greenland halibut Reinhardtius hippoglossoides, herring Clupea harengus, and capelin Mallotus villosus. Other species common to the region include Atlantic wolffish Anarhichas lupus, spotted wolffish Anarhichas minor, and flatfish such as plaice Pleuronectes platessa, dab Limanda limanda, and long rough dab Hippoglossoides platessoides. Starry ray Amblyraja (= Raja) radiata is a common elasmobranch. The feeding ecology of these species in Icelandic waters has been reasonably well documented, especially for cod and its major prey, capelin (Pálsson, 1983; Magnússon and Pálsson, 1989; Pálsson and Björnsson, 1993; MRI, 1997). Those studies generally focused on relating feeding habits to single factors. Some sources of variation in diet composition, such as fish size, and regional and temporal factors, have been identified and described for some species (Pálsson, 1983; MRI, 1997), and there is a need to summarize that knowledge by applying a new approach. Better understanding of the feeding patterns (diet composition, sources of variability, trophic interactions) of the major fish predators is crucial for determining their roles in the Icelandic marine ecosystem.
Feeding habits in fish may vary within wide bounds on temporal and spatial scales. It is of interest to identify those environmental factors which, along with ontogenetic mechanisms, largely determine diets of fish, and to obtain an overview of how diets vary with the environment (Hovde et al., 2002). Fish size, maturity, and condition, season, bottom depth, latitude, longitude, and habitat type are among potential factors influencing fish diets. They reflect or relate to ontogenetic changes, the physiological status of predators, seasonal variation in temperature and prey abundance, and spatial distribution of prey in the environment. However, the possibility of using these factors in an analysis is limited by the availability of data, and it is seldom that data will be available for all these factors. Furthermore, it would be desirable to quantify these effects and to rank them according to their relative importance. Multivariate analyses can be a useful tool in such analyses because they can elucidate general patterns in feeding ecology of the fish community, particularly when based on extensive and representative sampling. Such patterns are often hidden by other sources of variation (ter Braak and Verdonschot, 1995) and may be difficult to determine in a single-factor perspective, especially when the sampling design is unbalanced with respect to influential variables (Hovde et al., 2002).
There are many examples where multivariate techniques have been applied to study fish feeding ecology. Høines and Bergstad (1999) studied ontogenetic and seasonal variation in diets of gadoid fish in waters off southwestern Norway. De Crespin de Billy et al. (2000) proposed a multivariate method, derived from principal component analysis, to investigate fish diets, and applied it to brown trout Salmo trutta diets. Feeding habits in the northwest Atlantic were studied specifically for hake (Garrison and Link, 2000), Atlantic cod (Link and Garrison, 2002), and flatfish (Link et al., 2002). Hovde et al. (2002) examined spatial, temporal, and ontogenetic patterns in the diet of Northeast Arctic Greenland halibut. Trenkel et al. (2005) investigated the relationship between the presence of selected prey and environmental factors in the Celtic Sea fish community. These studies often corroborate known facts such as shifts in diets with size, but they also reveal regional and seasonal differences as well as changes in diets across time-series. Some of them also assess the degree of intra- and interspecific overlap in diets (Garrison and Link, 2000; Link and Garrison, 2002; Høines and Bergstad, 2002).
The objectives of the present study are: (i) to examine the feeding habits of selected key fish predators in Icelandic waters (individually for each species) across wide size ranges and across broad spatial and seasonal scales, (ii) to assess the relative importance of different factors influencing the feeding habits, and (iii) to compare and contrast the diets of the major predators in a synthetic manner. These objectives are addressed using multivariate techniques.
|
Material and methods |
|---|
|
TopIntroductionMaterial and methodsResultsDiscussionReferences |
|---|
Data collection
The stomach data for the study were collected during groundfish surveys in 1992 in the waters around Iceland by the Marine Research Institute (Figure 1). That year's sampling effort was substantially increased in terms of sampled stations, species, and stomachs (Pálsson and Björnsson, 1993). The surveys took place during three seasons: February–March, July–August, and November–December, with the bulk of samples (96%) taken in March, July, and November. Samples were collected from depths between 20 and 823 m, almost all (99%) within the 500-m isobath. There were in total 854 tows in which stomach samples were collected. The methods of sampling and stomach contents analysis have been described in detail by Pálsson (1983) and Pálsson et al. (1989).
|
The stomach data for 1992 were based on aggregate stomach samples (up to ten stomachs of fish from a length group per sample). Those samples were taken for a number of length groups (Pálsson, 1983). More than 14 000 aggregate samples were examined for that year. For each prey item in a stomach sample, the number and the weight were recorded (the latter to the nearest 0.001–0.1 g, depending on the size of the prey items). In most cases, the prey were identified to species level (Pálsson, 1983). In addition to data on stomach content and fish size (total length), date, location (latitude and longitude), and bottom depth were known for each tow.
Predators and prey categories
Among the fish sampled, 35 predator species were identified. For the purpose of this study, only the eight best-sampled species were selected (Table 1). Most of them are demersal; only cod and redfish are classified as benthopelagic and pelagic, respectively (according to species characteristics in FishBase; Froese and Pauly, 2006). Although classified as a demersal species, saithe has a highly pelagic nature (Jónsson, 1983; Pálsson, 1983). Prey items for each predator were grouped into 11 categories: (i) polychaetes, (ii) molluscs excluding cephalopods (mainly gastropods and bivalves), (iii) cephalopods, (iv) zooplankton (excluding euphausiids), (v) non-decapod benthic crustaceans (mainly amphipods and isopods), (vi) euphausiids, (vii) decapods, (viii) echinoderms, (ix) unclassified benthos (e.g. actinarians, poriferans, bryozoans, and unidentified benthos), (x) fish, and (xi) other and unidentified prey. Those prey items, species or higher taxa that comprised a considerable portion of the diets (on average >5% by weight for a given predator species) were singled out as additional prey categories.
|
Diet analyses
Diets of the eight selected predators were analysed using canonical correspondence analysis (CCA; ter Braak, 1986). It is a direct ordination technique designed for analysis of the relationships between biological assemblages of species and their environment (ter Braak and Verdonschot, 1995). CCA generally assumes that species have unimodal distribution (i.e. one optimal condition) along environmental gradients (Legendre and Legendre, 1998), but some species may have their optima outside the environmental region actually sampled and their response function is monotonic decreasing or increasing (ter Braak and Verdonschot, 1995).
In the present study, CCA assessed the multivariate diet response of each predator species to a number of explanatory variables. Prey data were converted to relative values (% of diet by weight) for each sample and subsequently arcsine transformed, which is appropriate for percentages and proportions (Zar, 1996), to normalize the data and to reduce heterogeneity of variance. It should be emphasized, however, that CCA is considered robust to assumption violations such as non-normality or lack of unimodality (ter Braak, 1986; ter Braak and Verdonschot, 1995). Fish size, latitude, longitude, and bottom depth were regarded as continuous variables, whereas season was treated as a categorical variable (with levels corresponding to the seasons when the surveys took place). The latter variable was recoded as a set of dummy variables (Legendre and Legendre, 1998).
The important explanatory variables were found using a forward stepwise selection procedure (ter Braak and Verdonschot, 1995). The criterion for inclusion of a variable in the model was the amount of variation explained by the variable in a partial CCA (ter Braak and Verdonschot, 1995), with the previously selected variables being treated as covariables. Only k – 1 dummy variables created from the categorical variable "season" with k categories (in the present study, k = 2 or 3) could enter the model because of dependencies among the k dummy variables (Zar, 1996). Each variable to be included in the model was tested for significance (at the 5% significance level) using a permutation test (ter Braak and Verdonschot, 1995; Legendre and Legendre, 1998). CCA, partial CCA, and permutation tests were performed using the Community Ecology Package "vegan" written by Oksanen (2005), which is an extension to the statistical package R (Ihaka and Gentleman, 1996).
The results of CCA were presented in the form of ordination diagrams containing the continuous explanatory variables plotted as arrows along with points for prey categories. The categorical variable "season" was represented by centroids of the samples belonging to each category (Legendre and Legendre, 1998). The arrows of the quantitative explanatory variables indicate the direction of maximum change of these variables across the diagram. The projection of prey categories onto an arrow gives an approximation of their weighted averages with respect to the variable. The results of CCA were verified in two ways: (i) changes in diet composition for the pooled data were examined individually along each gradient, and (ii) plots of diet distribution around Iceland were made separately for different seasons and size groups of predators (and thus for more homogenous data subsets). Only the most evident dependencies and trends, particularly those apparent from the CCA diagrams and plots of spatial distribution, were taken into consideration. Owing to space considerations, only CCA diagrams and selected univariate plots (to aid interpretation of the CCA results) are shown here.
Non-metric multi-dimensional scaling (NMDS) and hierarchical clustering were performed using PRIMER 5 (PRIMER-E Ltd, Plymouth, UK) to assess the degree of similarities in diets between the eight fish species. In the NMDS method, an optimal configuration of samples is constructed in a specified number of dimensions (in this case, in a two-dimensional space), based on relative values of similarity between the samples. The extent to which the distances among the scatter points deviate from the matrix input (i.e. the level of distortion) is measured by "stress" (Clarke and Warwick, 2001). The fish were grouped into three (cod) or two (other species) size classes. The species size classes were "observations" and the 11 main prey categories and capelin, as an additional category, were "variables". The analysis was done with arcsine transformed data and based on Bray–Curtis similarities. Groups of predators were identified at an arbitrarily chosen similarity level (60%) from clusters obtained from hierarchical group-average clustering. The latter method also served to verify the adequacy of the NMDS ordination (Clarke and Warwick, 2001).
|
Results |
|---|
|
TopIntroductionMaterial and methodsResultsDiscussionReferences |
|---|
Cod
Cod fed predominantly on capelin and other fish. Other important prey included the northern shrimp Pandalus borealis, other decapods, euphausiids (mainly Meganyctiphanes norvegica), and amphipods (Figure 2a). CCA explained 8% of the total variation in the diet data. The first two canonical axes accounted for 70% of the constrained (explainable) variation.
|
The most important explanatory variable was fish size. Small cod (
30 cm) preyed on amphipods, polychaetes, and euphausiids (Figures 2a and 3a). The euphausiid M. norvegica was consumed at roughly the same rates by a wide range of sizes of cod. Medium-sized cod (30–80 cm) consumed P. borealis, other decapods, and capelin. Fish, mainly capelin, ammodytids, Sebastes spp., and gadids were the major components of the diet of large cod (>80 cm), and were supplemented with smaller prey such as decapods and M. norvegica.
|
Capelin were consumed primarily in March (when the species constituted 40% of the diet), predominantly in the northern areas (Figures 2a and 3a). By contrast, other fish were important as food in the second half of the year, and their importance (especially that of Sebastes spp. and gadids) was higher in the southern areas. Also, the importance of decapods other than P. borealis decreased northwards. The proportion of P. borealis in the diet of cod was highest in autumn. The diet of cod was generally more diverse in the second half of the year and at lower latitudes.
Decapods peaked in the cod diet at medium (150–300 m, P. borealis) and shallow depths (up to 150 m, other decapods; Figures 2a and 3a). Euphausiids (including M. norvegica) and amphipods tended to increase in importance with increasing depth. Fish prey as a whole (including capelin) tended to be more common in diets at shallow and medium depths. Ammodytids (an important component of the category "other fish") were consumed almost exclusively in shallower waters, whereas Lycodes spp. and argentinids tended to increase in importance with increasing depth (not shown).
Haddock
The most important prey items of haddock were polychaetes. Other important prey included ophiuroids, amphipods, capelin, euphausiids, fish other than capelin, benthic molluscs (excluding cephalopods), and echinoderms other than ophiuroids (mainly echinoids; Figure 2b). The CCA ordination explained 9% of the total variation in the haddock diet. The first two canonical axes showed 68% of the constrained variation.
Fish size was the most important explanatory variable. The most apparent shift in diet across size was from small benthic crustaceans (mainly amphipods), euphausiids, and polychaetes (predominant in the diet of small haddock) to ophiuroids, benthic molluscs, and decapods (in medium-sized haddock), and fish (including capelin) and echinoderms other than ophiuroids (in large haddock; Figures 2b and 3b).
As in cod, the importance of capelin in the diet of haddock was strongly correlated with season, with a peak in March (15% of the diet) and almost zero consumption in the second half of the year (Figures 2b and 3b). This prey was most common in the western and northwestern areas. Other fish (with a considerable component of ammodytids) were more important in the second half of the year than in March, and at lower latitudes than capelin. The proportion of fish (including capelin) in the diet of haddock decreased sharply to a depth of approximately 200 m, while the proportion of amphipods and euphausiids tended to increase with increasing depth.
Saithe
The diet of saithe was almost entirely fish (mainly ammodytids and capelin) and euphausiids (mainly M. norvegica; Figure 2c). CCA explained a relatively high proportion of the total variation (16%). The first two canonical axes accounted for 80% of the constrained variation.
The most important explanatory variable was season or, more specifically, its summer component. Ammodytids dominated the diet of saithe in summer (73% of the diet; Figures 2c and 3c). They were also of some importance in autumn (15%), but were not found in March. In contrast, capelin were most important in March (44%). Fish other than ammodytids and capelin were most common in saithe diet in autumn. Depth was the next important determinant of saithe diet. Ammodytids were predominantly preyed upon in shallower waters (up to 150 m) off the southwest coast. The proportion of capelin in saithe diet tended to decrease with increasing depth. Euphausiids (including M. norvegica) were absent from the saithe diet in shallow waters (to 100 m). Their importance increased with increasing depth (to 79% at depths >400 m).
Although fish size was selected as a significant explanatory variable, its importance was low compared with other variables (Figure 2c, Table 2). The general trends were that the proportion of euphausiids decreased and that of fish increased with increasing predator size (Figure 3c).
|
Redfish
The main components of redfish diet were zooplankton (with a large proportion of Calanus hyperboreus), euphausiids (mainly M. norvegica and Thysanoessa inermis), capelin, and other fish (Figure 2d). CCA explained 9% of the total variation in the diet composition of redfish. The two first axes accounted for 69% of the constrained variation.
Fish size was the major factor determining the redfish diet. Zooplankton dominated the diet of the smallest fish, and its importance decreased with increasing fish size (Figures 2d and 3d). The euphausiid T. inermis followed by M. norvegica were common in the diet of medium-sized redfish. The proportion of fish (including capelin) in the diet of redfish increased with increasing predator size.
The proportion of M. norvegica and T. inermis in the diet of redfish was higher in the second half of the year than in March (Figures 2d and 3d), whereas the opposite was found for other euphausiids. Also, consumption of capelin and C. hyperboreus was greatest in March. Zooplankton, particularly C. hyperboreus, were more important in the northern areas.
Atlantic wolffish
The diet of Atlantic wolffish was mainly hard-shelled benthic invertebrates such as echinoderms, decapods, and benthic molluscs (mainly bivalves and gastropods; Figure 2e). Only 6% of the total variation was explained by the explanatory variables in CCA. The first two canonical axes accounted for 82% of the constrained variation.
Fish size was the main explanatory variable. Ophiuroids were mainly eaten by small fish (up to 40% of the diet in fish 20 cm) and their importance decreased with increasing predator size (Figures 2e and 3e). By contrast, echinoids peaked in the diet of larger fish. The peak for other (and unidentified) echinoderms was between the peaks for ophiuroids and echinoids. Decapods were eaten by a wide range of medium sizes of Atlantic wolffish. Benthic molluscs (bivalves and gastropods) tended to increase in importance with increasing fish size and constituted almost half (48%) the diet of the largest fish (>80 cm). Fish (including capelin) were, on the whole, of relatively little importance as prey, except for the largest size classes of the predator (24% of the diet in fish >80 cm).
Following fish size, latitude was the next important explanatory variable. Ophiuroids and "other echinoderms" (unidentified and other than echinoids) increased in importance from south to north, whereas echinoids and benthic molluscs were more common in the southern habitats (Figures 2e and 3e). The depth range for Atlantic wolffish was relatively small (most stomach samples were taken shallower than 400 m). The proportion of echinoids and bivalves in the diet decreased with increasing depth. Decapods peaked in shallow waters (at about 100-m depth) and ophiuroids at medium depth (150–300 m). At greater depth, the diet was dominated (56% at depths >200 m) by echinoderms (all categories). Fish were mainly eaten in shallower waters (at depths <200 m).
The diet of Atlantic wolffish did not vary much between seasons (Figure 2e, Table 2). Fish were of some importance as prey in March, but unimportant in the second half of the year (Figure 3e).
Starry ray
Starry ray preyed mainly on amphipods, decapods (with a large component of P. borealis), polychaetes, capelin, other fish, and euphausiids (Figure 2f). CCA explained 8% of the total variation. The first two canonical axes showed 84% of the constrained variation.
Fish size was the most important variable determining the diet composition. Amphipods were a predominant prey category in small starry ray (59% of the diet of fish 20 cm), and their importance decreased with increasing predator size (Figures 2f and 3f). Polychaetes, euphausiids, and decapods (including P. borealis) were mainly eaten by medium-sized fish. Fish prey dominated the diet of large starry ray (53% of the diet of fish >50 cm).
Data for starry ray were available only for March and July. Euphausiids were more common in the diet in July than in March (Figures 2f and 3f). Capelin were much more important in March than in July. Other prey categories did not vary much in importance between the two seasons.
Amphipods tended to increase in importance northwards (Figures 2f and 3f). The proportion of euphausiids in the diet of starry ray increased with increasing depth.
Spotted wolffish
The diet of spotted wolffish was less diverse than that of Atlantic wolffish and consisted almost exclusively of echinoderms (with a large proportion of ophiuroids) and fish (Figure 2g). Capelin were of little importance (just 1% of the diet and 10% of the "fish" category). CCA explained 14% of the total variation. The first two canonical axes accounted for 91% of the constrained variation.
Fish size had the greatest effect on diet composition. The proportion of echinoderms was high in small size classes (90% of the diet in fish 40 cm) and decreased, while that of fish prey increased, with increasing predator size (Figures 2g and 3g).
No difference was found in the diet composition between summer (July) and autumn (November) (Figures 2g and 3g); adding either of these two seasons to the model resulted in non-significance. Ophiuroids were much more common in the July and November diet than in March. In contrast, other echinoderms were most common in March. The proportion of fish in the diet of spotted wolffish did not vary much throughout the year, but it was highest in March.
The importance of ophiuroids increased from south to north; they were important prey in the north and northwest. Fish were mainly preyed upon in shallower waters (54% of the diet at depths <100 m), and ophiuroids at medium and greater depths.
Long rough dab
The diet of long rough dab was variable and consisted of ophiuroids, polychaetes, decapods, bivalves, capelin, other fish, benthic crustaceans (other than decapods, mainly amphipods), and euphausiids (Figure 2h). CCA explained 8% of the total variation. The two main canonical axes represented 75% of the constrained variation.
Fish size was the most important determinant of the diet. In general, the proportion of polychaetes, bivalves, and non-decapod benthic crustaceans steadily decreased, whereas that of ophiuroids and fish steadily increased with increasing predator size (Figures 2h and 3h). The diet of the largest long rough dab (>40 cm) was dominated by ophiuroids and fish (58% and 24%, respectively). Decapods were consumed largely by medium-sized fish.
Euphausiids and capelin were mainly preyed upon in March (Figures 2h and 3h). The proportion of ophiuroids and fish other than capelin increased between March and December.
The diet of long rough dab varied considerably with depth. The proportion of all fish prey was high in shallow waters (49% at depths <50 m), but it decreased sharply with increasing depth (Figures 2h and 3h). However, there was a slight increase in the occurrence of capelin in the diet of long rough dab taken at depths greater than 200 m. Polychaetes and bivalves were most common in diets in shallower waters (with a peak at about 100 m). Ophiuroids tended to increase in importance with increasing depth. Euphausiids dominated the diet of long rough dab over the deeper shelf (59% at depths >400 m). Ophiuroids increased in importance from south to north.
Explanatory variables
Fish size was the most important explanatory variable for most of the predator fish studied (Table 2). Only for saithe was it of much less importance. Season, or more specifically, March, was the next most explanatory variable, primarily because of the markedly increased proportion of capelin in diets then.
After fish size and March, depth was the next most important explanatory variable across species. It was particularly important for saithe and starry ray, reflecting the increasing dietary presence of euphausiids with increasing depth (Figure 3a–c, f, h). Ammodytids, which were consumed in great quantity by saithe, also varied greatly in importance with depth (Figure 3c).
The next important determinant of diet composition was latitude. It was of great importance for Atlantic wolffish, reflecting pronounced trends in the distribution of the main prey categories of this predator along the latitude gradient (Figure 3e).
Longitude and the two remaining seasons, summer (July–August) and autumn (November–December), were generally of far less importance than the first four explanatory variables. The diets varied relatively little between summer and autumn (compared with March). Only for saithe was summer more important as an explanatory variable than March, owing to the greatly increased consumption of ammodytids in July and August (Figure 3c). Longitude was important for the two wolffish species, but changes in feeding patterns along this gradient were in general difficult to identify.
Feeding patterns
Distinct feeding patterns could be distinguished among the main eight predators, with some changes occurring throughout the year. The cluster analysis separated the predators into three groups in March (Figure 4a). The first one was formed by the two wolffish species, long rough dab and haddock. This group consumed mainly echinoderms and smaller quantities of fish, benthic molluscs, polychaetes, and decapods. The second group was formed by starry ray, cod, and saithe. For those predators, fish and crustaceans (such as amphipods, decapods, and euphausiids) were the major prey. The third group was represented by redfish. In March, this predator preyed mainly on zooplankton, euphausiids, and fish.
|
In summer (July and August), four groups were differentiated (Figure 4b). Haddock and small long rough dab joined the second group from March. The additional group was formed by saithe. The first group continued to represent "echinoderm feeders" for which benthic molluscs, and to a lesser degree fish, were supplementary prey. The second group was formed by predators preying predominantly on fish, crustaceans, and polychaetes. Redfish, forming the third group, continued to prey on zooplankton, euphausiids, and fish. Saithe, representing the fourth group, fed predominantly on fish (mainly ammodytids) with a supplement of euphausiids.
In autumn (November and December), three groups were distinguished (Figure 4c; note the lack of data for starry ray in this season). The first group was formed by the same predators as in March, except for small spotted wolffish, which formed another predator category. Also the feeding pattern was similar, with echinoderms as the dominant prey supplemented with fish, polychaetes, benthic molluscs, and decapods. The second group was also similar to that of March, consisting of predators that consumed mainly fish and crustaceans. In autumn, euphausiids were far more important in the diet of small cod than in the diets of medium-sized and large cod. The diet of redfish then was more similar to that of cod and saithe in that it contained considerably less zooplankton and more decapods and other benthic crustaceans than in March. Small spotted wolffish, forming the third group, consumed predominantly echinoderms supplemented with fish, with practically no polychaetes or molluscs.
With data averaged over all three seasons, two large groups were found (Figure 4d): "echinoderm feeders" (with a supplement of fish, polychaetes, benthic molluscs, and smaller quantities of crustaceans) and "crustacean/fish feeders". The first group was represented by spotted wolffish, Atlantic wolffish, long rough dab, and haddock, the two wolffish species (especially spotted wolffish) being more specialized echinoderm feeders. The second group was represented by cod, saithe, redfish, and large starry ray. The diet of large starry ray was similar to that of small cod in that it contained relatively large proportions of crustaceans (mainly decapods) and polychaetes. Large cod and saithe were similar in that they were both predominantly piscivorous. The similarity between saithe and redfish diets resulted mainly from a large proportion of euphausiids, whereas the dissimilarity was due to a particularly high proportion of zooplankton (other than euphausiids) in the redfish diet. Small starry ray, forming a separate predator category, consumed large quantities of crustaceans (mainly amphipods), supplemented with polychaetes, some fish prey (much less than in the diet of large starry ray and small cod), and with practically no echinoderms that were common in the diet of small haddock and long rough dab.
|
Discussion |
|---|
|
TopIntroductionMaterial and methodsResultsDiscussionReferences |
|---|
The study has revealed general patterns in the feeding ecology of the major predators in Icelandic waters based on extensive data collected in a single year. Diet composition showed, in general, great variation. Within species, diets varied mainly with fish size, but also across spatial and seasonal scales. Distinct feeding guilds were apparent among the fish species under study.
The effectiveness of CCA in the present study, in terms of the percentage of explained variation (6–16%), was comparable with that in the studies conducted by Garrison and Link (2000), Link and Garrison (2002), and Link et al. (2002), i.e. 8–18%. Hovde et al. (2002) found high proportions (46% and 30% in ordination on stations and individuals, respectively) of total variation in diet composition of Greenland halibut in the Barents Sea explained by explanatory variables (longitude, latitude, depth, temperature, sampling period, predator length, weight, sex, and maturity stage) and their combinations. However, the study area and data analysis in that study differed from those in the present one in many ways, e.g. with respect to area size and characteristics, prey resolution, and prey data standardization. Although Greenland halibut constitute an important fish stock in Icelandic waters, the number of stomach samples collected for this species in 1992 was too low to conduct a CCA and to make meaningful comparisons with the feeding patterns of Greenland halibut in the Barents Sea.
The relatively small proportion of total variation that can be explained in a CCA is not surprising. Although the diet generally reflects the availability of different prey categories in the environment, fish sampled at the same time and place can have significantly different stomach contents (Ringler, 1983). A large amount of noise is present in this case, reflecting a variety of effects such as physiological status and competition among predators, as well as size, availability, and selectivity of prey (Ringler, 1983; Wootton, 1992). Other factors, such as localized oceanographic conditions or processes, may also influence fish diets (Link and Garrison, 2002). The influence of aggregating stomachs, as was the case in the present study, on the results of the CCA is not clear. Although aggregating stomachs is considered a common space- and time-saving approach for sampling large numbers of fish stomachs, stomachs of individual fish are more informative (Stefánsson and Pálsson, 1997). Stomach data have been recorded, in Icelandic waters, for individual fish since 1993.
CCA proved to be a useful tool that can elucidate general patterns in feeding habits and indicate those variables that importantly determine diet composition. The present findings are generally consistent with previous results on the feeding habits of fish in Icelandic waters, but they also reveal some other unknown patterns and show the relative importance of the different determinants of diet composition. The explanatory variables considered for inclusion in the CCA model (fish size, season, depth, latitude, and longitude) were generally significant, with varying importance for different predators. As the study of Hovde et al. (2002) shows, more factors can be considered in future studies. Habitat type is one factor that may explain much of the variation in diets, but for which data were missing in the present study. Other factors that potentially may explain some variation in diet composition include interspecific competition, and the degree of temporal and spatial overlap between the predator and its major prey (Hovde et al., 2002).
An important factor determining the effectiveness of CCA in explaining the variation in diets is taxonomic resolution. Ideally, this resolution should not change significantly across gradients (e.g. with time). This was not entirely avoided in this study. Varying prey resolution may bias the results.
CCA revealed that the relative importance of different food categories changed with fish size. The most common shift was from small (usually soft-bodied, e.g. zooplankton, polychaetes, ophiuroids) to larger invertebrates (often with hard structures, e.g. decapods, echinoids, benthic molluscs) and fish prey. In some predators (particularly cod and saithe), piscivory was a dominant feeding habit in the largest fish. Saithe was unique among the predators examined in that its diet varied little with size. That could be explained by the fact that the smallest size classes of this species were not represented. Small saithe inhabit inshore waters (Froese and Pauly, 2006), but sampling for this study was mainly offshore. In addition, the species is semi-pelagic, and the two main components in the diet of saithe, euphausiids and fish, are consumed by a wide range of sizes. The ontogenetic shift is the natural result of morphological changes that accompany growth; as fish grow, they can handle bigger prey (Wootton, 1992). The general pattern of changes in diet composition with increasing predator size found in this study has also been observed in earlier studies of fish predators in Icelandic waters (Pálsson, 1983). Our results also correspond well with those reported by Høines and Bergstad (1999), who found a strong size-related feeding pattern in cod and haddock and no significant pattern in saithe in waters off southwestern Norway. Size was also the most important factor influencing the diets of hake, cod, and flatfish in the northwest Atlantic (Garrison and Link, 2000; Link and Garrison, 2002; Link et al., 2002). In contrast to these general findings, Hovde et al. (2002) found that spatial and temporal factors were most influential on the variation in diet composition of Northeast Arctic Greenland halibut, and that biotic variables (including predator size) were of less importance.
The observed spatial and seasonal variability in diets seems to have reflected patterns of prey availability. For instance, most predators exhibited more marked seasonal differences in diet composition between March and the second half of the year than between summer and autumn. This is largely due to the seasonality of capelin occurrence. In March, large quantities of capelin are recorded around Iceland, including large spawning concentrations off the southwest and west coasts (Pálsson and Björnsson, 1993). Nevertheless, capelin spend most of their life cycle north and northwest of Iceland (Pálsson and Björnsson, 1993; Vilhjálmsson, 1994), or even farther north, where the main feeding grounds are located (Vilhjálmsson, 1994). Concentrations of sandeels (which were important prey of saithe in summer) are seen in summer in shallower waters off the south, southwest, and west coasts off Iceland (Jónsson, 1983). Link and Garrison (2002) compared the diet of northwest Atlantic cod with the spatio-temporal distribution of its prey in the ecosystem, and found that most major prey species were consumed when they were abundant and their distribution overlapped with that of cod. Similarly, Trenkel et al. (2005) found that the frequency of fish prey in stomachs of fish predators in the Celtic Sea was on a spatial scale in agreement with the density-distribution patterns of the prey. It would be interesting to explore in further research how spatial and temporal factors affect the strength of predator–prey dependencies in the Icelandic marine ecosystem.
Among the fish predators examined in the present study, two main feeding guilds could be distinguished: (i) species preying mainly on echinoderms supplemented with fish and some other benthic invertebrates (such as polychaetes and benthic molluscs), and (ii) species preying mainly on crustaceans (benthic, planktonic, and nektonic) and fish. This gross division was based on pooled data for all areas. It could be revealing to study similarities in diets not only between different size groups of predators, in different seasons or averaged over all seasons (as in this study), but also separately for different habitats. A more detailed knowledge of intra- and interspecific dietary overlap (Høines and Bergstad, 1999, 2002), coupled with a knowledge of prey and predator abundances, could help to identify possible competitive interactions in the ecosystem.
In addition to environmental and ontogenetic factors, fishing may have an important effect on the trophic ecology of fish. Fishing may directly affect fish populations, but it may also have an impact on benthic habitats and communities (Frid et al., 1999; Rijnsdorp and Vingerhoed, 2001), and thus alter trophic relationships (e.g. Link et al., 2002). It would be revealing to study changes in diets on longer temporal scales using time-series (Hanson and Chouinard, 2002), particularly in relation to fishing effort. A better understanding of the joint impact of fishing and biological interactions is necessary for development and application of multispecies predictive models towards more effective fisheries management.
|
Acknowledgements |
|---|
We thank Ólafur K. Pálsson, Jason Link, Lorna Taylor, the Journal editor, and an anonymous referee for their insightful and constructive comments on earlier versions of this manuscript.
|
References |
|---|
|
TopIntroductionMaterial and methodsResultsDiscussionReferences |
|---|
-
Clarke K.R. and Warwick R.M. (2001) Change in Marine Communities: an Approach to Statistical Analysis and Interpretation 2nd edn (PRIMER-E, Plymouth, UK) 172 pp.
De Crespin de Billy V., Doledec S., Chessel D. (2000) Biplot presentation of diet composition data: an alternative for fish stomach contents analysis. Journal of Fish Biology 56:961–973.[Web of Science]
Frid C.L.J., Clark R.A., Hall J.A. (1999) Long-term changes in the benthos on a heavily fished ground off the NE coast of England. Marine Ecology Progress Series 188:13–20.[Web of Science]
FishBase (2006) World Wide Web electronic publication www.fishbase.org version (02/2006).
Garrison L.P. and Link J.S. (2000) Diets of five hake species in the northeast United States continental shelf ecosystem. Marine Ecology Progress Series 204:243–255.[Web of Science]
Gudmundsson K. (1998) Long-term variation in phytoplankton productivity during spring in Icelandic waters. ICES Journal of Marine Science 55:635–643.
Hanson J.M. and Chouinard G.A. (2002) Diet of Atlantic cod in the southern Gulf of St Lawrence as an index of ecosystem change, 1959–2000. Journal of Fish Biology 60:902–922.[Web of Science]
Høines Å.S. and Bergstad O.A. (1999) Resource sharing among cod, haddock, saithe and pollack on a herring spawning ground. Journal of Fish Biology 55:1233–1257.[Web of Science]
Høines Å.S. and Bergstad O.A. (2002) Food partitioning by flatfishes on a herring spawning ground. Sarsia 87:19–33.[CrossRef][Web of Science]
Hovde S.C., Albert O.T., Nilssen E.M. (2002) Spatial, seasonal and ontogenetic variation in the diet of Northeast Arctic Greenland halibut (Reinhardtius hippoglossoides). ICES Journal of Marine Science 59:421–437.
Ihaka R. and Gentleman R. (1996) R: a language for data analysis and graphics. Journal of Computational and Graphical Statistics 5:299–314.[CrossRef]
Jónsson G. (1983) Íslenskir fiskar. Fishes of Iceland(Fjölvaútgáfan, Reykjavík) 519 pp. (in Icelandic).
Legendre P. and Legendre L. (1998) Numerical Ecology 2nd English edn (Elsevier Science, Amsterdam) xv + 853 pp.
Link J.S., Bolles K., Milliken C.G. (2002) The feeding ecology of flatfish in the Northwest Atlantic. Journal of Northwest Atlantic Fishery Science 30:1–17.
Link J.S. and Garrison L.P. (2002) Trophic ecology of Atlantic cod Gadus morhua on the northeast US continental shelf. Marine Ecology Progress Series 227:109–123.[Web of Science]
Magnússon K.G. and Pálsson Ó.K. (1989) Trophic ecological relationships of Icelandic cod. Rapports et Procès-Verbaux des Réunions du Conseil International pour l'Exploration de la Mer 188:206–224.
Malmberg S.A. and Kristmannsson S.S. (1992) Hydrographic conditions in Icelandic waters, 1980–1989. In Dickson R.R., Mälkki P., Radach G., Saetre R., Sissenwine M.P. (Eds.). Hydrobiological Variability in the ICES Area, 1980–1989 195: pp. 76–92 ICES Marine Science Symposia.
MRI. (1997) Fjölstofnarannsóknir 1992–1995. Multispecies research 1992–1995. Hafrannsóknastofnun Fjölrit 57: 413 pp. (in Icelandic).
Oksanen J. (2005) Multivariate Analysis of Ecological Communities in R: vegan tutorial. World Wide Web electronic publication www.cc.oulu.fi/
jarioksa/opetus/metodi/vegantutor.pdf.
Pálsson Ó.K. (1983) The feeding habits of demersal fish species in Icelandic waters. Rit Fiskideildar 7:1–60.
Pálsson Ó. K. and Björnsson H. (1993) Spatial distributions of stomach content data of Icelandic cod. ICES 1993/CCC Symposium/No. 33. 32 pp.
Pálsson Ó.K., Jónsson E., Schopka S.A., Stefánsson G., Steinarsson B.Æ. (1989) Icelandic groundfish survey data used to improve precision in stock assessment. Journal of Northwest Atlantic Fishery Science 9:53–72.
Rijnsdorp A.D. and Vingerhoed B. (2001) Feeding of plaice Pleuronectes platessa L. and sole Solea solea (L.) in relation to the effects of bottom trawling. Journal of Sea Research 45:219–229.[CrossRef]
Ringler N.H. (1983) Variation in foraging tactics of fishes. In Noakes D.L.G., Lindquist D.G., Helfman G.S., Ward J.A. (Eds.). Predators and Prey in Fishes(Junk, The Hague) pp. 159–171 (cited in Wootton, 1992).
Stefánsson G. and Pálsson Ó.K. (1997) Statistical evaluation and modelling of the stomach contents of Icelandic cod (Gadus morhua). Canadian Journal of Fisheries and Aquatic Sciences 54:169–181.
ter Braak C.J.F. (1986) Canonical correspondence analysis: a new eigenvector technique for multivariate direct gradient analysis. Ecology 67:1167–1179.[CrossRef][Web of Science]
ter Braak C.J.F. and Verdonschot P.F.M. (1995) Canonical correspondence analysis and related multivariate methods in aquatic ecology. Aquatic Sciences 57:255–289.[CrossRef][Web of Science]
Trenkel V.M., Pinnegar J.K., Dawson W.A., du Buit M.H., Tidd A.N. (2005) Spatial and temporal structure of predator–prey relationships in the Celtic Sea fish community. Marine Ecology Progress Series 299:257–268.[Web of Science]
Vilhjálmsson H. (1994) The Icelandic capelin stock. Rit Fiskideildar 13: 281 pp.
Wootton R.J. (1992) Ecology of Teleost Fishes(Chapman and Hall, London) 404 pp.
Zar J.H. (1996) Biostatistical Analysis 3rd edn (Prentice Hall, Upper Saddle River, NJ) 662 pp.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  H. Armannsson, S. Th. Jonsson, J. D. Neilson, and G. Marteinsdottir Distribution and migration of saithe (Pollachius virens) around Iceland inferred from mark-recapture studies ICES J. Mar. Sci., July 1, 2007; 64(5): 1006 - 1016. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||