Abstract

Lundström, K., Hjerne, O., Lunneryd, S-G., and Karlsson, O. 2010. Understanding the diet composition of marine mammals: grey seals (Halichoerus grypus) in the Baltic Sea. – ICES Journal of Marine Science, 67: 1230–1239.

Dietary studies are important in understanding the ecological role of marine mammals and in formulating appropriate management plans in terms of their interactions with fisheries. The validity of such studies has, however, often been compromised by unrepresentative sampling procedures, resulting in false weight being given to external factors seeming to influence diet composition. The bias caused by non-random sampling was examined, using canonical correspondence analysis to assess how the prey species composition in digestive tract samples of Baltic grey seals (Halichoerus grypus) was related to spatial, temporal, and demographic factors and to whether the samples were collected in association with fishing gear or not (“sampling condition”). Geographic region explained the largest fraction of the observed variation, followed by sampling condition, age group, and year. Season and gender were not statistically significant. Segregation of the two age categories “pups” and “juveniles–adults”, and the two geographic categories “Baltic proper” and “Gulf of Bothnia” are proposed to estimate the diet and fish consumption of the Baltic grey seal population as a whole. Atlantic herring was the most commonly recovered prey item in all areas and age groups, followed by European sprat in the south, and common whitefish in the north. Pups had eaten relatively more small non-commercial species than older seals.

Introduction

To describe prey choice and food consumption of seals and other marine mammals, it is important to be aware of how different factors may influence the composition of the diet. Dietary differences between locations can reflect local variations in the abundance of prey species, for instance because of different oceanographic conditions (Benoit and Bowen, 1990; Hammond et al., 1994; Hammill et al., 2007). Variations in diet between years can be caused by long-term changes in the prey species composition, and seasonal changes will reflect variations on a shorter time-scale owing to spawning and migration of the prey species (Prime and Hammond, 1990; Bowen and Harrison, 1994, 2007; Hammond et al., 1994), changes in prey density and behavioural patterns of the seals, such as reproduction and moulting. Seal pups are to some extent limited in what prey they can catch owing to inexperience and incomplete development of hunting behaviour and physiology, e.g. diving capacity (Noren et al., 2005), suggesting the likelihood of age-related differences in prey choice. Dietary variations between the sexes may be caused by differences in energy requirements, niche utilization, and trade-offs between feeding and other activities that differ between females and males, as reviewed in Beck et al. (2007). Sampling condition, i.e. whether or not the samples were collected in association with fishing gear, may also influence the species composition of the dietary samples; for example, seals retrieved from fishing gear may contain relatively more of the species caught in that fishery (Pierce et al., 1991).

In our research, we searched for dietary patterns and examined the prey species composition of individual grey seals (Halichoerus grypus) in relation to foraging location, year, time of year, age and sex of the animal, and the sampling condition. Based on this analysis, in which the risk of bias in dietary composition data attributable to non-random sampling is estimated using canonical correspondence analysis (CCA), we answered questions such as what population characters are needed to estimate the dietary composition of the seal population, and whether the prey species composition in seals entrapped in fishing gear is in fact biased towards fish species targeted by the gear in question. The methodology described can be applied elsewhere in dietary studies, not only with seals.

The population size of grey seals in the Baltic Sea recovered from only a few thousand in the 1970s (Hårding and Härkönen, 1999) to some 21 000 in 2004 (Hiby et al., 2007). The growing numbers of seals have led to more interactions between seals and fisheries, and the annual cost of seal damage to fishing gear and catches was estimated in 2004 to exceed €5 million (Westerberg et al., 2008). Grey seals and coastal fisheries compete, at least to some extent, for the same resource. Accurate information about the composition of the diet and the factors influencing what is being consumed is a prerequisite in quantifying this indirect interaction and in understanding the ecological role of the Baltic grey seal in general.

Earlier comprehensive studies of the diet of grey seals in the Baltic Sea are limited and were based primarily on material collected in the 1960s and 1970s, mainly from the Baltic proper (Söderberg, 1972, 1975). A more recent study presents the prey species composition in the diet of 145 seals collected from both the Baltic proper and the Gulf of Bothnia between 2001 and 2004 (Lundström et al., 2007). Since then, more seals have been collected and more-detailed information of the sampled seals has been made available, making a thorough assessment possible. Furthermore, new and improved relationships between otolith size and fish size for a number of prey species have been documented; these have been used in generating the estimates of dietary composition presented here. Whereas Lundström et al. (2007) concentrated on describing a methodology for compensating the effects of digestive erosion when analysing stomach and intestine contents, our main focus was on assessing the effects of various external factors that could affect dietary estimates.

Material and methods

Stomachs and intestines were examined from 299 grey seals collected between 2001 and 2005. The samples, collected by commercial fishers and seal hunters, were made available by the Swedish Museum of Natural History, which collects tissue samples from Baltic seals as part of the Swedish marine environmental monitoring programme. All samples were stored in plastic bags at −20°C until examination.

Explanatory factors hypothesized as influencing the composition of the diet of individual seals were defined as geographic region [ICES Subdivisions (SD) 27, 29, 30, and 31], year (2001–2005), season (quarter of the year), age group (see below), gender, and sampling condition. Sampling condition was dependent on whether samples were associated with one of five different fishing gear types (entrapped in or shot next to) or were not associated with fishing gear. The gear types were (i) set traps and nets for Atlantic salmon (Salmo salar), sea trout (Salmo trutta), and common whitefish (Coregonus lavaretus), (ii) set traps for European eel (Anguilla anguilla), (iii) flatfish nets, (iv) Atlantic cod (Gadus morhua) nets, and (v) nets and trawls for Atlantic herring (Clupea harengus). Seals that did not contain any prey (n = 52) or with an incomplete set of factors defined (n = 37) were discarded, along with samples from the first quarter of the year, owing to their scarcity (n = 2), resulting in a dataset of 208 seals for the subsequent analyses. Seal age was determined from longitudinal sections of the canine teeth (Hewer, 1964), and the seals were divided into the following age groups: pups (<2 years old), juveniles (≥2 but <5 years old), and adults (≥5 years old). Each explanatory factor was split into two or more categories, e.g. geographic region into the four categories ICES SD27, 29, 30, and 31, coded as dummy variables with values of either 1 or 0. To assess the relative distribution of seals between the categories and how they were related, we used a principal component analysis (PCA) of the correlation matrix of the chosen categories (Lepš and Šmilauer, 2003).

The recovery and identification of digestive tract contents followed the procedures described in Lundström et al. (2007). To assess the sizes and numbers of prey individuals consumed, we used the mean of the values obtained from two different methods of estimation. One was based on all hard parts recovered, and the other was based only on otoliths recovered but including compensation for the assumed losses and size reduction in otoliths through digestion, using size and numerical correction factors (SCFs and NCFs, respectively). Erosion-class-specific SCFs (Tollit et al., 1997; Leopold et al., 1998) were used to estimate the original uneroded size of the ingested otoliths. The SCFs were calculated as the ratio between the average otolith size in erosion class 1 (minimally eroded) and erosion class 2 (obvious signs of erosion) for class 2 otoliths, and as the average otolith size ratio between classes 1 and 3 (highly eroded) for class 3 otoliths. We used known relationships between fish and otolith size to estimate the prey length and weight. The regression equations for Atlantic cod, common whitefish, Atlantic herring, and European sprat (Sprattus sprattus) were calculated from fish sampled in the Baltic Sea (Table 1). For Atlantic salmon, we used Härkönen (1986), whereas Leopold et al. (2001) was used for other species. To compensate for complete loss of otoliths, we used NCFs based on the relationship between size and recovery rate of otoliths from feeding experiments on captive seals. The NCFs (defined as the inverse of the recovery rate) were calculated from the average width of the size-corrected otoliths of each species in our dataset. For more details on our application of SCFs and NCFs, see Lundström et al. (2007). From the estimated individual weights and numbers of prey items, we estimated the total biomass of each prey species in each seal, and this was used as input in the multivariate analysis described below. Frequency of occurrence for a prey species was calculated as the number of seals containing the species, using all hard-part structures, in relation to the total number of seals containing prey.

Table 1.

Regression equations relating otolith width (OW) to total fish length (FL) and fish weight (FW) based on n individuals, valid for the Baltic Sea.

 FL (mm) = a + b × OW (mm)
 
FW (g) = c × OW (mm)d
 
Range
 
  
Species a b r2 c d r2 FL (mm) FW (g) n ICES SD 
Atlantic cod −81.44 78.82 0.96 0.54 3.84 0.97 120–870 12–6 100 107 25, 28 
Common whitefish −119.31 132.71 0.78 0.71 4.87 0.81 137–573 17–2 006 285 27, 30–31 
Atlantic herring −24.82 111.45 0.85 4.13 3.43 0.84 70–265 3–111 251 25, 27–31 
European sprat 3.45 97.21 0.80 6.17 2.45 0.78 65–140 2–17 149 25, 27–29 
 FL (mm) = a + b × OW (mm)
 
FW (g) = c × OW (mm)d
 
Range
 
  
Species a b r2 c d r2 FL (mm) FW (g) n ICES SD 
Atlantic cod −81.44 78.82 0.96 0.54 3.84 0.97 120–870 12–6 100 107 25, 28 
Common whitefish −119.31 132.71 0.78 0.71 4.87 0.81 137–573 17–2 006 285 27, 30–31 
Atlantic herring −24.82 111.45 0.85 4.13 3.43 0.84 70–265 3–111 251 25, 27–31 
European sprat 3.45 97.21 0.80 6.17 2.45 0.78 65–140 2–17 149 25, 27–29 

The samples were collected in the specified ICES SD in 2004 (see Figure 1).

We used eigenanalysis-based multivariate ordination techniques to examine the relationship between the diet and the explanatory factors. All ordinations were carried out with log(x + 1)-transformed biomasses to reduce the effects of extreme values and to approximate normally distributed data and homogeneity of variances better.

Detrended correspondence analysis (DCA) was used to examine the heterogeneity of the diet data and to determine whether ordination models based on linear or unimodal species responses to underlying variables were most appropriate (Lepš and Šmilauer, 2003).

CCA is a direct ordination technique where the ordination of a response matrix is constrained by an explanatory matrix. The technique is commonly used to examine community structure, ecological gradients, and distributions of organisms, but it is rarely implemented for dietary studies, especially among mammals. In the literature, there are some examples of studies that have used CCA to examine the dietary patterns and trophic ecology of fish (e.g. Magnan et al., 1994; Orr and Bowering, 1997; Link and Garrison, 2002; Cocheret de la Moriniere et al., 2003; Pouilly et al., 2003), but only a few have used the technique to analyse the diets of seals (Labansen et al., 2007; Lundström et al., 2007). We used CCA to model the relative quantities of prey species and to provide a measure of the extent of variation in the diet data that could be accounted for by the suggested explanatory factors (ter Braak, 1986). The null hypothesis of independence between the prey species composition and the proposed explanatory factors was tested statistically using CCA with forward selection and Monte Carlo permutation tests (n = 5000). In the forward selection, categories of the explanatory factors were chosen, one at a time, based on how strongly they explained the residual variation in the diet data. Categories were included if their addition improved the explanation of the diet data significantly (p < 0.05). When one of the categories (e.g. pups) of an explanatory factor (e.g. age group) was found to be significant, we added the remaining categories of the same factor (e.g. juveniles and adults) before testing the next factor. The fraction of the prey species variation described by each explanatory factor was assessed using partial CCA (pCCA), where the categories of the factor of interest were the only explanatory variables used, and the categories of the remaining factors were defined as covariables to account for their effects (ter Braak, 1988). The prey species variation described by one factor can be divided into one part that is described uniquely by the factor in question and another part that is described jointly, in conjunction with the other factors (Borcard et al., 1992; Lepš and Šmilauer, 2003). Only prey taxa with a goodness-of-fit ≥2.5% were included in the pCCA ordination plots. By removing taxa with a poor fit, we excluded prey species whose occurrence was not well explained by our categories and focused on those whose occurrence actually differed with respect to each factor. The influence of individual categories on the diet was investigated by pairwise selection of categories within each factor and tested statistically with Monte Carlo permutation tests (n = 5000).

The multivariate statistical computations used to examine the dietary patterns were carried out using the programme CANOCO 4.5 (ter Braak and Šmilauer, 2002). The function “down-weighting of rare species” was selected to prevent rare species from distorting the analyses by giving them less weight, yet still keeping them in the analysis.

Finally, we estimated the average proportions of the total biomass consumed accounted for by each prey species and used bootstrapping to assess the uncertainties around the estimates. Randomly, n samples were selected with replacement from a set of n seals. This procedure was iterated 2000 times, and bias-corrected 95% confidence limits around the mean proportion were calculated for the different species (Haddon, 2001).

Results

The geographic distribution of seals included in the analyses was relatively even between the southern (ICES SD25–29, Baltic proper) and northern (ICES SD30 and SD31, Gulf of Bothnia) Baltic Sea, but the samples were dominated by seals from SD27 and SD30 (Figure 1). The samples from SD25 (n = 2) and SD28 (n = 6) were added to SD27 in the analyses because of the small numbers for that subdivision and because the samples were collected close to SD27.

Figure 1.

Distribution of the analysed seals among the different geographic regions ICES SD25–31. SD25–29 constitute the Baltic proper and SD30 and SD31 the Gulf of Bothnia.

Figure 1.

Distribution of the analysed seals among the different geographic regions ICES SD25–31. SD25–29 constitute the Baltic proper and SD30 and SD31 the Gulf of Bothnia.

The seasonal distribution of samples varied between years; fewer than 5% of the samples were collected in 2001, whereas the other years were better represented (Figure 2). Both males (n = 112) and females (n = 96), as well as the three age groups (78 pups, 60 juveniles, and 70 adults) were well represented (Table 2, Figure 2).

Table 2.

Distribution of seals in relation to sampling condition, geographic region (ICES SD) and age group.

Sampling condition Total SD27 SD29 SD30 SD31 PUP JUV ADU 
Salmonid geara 39 27 15 11 13 
Eel traps 38 37 12 22 
Flatfish nets 25 17 18 23 
Atlantic cod nets 13 10 12 
Atlantic herring nets or trawls 11 10 
From fishing gear 126 73 13 37 63 38 25 
Not from fishing gear 82 17 43 20 15 22 45 
Sampling condition Total SD27 SD29 SD30 SD31 PUP JUV ADU 
Salmonid geara 39 27 15 11 13 
Eel traps 38 37 12 22 
Flatfish nets 25 17 18 23 
Atlantic cod nets 13 10 12 
Atlantic herring nets or trawls 11 10 
From fishing gear 126 73 13 37 63 38 25 
Not from fishing gear 82 17 43 20 15 22 45 

PUP refers to pups (<2 years old), JUV to juveniles (≥2 but <5 years old), and ADU to adults (≥5 years old).

aSet traps or nets for Atlantic salmon, sea trout, or common whitefish.

Figure 2.

Temporal distribution of samples over years and quarters of the year, by gender.

Figure 2.

Temporal distribution of samples over years and quarters of the year, by gender.

Approximately 60% of the seals were sampled in association with fishing gear, and most of those were entrapped in gear targeting Atlantic salmon, sea trout, common whitefish, and European eel (Table 2). Entrapped seals dominated the samples from areas SD27 and SD29, but in SD30, and particularly in SD31, seals retrieved by hunters were most used (Table 2).

In the PCA plot (Figure 3), the length of an arrow is a measure of the importance of the specific category of the explanatory factor, whereas the angle between two arrows indicates the correlation between those categories. Positively related categories point in the same direction, negatively related categories point in opposite directions, and unrelated categories are perpendicular to each other. Most of the pups sampled came from the southern Baltic (SD27–29) and most of the adults from the northern Baltic (SD30 and SD31; Figure 3). Pups were positively correlated with gear targeting flatfish and Atlantic cod, whereas juveniles were positively correlated with eel traps. Adults were positively correlated both with Atlantic herring gear and with the absence of fishing gear (Figure 3).

Figure 3.

PCA plot on the correlation matrix of the categories of explanatory factors. The categories are ICES SD27, SD29, SD30, and SD31 for the geographic region, pups (PUP, <2 years old), juveniles (JUV, ≥2 but <5 years old), and adults (ADU, ≥5 years old) for the age group, females (F) and males (M) for the gender, 2001–2005 for the year and Q2–Q4 for the quarter of the year. The sampling condition categories are salmonid gear (set traps and nets for Atlantic salmon, sea trout, and common whitefish; SALMO), set traps for European eel (EEL), flatfish nets (FLAT), Atlantic cod nets (COD), nets and trawls for Atlantic herring (HERR), and not collected from fishing gear (NO GEAR).

Figure 3.

PCA plot on the correlation matrix of the categories of explanatory factors. The categories are ICES SD27, SD29, SD30, and SD31 for the geographic region, pups (PUP, <2 years old), juveniles (JUV, ≥2 but <5 years old), and adults (ADU, ≥5 years old) for the age group, females (F) and males (M) for the gender, 2001–2005 for the year and Q2–Q4 for the quarter of the year. The sampling condition categories are salmonid gear (set traps and nets for Atlantic salmon, sea trout, and common whitefish; SALMO), set traps for European eel (EEL), flatfish nets (FLAT), Atlantic cod nets (COD), nets and trawls for Atlantic herring (HERR), and not collected from fishing gear (NO GEAR).

The 247 seals with prey remains in our complete dataset contained a total of 13 630 otoliths from 23 different taxa. From the 208 seals with a complete set of explanatory variables, 12 103 otoliths from 22 prey taxa were identified (Table 3). Rare prey species, those in <1% of the samples (n = 6), were discarded from further analyses, leaving 16 prey taxa for subsequent analysis (Table 3). The total weight of the rare prey excluded was estimated as <0.08% of the total biomass consumed, whereas the dominant prey species, Atlantic herring, European sprat, and common whitefish, contributed an estimated 84.7% of the total. The number of prey species consumed by individual seals was generally low. A single prey taxon was observed in 41% of the samples, and two and three prey taxa were found in 34% and 17% of the samples, respectively.

Table 3.

Occurrence of prey taxa.

Prey Taxon Frequency of occurrence (%) 
Atlantic herring Clupea harengus 85 
European sprat Sprattus sprattus 30 
Common whitefish Coregonus lavaretus 17 
Sandeel Ammodytes spp. 12 
Viviparous blenny Zoarces viviparus 
Atlantic cod Gadus morhua 
Cyprinids Cyprinidae 
Atlantic salmon and sea trout Salmo spp. 
Flatfish Pleuronectiformes 
Perch Perca fluviatilis 
European eel Anguilla anguilla 
Smelt Osmerus eperlanus 
Sculpins Cottidae 
Gobids Gobiidae 
Burbot Lota lota 
Pike Esox lucius 
Rare prey found in <1% of the seals excluded from the analysis 
 Ruffe Gymnocephalus cernuus 
 Garpike Belone belone  
 Rock gunnel Pholis gunnellus  
 Zander Sander lucioperca  
 Lampreys Petromyzontidae  
 Benthic isopod Saduria entomon  
Prey Taxon Frequency of occurrence (%) 
Atlantic herring Clupea harengus 85 
European sprat Sprattus sprattus 30 
Common whitefish Coregonus lavaretus 17 
Sandeel Ammodytes spp. 12 
Viviparous blenny Zoarces viviparus 
Atlantic cod Gadus morhua 
Cyprinids Cyprinidae 
Atlantic salmon and sea trout Salmo spp. 
Flatfish Pleuronectiformes 
Perch Perca fluviatilis 
European eel Anguilla anguilla 
Smelt Osmerus eperlanus 
Sculpins Cottidae 
Gobids Gobiidae 
Burbot Lota lota 
Pike Esox lucius 
Rare prey found in <1% of the seals excluded from the analysis 
 Ruffe Gymnocephalus cernuus 
 Garpike Belone belone  
 Rock gunnel Pholis gunnellus  
 Zander Sander lucioperca  
 Lampreys Petromyzontidae  
 Benthic isopod Saduria entomon  

The gradient lengths of the first two ordination axes (4.3 and 5.4, respectively), obtained by DCA, indicated that ordination models using unimodal species responses would fit the data adequately, so CCA was used.

Our full set of explanatory factors accounted for 18.8% of the total prey species variation. Geographic region was the main factor, followed by sampling condition and age group (Table 4). A part of the variation explained by each factor was also explained by other factors, so the sum of the variations explained by the individual factors exceeded the variation explained by all factors. In the forward selection, the factors determined to be statistically significant were geographic region, age group, sampling condition, and year, whereas gender and season did not contribute significantly as explanatory factors (Table 4). Together, these four factors explained 17.7% of the prey species variation.

Table 4.

Diet variation accounted for by the explanatory factors and p-values from forward selection.

Explanatory factor % explained p-value 
Geographic region 9.1 <0.001 
Year 4.8 0.006 
Season 2.3 0.09 
Age group 6.7 <0.001 
Sex 0.7 0.9 
Sampling condition 8.0 0.001 
All factors together 18.8  
Explanatory factor % explained p-value 
Geographic region 9.1 <0.001 
Year 4.8 0.006 
Season 2.3 0.09 
Age group 6.7 <0.001 
Sex 0.7 0.9 
Sampling condition 8.0 0.001 
All factors together 18.8  

The p-values are valid for the category of each factor with the greatest explanatory power, and the remaining categories were added before the next factor was tested.

In the pCCA plots (Figure 4a–d), a prey species symbol close to a specific factor category symbol indicates a larger average weight of that prey species in seals belonging to that specific category, compared with prey species farther away. The closer the two categories are to each other, the more similar are the estimated diets in seals belonging to those categories.

Figure 4.

pCCA plots with the explanatory factors geographic region (ICES SDs), age group, sampling condition, and year in panels (a), (b), (c), and (d), respectively. The factors not used as explanatory were defined as covariables. The size of the symbols corresponds to the goodness-of-fit of the prey taxa. Only taxa with a goodness-of-fit ≥2.5% are included.

Figure 4.

pCCA plots with the explanatory factors geographic region (ICES SDs), age group, sampling condition, and year in panels (a), (b), (c), and (d), respectively. The factors not used as explanatory were defined as covariables. The size of the symbols corresponds to the goodness-of-fit of the prey taxa. Only taxa with a goodness-of-fit ≥2.5% are included.

The dietary patterns of seals collected from SD27 differed from those from both SD30 (p = 0.0002, f-ratio = 5.35) and SD31 (p = 0.0046, f-ratio = 2.62), but between SD30 and SD31 there were no significant differences (p = 0.91, f-ratio = 0.37). The prey species composition in SD29 did not differ significantly from any of the other areas, but the permutation tests suggested that it was more similar to the diet in SD27 (p = 0.33, f-ratio = 1.14) than to that in SD30 (p = 0.12, f-ratio = 2.03) and SD31 (p = 0.17, f-ratio = 1.72). Among the seals collected from SD27, European sprat and sandeel were relatively more abundant than among the seals from SD30 and SD31, where Atlantic herring and, to a lesser degree, common whitefish were relatively more abundant (Figure 4a).

The estimated prey species composition of the diet of pups differed significantly from that of both juveniles (p = 0.035, f-ratio = 1.93) and adults (p = 0.0004, f-ratio = 3.48), but there was no significant difference between the diets of juveniles and adults (p = 0.07, f-ratio = 1.76). Among the pups, viviparous blenny (Zoarces viviparus) and sandeel were relatively more abundant as prey items than among the older seals. Atlantic cod seemed to be a more characteristic prey item among juveniles, whereas burbot (Lota lota), cyprinids, Atlantic salmon, and sea trout seemed to be more common in the adult diet (Figure 4b). The true importance of burbot is most likely moderate, because just four specimens were recorded in our study, though all of them consumed by adult seals.

The prey species composition in seals collected in 2003 differed significantly from that in seals collected in 2002 (p = 0.0362, f-ratio = 1.84), 2004 (p = 0.0054, f-ratio = 2.24), and 2005 (p = 0.0056, f-ratio = 2.48). Moreover, the prey species composition in seals collected in 2004 differed from that in those collected in 2005 (p = 0.003, f-ratio = 2.78). The relative abundance of cyprinids was greater among the seals collected in 2003 than in other years, whereas the diets of seals collected in 2002 were characterized by Atlantic salmon, sea trout, Atlantic cod, and European sprat. Atlantic cod, Atlantic salmon, and sea trout also seemed to be relatively more abundant in the diet of seals collected in 2005 than in seals collected in 2004 (Figure 4d).

The prey species composition in seals not collected from fishing gear differed significantly from that in seals collected from salmonid gear (p = 0.0074, f-ratio = 2.44) and eel traps (p = 0.0362, f-ratio = 1.98). Otherwise, no significant differences were observed (p-values > 0.05). The relative abundance of Atlantic herring seemed to be less among seals collected from salmonid gear and eel traps than from seals collected other than from fishing gear, whereas the relative abundances of Atlantic salmon, sea trout, European eel, and flatfish were less related to sampling condition (Figure 4c).

In light of these results, we propose the division of the samples into only two age groups, pups vs. juveniles and adults, and two geographic regions, Baltic proper vs. Gulf of Bothnia, as a straightforward means of presenting the estimated diets (commented on further below).

Atlantic herring constituted the largest fraction of the biomass consumed in all age groups and all geographic regions, followed by European sprat in the Baltic proper and common whitefish in the Gulf of Bothnia (Table 5). Among the samples from the Baltic proper, viviparous blenny and sandeel followed among pups, and Atlantic cod and cyprinids among older animals (Table 5). In the samples from the Gulf of Bothnia, the most common prey species after Atlantic herring and common whitefish were sculpins (Cottidae) and viviparous blenny among pups, and Atlantic salmon, sea trout, and cyprinids among juveniles and adults (Table 5). Atlantic herring seemed to be more important among juveniles and adults in the Gulf of Bothnia than in the Baltic proper.

Table 5.

Average fractions of total biomass consumed accounted for by each prey taxon, with bootstrapped CI, in (top) the Baltic proper and (bottom) the Gulf of Bothnia.

 Pups
 
Juveniles and adults
 
Prey Mean 95% CI Mean 95% CI 
Baltic proper 
 Atlantic herring 0.593 0.518–0.668 0.374 0.281–0.491 
 European sprat 0.202 0.148–0.268 0.190 0.112–0.288 
 Sandeel 0.034 0.019–0.057 0.004 0.001–0.008 
 Viviparous blenny 0.067 0.023–0.127 0.011 0.001–0.039 
 Atlantic cod 0.025 0.001–0.068 0.104 0.054–0.171 
 Cyprinids 0.019 0.003–0.047 0.096 0.038–0.174 
 Atlantic salmon and sea trout Absent Absent 
 Flatfish 0.030 0.006–0.070 0.021 0–0.082 
 Perch 0.006 0–0.022 0.021 0.002–0.058 
 European eel 0.007 0–0.023 0.064 0.020–0.128 
 Smelt 0.012 0.002-0.032 0.012 0–0.061 
 Sculpins Absent Absent 
 Gobids <0.001 0–0.001 <0.001 0.00–0.00 
 Burbot Absent 0.003 0–0.005 
 Pike Absent 0.024 0–0.081 
Gulf of Bothnia 
 Atlantic herring 0.581 0.406–0.727 0.712 0.627–0.789 
 European sprat 0.017 0.004–0.045 0.009 0.001–0.025 
 Common whitefish 0.135 0.048–0.253 0.133 0.085–0.197 
 Sandeel 0.012 0.001–0.033 0.001 0–0.004 
 Viviparous blenny 0.054 0.004–0.179 0.014 0.001–0.046 
 Atlantic cod Absent Absent 
 Cyprinids 0.014 0–0.041 0.023 0.004–0.059 
 Atlantic salmon and sea trout 0.020 0–0.040 0.081 0.040–0.138 
 Flatfish Absent 0.003 0–0.008 
 Perch 0.023 0–0.064 0.015 0.002–0.046 
 European eel 0.046 0–0.231 Absent 
 Smelt 0.001 0–0.002 <0.001 0–0.002 
 Sculpins 0.097 0.018–0.214 0.001 0–0.005 
 Gobids Absent Absent 
 Burbot Absent 0.006 0–0.017 
 Pike Absent 0.002 0–0.004 
 Pups
 
Juveniles and adults
 
Prey Mean 95% CI Mean 95% CI 
Baltic proper 
 Atlantic herring 0.593 0.518–0.668 0.374 0.281–0.491 
 European sprat 0.202 0.148–0.268 0.190 0.112–0.288 
 Sandeel 0.034 0.019–0.057 0.004 0.001–0.008 
 Viviparous blenny 0.067 0.023–0.127 0.011 0.001–0.039 
 Atlantic cod 0.025 0.001–0.068 0.104 0.054–0.171 
 Cyprinids 0.019 0.003–0.047 0.096 0.038–0.174 
 Atlantic salmon and sea trout Absent Absent 
 Flatfish 0.030 0.006–0.070 0.021 0–0.082 
 Perch 0.006 0–0.022 0.021 0.002–0.058 
 European eel 0.007 0–0.023 0.064 0.020–0.128 
 Smelt 0.012 0.002-0.032 0.012 0–0.061 
 Sculpins Absent Absent 
 Gobids <0.001 0–0.001 <0.001 0.00–0.00 
 Burbot Absent 0.003 0–0.005 
 Pike Absent 0.024 0–0.081 
Gulf of Bothnia 
 Atlantic herring 0.581 0.406–0.727 0.712 0.627–0.789 
 European sprat 0.017 0.004–0.045 0.009 0.001–0.025 
 Common whitefish 0.135 0.048–0.253 0.133 0.085–0.197 
 Sandeel 0.012 0.001–0.033 0.001 0–0.004 
 Viviparous blenny 0.054 0.004–0.179 0.014 0.001–0.046 
 Atlantic cod Absent Absent 
 Cyprinids 0.014 0–0.041 0.023 0.004–0.059 
 Atlantic salmon and sea trout 0.020 0–0.040 0.081 0.040–0.138 
 Flatfish Absent 0.003 0–0.008 
 Perch 0.023 0–0.064 0.015 0.002–0.046 
 European eel 0.046 0–0.231 Absent 
 Smelt 0.001 0–0.002 <0.001 0–0.002 
 Sculpins 0.097 0.018–0.214 0.001 0–0.005 
 Gobids Absent Absent 
 Burbot Absent 0.006 0–0.017 
 Pike Absent 0.002 0–0.004 

Discussion

Investigations of the diets of marine mammals based on contents of the digestive tract are liable to several sources of error, some linked to how well the prey remains represent the actual diet. Digestive erosion in the gastro-intestinal tract decreases the size of otoliths, and complete erosion of hard parts results in reduced recovery rates of faster-eroding specimens. When the prey consists of larger fish such as salmon, trout, and cod, the seals may only take bites out of the flesh, so taking no identifiable hard parts to be found in the digestive tract and leading to an underestimate of the quantity of these larger prey fish consumed. Other possible errors are linked to sampling strategies. Random sampling is seldom or never achieved in marine mammal studies, where most samples are collected from fishing gear set in a specific area and for a specific catch. Even hunted animals are seldom collected randomly, because hunting also takes place in areas where the shot animals can be retrieved, typically close to shore or on sea ice. Our samples were collected opportunistically rather than randomly, which made our material unbalanced regarding correlations between factors. For example, most seals were hunted in the north and entrapped in the south, so sampling condition could to a large extent be explained by area. However, by analysing the effect of one factor at a time and accounting for the effects of the other factors, we reduced this bias and were able to reveal the explicit effect of each factor on the dietary estimates.

The remains of different prey species are influenced by digestive erosion to differing extents, and if this fact is not taken into consideration, the estimates of both prey size and prey species composition may be biased (Jobling and Breiby, 1986; Harvey, 1989; Tollit et al., 1997; Bowen, 2000; Christiansen et al., 2005; Grellier and Hammond, 2006). As other studies have done, we back-calculated the uneroded size of the otoliths, using SCFs based on the ratios of mean eroded otolith size to mean uneroded otolith size (Harvey, 1989; Prime and Hammond, 1990; Hammond et al., 1994; Tollit et al., 1997; Grellier and Hammond, 2006). We used also other prey remains than otoliths, along with appropriate NCFs, to compensate for the likelihood of complete digestion of some otoliths, a method used before to increase the recovery of prey data (Harvey, 1989; Tollit et al., 1997, 2007; Browne et al., 2002; Orr et al., 2004; Grellier and Hammond, 2006).

How well the estimated dietary composition relates to spatial, temporal, and demographic variations and sampling condition is important to know when making assessments of the diet. We calculated that, taken together, the factors geographic region, sampling condition, age group, and year, in that order, explained a statistically significant portion (17.7%) of the variation in prey species composition among the seals analysed. Low levels of explained variation in species data are common in ecological studies, but important patterns can nevertheless be found (ter Braak and Šmilauer, 2002). The high level of unexplained prey species variation remaining is probably caused by a large portion of natural variation, but may also depend on some additional factor(s) not considered.

CCA is a robust method that is little affected by skewed species distributions, high frequencies of zero observations, or deviations from unimodal species responses to the explanatory variables (Palmer, 1993; ter Braak and Verdonschot, 1995), all of which conditions were present in our data. The factors selected in the forward selection do not represent the only possible choice of factors, but a sufficient subset that explains a significant part of the data. However, because of the limited number of factors examined and because the factors selected in the forward selection also explained the largest portion of prey species variation (Table 4), we think that these factors constituted an adequate subset. For further analysis, e.g. average dietary composition and prey consumption for the entire seal population, we are primarily interested in excluding factors that do not influence the seal diet. Therefore, it could be argued that the significance level (pcrit = 0.05) is rather low, but this is at least partly counteracted by the increased risk of type I error as a result of the multiple statistical tests used during the forward selection.

Grey seals have been reported to consume a range of species, taking advantage of locally and seasonally abundant prey, and the diet is generally determined by the availability of potential prey species. Owing to the distinct salinity gradient in the Baltic Sea, with higher salinity in the south, and accompanying changes in prey species composition (Ojaveer et al., 1981), geographic variations in the diet were expected. Differences in diet between regions in the Baltic Sea have indeed been observed in previous studies, both directly (Söderberg, 1972; Lundström et al., 2007) and indirectly, based on fatty acid composition in pups (Karlsson, 2003). Geographic variations have also been found in grey seal diet studies from other areas (Bowen and Harrison, 1994; Hammond et al., 1994; Mikkelsen et al., 2002; Hammill et al., 2007). Nevertheless, we found Atlantic herring to be the most common prey all over the Baltic Sea, followed by European sprat (marine origin) in the south and common whitefish (freshwater origin) in the north. Although we separated our samples into smaller geographic regions than Lundström et al. (2007), our results imply that dividing the Baltic Sea into two dietarily significantly different areas, the Baltic proper and the Gulf of Bothnia, as done by Lundström et al. (2007), is sufficient.

We found support for annual variations in the diet, as has also been documented from other areas (Walton and Pomeroy, 2003; Bowen and Harrison, 2007; Hammill et al., 2007). However, we found no reason to believe that there were any large changes among the prey populations during the limited study period covered by our samples, and we believe that any changes in seal diet attributable to changing availability of prey species are more likely to be seen on a decennial than an annual scale. Additionally, the year factor described a relatively small part of the prey species variation. We therefore suggest pooling samples from different years and treating the present study as a single period. Looking at a longer time-scale, our study confirms the observations by Lundström et al. (2007) that the importance of Atlantic herring and European sprat in the diet has increased and the importance of Atlantic cod has decreased since the late 1960s and early 1970s (Söderberg, 1972, 1975).

Contrary to the results of several studies from other areas (Prime and Hammond, 1990; Bowen and Harrison, 1994; Hammond et al., 1994; Hauksson and Bogason, 1997; Beck et al., 2007; Hammill et al., 2007), but in line with previous studies from the Baltic Sea (Söderberg, 1972; Lundström et al., 2007), we found no support in our material for the notion that seasonal factors can affect dietary composition, when the effects of the other factors were accounted for.

Consistent with previous studies of grey seals from the Baltic Sea (Söderberg, 1972; Lundström et al., 2007), we found support for age-related variations in the diet, except for Atlantic herring, common whitefish, and European sprat, which were common prey among all age groups. The prey species composition of pup diets differed from that of juveniles and adults and was characterized by there being more small, non-commercial species such as sandeel and viviparous blenny. The prey species composition of juveniles and adults, on the other hand, was characterized by more Atlantic cod, cyprinids, Atlantic salmon, and sea trout. Prey species variation with age has also been observed in previous studies (Hauksson and Bogason, 1997; Mikkelsen et al., 2002; Beck et al., 2007), but not in others (Murie and Lavigne, 1992; Bowen et al., 1993). Possible explanations for the age-related differences found may be that seals of different age forage in different habitats or that it is easier for pups to catch and handle small fish than larger, perhaps faster, fish.

Differences in habitat use and diving behaviour between female and male grey seals have been reported by Beck et al. (2003), Austin et al. (2006), and Breed et al. (2006). According to Hauksson and Bogason (1997), Mikkelsen et al. (2002), and Beck et al. (2007), the diet differs on a gender-related basis, but neither Bowen et al. (1993) nor Hammill et al. (2007) considered there to be gender-related differences. We found no evidence of differences in dietary composition between female and male grey seals, agreeing with previous observations from the Baltic (Söderberg, 1972; Lundström et al., 2007).

Söderberg (1972) suggested that the diet of seals collected from fishing gear in the Baltic might be biased towards the species caught in the gear. In the study by Lundström et al. (2007), no differences in prey species composition were found among seals caught in fishing gear and seals collected elsewhere. However, that study pooled all seals collected from any fishing gear, disregarding the type of gear, so any effect on the diet attributable to the use of a particular gear would have been obscured. When categorizing the entrapped seals in our study according to the type of fishing gear in which they were caught, we found that the prey species composition in seals not collected from fishing gear differed significantly from the diet of seals collected from both eel traps and salmonid gear. Our conclusion that the sampling condition actually does seem to affect seal diet was consistent with the results of Pierce et al. (1991). The main factor behind the differences found in our study seemed to be the relative increase in the abundance of Atlantic herring in samples not collected from fishing gear, rather than which species the gear was primarily targeting. The diet assessed from seals obtained from salmonid gear and eel traps might therefore underestimate the abundance of Atlantic herring compared with that in seals collected elsewhere. The abundance in the diet of other commercially important fish, such as Atlantic salmon, sea trout, European eel, and flatfish, did not seem to be influenced by sampling condition to the same extent (Figure 4c). That the diet of seals entrapped in fishing gear differs from seals collected elsewhere does not necessarily mean that the gear is used as a food source for species not easily caught elsewhere. The diet of seals entrapped in fishing gear might simply reveal species that are common in the area. If this is the case, the diet of seals entrapped in gear targeting Atlantic salmon, sea trout, common whitefish, and European eel might simply be more representative of seals that forage inshore, whereas the diet of seals collected from Atlantic cod and flatfish nets, or shot at distant skerries or on the ice, might be more representative of offshore foraging. What samples are actually being collected from fishing gear may also depend on the size of the seals and the willingness of individual fishers to report and deliver the seals. Larger seals caught in nets and set traps may be too heavy to be hauled aboard, or the weight of the seal may tear the net, causing the seal to be lost, with a resultant bias towards the recovery of more small seals.

Concluding, temporal changes in diet attributable to changes in prey availability seem unlikely to be detectable in our material during this relatively short time-span, and we propose that the collection period be regarded as a single period. Moreover, pooling the data for several years increases the power of the age- and area-related differences. We found that the occurrence of some prey species in the diet depended on the sampling condition, and we emphasize the importance of being aware of not only where and when, but also how, the samples are collected. In our material, the greatest risk of bias seemed to be in underestimating the abundance of Atlantic herring in the diet when analysing samples collected from eel traps and salmonid gear.

For future diet studies, we propose that effort be made to achieve a balanced and representative sampling of seals in terms of age group, area, and sampling condition, taking into account both latitudinal and inshore–offshore gradients in prey species composition. We acknowledge this is difficult to achieve in practice, but being aware of the problem with biased sampling can decrease the risk of interpreting the diet deduced from such sampling as representative of the entire population. Balanced sampling in relation to gender and season seems to be less important, but is nevertheless desirable, if possible. To be able to evaluate how diets relate to possible temporal fluctuations, samples need to be collected continuously during a sufficient period or in different periods. To use diet studies such as this to estimate the total prey consumption of the Baltic grey seal population, conclusions about the dietary composition itself would need to be combined with data on population structure, seasonal energy intake, spatial distribution, and temporal variations. Furthermore, digestive tract contents only offer evidence of the most recent few meals before capture, based on identifiable hard-part prey remains eaten by the seals investigated, and do not give an integrated picture of the food intake over longer time-scales. Estimates of diet based on gastro-intestinal contents should therefore be complemented with alternative study methods, such as analyses of faecal contents, fatty acids, and stable isotopes.

We estimated the diet composition following the suggested grouping of the samples into two geographic areas (Baltic proper and Gulf of Bothnia) and two age groups (pups and older seals). Our sample of 208 seals most likely gives a reasonable picture of the occurrence of the common prey species (Table 5), but the large confidence intervals around the rarer prey species demonstrate that it is insufficient to describe the diet in detail. Nevertheless, our analyses offer an example of how the influence of different factors on the diet can be determined and how biases attributable to non-random sampling can be avoided and accounted for. This is essential for adequately estimating the consumption of prey and for assessing the ecological role of Baltic grey seals, including possible competition with commercial fisheries.

Acknowledgements

The study was financed by the EU project FRAP (Framework of Biodiversity Reconciliation Action Plans, www.frap-project.net), the Swedish Board of Fisheries, and the Stockholm Marine Research Centre. We thank K. Alexandersson and M. Boström (Institute of Coastal Research, Swedish Board of Fisheries) for analysis of prey remains. Grey seal samples were kindly made available by the staff at the Department of Contaminant Research, Swedish Museum of Natural History, in particular by C. Moraeus, M. Molnár-Veress, and A. Roos. Two anonymous reviewers provided valuable comments on the manuscript, and Graham Timmins improved the language.

References

Austin
D.
Bowen
W. D.
McMillan
J. I.
Boness
D. J.
Stomach temperature telemetry reveals temporal patterns of foraging success in a free-ranging marine mammal
Journal of Animal Ecology
 , 
2006
, vol. 
75
 (pg. 
408
-
420
)
Beck
C. A.
Bowen
W. D.
McMillan
J. I.
Iverson
S. J.
Sex differences in diving at multiple temporal scales in a size-dimorphic capital breeder
Journal of Animal Ecology
 , 
2003
, vol. 
72
 (pg. 
979
-
993
)
Beck
C. A.
Iverson
S. J.
Bowen
W. D.
Blanchard
W.
Sex differences in grey seal diet reflect seasonal variation in foraging behaviour and reproductive expenditure: evidence from quantitative fatty acid signature analysis
Journal of Animal Ecology
 , 
2007
, vol. 
76
 (pg. 
490
-
502
)
Benoit
D.
Bowen
W. D.
Seasonal and geographic variation in the diet of grey seals (Halichoerus grypus) in eastern Canada
Canadian Bulletin of Fisheries and Aquatic Sciences
 , 
1990
, vol. 
222
 (pg. 
215
-
226
)
Borcard
D.
Legendre
P.
Drapeau
P.
Partialling out the spatial component of ecological variation
Ecology
 , 
1992
, vol. 
73
 (pg. 
1045
-
1055
)
Bowen
W. D.
Reconstruction of pinniped diets: accounting for complete digestion of otoliths and cephalopod beaks
Canadian Journal of Fisheries and Aquatic Sciences
 , 
2000
, vol. 
57
 (pg. 
898
-
905
)
Bowen
W. D.
Harrison
G.
Seasonal and interannual variability in grey seal diets on Sable Island, eastern Scotian Shelf
NAMMCO Scientific Publications
 , 
2007
, vol. 
6
 (pg. 
123
-
134
)
Bowen
W. D.
Harrison
G. D.
Offshore diet of grey seals Halichoerus grypus near Sable Island, Canada
Marine Ecology Progress Series
 , 
1994
, vol. 
112
 (pg. 
1
-
11
)
Bowen
W. D.
Lawson
J. W.
Beck
B.
Seasonal and geographic variation in the species composition and size of prey consumed by grey seals (Halichoerus grypus) on the Scotian Shelf
Canadian Journal of Fisheries and Aquatic Sciences
 , 
1993
, vol. 
50
 (pg. 
1768
-
1778
)
Breed
G. A.
Bowen
W. D.
McMillan
J. I.
Leonard
M. L.
Sexual segregation of seasonal foraging habitats in a non-migratory marine mammal
Proceedings of the Royal Society of London, Series B: Biological Sciences
 , 
2006
, vol. 
273
 (pg. 
2319
-
2326
)
Browne
P.
Laake
J. L.
Delong
R. L.
Improving pinniped diet analyses through identification of multiple skeletal structures in fecal samples
Fishery Bulletin US
 , 
2002
, vol. 
100
 (pg. 
423
-
433
)
Christiansen
J. S.
Gamst Moen
A-G.
Hansen
T. H.
Nilssen
K. T.
Digestion of capelin, Mallotus villosus (Muller), herring, Clupea harengus L., and polar cod, Boreogadus saida (Lepechin), otoliths in a simulated seal stomach
ICES Journal of Marine Science
 , 
2005
, vol. 
62
 (pg. 
86
-
92
)
Cocheret de la Moriniere
E.
Pollux
B. J. A.
Nagelkerken
I.
van der Velde
G.
Diet shifts of Caribbean grunts (Haemulidae) and snappers (Lutjanidae) and the relation with nursery-to-coral reef migrations
Estuarine, Coastal and Shelf Science
 , 
2003
, vol. 
57
 (pg. 
1079
-
1089
)
Grellier
K.
Hammond
P. S.
Robust digestion and passage rate estimates for hard parts of grey seal (Halichoerus grypus) prey
Canadian Journal of Fisheries and Aquatic Sciences
 , 
2006
, vol. 
63
 (pg. 
1982
-
1998
)
Haddon
M.
Modelling and Quantitative Methods in Fisheries
2001
Boca Raton, FL
Chapman and Hall/CRC
Hammill
M. O.
Stenson
G. B.
Proust
F.
Carter
P.
McKinnon
D.
Feeding by grey seals in the Gulf of St Lawrence and around Newfoundland
NAMMCO Scientific Publications
 , 
2007
, vol. 
6
 (pg. 
135
-
152
)
Hammond
P. S.
Hall
A. J.
Prime
J. H.
The diet of grey seals around Orkney and other island and mainland sites in north-eastern Scotland
Journal of Applied Ecology
 , 
1994
, vol. 
31
 (pg. 
340
-
350
)
Hårding
K. C.
Härkönen
T. J.
Development in the Baltic grey seal (Halichoerus grypus) and ringed seal (Phoca hispida) populations during the 20th century
Ambio
 , 
1999
, vol. 
28
 (pg. 
619
-
625
)
Härkönen
T.
Guide to the Otoliths of the Bony Fishes of the Northeast Atlantic
1986
Hellerup, Denmark
Danibu ApS
pg. 
256
 
Harvey
J. T.
Assessment of errors associated with harbour seal (Phoca vitulina) faecal sampling
Journal of Zoology (London)
 , 
1989
, vol. 
219
 (pg. 
101
-
111
)
Hauksson
E.
Bogason
V.
Comparative feeding of grey (Halichoerus grypus) and common seals (Phoca vitulina) in coastal waters of Iceland, with a note on the diet of hooded (Cystophora cristata) and harp seals (Phoca groenlandica)
Journal of Northwest Atlantic Fishery Science
 , 
1997
, vol. 
22
 (pg. 
125
-
135
)
Hewer
H. R.
The determination of age, sexual maturity, longevity and a life-table in the grey seal (Halichoerus grypus)
Proceedings of the Zoological Society of London
 , 
1964
, vol. 
142
 (pg. 
593
-
624
)
Hiby
L.
Lundberg
T.
Karlsson
O.
Watkins
J.
Jussi
M.
Jussi
I.
Helander
B.
Estimates of the size of the Baltic grey seal population based on photo-identification data
NAMMCO Scientific Publications
 , 
2007
, vol. 
6
 (pg. 
163
-
175
)
Jobling
M.
Breiby
A.
The use and abuse of fish otoliths in studies of feeding habits of marine piscivores
Sarsia
 , 
1986
, vol. 
71
 (pg. 
265
-
274
)
Karlsson
O.
Population structure, movements and site fidelity of grey seals in the Baltic Sea
2003
Sweden
PhD thesis, Department of Zoology, Stockholm University
Labansen
A. L.
Lydersen
C.
Haug
T.
Kovacs
K. M.
Spring diet of ringed seals (Phoca hispida) from northwestern Spitsbergen, Norway
ICES Journal of Marine Science
 , 
2007
, vol. 
64
 (pg. 
1246
-
1256
)
Leopold
M. F.
Van Damme
C. J. G.
Philippart
C. J. M.
Winter
C. J. N.
Otoliths of North Sea fish: fish identification key by means of otoliths and other hard parts
2001
Amsterdam
World Biodiversity Database CD-ROM Series. Expert Center for Taxonomic Identification (ETI)
Leopold
M. F.
Van Damme
C. J. G.
van der Veer
H. W.
Diet of cormorants and the impact of cormorant predation on juvenile flatfish in the Dutch Wadden Sea
Journal of Sea Research
 , 
1998
, vol. 
40
 (pg. 
93
-
107
)
Lepš
J.
Šmilauer
P.
Multivariate Analysis of Ecological Data using CANOCO
2003
Cambridge, UK
Cambridge University Press
Link
J. S.
Garrison
L. P.
Trophic ecology of Atlantic cod Gadus morhua on the northeast US continental shelf
Marine Ecology Progress Series
 , 
2002
, vol. 
227
 (pg. 
109
-
123
)
Lundström
K.
Hjerne
O.
Alexandersson
K.
Karlsson
O.
Estimation of grey seal (Halichoerus grypus) diet composition in the Baltic Sea
NAMMCO Scientific Publications
 , 
2007
, vol. 
6
 (pg. 
177
-
196
)
Magnan
P.
Rodriguez
M. A.
Legendre
P.
Lacasse
S.
Dietary variation in a fresh-water fish species: relative contributions of biotic interactions, abiotic factors, and spatial structure
Canadian Journal of Fisheries and Aquatic Sciences
 , 
1994
, vol. 
51
 (pg. 
2856
-
2865
)
Mikkelsen
B.
Haug
T.
Nilssen
K. T.
Summer diet of grey seals (Halichoerus grypus) in Faroese waters
Sarsia
 , 
2002
, vol. 
87
 (pg. 
462
-
471
)
Murie
D. J.
Lavigne
D. M.
Growth and feeding habits of grey seals (Halichoerus grypus) in the northwestern Gulf of St Lawrence, Canada
Canadian Journal of Zoology
 , 
1992
, vol. 
70
 (pg. 
1604
-
1613
)
Noren
S. R.
Iverson
S. J.
Boness
D. J.
Development of the blood and muscle oxygen stores in gray seals (Halichoerus grypus): implications for juvenile diving capacity and the necessity of a terrestrial postweaning fast
Physiological and Biochemical Zoology
 , 
2005
, vol. 
78
 (pg. 
482
-
490
)
Ojaveer
E.
Lindrooth
A.
Bagge
O.
Lehtonen
H.
Toivonen
J.
Voipio
A.
Fish and fisheries
The Baltic Sea
 , 
1981
Amsterdam
Elsevier Oceanography Series, 30. Elsevier
(pg. 
275
-
350
418 pp.
Orr
A. J.
Banks
A. S.
Mellman
S.
Huber
H. R.
Delong
R. L.
Brown
R. F.
Examination of the foraging habits of Pacific harbor seal (Phoca vitulina richardsi) to describe their use of the Umpqua River, Oregon, and their predation on salmonids
Fishery Bulletin US
 , 
2004
, vol. 
102
 (pg. 
108
-
117
)
Orr
D. C.
Bowering
W. R.
A multivariate analysis of food and feeding trends among Greenland halibut (Reinhardtius hippoglossoides) sampled in Davis Strait, during 1986
ICES Journal of Marine Science
 , 
1997
, vol. 
54
 (pg. 
819
-
829
)
Palmer
M. W.
Putting things in even better order: the advantages of canonical correspondence analysis
Ecology
 , 
1993
, vol. 
74
 (pg. 
2215
-
2230
)
Pierce
G. J.
Boyle
P. R.
Diack
J. S. W.
Digestive tract contents of seals in Scottish waters: comparison of samples from salmon nets and elsewhere
Journal of Zoology (London)
 , 
1991
, vol. 
225
 (pg. 
670
-
676
)
Pouilly
M.
Lino
F.
Bretenoux
J. G.
Rosales
C.
Dietary–morphological relationships in a fish assemblage of the Bolivian Amazonian floodplain
Journal of Fish Biology
 , 
2003
, vol. 
62
 (pg. 
1137
-
1158
)
Prime
J. H.
Hammond
P. S.
The diet of grey seals from the south-western North Sea assessed from analyses of hard parts found in faeces
Journal of Applied Ecology
 , 
1990
, vol. 
27
 (pg. 
435
-
447
)
Söderberg
S.
Sälens födoval och skadegörelse på laxfisket i Östersjön
1972
Undersökning utförd på uppdrag av Svenska Ostkustfiskarenas Centralförbund
pg. 
60
 
Söderberg
S.
Feeding habits and commercial damage of seals in the Baltic
Proceedings of the Symposium on the Seal in the Baltic, Lidingö, Sweden, 4–6 June 1974
 , 
1975
Stockholm
National Swedish Environment Protection Agency
(pg. 
66
-
78
)
ter Braak
C. J. F.
Canonical correspondence analysis: a new eigenvector technique for multivariate direct gradient analysis
Ecology
 , 
1986
, vol. 
67
 (pg. 
1167
-
1179
)
ter Braak
C. J. F.
Bock
H. H.
Partial canonical correspondence analysis
Classification and Related Methods of Data Analysis
 , 
1988
Amsterdam
Elsevier Science
(pg. 
551
-
558
)
ter Braak
C. J. F.
Šmilauer
P.
CANOCO Reference Manual and CanoDraw for Windows Users Guide: Software for Canonical Community Ordination (version 4.5)
2002
Ithaca
Microcomputer Power
ter Braak
C. J. F.
Verdonschot
P. F. M.
Canonical correspondence analysis and related multivariate methods in aquatic ecology
Aquatic Sciences
 , 
1995
, vol. 
57
 (pg. 
255
-
289
)
Tollit
D. J.
Heaslip
S. G.
Barrick
R. L.
Trites
A. W.
Impact of diet-index selection and the digestion of prey hard remains on determining the diet of the Steller sea lion (Eumetopias jubatus)
Canadian Journal of Zoology
 , 
2007
, vol. 
85
 (pg. 
1
-
15
)
Tollit
D. J.
Steward
M. J.
Thompson
P. M.
Pierce
G. J.
Santos
M. B.
Hughes
S.
Species and size differences in the digestion of otoliths and beaks: implications for estimates of pinniped diet composition
Canadian Journal of Fisheries and Aquatic Sciences
 , 
1997
, vol. 
54
 (pg. 
105
-
119
)
Walton
M.
Pomeroy
P.
Use of blubber fatty acid profiles to detect inter-annual variations in the diet of grey seals Halichoerus grypus
Marine Ecology Progress Series
 , 
2003
, vol. 
248
 (pg. 
257
-
266
)
Westerberg
H.
Lunneryd
S. G.
Fjälling
A.
Wahlberg
M.
Reconciling fisheries activities with the conservation of seals throughout the development of new fishing gear: a case study from the Baltic fishery—grey seal conflict
American Fisheries Society Symposium
 , 
2008
, vol. 
49
 (pg. 
1281
-
1292
)