Correlations between pelagic distribution of Common 
and Brunnich’s Guillemots and their prey in the 
Barents Sea 
W E L L  EINAR ERIKSTAD, TRULS MOUM A N D  WIM VADER 

Erikstad, K. E., Moum, T. & Vader, W. 1990: Correlations between pelagic distribution of Common and 
Briinnich’s Guillemots and their prey in the Barents Sea. Polar Research 8 ,  77-87. 

Correlations between guillemots (Including Common Guillcmots Uria aalge and Briinnich’s Guillemots U .  
lomuia) and their prey (divided into five prey categories, capelin Mallatus villosus, herring CIupea harengus, 
polar cod Boreogadus saida, plankton, and a mixture of other prey species) at two depths (10-100m and 
100-200 rn) were estimated along an extended transect of 3,060 nautical miles (5,667 km) in the Barents 
Sea in April/May 1986. Spatial concordance was highest during daylight hours when the largest number 
of birds were seen on water (presumably feeding birds). Capelin was the single prey category which was 
most often associated with birds but no single prey category could alone explain the distribution of birds. 
Although only a small fraction of guillemots could be identified to species, there was some evidence that 
capelin were of greater importance lo Common than t o  Briinnich’s Guillemots. Overall correlation between 
birds and total prey density was statistically significant at the smallest scale of 5 nautical miles ( n m . ) .  The 
removal of herring from the calculations increased the strength of the correlation. The depth at which prey 
was located had little effect on the distribution of birds. The correlation between birds and prey was scale 
dependent, and reached a maximum at 90 n.m., although there seemed t o  be some upper threshold in the 
coefficient at c. 40 n.m. Numerical concordance (including only 5 n.m. periods wherc both prey and birds 
were present) was significant at the 5 n.m. scale but was higher for high density than for low density prey 
patches. The results are discussed in relation to thc few similar studies in other oceans and in relation to 
the severe reduction of important prey species in the Barents Sea. 

Kjell Einar Erikstad, Norwegian Institute for Nature Research, T r o m p  Museum, University of Tromsm, N -  
9OOO Trams@, Norway; TruLF Moum and Wim Vader, Zoology and Marine Biological Departmenls, T r o m p  
Museum, Uniuersiv of Trams@, N-Woo T r a m @ ,  Norway; November 1989 (revised January 1990). 

Feeding seabirds aggregate in large patches with 
chord lengths of several kilometres (Schneider & 
Duffy 1985; Schneider & Piatt 1986). One likely 
determinant of these aggregations is that food 
aggregates at a similar scale (Hunt & Schneider 
1987; Schneider & Piatt 1986). Two main 
approaches have been applied to test the ‘food 
aggregation hypothesis’ (see Hunt & Schneider 
1987, Hunt this volume and Schneider this volume 
for reviews): 

1. To look for correlations between aggre- 
gations of birds and the occurrence of fronts. 
This method is based on the observations that 
biological activity increases in areas of water mass 
boundaries (e.g. F’ingree et al. 1975; Emery et al. 
1973). 

2. To directly correlate numbers of seabirds 
with the abundance of their prey (Obst 1985; 
Schneider & Piatt 1986; Heinemann et al. 1989). 

So far numerous studies have related pelagic 
seabird abundance to the physical properties of 
the ocean (see Hunt & Schneider 1987 and 

Schneider this volume for reviews), but only a 
few have correlated the distribution of birds with 
their prey (see Hunt & Schneider 1987 and Hunt 
this volume for reviews). In general, the former 
show that birds aggregate at fronts, but that fronts 
explain only a very low fraction of the variance 
in avian abundance. The few studies of the 
relationships between seabirds and their prey 
have shown that the strength of correlation is 
highly variable (Hunt this volume) and that the 
correlation between birds and prey is stronger at 
large than at small scales (Schneider & Piatt 1986; 
Heinemann et al. 1989). The relatively low vari- 
ance of avian abundance explained by fronts may 
be, as pointed out by Schneider (this volume), 
due to the causal chain linking seabirds to fronts, 
namely that prey are linked to fronts whereas 
birds are linked to prey. 

Spatial correlation of marine birds with prey is 
highly variable, with highest coefficients for larger 
spatial scales (c. <10 km) and for prey assem- 
blages dominated by a single type of prey 



78 K .  E .  Erikstad, T. M o u m  & W. Vader 

(Schneider & Piatt 1986; Heinemann et al. 1989). 
The correlation with assemblages of several prey 
appears to be low in guillemots (Woodby 1984). 
However, Woodby (1984) did not look at cor- 
relation over several spatial scales. 

This study analyses the correlation of guille- 
mots with prey assemblages over a range of spatial 
scales. It also compares correlation of guillemots 
with individual prey species to correlation with 
the prey assemblage at the same spatial scale. 

Study area 

Observations were made from R/V 'Eldjarn' in 
the Barents Sea from 29 April to 15 May 1986 
along an extended transect of 3,060 nautical miles 
(5,667 km) in an area ranging from the Norwegian 
coast northward to the ice border at approxi- 
mately 75"N, and from west of Bjeirn~ya to west 
of Novaya Zemlya (Fig. 1). 

The area is a shallow continental shelf with a 
mean depth of 230 m and few areas deeper than 
400-500m (e.g. Loeng 1989). There are three 
main water masses in the Barents Sea (e.g. Midt- 
tun 1985; Loeng 1989; Midttun this volume): the 
Norwegian coastal current and the Atlantic cur- 
rent which both flow into the Barents Sea from 
the southwest and the Arctic current system which 
brings arctic water in from the north. Where the 
Arctic and Atlantic water meet they mix and form 
the Polar front. 

The Barents Sea is an important nursery and 
feeding area for several commercial fish species 
such as cod Gadus morhua, polar cod Boreogadus 
saida, haddock Melanogrammus aeglefinus, her- 
ring Clupea harenglls, and capelin Mallotus 
uillosus. Capelin is an important prey for both 
breeding and non-breeding guillemots in the area 
(Belopol'skii 1957; Furness & Barrett 1985; 
Erikstad & Vader 1989). During the non-breeding 

season, Briinnich's Guillemots also prey heavily 
on both gadoid fish and crustaceans (Erikstad in 
press). 

Material and methods 
Bird Observations 

Birds were counted from the top of the vessel's 
bridge (10 m above sea level) in 300 m transects 
and 10 min. blocks on one side of the ship (Tasker 
et al. 1984). The speed of the ship was about 10 
knots, such that counts covered an area of about 
0.92 km2 in 10 min. The total number of 10 min. 
blocks counted was 1,143, covering an area of 
1051.6 km2. Bird counts were made during the 
day, usually with a break between 2200-0200 hrs., 
and also with breaks when the boat was trawling. 

No attempts were made to count the whole 
seabird community since some species, e.g. gulls, 
Fulmars Fulmarus glacialis, and Kittiwakes Rissa 
tridactyla were strongly affected by the trawling 
activity of the ship (Erikstad et al. 1988). We 
limited the observations to alcid species and bird 
counts followed the standard procedure suggested 
by Tasker et al. (1984). Flying birds were, 
however, counted continuously and not by divid- 
ing the 10 min. transect into shorter sections. The 
procedure suggested by Tasker et al. (1984) mini- 
mises the problem of repeatedly counting birds 
circling around the ship. However, because flying 
alcids only cross the transects and are not affected 
by the presence of the ship, such effects could be 
disregarded. Bird numbers are probably under- 
estimated since diving birds are not seen and some 
distant birds may be overlooked. 

Common and Briinnich's Guillemots were dif- 
ficult to separate at a distance, and only those 
seen close to the ship were identified to species. 

15' 20' 25' 30° JS' 400 450 -'p" Estimates of prey densities 

1 
740 I 
730 I 

70' 

-1 1 -  -\ L- 

The ship made quantitative acoustic/trawl survey 
estimates of the densities of different fish species 
in the area, according to standard methods. The 
instrumentation on board R/V 'Eldjarn' included 
Simrad EK 400, a digital Scase echo-integrator 
and a Nord-10 computer for data storage and 
analysis, including echo-integration which aggre- 
gated data every 5 nautical miles (n.m.). 

The echoes produce traces often characteristic Fig. 1. Map of the study area showing the transect surveyed 
and the approximate position of the ice edge in April/May 1986. for different prey species or prey categories. On 



Common and Brunnich’s Guillemots and their prey 79 

this basis the integrator values were assigned to 
different prey categories. This procedure was 
verified by trawl net samples every 3 M O  n.m. 
or in areas of high integrator values. The prey 
categories were capelin, herring, polar cod, 
plankton and mixture. The last category groups 
together several fish species, the most numerous 
of which were cod, haddock and redfish Sebastes 
rnarinuslS. mentellu. The different prey cate- 
gories were also assigned t o  two depth cate- 
gories, 10-100 m and 100-200 m. 

Data analysis 

Prey densities were measured to a minimum scale 
of 5 n.m., whereas seabirds were counted in tran- 
sects of 10 min. blocks. To enable comparison of 
seabird and prey data at the minimum scale of 
5 n.m., we used two methods. First, maps were 
drawn of prey and bird distributions, using a grid 
computer programme which could be varied in 
size from squares of 20 x 20 km upwards. 
Secondly, to describe the patchiness and distri- 
bution of birds and prey along the continuous 
transect, we used the maximum number of birds 
seen within each 5 n.m. period as an index of bird 
density. Regressing total bird numbers against 
maximum numbers within 30 min. periods (c. 
5n.m. when the speed of the ship is 10 knots) 
showed that nearly all variation in avian density 
could be explained by this index (r2 = 0.99). 

The acoustic values of different prey species 
were used as an index of density only, since more 
information is needed for estimating absolute fish 
density (Johannesson & Mitson 1983). The size 
distribution of different fish species is a very 
important factor in determining the suitability 
of prey to seabirds (Swennen & Duiven 1977; 
Hulsman 1981). 

We used a logarithmic scale in the presentation 
of data and in several analyses we also ranked the 
data in order to meet the assumptions of the 
statistical tests. 

Spatial scaling 

To examine the importance of spatial scale on 
the correlation between seabird density and prey 
density we analysed the seabird and prey density 
estimates from the 5 n.m. periods and then the 
mean densities were calculated for larger scales 
by aggregating data over an increasing number of 
5 n.m. periods (2-20). 

The day time counts of birds provided sampling 
units separated both in space and time. We used 
all segments longer than 10 X 5n.m. periods (a 
total of 17 segments, covering a range from 10 to 
27 x 5 n.m. periods), and tested for correlation 
between mean prey and bird density. 

Statistical testing 

The spatial association between guillemots and 
prey (defined as the degree to which seabird and 
prey present are related) was calculated at the 
5 n.m. scale. We tested the presence/absence 
data, and tested for differences in spatial associ- 
ations at different times of the day (4 h blocks), 
using chi-square tests. 

To estimate the overall correlation between 
seabirds and prey, we calculated Spearman rank 
correlations. The data could not be described by 
standard statistical models (see also Schneider & 
Piatt 1986; Heinemann et al. 1989), so random- 
ization tests were used to evaluate statistical 
significance. Correlation between seabirds and 
prey at the 5 n.m. scale was tested for significance 
by comparing the observed correlation at this 
scale, r(5) to one hundred values of the same 
statistic obtained by randomizing 5 n.m. prey data 
with respect to location. A Turbo-Pascal random 
generator was used t o  obtain random permutation 
of prey data. Bird data were not randomized in 
this test. 

We tested the two following hypotheses 
(see also Schneider & Piatt 1986): 1) the null 
hypothesis, H,:r(5) = 0, was rejected if an 
observed value of r(5) was greater than 95 of 100 
values of r(5) from randomization data, and 2) 
the null hypothesis (scale independent corre- 
lation), Ho: r(5) = r(F), was rejected in favour of 
the alternative hypothesis, HA:r(5) <r(F), if 95 
of 100 randomized values of r(F) were greater 
than the observed r(5) value. F is spatial scale, 
taking on values of 2-20 periods of 5 n.m. Since 
densities above the 5 n . m .  scale are means of 
densities pooled within larger sampling units, we 
tested the correlations at larger scales against the 
underlying 5 n.m. scale (see Heinemann et al. 
1989). 

The numerical correlation (defined as the rank 
correlation between seabird numbers and prey 
densities where both are present) was estimated at 
the 5 n.m. scale only. Since numerical association 
may manifest itself only above some threshold in 
prey density (Piatt 1987; Heinemann et al. 1989), 



80 K. E. Erikstad, T. Moum C% W .  Vader 

160 - 

we analysed ‘low density’ and ‘high density’ 
patches separately. ‘High density’ prey patches 
were selected according to discontinuities in the 
distribution of prey densities and in such a way 
that an appropriate number of 5 n . m .  periods 
was obtained for statistical testing. 

mar 618 
Capelin 

Results 

240 - 1  

Bird densities 

A total of 8,726 birds were seen within the 1,143 
10 min. periods, giving a total of 7.6 birds per km2. 
Briinnich’s Guillemots and Common Guillemots 
were by far the most common alcid species, and 
totalled 90.8% of all birds seen. Only 38.4% of 
the guillemots seen could be identified to species, 
but the ratio of those identified showed that Briin- 
nich’s Guillemots were by far the most numerous 
(88.3%). Other auks seen were (in decreasing 
order of abundance): Little Auk Alle alle (5.6%), 
Puffin Fratercula arctica (3.1%), Razorbill A k a  
torda (0.1 %) and Black Guillemot Cepphus grylle 
(0.06%). Because of their small number, these 
four species were left out of the calculations 
below. 

Of the total number of Brunnich’s and Com- 
mon Guillemots seen (N = 7,919), 61.2% were 
seen on water and 38.8% flying. In the cal- 
culations of the correlations between birds and 
prey only birds seen on the water were 
considered, since flying birds are not necessarily 
associated with the habitat and their numbers 
can not be expected to be closely linked to the 
presence of food. 

Mixture 
max2218 

Spatial distribution of guillemots and prey 

Both birds and prey showed a clear non-random 
distribution where low density patches alternated 
with high density patches along the transect (Fig. 
2). Of the 388 five nautical mile periods where we 
had both bird and prey data, birds were present in 
275 (70.9%) periods, prey present in 243 (62.6%) 
and both birds and prey in 173 out of 275 periods 
(62.6%). 

The degree of spatial association of birds and 
prey varied during the day (Table 1, 2 = 27.2, 
df = 4, P < 0.001). It was highest during the day 
(1000-1400,75%) and lowest at night (0200-0600, 

The proportion of birds seen on water (and 
which were presumably feeding (Table 1)) varied 

21 %) . 

8 0 -  

40 -’ 

0 -  I 1.l 1, Id 

0 sbloolSo2i)O250340~400450500556 Bw 

5 neutlcal mile periods 
Fig. 2. Number of Common and Briinnichs Guillcmots and 
density of different prey categories along a transect in the 
southern Barents Sea. 

throughout the day and night. The maximum 
occurred when prey and birds were most closely 
correlated (1000-1400). 

The overall spatial associations of birds and 
prey irrespective of time of day were low for 
single prey categories (range from 8% in herring 
to 33.8% for capelin (Table 2)). Between 1000 
and 1400 hrs., the association of capelin increased 
to 58.9%, which differed significantly (2 = 



Common and Briinnich’s Guillemnts and their prey 81 

Table I .  Percentage of Common and Briinnich’s Guillemots seen in the southern Barents Sea on water (N = total number of birds 
seen) and the percentage of 5 nautical miles with birds that also had prey present (N = total number of 5 mile blocks with birds 
within this time period) during 4-h time blocks from 0200-2200 local time. 

Time of day 
~~ ~ - 

0001-0400 0401-0800 0801-1200 1201-1600 1601-2000 

(N = 748) (N = 1,081) (N = 3,082) ( N  = 1,305) (N = 1,501) 

(N = 19) (N = 72) (N = 55) (N = 64) (N = 64) 
Birds on water (To) 15.6 54.0 85.9 49.8 46.3 

5 mile periods with 21.1 37.5 70.5 65.7 60.9 
birds and prey (%) 

12.4, df = 1, P < 0.001) from when all time 
periods were considered, but the association with 
other single prey categories remained very low. 

The low spatial association between birds and 
single prey categories is also obvious from Figs. 
2 and 3, since birds were distributed in patches 
along the whole transect while each prey category 
was restricted to smaller areas with partly non- 
overlapping distributions. Most illustrative are 
the distributions of the schooling species capelin, 
herring and polar cod, which overlapped very 
little spatially. Capelin were most common in the 
northwestern part of the study area, while herring 
and polar cod were distributed to the south and 
east, respectively (Fig. 3). 

Overall correlation and spatial scaling 

Since it is obvious from inspection of the data 
that no single prey category could explain the 
distributional patterns of birds, we carried out an 
exploratory analysis of overall correlation 
between birds and prey first by combining all five 
prey categories at both depths and then stepwise 
removing different combinations of prey cate- 
gories and depths to evaluate combinations where 
we achieved the highest significant correlation. 

At the 5 n.m. scale there was a statistically 
significant correlation between birds and total 
prey density at depths 100-200m and 10-200m 
(Fig. 4). The coefficients were low (0.14-0.20) 
but could be increased somewhat ( O . l t M . 2 5 )  by 

aggregating data on the 10 n.m. scale. Coefficients 
were not significantly higher at larger scales, sug- 
gesting an absence of scale dependent correlation. 

The removal of herring from the calculations 
increased the strength of correlation. At the 
5 n m .  scale, correlations ranged from 0.22-0.29 
but were again only signficant at depths 100- 
200 m and 1&200 m. Maximum significant cor- 
relation coefficient was achieved at 100-200 m on 
a scale of 90 n.m. (0.54) and at 10-200 m on a 
scale of 80 n.m. (0.50). According to Fig. 4, there 
appears to be some upper threshold in the coef- 
ficients reached around a scale of 40 n.m.(r-values 
of about 0.43). Above 40 n m .  the r-values fluc- 
tuated markedly, although maximum significant 
correlation coefficients were achieved at 80 and 
90n.m. for depths 100-200m and 10-200m, 
respectively. The highest correlation was 
achieved when using the three prey categories 
capelin/1@-100 m, mixture/10-100 m and plank- 
ton/10-200 m. These coefficients were signi- 
ficant at all scales, and reached a maximum at 
90 n.m. (r = 0.64). 

Overall correlations between seabird and prey 
density means at the ‘segment’ periods (Fig. 5) 
were strong (r = 0.63). This correlation coef- 
ficient also differed significantly from the under- 
lying 5 n.m. correlation. 

Numerical correlation 

Correlations between birds and prey at the 5 n.m. 

Table 2. The presence of different prey categories (%) in 5 mile periods where birds were present according to different times of 
day. N = number of 5 n.m. periods. 

Capelin Herring Polar cod Plankton Mixture Total 

All times of day (N = 275) 33.8 8.0 15.6 12.4 17.4 62.9 
Day time (1001-1400) (N = 56) 58.9 7.0 17.9 21.4 28.6 75.0 



82 K .  E .  Erikstad, T. Mourn & W .  Vader 

a 10 
a 5  

. Common o%y 
0 1 Guillemot 0. 

10 
m 5 Capelln 

10 
0 5  

20 
m 10 Polar Cod 

Plankton 

. 10 
5 Mixture 

Fig. 3. The distribution of Common and Brunnich's Guillemots and different prey categories in the southern Barents Sea. Filled 
circles indicate numbcr of birds/kmz o r  m e a n  acoustic signal strength/km2 (for diffcrcnt prcy categories). 

scale when both were present were significant 
(r = 0.42). The correlations were much higher for 

Discussion 
high density prey patches (r = 0.71) than for low 
density patches (r = 0.35), suggesting that Spatial distribution of birds and prey 

numerical concordance between birds a i d  prey The association of birds and prey showed a clear 
may first manifest itself when a threshold is die1 pattern, and was highest during day time 
reached (Fig. 6). These high density prey patches when the largest number of birds was seen on 
covered only 5.6% of the area studied but the water (presumably feeding birds). Accord- 
included as much as 58.2% of all birds seen. ingly, the time of day (die1 feeding rhythm) is an 



Common and Brunnich's Guillemots and their prey 83 

0.6 

0.5 

0.4 

0.5 o.6 I A 

C 
- 

/. - e ,.-@4 
. ' 

0.0 ' 
0.6 r 

0.3 

0.2 

0.1 

0.2 1 " 

- 

- 

0.1 , 
0.0 

0.7 I 

O : l O - l W m d e p t h  
A :  100 -200 m depth 
0 : 10 - 200 m depth 

0.0 ' 
0 4 8 12 16 20 

Frame size (5 n.m.) 

Fig. 4. Correlations between guillemots Uriu spp. and prey in 
the southern Barents Sea at different depths as a function of 
measurement distance (5 nautical mile periods). A .  Guillemots 
and total prey density. B. Guillemots and prey when herring is 
rcmoved from the calculations. C. Guillemots and the prey 
categories which gave the highest correlation cocfficients (cap- 
elin at depth 1&100m, plankton at both depths, and mixture 
at dcpth 1&100 m). Statistical significance is indicated by solid 
symbols (p < 0.05). 

important factor which has to be considered when 
evaluating the correlations between birds and 
prey. Birds on the water include both resting and 
feeding birds (which are difficult to differentiate), 
while flying birds are presumably searching for 

food either by looking for other feeding birds 
or by spotting surface prey, and are therefore 
expected to be less associated with prey. Woodby 
(1984), who also addressed this problem in the 
southeastern Bering Sea, noted a similar diurnal 
pattern in the number of guillemots seen on water 
but he did not evaluate the degree of association 
between birds and prey at different times of the 
day. 

There was no single prey category which alone 
could explain the distributional patterns of birds. 
This agrees with recent studies of the diets of 
seabirds from the Barents Sea where Briinnich's 
Guillemots may feed heavily on both capelin, cod, 
and crustacea during the non-breeding season 
(Erikstad & Vader 1989; Erikstad in press). 

There is less information on the diet of Com- 
mon Guillemots but they have been shown to 
prey on capelin close to the breeding colonies 
during the prelaying season (Erikstad & Vader 
1989). From the open sea there is to date no 
study, but there is a general acceptance that they 
rely heavily on capelin also at this time of the year 
(Barrett 1979; Vader et al. in press), a suggestion 
which is in accordance with diet studies from the 
breeding colonies in this study area (Furness & 
Barrett 1985). Unfortunately we were not able to 
evaluate the associations between capelin and the 
two guillemot species separately, as only a small 
fraction of birds of the two species could be dif- 
ferentiated. However, there is some evidence that 
Common Guillemots may be more strongly 
associated with capelin than Briinnich's Guille- 
mots. First, capelin was the single prey species 
which was most common in areas where birds 
were located. Secondly, the distribution of the 
few Common Guillemots identified was similar 
to the distribution of capelin which covered the 
northwestern part of the transect system (Fig. 3). 
Briinnich's Guillemots were distributed through- 
out the area investigated. In the northwestern 
part they overlapped with the Common Guillemot 
but in the eastern part Briinnich's Guillemots 
were found alone within the range of polar cod 
but outside the capelin area. 

Herring was the only prey category which was 
less associated with birds than initially expected. 
At first sight this may seem contradictory since 
herring is known to have been an important 
nourishment for both guillemot species during 
summer in the eastern parts of the Barents Sea 
(Belopol'skii 1957). However, the herring avail- 
able in this study (mean length c. 20 cm with only 



84 K .  E .  Erikstad, T. Mourn & W ,  Vader 

rs= 0.63 
p < 0.01 

. 1 . 
1 . . 

8 .  . . . . 

. 

. 

-1 0 1 2  3 4 5 6 7 
Prey density index (log.) 

Fig. 5 .  Overall correlation between mean density of guillemots 
and prey at the ‘segment’ intervals. 

6 1  I 

r. -0.35 
5 p< 0.01 . i . 

0 L l - - l - - . - . - . - . 8 - * . l - ~  
0 1 2 3 4 5 

Prey density index (log. ) 

r. =0.71 1 p<o.o1 

L, 
6 7 8 

Prey denslty lndex(log.) 

0 .  
5 

Fig. 6. Numerical correlation (where both birds and prey were 
present) between guillemots and prey in the southern Barents 
Sea at low (A) and high (B) density prey patches. 

avery few specimens less than 18 cm, unpublished 
data from the Norwegian Fishery Institute) may 
have been too large for the guillemots to feed on. 
Herring of 20cm length have a body height of 
C. 45 mm. Although Common Guillemots are able 
to swallow fish of this size, Swennen & Duiven 
(1977) found that they prefer much smaller (c. 
12 cm) fish. This result agrees with the observed 
size range of other fish fed upon by the two 
guillemot species in the Barents Sea. Erikstad & 
Vader (1989) showed that Common and 
Brunnich’s Guillemots feed on capelin of c. 15 cm 
during the prelaying season, which is also con- 
sistent with studies from the breeding season 
(Furness & Barrett 1985; Barrett & Furness 
unpublished). Brunnich’s Guillemots feed on 8- 
12 cm cod and polar cod during winter (Erikstad 
in press), 

The capelin recorded in this study had a mean 
size of 15cm (unpublished data from the 
Norwegian Fishery Institute), suggesting that 
most of the capelin available were within the 
appropriate size range for guillemots. We know 
nothing about the size range of the other prey 
categories present in this study. 

Spatial scaling and overall correlation 

At the smallest scale of 5 n.m. there was a signifi- 
cant, albeit low, positive correlation between bird 
and total prey density. The correlation coefficient 
increased when herring was removed from the 
calculation. At spatial scales greater than 5 n.m. 
the correlations increased and reached a maxi- 
mum at 80n.m. (145 km). However, there 
appears to be some upper threshold in the cor- 
relation reached at about 3 W O  n.m. (54 km). 
The highest correlation coefficient was achieved 
when using the prey categories capelin/1@100 m, 
plankton/10-200 m and mixture/1@100 m. It is 
difficult to make any reasonable interpretation of 
this result. Presumably it is accidental and caused 
by some intercorrelations of different prey cate- 
gories. 

The depth at which different prey categories 
were distributed had little effect on the strength 
of the correlation. In fact the coefficient was 
somewhat higher for the depth range lO(r200 m 
than for 10-100 m. That birds seem to track prey 
independent of depth may be a result of several 
mechanisms. Guillemots can dive to depths 
greater than 100 m (Piatt & Nettleship 1985; 
Burger & Simpson 1986) even though they may 



Common and Briinnich’s Guillemots and their prey 85 

feed in shallower water (Barrett & Furness 
unpublished). Moreover, young year classes of 
important prey categories in the Barents Sea, 
such as capelin and cod, show diurnal vertical 
migration patterns and are found near the surface 
during the day (e.g. Belstad et al. 1975). This 
pattern is consistent with the diurnal variation in 
birds seen on the water in this study. To draw firm 
conclusions about the importance of the depth of 
prey, a detailed study of patches of feeding birds 
in relation to prey depth is needed. 

The present study is one of four which have 
documented a positive correlation between bird 
and prey density (Schneider & Piatt 1986; Piatt 
1987; Heinemann et al. 1989). The results pre- 
sented here are also consistent with the other 
three studies in that the degree of correlation was 
higher at larger spatial scales. Schneider & Piatt 
(1986) described three different mechanisms 
which may cause reduced correlation at a smaller 
scale: 1) local avoidance of the dense prey 
patches, 2) differences in locomotory capacities 
of birds and prey and 3) ‘sit and wait predation’ 
by birds within extended aggregations of prey. 
Local avoidance of dense prey patches seems 
unlikely in this study, since there was a high 
numerical correlation (where both birds and prey 
were present) between bird and prey density. We 
know less about the other two mechanisms but 
differences in the locomotory capacity of birds 
and prey may also be of importance in this study. 
Capelin, which is presumably the most important 
prey for guillemots, migrate from the northern 
Barents Sea to the coast of Norway and Mur- 
mansk (Jangaard 1974; SEtre & G j o s ~ t e r  1975) 
at the time of year this study was carried out. 
Rapid swimming by capelin schools may make it 
difficult for birds to track them, especially at night 
when they are not feeding (see also Schneider & 
Piatt 1986). 

Heinemann et al. (1989) mention two other 
factors which may explain the lower correlation 
at small scales: 1) Most prey patches are unex- 
ploited at any given time and 2) a small number 
of prey patches are disproportionately important 
to birds. We know nothing about the exact num- 
ber of prey and bird patches in this study but at 
the 5 n.m. scale there is no apparent excess of 
prey patches (Fig. 2). Such a low number of 
prey patches in relation to bird patches may be a 
consequence of the severely reduced stocks of 
three important fish species in the study area, 
namely herring, capelin and cod (Hamre 1986, 

1988; Bergstad et al. 1987). This reduction in 
prey abundance has also recently led to dramatic 
decreases in the breeding populations of both 
Common and Briinnich’s Guillemots in the area 
(Vader et al. in press). 

That some prey patches are more important to 
birds than others does not seem to be justified in 
this study. In fact there was a very close positive 
correlation between the sizes of prey and bird 
patches above some threshold (numerical con- 
cordance), suggesting that birds are very good at 
detecting prey and presumably also at evaluating 
the relative value of prey patches when prey avail- 
ability is low. The results presented by Piatt (1987) 
are interesting in this context. In a study on Com- 
mon Guillemots and capelin from Witless Bay, 
Newfoundland, he found evidence for an upper 
threshold in prey density. Maximum correlation 
between guillemots and capelin was achieved at 
an intermediate density of capelin. Piatt (1987) 
suggested that the capture rate of prey by birds 
levelled out at some threshold and that birds could 
not increase their foraging success further by sel- 
ecting patches above this level. The general prey 
density in the Barents Sea was possibly at such 
a low level that this upper threshold was not 
reached. 

Strength of correlation 

The strength of correlation between prey and 
pelagic seabird density varies greatly in different 
studies from very low (68%) to very high 
explained variances (above 90%, Schneider & 
Piatt 1986; Heinemann et al. 1989). Although few 
studies have been carried out, there appears to 
be some general trend in the results obtained, 
as pointed out by Hunt (this volume). Studies 
focusing on diurnally foraging alcids and capelin 
have been more successful in finding positive cor- 
relations (Schneider & Piatt 1986; Piatt 1987) 
than those focusing on seabirds and krill (Obst 
1985; Heinemann et al. 1989). As pointed out by 
Hunt (this volume), the majority of krill special- 
ists take their prey at night, whereas birds have 
to be counted during the day. He also mentions 
the possibility that planktivorous birds are less 
able to locate the densest prey patches or, alterna- 
tively, that they may not need to seek the densest 
patches if background densities or the frequency 
of micro-patches of prey at high densities are 
sufficiently great (Hunt this volume). 

When comparing results from different studies, 



86 K .  E .  Erikstad, T .  Mourn & W .  Vader 

the use of different correlation techniques should 
be taken into account. Schneider & Piatt (1986), 
who have found the highest coefficients (0.88- 
l . O ) ,  used the product moment correlation coef- 
ficient. In this study we estimated the Spearman 
rank correlation. Calculating the product moment 
correlation on our data gives an r-value of 0.86 
at the 5 n.m. scale, whereas the Spearman rank 
correlation at this scale was only 0.23. However, 
we felt that the Spearman statistic was more 
appropriate, since the correlation was driven by 
a few large patches of birds and prey. 

Schneider & Piatt (1986) observed that the 
correlation between birds and prey was only 
intermittent and repeated surveys were necessary 
for detection of correlations (see also Safina & 
Burger 1985, 1988). They found that the cor- 
relation of birds and capelin was significant in 
only 6 out of 12 transects ranging in length from 
15 to 30km. With this in mind, the r-values 
obtained in this study along one long transect of 
nearly 6,000 km are high. The severe reduction of 
important prey species in the Barents Sea (Hamre 
1986, 1988; R ~ t t i n g e n  this volume) may have 
contributed to this result. A strong reduction in 
populations of prey species may also reduce the 
number of prey patches and make it more difficult 
for birds to track prey at sea. This may explain 
the mass starvation of Common Guillemots in the 
study area during the winter of 1987 (Vader et al. 
in press), when the capelin stock was at its lowest 
level (Hamre 1988). However, such a reduction 
in the number of prey patches may strengthen the 
correlations between birds and prey, since the 
number of prey patches which at any time are 
unexploited may be low. Hence, it can be hypo- 
thesized that correlations between birds and prey 
should decrease when prey density increases. 
Both the capelin and herring stocks are slowly 
recovering in the Barents Sea (Hamre 1988), 
which gives us the unique opportunity to test this 
hypothesis by repeating counts along such an 
extended transect during the years of recovery. 

Conclusion and management 
implications 
Hunt (this volume) mentions three assumptions 
which are implicit in the hypothesis that predators 
should aggregate where prey are most abundant: 
1) There is a close coupling between prey abun- 
dance and prey availability, 2) the predators are 

good at detecting and evaluating the relative value 
of prey patches and 3) predators do not deplete 
prey patches prior t o  measurement of prey abun- 
dance. 

All of these assumptions seem to be reasonably 
well justified in this study. Birds and prey were 
closely coupled especially during the day, and 
the high numerical correlation between birds and 
prey suggests that birds were both effective in 
locating and evaluating the relative value of prey 
patches. A high numerical correlation also sug- 
gests that the birds had not depleted the prey 
patches severely before measurement. Accord- 
ingly, at a scale of 3WO n.m. the distribution of 
birds could be predicted fairly well, using data on 
prey distribution and abundance. In fact about 
60% of all birds, which were located within only 
5% of the study area, could be closely predicted 
by extremely high values of prey abundance. 

Predicting the distributional patterns of pelagic 
seabirds using standard acoustic prey density esti- 
mates has important management implications in 
areas such as the Barents Sea, where pelagic fish 
species are heavily exploited and where both the 
distribution and density of fish species are closely 
monitored. However, before drawing any final 
conclusion about the usefulness of standard prey 
data in predicting seabird abundance, further 
studies are needed. Repeated surveys should be 
carried out in years when prey density increases 
(see discussion above) and further studies should 
also aim at estimating the predictability of prey 
patches within larger areas of the ocean. 

Acknowkdgemenfs. ~ We are grateful for the help and co- 
operation of the captain, crew and scientists on board R/V 
‘Eldjarn’. We also thank V .  Bakken, K .  0. Jacobsen and I .  
Johnsen for help with the bird observations, H .  M .  Iversen for 
programming and the Fisheries Research Institute in Bergen 
for letting us use their acoustic data on different fish species. 
David Schneider and Rob Barrett kindly made comments on 
an earlier draft of this paper and Rob Barrett also corrected 
the English. This study was part of a larger project (FOBO), 
studying the pelagic distribution of seabirds and their prey in 
the Barents sea. The study was financed by the Norwegian 
Research Council for Science and the Humanities and the o i l  
companies operating in the Barents Sea area (OKN). 

References 
Barrett, R .  T. 1979: Small oil spill kills 10-20,000 seabirds in 

Belopol’skii, L. 0. 1957: Ecology of sea colony birdr of the 
northern Norway. Mar. Poll. Bull. 10, 253-255. 

Barents Sea (Engl. transl. 1961, IPST, Jerusalem). 



Common and Briinnich’s Guillemots und their prey 87 

Belstad, A. K., Nakken, 0. & Smestad, 0. M. 1975: Inves- 
tigations on die1 vertical migration of 0-group fish in the 
Barcnts Sea. FiskDir. Skr. Ser. HauUnders. 16, 229-244. 

Bergstad, 0. A , ,  Jorgensen, T. & Dragesund, 0. 1987: Life 
history and ecology of thc gadoid resources of the Barents 
Sea. J .  Fish. Res. 5 ,  119-161. 

Burger, A .  E .  & Simpson, M. 1986: Diving depths of Atlantic 
Puffins and Common Murres. Auk 103, 828-830. 

Emcry, K. O., Milliman, J .  D. & Uchupi, E. 1973: Physicdl 
properties and suspended matter of surface waters in the 
southeastern Atlantic Ocean. J .  sedim. Petrol. 43, 822-837. 

Erikstad, K. E. in press: Winter diets of 4 seabird species in 
the Barents Sea after a crash in the capelin stock. Polar Biol. 

Erikstad, K. E . ,  Bustnes, J. 0. & Jacobscn, 0. 1988: Duration 
of ship-following by Kittiwakes Rissa tridactyla in the Barents 
Sea. Polar Research 6 ,  191-194. 

Erikstad, K. E. & Vader, W. 1989: Capelin selection by Com- 
mon and Briinnich’s Guillemots during the prelaying season. 
Ornis Scand. 20, 151-155. 

Furness, R .  W. & Barrett, R. T. 1985: The food requirements 
and ecological rclationships of a seabird community in North 
Norway. Ornis Scund. 16, 305-313. 

Hamre, J. 1986: Bestands- og forvaltningspolitikk for lodde og 
sild i 1979 og 1980-irene. Seminar om Barentshuuets ressur- 
ser. Rapp. Norges FiskarladFiskeridir. Haufors- 
kningsinsfitutt, 72-03. 

Hamrc, J 198X: Some aspects of the interrelation between the 
Herring in the Norwegian Sea and the stocks of Capelin and 
Cod in the Barents Sea. Unpubl. ICES paper C . M .  19881 
H : 4 2 .  I5 pp. 

Heinemann, D . ,  Hunt, G .  L . ,  Jr. & Evcrson, I .  1989: Thc 
distributions of marine avian predators and their prey, 
Euphausia superha, in the Branfield Strait and southern 
Drake Passage, Antarctica. Mar. Ecol. Prog. Ser. 58, 3-16, 

Hulsman, K. 1981: Width of gape as determinants of size of 
prey eaten by terns. Emu 81, 29-32. 

Hunt, G .  L., Jr. 1990: The pelagic distribution of marine birds 
in a heterogeneous environment. Polar Research 8, 43- 
54. 

Hunt, G. L.. Jr. & Schneider, D. C. 1987: Scdlc-dependent 
processes in physical and biological environment of marine 
birds. Pp. 7-41 in Croxall, J. P. (en.): Seabird,: feeding 
biology and role in marine ecosystemc.. Cambridge University 
Press, Cambridge. 

Jangaard, P. M. 1974: The c d p e h  (Mallotus uillosus): biology, 
distribution, exploitation, utilization and composition. Bull. 
Fish. Res. Bd. Can. 186, 1-70. 

Johannesson, K. A .  & Mitson, R. B. 1983: Fisheries acoustics. 
A practical manual for aquatic biomass estimation. F A 0  Fish. 
Tech. Pap. 240. 249 pp. 

Lneng, H. 1989: Ecological features of the Barents Sea. Pp. 

327-365 in Rey, L. & Alexander, V. (cds.): Proc. 6th conf. 
comitd arct. int. May 1985. 

Midttun, L. 1985: Formation of dense bottom water in the 
Barents Sea. Deep Sea Research 32, 1233-3244. 

Midttun, L. 1990: Surface temperatures of the Barcnts Sea. 
Polar Research 8 ,  11-16. 

Obst, B. S. 1985: Densities of Antarctic seabirds at sea and the 
presence of the Krill Euphausia superba. Auk 102,54&549. 

Piatt, J. F. 1987: Behauioural ecology of Common Murre and 
Atlantic Puffin predation on cupelin; implications for p o p u -  
lation ecology. Ph.D. thesis, Department of Biology, Mem- 
orial University of Newfoundland, St. John’s. 

Piatt, J .  F. & Nettleship, D. N .  1985: Diving depths of four 
alcids. Auk 102; 293207. 

Pingree, R .  D . ,  Pugh, P. R . ,  Holligan, P. M. & Foster, G. R. 
1975: Summer phytoplankton blooms and red tides along 
tidal fronts in the castern Barents Sea. J .  Geophys. Res. 83, 
4659-4667. 

Rottingen, I .  1990: Changes in the distribution and abundance 
of Norwegian spring spawning herring and Barcnts Sea 
capelin. Polar Research 8 ,  33-42. 

Safina, C. & Burger, J .  1985: Common tern foraging. Seasonal 
trends in prcy fish densities and competition with bluefish. 
Ecology 6 6 ,  1457-1463. 

Safina, C .  & Burger, J .  1988: Ecological dynamics among prey 
fish, bluefish, and foraging Common Terns in an Atlantic 
coastal system. In Burger, J .  (ed.): Seabirds and other marine 
uertebrates. Columbia Univcrsity Press, New Y o r k .  

Schneider, D. C. 1990: Seabirds and fronts: a brief overview. 
Polar Reseurch 8 ,  17-21. 

Schneider, D .  C. & Duffy, D .  C. 1985: Scalc-dependent varia- 
bility in seahird abundance. Mar. Ecol. Prog. Ser. 25, 211- 
218. 

Schneider, D .  C. & Piatt, J. 1986: Scale-dependent correlation 
of seabirds with schooling fish in a coastal ecosystem. Mar. 
Ecol. Prog. Ser. 32, 237-246. 

Swennen, C. & Duiven, P. 1Y77: Size of food objects of three 
fish-eating seabird species: Uriu ualge. A k a  tordu, and Fra- 
tercula arcticu (Aves, Alcidac). Neth. J .  Sea Res. 11, 92-98, 

Saetre, R .  & Gjoszter, J .  1975: Ecological investigations o n  thc 
spawning grounds of the Barents Sea capelin. FiskDir. Skr. 
Ser. HauUnders. 16, 203-227. 

Tasker, M .  L., Hope Jones, P., Dixon. T. & Blakc, B. F. 1984: 
Counting scabirds at sea from ships: A review of methods 
and a suggestion for a standard approach. Auk 101.567-577. 

Vader, W . ,  Barrett, R. T.. Erikstad, K. E. & Strann, K.-B. in 
press: Differcntial responses of Common and Thick-hilled 
Murres Uria spp. to a crash in the capelin stock in the southern 
Barents Sea. Studies in Auian B i d .  

Woodby, D .  A .  1984: The April distribution of murres and prey 
patches in the Bering Sea. Limnol. Oceanogr. 29, 181-188.