ICES J. mar. Sci., 53: 11–21. 1996
The diet of haddock (Melanogrammus aeglefinus L.) in the
Barents Sea during the period 1984–1991
Weimin Jiang and Terje Jørgensen
Jiang, W. and Jørgensen, T. 1996. The diet of haddock (Melanogrammus aeglefinus L.)
in the Barents Sea in the period 1984–1991. – ICES J. mar. Sci., 53: 11–21.
An analysis of the diet of haddock in the Barents Sea based on 9500 stomachs collected
between 1984 and 1991 is presented. The analysis was made on a quarterly basis for
10 cm fish length-classes. An overall 27% of the stomachs collected were empty and the
average wet weight of the stomach contents was low for all size-classes (generally
<10 g). By weight the prey categories Crustacea and Echinodermata made up an
overall 60% of the diet, and fishes 20%. The ontogenetic change in the diet was small
for haddock larger than 20 cm. Confidence limits for the estimated mean weight
percentages were calculated using both theoretical variance estimators and bootstrapping. Precision was generally poor, with coefficients of variation typically in the range
0.4–0.6.
? 1996 International Council for the Exploration of the Sea
Key words: haddock, diet composition, confidence estimates.
Received 24 January 1994; accepted 16 May 1995.
W. Jiang: Yellow Sea Fisheries Research Institute, 106, Nanjing Road, Qingdao 266071,
P.R. China. T. Jørgensen: Department of Fisheries and Marine Biology, University of
Bergen, HiB, 5020 Bergen, Norway.
Introduction
The need for a better understanding of marine ecosystems and for more optimal fisheries management
has led to the development of multi-species modelling
in an attempt to account for the interaction between
species. One important source of empirical information
for modelling is quantitative stomach analysis. Largescale efforts have recently been invested in stomach
sampling projects (e.g. the concerted sampling in the
North Sea in 1981 and 1991 (Anon., 1984, 1991). In
1984, the Institute of Marine Research in Bergen
initiated a stomach sampling programme in the
Barents Sea in connection with the development of a
multi-species model for the area (Mehl and Yaragina,
1991). The sampling effort was focused on cod (Gadus
morhua), the major predator in the system, and its
interaction with capelin (Mallotus villosus) (Mehl,
1989), but a large number of haddock stomachs were
also sampled.
Studies on the diet of haddock in different areas of the
North Atlantic have shown that it feeds on a variety of
prey organisms, with differences between localities, seasons, and years reflecting changes in prey availability
(e.g. Homans and Needler, 1946; Sonina, 1969; Jones,
1978; Langton and Bowman, 1980; Pálsson, 1983;
Villemarqué, 1985; Mattson, 1992).
1054–3139/96/010011+11 $12.00/0
In the Barents Sea, diet studies for haddock have
generally been qualitative (Sonina, 1969; Ponomarenko
and Ponomarenko, 1975; Ponomarenko et al., 1978;
Antipova et al., 1980; Kovtsova, 1988), earlier studies
showing that haddock prey heavily on capelin (Tseeb,
1957, 1958, 1960, 1964, cited by Sonina, 1969). To
provide more quantitative information, an analysis was
made of samples collected during the period from 1984–
1991. Ontogenetic, seasonal, and year-to-year variation
in the food composition were studied. Of special interest
was an assessment of the role haddock play as a
predator on capelin. Another focus of the study was to
assess the precision of the estimates of diet composition,
an aspect generally ignored in quantitative feeding
studies.
Material and methods
Data
The analysis is based on data for 9531 haddock
stomachs collected in the Barents Sea during the period
from 1984–1991 (Table 1). The data were collected by
the Institute of Marine Research, Bergen, Norway,
except for 574 stomachs collected in 1988 and 1989
by the Knipovich Polar Research Institute of Marine
Science (PINRO) in Murmansk, Russia.
? 1996 International Council for the Exploration of the Sea
12
W. Jiang and T. Jørgensen
Table 1. Number of stations sampled (n) and number of haddock stomachs collected in each year by quarter (Q) and length-class.
Length-class (cm)
Year
Q
N
10–19
20–29
30–39
40–49
>50
Total
1984
2
4
1
2
4
1
2
3
4
1
2
4
1
2
3
4
1
2
3
4
1
3
4
1
3
4
10
9
21
13
5
31
3
7
11
32
1
8
43
5
6
19
44
16
16
1
53
3
13
25
8
3
22
47
55
66
35
133
3
11
33
59
6
6
39
2
—
69
186
85
44
—
328
5
41
204
10
4
61
113
204
121
63
175
18
25
84
166
19
50
180
13
—
66
90
64
68
17
231
3
91
168
58
26
39
83
108
79
52
160
29
20
104
204
21
84
350
54
20
110
234
114
36
6
132
6
76
93
15
6
1
28
83
68
44
126
23
19
65
137
18
85
422
41
2
191
368
141
86
20
205
—
49
38
4
3
—
7
19
6
19
18
4
10
28
53
6
39
189
8
—
114
246
64
160
17
193
3
102
39
8
10
123
278
469
340
213
612
77
85
314
619
70
264
1180
118
22
550
1124
468
394
60
1089
17
359
542
95
49
406
1493
2174
2235
2267
1362
9531
1985
1986
1987
1988
1989
1990
1991
Total
The sampling area covered the area where the main
part of the haddock population was distributed (Dalen
et al., 1984; Hylen et al., 1985, 1986; Godø et al., 1987;
Hylen et al., 1988; Jakobsen et al., 1989). An example of
the location of sampling stations in the first quarter of
the year is given in Figure 1. The samples were not
equally distributed between season and years, more
samples being taken during the 1st quarter than during
the others (Table 1). For each sampling station the aim
was to collect five haddock stomachs for each 5 cm
length-group.
The stomachs were frozen as soon as possible after the
samples were taken, but those collected by PINRO were
preserved in 4% formalin. Stomachs from fish which
showed any evidence of regurgitation were discarded.
Successful identification of stomach content depended
on the type of prey and the stage of digestion, but the
prey were identified to the lowest possible taxonomic
level. The wet weight of each prey was measured to the
nearest milligram, but the length of the prey was generally not recorded. The stomach analyses were made in
the laboratory by personnel at the IMR (Norwegian
data) and PINRO (Russian data).
In order to obtain a general picture of the diet
of haddock, the prey were grouped into five major
taxonomic categories: Annelida, Mollusca, Crustacea,
Echinodermata and fish (Pisces). Prey not belonging to
any of these five major categories were placed in a
‘‘miscellaneous’’ group. The analyses were made on a
seasonal basis, and to study variation in food composition with fish size, five length-classes (10–19, 20–29,
30–39, 40–49 and >50 cm) were used. Due to the limited
number of stomachs, data from the whole area were
pooled.
The weight of prey, which could only be classified to a
higher taxonomic level than that used in the analysis,
was redistributed proportionally among the identified
prey categories at the taxonomic level below. This
procedure was repeated for successive taxonomic levels.
No analysis was made if the number of stomachs
containing food in a given length-group, quarter, or year
was less than five.
Analysis
Variability in the weight of stomach contents within a
predator length-class was analysed using the linear
model:
Wijkl =ì+Yi +Qij +Aijk +åijkl
(1)
where Wijkl is the wet weight of the stomach contents of
fish 1, in station k, in quarter j, and year i, ì is the
The diet of haddock in the Barents Sea
13
Figure 1. The location of the sampling stations (+) where the haddock were collected. Data for the 1st quarter of years 1985
to 1988.
average wet weight of the stomach content of haddock
in the length-class, Y is the effect of year, Q is the effect
of quarter of the year, A is the effect of station (incorporating the effects of area and time of the day), and å is
the deviation from the mean not accounted for by the
other factors.
The variance components of model (1) are difficult
to estimate because the data are severely unbalanced
(see Table 1; Searle, 1987). The approximate method
suggested by Pennington et al. (1981) was therefore
used to estimate the percentage contribution of each
factor in model (1) to the total variance in stomach
content weight. Empty stomachs were included in
the computation of the average stomach content
weight.
The contribution of each prey category to the diet of
haddock was estimated by taking the ratio of the wet
weight of the prey category to that of the total stomach
contents:
where: Wijk is the individual wet weight of stomach
content category j, for fish k, in size-class i, and ni is the
total number of stomachs containing food in size-class i.
The standard error of this ratio estimator is (Cochran,
1977, p. 33):
W. Jiang and T. Jørgensen
where the variables are the same as in equation (2) and
the finite population correction has been ignored. 95%
confidence limits for the estimated average weight percentages were also calculated by bootstrapping (Efron
and Tibshirani, 1993). The procedure was implemented
with a FORTRAN programme, using 2000 resamplings, and with each bootstrap sample the size of
the original sample. The confidence limits estimated
from equation (3) and by bootstrapping individual fish
use data pooled over hauls. The between-haul variation
is thereby ignored and the width of the confidence
intervals may become misleadingly small (Cochran,
1977). To explore this effect, stations were also used as
the unit for bootstrapping. This latter analysis was
confined to the 1st quarter, due to the small number of
stations for some of the other quarters (Table 1). As the
proportions of the various prey categories in the diet are
not independent, the confidence estimates are only
applicable to individual prey categories and not to the
simultaneous distribution of two or more.
Principal Component Analysis (PCA), a multivariate
ordination technique, was used to summarize possible
patterns in the composition of the haddock diet. The
programme CANOCO (Ter Braak, 1987) was used
to run the analysis. Arcsine-transformed weight percentages (Sokal and Rohlf, 1981) were used as input
data. To reduce the effect of rare prey species, the
‘‘downweighting’’ option of the program was used. A
preliminary analysis showed that the recommendations
for using PCA were fulfilled (the length gradient of the
1st axis in DCA was only 1.4 standard deviations; see
Ter Braak and Prentice (1988)).
For all statistical tests a significance level of 5% was
used.
Results
Variability in the weight of stomach content
The coefficient of variation (CV) of the stomach contents in each size-class (Fig. 2) was found to be independent of fish size (Kruskal-Wallis test, Bhattacharyya and
Johnson, 1977). Therefore, the average of the CVs over
all length-classes was used to estimate the variance
Coefficient of variation
14
2.5
2.0
1.5
1.0
0.5
0
1
2
4
3
5
Length class
6
Average CV
8
Figure 2. The coefficient of variation of average weight of
stomach content vs fish size (length-class). Key to lengthclasses: 1=10–19 cm, 2=20–24 cm, 3=25–29 cm, 4=30–34 cm,
5=35–39 cm, 6=40–44 cm, 7=45–49 cm, 8= >50 cm.
components of model (1). The calculations showed that
35% of the variance in stomach content weight was
between individuals caught in the same haul. The
remaining variance was partitioned approximately
equally between the factors area, season, and year
(Table 2).
Feeding intensity
Empty stomachs
The observed percentage of empty stomachs was high
and varied significantly with fish size, quarter, and year,
ranging from 0–66% (Fig. 3). Within a year, the proportion of empty stomachs within a predator lengthclass was significantly different between quarters in 22
out of 36 cases tested (Pearson chi-square test, Fienberg,
1977). Further analysis showed that this significance was
largely due to a higher frequency of empty stomachs in
the 1st quarters. Within a quarter, significant differences
in the proportion of empty stomachs between lengthclasses were found in 10 out of 24 cases tested (chisquare test). The significance was mainly caused by the
high proportion of empty stomachs among small fish
(10–19 cm, 20–29 cm). Within a length-class and quarter, the proportion of empty stomachs was generally
significantly different between years, except for the 3rd
quarter (chi-square test).
Table 2. The average coefficient of variation (CV) of stomach wet weight at the various levels and the
partition of variance among the factors of model (1).
Level
7
Source of variability
Proportion
Within a sample
Within a quarter
Within a year
1.137
1.460
1.751
Between individuals
Between hauls (area, time of day)
Quarter
35
22
25
Total
1.930
Year
18
The diet of haddock in the Barents Sea
15
Figure 3. The relationship between the percentage of empty stomachs and fish size. Data by quarter and year. Key to length-classes:
1=10–19 cm, 2=20–29 cm, 3=30–39 cm, 4=40–49 cm, 5= >50 cm. /=1st quarter; .=2nd quarter; 4=3rd quarter; -=4th
quarter.
Average weight of stomach content
Since the factor of quarter was the second most important source of variability in the weight of the stomach
contents, the analysis was made on a quarterly basis.
The average weight of the stomach contents generally
increased with fish size, although the weight was low for
all size-classes (generally <10 g, Fig. 4). Within a year
and size-class the weight of the stomach contents was
significantly different between quarters in 23 out of 36
cases tested (Kruskal-Wallis test). Generally, the weight
of the stomach contents was lowest in the first quarter.
For a given length-class and quarter, the stomach
contents weight was significantly different between
years in 6 out of 19 cases tested (Kruskal-Wallis test).
Particularly high values of stomach contents were
observed for the length group >50 cm in the 3rd quarter
of 1991.
Diet of haddock
Haddock in the Barents Sea was found to feed on a
wide variety of prey items, ranging from seaweed to
fishes. A total of 210 prey taxa were observed in the
haddock diet; however, only a few of these were found
to be of importance in the diet (for details see Jiang
(1992)).
Composition of the diet
The relative importance by weight of each of the six prey
categories to the diet of haddock is shown by year,
quarter, and length-class in Figure 5. The three most
important categories were Crustacea, Echinodermata,
and fishes. Combined, they accounted for 80% or more
of the diet. However, the relative contribution of each
category was highly variable. Fish were particularly
16
W. Jiang and T. Jørgensen
30
1984
1985
1986
1987
1988
1989
20
10
0
30
Mean weight (g)
20
10
0
30
20
10
0
30
(57.7)
1990
1991
20
10
0
20
40
60
80
0
Length
20
40
60
80
Figure 4. Mean weight of stomach content vs mean fish length in each of the 5 length-classes. Key as in Figure 3.
important in the diet in the 1st quarter of 1991,
Crustacea dominated in the 2nd quarter of 1988, and
Echinodermata dominated in the 4th quarter of 1989.
The more detailed analysis of food composition
showed that the Euphausiidae, Amphipoda, Hyperiidae,
and Pandalus borealis were the most commonly
occurring and important crustaceans (Jiang, 1992).
Ophiuroidea were dominant among the Echinodermata.
The most common fish prey were redfish (Sebastes sp.),
capelin (Mallotus villosus), and sandeels (Ammodytes
sp.). Sandeels were found to be the main component in
the diet of haddock in the 3rd quarter of 1991, whereas
they were almost absent in the other quarters and years.
All samples taken from 1984–1991 were combined
in an ordination to reveal dietary difference between
seasons (Fig. 6). The eigenvalues of the first four axes
were 0.59, 0.30, 0.07 and 0.04, respectively, indicating
that 89% of the variation in the data could be explained
by the first two ordination axes. Four groups, corresponding to each quarter, were separated in the predator
plot. In general, Crustacea and Annelida were more
important in the 3rd quarter than in the others, fish were
more important in the 2nd quarter, while Echinodermata and Mollusca were more important in the
4th quarter. The smallest haddock (10–19 cm) did not
conform to this pattern.
The analysis also showed a variation in diet composition with predator size (Fig. 6). In the 4th quarter, the
contribution of Echinodermata and Mollusca increased
and the contribution of Crustacea and fish decreased
with increasing predator size. In the 2nd quarter, the
proportion of Crustacea decreased and the proportion
of fish increased with increasing predator size. Moreover, the analysis showed that the difference in diet
composition was mainly between fish in the size-class
10–19 cm and bigger fish. Small haddock consumed
more Crustacea and Annelida, but less Echinodermata
and fish than bigger haddock.
Precision of estimated diet composition
Confidence limits as estimated from equation (3) and by
bootstrapping when individual fish were the units for
The diet of haddock in the Barents Sea
17
Quarter
1st
2nd
3rd
4th
100
1984
50
0
100
1985
50
0
100
1986
50
0
100
1987
Percent
50
0
100
1988
50
0
100
50
1989
0
100
50
1990
0
100
50
0
1991
1
2
3
4
5
1
2
3
4
5
1 2
Length class
3
4
5
1
2
3
4
5
Figure 5. The gravimetric food composition in percent by year, quarter and length-class. Prey were grouped into 6 main prey
categories. Key as in Figure 3.
re-sampling agreed closely (Table 3). Using stations as
the unit for bootstrapping generally resulted in wider
confidence limits than those obtained when individual
fish were the units for re-sampling (Table 3). The
precision of the estimated proportions of each prey
category was generally poor with CVs for the dominating prey categories of Crustacea, Echinodermata, and
fish, being typically in the range of 30–60%, while those
18
W. Jiang and T. Jørgensen
quarters, reflecting the relatively higher sampling effort
in this quarter.
Discussion
Figure 6. Ordination axes resulting from the Principal Component Analysis (PCA) of the stomach content data (gravimetric
data). For a given quarter and length-class data for all years
1984 to 1991 were pooled. (a) prey scores; (b) predator scores.
iq-j (j=1, . . ., 5; i=1, . . ., 4) marks the ordination of the data
for the j’th length-class in the i’th quarter.
of the less important prey categories such as Mollusca
and Annelida varied from 50–100%. As an example, the
estimates for the 30–39 cm length-class are shown in
Figure 7. The precision of the estimates in the first
quarter was generally higher than that in the other
Although the samples were collected within the main
distributional area of the Barents Sea haddock as
deduced from acoustic/trawl surveys in autumn and
winter, sampling was often confined to a smaller subarea
that varied between quarters and years. Moreover, in
some quarters, the sampling effort was low and all the
haddock were sampled from a few hauls (e.g. 2nd
quarter 1987, 4th quarter 1989). With local aggregations
of prey, this may lead to a distorted picture of the diet of
the haddock stock. Feeding in the trawl, regurgitation,
differential digestion rates, and post-capture digestion
are also possible sources of error in feeding studies, but
they are difficult to quantify and were not specifically
addressed in this study. The effect of redistribution of
indeterminate prey to lower taxa is likely to be a minor
source of error because the percentage by weight of
indeterminate prey was comparatively low (10%).
The choice of an index to describe the importance of
different prey in the diet will influence the results when
there is a large difference in prey size (Hyslop, 1980).
This is the case for haddock where fish prey tend to be
much larger than other prey. For example, in the first
quarter of 1986, only one capelin was found in the 70
stomachs containing food in predator length-class
10–19 cm. Nevertheless, capelin made up 46% of the
pooled stomach content by weight in these stomachs.
Overall, 27% of the stomachs in the present study
were empty. Compared with the studies by Villemarqué
(1985) in the North Sea (10% empty stomachs), and Du
Buit (1982) in the Celtic Sea (16% empty stomachs), this
value seems high and suggests that regurgitation has not
always been detected (Bowman, 1986). On the other
hand, Cranmer (1986) observed 39% empty stomachs in
samples of haddock from the North Sea.
The generally low food content of haddock stomachs
found in the present study is similar to the values found
by Jones (1978) for haddock around the Faroes and in
the North Sea, but considerably lower than those
reported by Villemarqué (1985) for haddock in the
North Sea. The higher values in the latter study may be
due to the comparatively greater contribution of fish in
the diet. The analysis of the average weight of stomach
content and the percentage of empty stomachs indicated
that the feeding intensity of haddock was lowest in the
first quarter while more intensive feeding took place in
the second, third, and fourth quarters of the year. In
contrast, Antipova et al. (1980) reported that the most
intensive feeding occurred late in the first quarter.
The present study confirmed earlier observations
by Kovtsova (1988) that Echinodermata, mainly
Ophiuroidea, are important prey for haddock in the
The diet of haddock in the Barents Sea
19
Table 3. Estimated diet composition by weight (%W) for haddock. Data for 1st quarter 1989. The 95% confidence intervals were
estimated using the normal approximation (point estimate&1.96 standard error) and bootstrapping (number of resamplings=2000), respectively. Bootstrap—1 is based on individual fish as the unit for re-sampling, while bootstrap—2 used station
(haul) as the unit for re-sampling. n is the total number of stomachs with identified prey in each length group. Prey category key:
1=Crustacea, 2=Echinodermata, 3=fish, 4=Mollusca, 5=Annelida, 6=others.
95% confidence interval
Length
class
(cm)
n
10–19
53
Normal
approximation
Bootstrap—1
Bootstrap—2
56.4
31.3
10.7
0.0
1.6
<0.1
33.0–85.4
7.7–57.5
0.0–31.6
29.3–83.5
2.9–59.5
0.0–30.3
26.0–89.5
6.7–61.6
0.0–31.6
0.0–4.9
0.0–0.2
0.0–4.9
0.0–0.2
0.0–7.3
0.0–0.3
Prey
category
W%
1
2
3
4
5
6
20–29
30
1
2
3
4
5
6
43.4
24.1
28.0
1.5
0.1
2.9
15.1–71.7
6.3–42.0
0.0–58.9
0.0–4.3
0.0–0.3
0.0–10.7
16.3–69.9
10.3–48.6
0.0–56.1
0.0–4.5
0.0–0.3
0.0–10.0
14.4–77.0
8.3–45.8
0.0–54.0
0.0–6.0
0.0–0.4
0.0–9.8
30–39
139
1
2
3
4
5
6
61.8
19.9
11.3
2.2
3.6
1.2
48.7–74.8
11.2–28.8
2.8–19.7
1.0–3.5
0.0–8.1
0.0–2.8
47.7–73.4
12.5–29.9
3.8–20.3
1.2–3.7
0.4–8.9
0.1–3.2
34.3–80.2
8.3–36.0
4.0–21.8
1.2–4.0
0.3–10.9
0.0–3.4
40–49
191
1
2
3
4
5
6
66.5
14.1
14.6
2.7
0.7
1.4
57.7–75.3
8.2–19.9
7.0–22.2
1.1–4.3
0.0–1.5
0.3–2.5
56.1–75.1
9.1–20.9
7.8–22.5
1.4–4.4
0.2–1.7
0.5–2.7
34.2–80.4
5.7–35.0
5.5–26.2
1.0–2.5
0.2–2.5
0.4–3.9
>50
135
1
2
3
4
5
6
60.0
24.5
9.3
2.3
1.8
2.1
47.0–73.0
14.9–34.0
3.3–15.4
0.8–3.7
0.1–3.6
0.4–3.9
45.8–71.5
15.2–34.8
4.1–16.5
1.0–4.1
0.4–3.9
0.7–4.2
31.1–79.6
10.0–46.1
3.1–20.9
0.6–5.1
0.3–5.1
0.5–5.4
Barents Sea. Overall, they made up 38% of the diet
during the period 1984–1991 (compared to 28% reported
by Kovtsova (1988) for the years 1980–1987). Kovtsova
reported that the importance of this category had
increased during the 1980s, but no such trend was
discernible in the present study. Studies in other regions
also showed that Ophiuroidea contributed a large proportion of the diet of haddock (Homans and Needler,
1946; Langton and Bowman, 1980; Pálsson, 1983).
Earlier studies in the Barents Sea showed that
haddock fed heavily on capelin (Antipova et al., 1980;
Tseeb, 1957, 1958, 1960, 1964, cited by Sonina, 1969).
For example, the maximum frequency of occurrence of
capelin in the first quarter of 1954 reached 79%
(Antipova et al., 1980). Based on the data from 1954 to
1958, Tseeb (1957, 1958, 1960, 1964, cited by Sonina,
1969) concluded that the diet of haddock was obviously
selective with preference for capelin. However, the
present study showed that haddock rarely preyed upon
capelin. The overall frequency of occurrence of capelin
in the stomach of haddock was only 1.4%, while the
highest frequency of occurrence was observed in the
2nd quarter of 1991 at 18%.
There are probably two major reasons for this discrepancy. First, the capelin stock abundance has been
low throughout the present study period, and especially
during 1986–1989 (Anon., 1993). The complete absence
of capelin in the haddock stomachs in 1987 and 1988
correspond to the two years when the capelin spawning
stock reached its lowest recorded level (Anon., 1993).
Second, the large quantities of capelin eggs in the
haddock diet in Tseeb’s study (Tseeb, 1960, cited by
Langton and Bowman, 1980), indicate disproportionate
sampling of haddock occurred on the nearshore spawning grounds for capelin (see Dragesund and Gjøsæter
(1988) for a general review of the Barents Sea capelin). It
20
W. Jiang and T. Jørgensen
Coefficient of variation
1.5
1.0
0.5
1984
1985
1986
1987
1988
1989
1990
12
47
43
98
139
88
34
189
51
19
80
41
116
14
21
58
79
44
63
44
61
33
0.0
1991
Quarter
n
Year
Figure 7. Coefficient of variation for the estimated weight percentages of each of the 6 prey categories in the diet. Data by quarter
and year for the 30–39 cm length-class. Standard deviations were calculated using equation (3). n is the number of stomach with
identified prey. /=Crustacea; 4=Echinodermata; -=Fish; ,=Mollusca; .=Annelida; 0=other.
is well documented that many capelin die after spawning
(Friðgeirsson, 1976) and dying or dead capelin may
easily be taken by haddock. With its typically ventral
mouth, haddock is generally assumed to have difficulties
catching large, fast-moving pelagic prey (Mattson,
1992). A closer inspection of those sampling locations in
the present study where haddock had eaten capelin,
showed that these, with few exceptions, were located
nearshore. Moreover, with the exception of one sample
collected in the 3rd quarter of 1988, haddock was found
to feed on capelin only during the first and (especially)
the second quarter, e.g. during the capelin spawning
season in March and April. During the 3rd and 4th
quarters the distributional areas of capelin and haddock
do not overlap to a large extent (Dragesund and
Gjøsæter, 1988; Ozhigin and Luka, 1985).
No other quantitative study of the diet of fishes giving
the precision of the calculated prey composition is
known to the authors. Nevertheless, multi-species
models have extensively used quantitative stomach data
for calculating species interaction (e.g. the MSVPA
model for the North Sea). If the results of this study are
indicative of the precision of estimates of food composition of fishes in general, the consumption estimates and
species interaction patterns that are deduced are likely to
be very approximate. Even when only major prey categories were used and sample sizes were comparatively
large (>50 per length-class), the precision was generally
poor (Table 3 and Fig. 7). The present study clearly
demonstrates that only major differences in the estimated prey compositions between predator length-
classes, quarters, and years are statistically significant.
Accompanying precision estimates should be an integral
part of any quantitative diet study. Bootstrapping is a
promising technique for obtaining such estimates,
especially when several sources of variation (between
fish, between stations etc) are present and theoretical
variance estimators are not readily available.
Acknowledgements
The authors would like to thank the Institute of Marine
Research, Bergen for allowing us to use their data. We
gratefully acknowledge the assistance and comments by
Bjarte Bogstad, Arild Folkvord, Agnes Christine
Gundersen, Åge Høines, and E. Mumtaz Tirasin. The
comments by the referees also considerably improved
the paper. The first author is indebted to the Norwegian
Agency for Development Cooperation and the ‘‘Bei
Dou’’ project for financial support to undertake his
MPhil studies at the University of Bergen.
References
Anon. 1984. Report of the meeting of the coordinators of the
stomach sampling project 1981. ICES CM 1984/G: 37.
Anon. 1991. Manual for the ICES North Sea Stomach
Sampling Project in 1991. ICES CM 1991/G: 3.
Anon. 1993. Report of the Atlanto-Scandian herring and
capelin Working Group. ICES CM 1993/Assess: 6.
Antipova, T. V., Ponomarenko, I. Ya., and Yaragina, N. A.
1980. Seasonal and year-to-year dynamics of occurrence
The diet of haddock in the Barents Sea
frequency of the main food objects for haddock and their
stomach fullness degree in the Barents Sea. ICES CM
1980/H: 21.
Bhattacharyya, G. K., and Johnson, R. A. 1977. Statistical
concepts and methods. John Wiley and Sons, New York.
Bowman, R. E. 1986. Effect of regurgitation on stomach
content data of marine fishes. Environmental biology of
fishes, 16: 171–181.
Cochran, W. G. 1977. Sampling techniques. 3rd edition.
J. Wiley and Sons, New York.
Cranmer, G. J. 1986. The food of haddock (Melanogrammus
aeglefinus) in the North Sea. ICES CM 1986/G: 86.
Dalen, J., Hylen, A., Jakobsen, T., Nakken, O., and Randa, K.
1984. Preliminary report of the investigations on young cod
and haddock in the Barents Sea during the winter 1984. ICES
CM 1984/G: 44.
Dragesund, O., and Gjøsæter, J. 1988. The Barents Sea. In
Continental shelves. Ecosystems of the world, 27, pp. 339–
361. Ed. by H. Postma and J. J. Zijlstra. Elsevier.
Du Buit, M. H. 1982. Essai d’évaluation de la preédation de
quelques téléostéens en Mer Celtique. Journal du Conseil
International pour l’Exploration de la Mer, 40(1): 37–46.
Efron, B., and Tibshirani, R. J. 1993. An introduction to the
bootstrap. Chapman and Hall, Inc. New York. 436 pp.
Fienberg, S. E. 1977. The analysis of cross-classified data. MIT
Press, Cambridge, MA.
Friðgeirsson, E. 1976. Observations on spawning behaviour
and embryonic development of the Icelandic capelin. Rit
Fiskideildar, 5(5): 1–33.
Godø, O. R., Hylen, A., Jacobsen, J. A., Jakobsen, T., Mehl,
S., Nedreaas, K., and Sunnanå, K. 1987. Estimates of stock
size of Northeast Arctic cod and haddock from survey data
1986/1987. ICES CM 1987/G: 37.
Homans, R. E. S., and Needler, A. W. H. 1946. Food of the
haddock. Proceedings of the Nova Scotian Institute of
Science, 5.
Hylen, A., Jakobsen, T., Nakken, O., Nedreaas, K., and
Sunnanå, K. 1985. Preliminary report of the Norwegian
investigations on young cod and haddock in the Barents Sea
during the winter 1985. ICES CM 1985/G: 68.
Hylen, A., Jakobsen, T., Nakken, O., Nedreaas, K., and
Sunnanå, K. 1986. Preliminary report of the Norwegian
investigations on young cod and haddock in the Barents Sea.
ICES CM 1986/G: 76.
Hylen, A., Jacobsen, J. A., Jakobsen, T., Mehl, S., Nedreaas,
K., and Sunnanå, K. 1988. Estimates of stock size of
Northeast Arctic cod and haddock, Sebastes mentella and
Sebastes marinus from survey data, winter 1988. ICES CM
1988/G: 43.
Hyslop, E. J. 1980. Stomach contents analysis – a review of
methods and their application. Journal of Fish Biology, 17:
411–429.
Jakobsen, T., Mehl, S., Nedreaas, K., Nakken, O., and
Sunnanå, K. 1989. Estimates of stock size of Northeast
Arctic cod and haddock, Sebastes mentella and Sebastes
marinus from survey data, winter 1989. ICES CM 1989/G:
42.
Jiang, W. 1992. Food of haddock (Melanogrammus aeglefinus)
in the Barents Sea from 1984 to 1991. MPhil thesis,
University of Bergen, Norway.
Jones, R. 1978. Estimates of the food consumption of haddock
(Melanogrammus aeglefinus) and cod (Gadus morhua).
21
Journal du Conseil International pour l’Exploration de la
Mer, 38: 18–27.
Kovtsova, M. V. 1988. Fatness dynamics and feeding peculiarities of haddock (Melanogrammus aeglefinus) in the Barents
Sea. ICES CM 1988/G: 32.
Langton, R. W., and Bowman, R. E. 1980. Food of fifteen
Northwest Atlantic gadiform fishes. NOAA Technical
Report NMFS SSRF-740. Department of Commerce, USA.
Mattson, S. 1992. Food and feeding habits of fish species
over a soft sublittoral bottom in the Northeast Atlantic. 3.
Haddock (Melanogrammus aeglefinus (L.)) (Gadidae). Sarsia,
77: 33–45.
Mehl, S. 1989. The Northeast Arctic cod stock’s consumption
of commercially exploited prey species in 1984–1986.
Rapports et Procès-Verbaux des Réunions du Conseil
International pour l’Exploration de la Mer, 188: 185–205.
Mehl, S., and Yaragina, N. A. 1991. Methods and results in
the joint PINRO-IMR stomach sampling program. In
Interrelations between fish populations in the Barents
Sea. Proceedings of the fifth PINRO-IMR symposium,
Murmansk, 12–16 August 1991, pp. 5–16. Ed. by B. Bogstad
and S. Tjelmeland. Institute of Marine Research, Bergen,
Norway.
Ozhigin, V. K., and Luka, G. I. 1985. Some peculiarities of
capelin migration depending on thermal conditions in the
Barents Sea. In The Proceedings of the Soviet-Norwegian
symposium on the Barents Sea capelin, pp. 135–148. Ed. by
H. Gjøsæter. Institute of Marine Research, Bergen, Norway.
Pálsson, O. K. 1983. The feeding habits of demersal fish species
in Icelandic waters. Rit Fiskideildar, 7: 1–60.
Pennington, M., Bowman, R., and Langton, R. 1981. Variability of the weight of stomach contents of fish and its implications for food studies. In Gutshop ’81, pp. 2–7. Ed. by
G. M. Cailliet and C. A. Simenstad. Washington Sea Grant
Publication, University of Washington, Seattle.
Ponomarenko, V. P., and Ponomarenko, I. Ya. 1975. Consumption of the Barents Sea capelin by cod and haddock.
ICES CM/1975 F: 10.
Ponomarenko, V. P., Ponomarenko, I. Ya., and Yaragina,
N. A. 1978. Consumption of the Barents Sea capelin by cod
and haddock in 1974–1976. ICES CM 1978/G: 23.
Searle, S. R. 1987. Linear models for unbalanced data. John
Wiley and Sons, New York.
Sokal, R. S., and Rohlf, F. J. 1981. Biometry. Second edition.
Freeman and Co., New York.
Sonina, M. A. 1969. Biology of the Arcto-Norwegian haddock
during 1927–1965. Trudy polyar. naucho-issled. Inst. morsk.
rijb. Kohz. Okeanogr., 26: 3–111 and 115–124. Fisheries
Research Board of Canada Translation Series. No. 1924.
Ter Braak, C. J. F. 1987. CANOCO a FORTRAN program for
canonical community ordination by (partial) (detrended)
(canonical) correspondence analysis, principal components
analysis and redundancy analysis (version 2.1). Agricultural
Mathematics Group. Wageningen, The Netherlands.
Ter Braak, C. J. F., and Prentice, I. C. 1988. A theory of
gradient analysis. Advances in ecological research, 18:
271–317.
Villemarqué, J. Hersert de la, 1985. Rapport préliminaire sur
l’analyse des estomacs d’églefins récoltés en 1981 dans le
cadre du programme d’échantillonnage d’estomacs de
poissons en Mer du Nord. ICES CM 1985/G: 39.