Polar Biology

, 32:569

Identifying patterns in the diet of mackerel icefish (Champsocephalus gunnari) at South Georgia using bootstrapped confidence intervals of a dietary index

Authors

    • British Antarctic SurveyNatural Environment Research Council
    • Fisheries Research Services, Marine Laboratory
  • Martin A. Collins
    • British Antarctic SurveyNatural Environment Research Council
  • Richard Mitchell
    • British Antarctic SurveyNatural Environment Research Council
  • Mark Belchier
    • British Antarctic SurveyNatural Environment Research Council
Original Paper

DOI: 10.1007/s00300-008-0552-7

Cite this article as:
Main, C.E., Collins, M.A., Mitchell, R. et al. Polar Biol (2009) 32: 569. doi:10.1007/s00300-008-0552-7

Abstract

Ontogenetic, inter-annual and regional variations in diet were investigated for mackerel icefish, Champsocephalus gunnari, in three successive summer seasons around South Georgia. Stomach contents from 2239 C. gunnari (130–560 mm total length) were examined. A bootstrapping technique was used to calculate confidence intervals for an index of relative importance of prey categories (% IRIDC). Diet varied significantly between years and age classes but there was little regional difference in diet. In general, diet was dominated by krill, Euphausia superba and by the amphipod Themisto gaudichaudii. Smaller (younger) fish tended to prey on a higher proportion of T. gaudichaudii and small euphausiids such as Thysanoessa sp. and took smaller quantities of E. superba. In a season of poor krill availability (summer of 2003–2004) the proportion of krill in the diet, stomach fullness and fish condition (indicated by length–weight relationships) were significantly lower than in the other summer seasons. A large reduction (>80%) in the estimated annual (2005) biomass of the C. gunnari stock directly followed the season of poor krill availability. This decline was largely because of mortality of 2+ and 3+ fish, which were more krill dependent than 1+ fish. Younger fish appear to have survived, leading to an increase in the estimated population biomass in 2006.

Keywords

ChannichthyidaeDietary indexFeeding ecologyKrillEuphausia superbaScotia Sea, interannual variability, fish stocks

Introduction

A prerequisite for ecosystem based fisheries management (EBFM) is an understanding of variability in exploited stocks in relation to natural variability in the system. There are various perspectives on EBFM (Browman and Stergiou 2004) and the consideration of ecosystem components associated with fished species should ultimately aid sustainable exploitation of living resources. Although EBFM may be difficult to implement technically (Frid et al. 2006) widespread declines in fish stocks have provided impetus for its adoption (FAO 2003). Due consideration of the ecosystem approach therefore requires an understanding of the biology and ecology of target species. It is also essential to monitor species interactions over time and across different locations. A fundamental part of this is an understanding of variability in the diet of target species, and the potential effects of this variability on the population (Marasco et al. 2007).

The productive seas around the sub-Antarctic island of South Georgia, SW Atlantic (Fig. 1) provide a rich feeding ground for Antarctic krill (Euphausia superba Dana, hereafter referred to as krill) which is an important component of the diet of large populations of vertebrate predators such as penguins, seals and fish (Atkinson et al. 2001). Commercial fisheries for krill, mackerel icefish (Champsocephalus gunnari Lönnberg) and Patagonian toothfish (Dissostichus eleginoides Smitt) (Agnew 2004) also operate around the island.
https://static-content.springer.com/image/art%3A10.1007%2Fs00300-008-0552-7/MediaObjects/300_2008_552_Fig1_HTML.gif
Fig. 1

The study area illustrating the different regions identified in the text, with Champsocephalus gunnari bottom trawl catches (kgs) and catch-weighted length frequencies of Champsocephalus gunnari from each survey

At South Georgia mackerel icefish have been commercially exploited since the late 1960s. Landings from the bottom trawl fishery peaked at ~120,000 tonnes in 1983 (Anon 1990a, b). Poor catches in the 1990–1991 season led to the closure of the fishery by the Commission for the Conservation of Antarctic Marine Living Resources (CCAMLR) (Agnew 2004). The fishery was re-opened as a pelagic trawl fishery for the 1992–1993 fishing season, but with the necessary consideration of the ecological importance of C. gunnari as prey of higher predators such as Antarctic fur seals (Arctocephalus gazella Peters, hereafter termed fur seals) (Reid and Arnould 1996; Reid et al. 2005) and gentoo penguins (Pygoscelis papua Forster) (Croxall et al. 1984) and as a significant predator of krill (Kock et al. 1994; Everson et al. 1999). This pivotal ecological role of C. gunnari has necessitated the setting of conservative catch limits and the adoption of the highly precautionary management approach advocated by the CCAMLR (Constable et al. 2000).

Previous investigations into the diet of mackerel icefish found that krill was the main prey species (Kock et al. 1994; Barrera-Oro et al. 1998). Other important prey included the hyperiid amphipod Themisto gaudichaudii, mysids (mainly Antarctomysis maxima) and euphausiids of the genus Thyssanoessa. Prey switching from krill to T. gaudichaudii has been detected previously in a year of low krill abundance (Kock et al. 1994). Considerable inter-annual variability in mackerel icefish abundance that cannot be attributed to fishing has been linked to variability in the abundance of krill (Kock et al. 1994; Everson et al. 1999) indicating that C. gunnari are highly dependent on this food resource. In years of high krill abundance the condition of mackerel icefish is better and more adults reach spawning condition (Everson et al. 1997, 2000). Krill are an important prey of both mackerel icefish and fur seals and in years of low krill abundance, mackerel icefish populations potentially suffer both from a shortage of food and from increased predation from fur seals (Everson et al. 1999). Evidence of a long-term, climate-linked reduction in krill populations in the Scotia Sea (Atkinson et al. 2004) provides additional impetus to understand and quantify the diet of krill dependent predators such as C. gunnari so that the management of these key components of the South Georgia ecosystem can continue in an ecosystem-based framework.

Here we used an index of relative importance (IRI, Pinkas et al. 1971; Cortes 1997) to examine the ontogenetic, inter-annual and geographical variation in the diet of C. gunnari at South Georgia and consider how variation in diet relates to inter-annual variation in condition and food availability. The use of a bootstrapping technique to generate confidence intervals for the index of relative importance in each sample group enabled a robust statistical comparison of diets between years, age classes and geographical regions for C. gunnari.

Methods

Sampling locations and collection

The study area of the South Georgia shelf (Fig. 1) is situated between the Polar Front and the Southern Antarctic Circumpolar Current Front, downstream of the Antarctic Peninsula in the Antarctic Circumpolar Current (see Murphy et al. 2007b). Samples of C. gunnari were collected by bottom and midwater trawls during January in 2004, 2005 and 2006. The surveys were all undertaken on the FV Dorada in the area of South Georgia and Shag Rocks. Bottom trawling used a commercial sized otter trawl (FP-120; see Everson et al. 1999) which was fished for 30 min on the sea floor at a speed of 3–4 knots with a headline height of 4–6 m, a wingspread of approximately 18 m and with a cod-end mesh size of 40 mm. Trawl stations were arranged in a random, stratified design split into grid squares and depth strata (<150 m; 151–250 m; >250 m). All trawling took place during daylight.

In addition to bottom trawling a series of acoustic transects were undertaken on the shelf to the northwest of South Georgia. Midwater fish aggregations, putatively identified as mackerel icefish, were targeted using an International Young Gadoid Pelagic Trawl (IYGPT) net. The IYGPT is a pelagic otter trawl, which was towed at 2.5–3 knots, with a horizontal opening of 12 m, a vertical opening of 7 m and with a 10 mm cod-end mesh.

Fish were sorted into species and the total catch of C. gunnari was weighed on motion compensated marine scales. Sub-samples of C. gunnari (~200 per trawl) were collected from all trawls and were measured [total length (TL) to the nearest 10 mm below]. Stomachs were removed from sub-samples of fish and immediately frozen at −20°C, for later processing. Stomach fullness was recorded on a scale of 0–4, judged subjectively, with 0 being empty, 1: <25% full; 2: 25–50% full; 3: 51–75% full and 4: >75% full. Gutted weight (with liver, viscera and gonads removed) was then measured for each fish (nearest g). Accurate weights could not be obtained at sea for small fish. The number of stomach samples obtained for each sample group that was pooled for analysis of diet composition is indicated in Table 1.
Table 1

The numbers of C. gunnari stomach samples that were pooled for comparisons of % IRIDC between year classes, years and regions

% IRIDC comparison

Pooled samples

Excluded samples

n

Year classes

Regions, years

No samples excluded

165 (age 1+)

499 (age 2+)

942 (age 3+)

640 (age 4+)

Years (age-class specific)

Regions

No samples excluded

 

2004

2005

2006

Age 1+

26

70

69

Age 2+

137

60

302

Age 3+

192

250

500

Age 4+

229

106

305

Regions

Years, 2+ fish, 3+ fish

1+ fish, 4+ fish

220 (NE)

575 (NW)

146 (SE)

276 (SW)

224 (SR)

Region boundaries indicated in Fig. 1

NE northeast region, NW northwest region, SE southeast region, SW southwest region, SR Shag Rocks region

Laboratory processing

In the laboratory stomachs were thawed and the total contents were weighed prior to being sorted into species or species groups. Contents were identified to the lowest taxonomic level possible using published guides (e.g. Gon and Heemstra 1990; Boltovskoy 1999) and reference collections. Partially digested fish were identified, where possible, from sagittal otoliths using reference material and published guides (Hecht 1987; Reid 1996). Sorted contents were counted and weighed (to the nearest 0.01 g). Total length of individual prey items was measured where digestion permitted. Any fresh prey items that had visibly undergone no digestion were thought to constitute net feeding and were excluded from the data.

Data processing

For each demersal trawl mackerel icefish density was estimated using the mass of the catch, distance trawled and horizontal opening of the net. To estimate the stock biomass, the survey area was divided up into five geographical regions around Shag Rocks (SR) and mainland South Georgia (NE, NW, SE, SW South Georgia; see Fig. 1). To correct for different sampling intensity in different strata the following correction was applied to the raw haul density (D) data:
$$ D_{\text{C}} = D \times \frac{{A_{\text{S}} }}{{A_{\text{T}} }} \times \frac{{H_{\text{T}} }}{{H_{\text{S}} }} $$
where DC = corrected density; AS = stratum area; AT = total area; HT = total number of hauls and HS number of hauls in that stratum. Sea floor areas for each stratum (0–300 m) were derived from a recent swath bathymetry survey of the South Georgia shelf (Fretwell et al. 2008). Biomass estimates were obtained by multiplying the mean corrected density by the total sea-floor area. Confidence intervals were determined by bootstrap resampling (Efron and Tibshirani 1993) of the corrected densities using the statistical programming language R (R Development Core Team 2008).
Catch weighted length frequencies were plotted for each year of data (Fig. 2) and were used to define year classes by grouping fish of lengths clustered around peaks (see North 2005). Year classes were given putative ages from 1+ to 4+. For regional analyses the sampling area was split up into the five geographical regions described above.
https://static-content.springer.com/image/art%3A10.1007%2Fs00300-008-0552-7/MediaObjects/300_2008_552_Fig2_HTML.gif
Fig. 2

Catch-weighted length frequencies of Champsocephalus gunnari from the South Georgia and Shag Rocks shelves (2004–2006) showing the separation of putative age classes

Diet analysis

Variation in icefish diet between age classes, years and geographical regions was investigated with an index of relative importance (IRI) of prey (Pinkas et al. 1971; Cortes 1997). This was calculated from percent number (% N) percent mass (% M) and percent frequency of occurrence (% F) of prey items identified from stomach contents using the equation given below. Percent mass was based on the wet weight of the prey found in the stomach and not on reconstituted mass. S = number of stomachs containing food remains.
$$ {\text{Percent mass}}:\,\%\, M_{i} = \frac{{M_{i} }}{{\sum\nolimits_{i = 1}^{n} {M_{i} } }} \times 100 $$
$$ {\text{Percent frequency of}}\,{\text{occurrence}}:\,\%\, F_{i} = \frac{{F_{i} }}{S} \times 100 $$
$$ {\text{Percent number}}:\,\%\, N_{i} = \frac{{N_{i} }}{{\sum\nolimits_{i = 1}^{n} {N_{i} } }} \times 100 $$
$$ {\text{Percent index of}}\,{\text{relative importance}}:\,\%\, {\text{IRI}}_{i} = \frac{{(\%\; N_{i} + \%\; M_{i} ) \times \%\; F_{i} }}{{\sum\nolimits_{i = 1}^{n} {(\%\; N_{i} + \%\; M_{i} \times \%\; F_{i} )} }} \times 100 $$

The % IRI was calculated for individual components of diet, i.e. to the lowest taxon. The % IRIDC, based on diet categories (Bethea et al. 2007) was calculated in order to carry out comparisons of C. gunnari diet. This grouped prey according to the following categories: krill; T. gaudichaudii; Thyssanoessa spp.; mysids; other crustacea (including other euphausiids, amphipods and decapods); fish and others (including salps & molluscs).

The % IRIDC was calculated in this way for C. gunnari to indicate importance of prey items for fish pooled by year, age class or geographical region (see Table 1). The 95% confidence intervals were calculated for mean % IRIDC values by using a bootstrapping technique implemented in R, and a percentile method for estimating the confidence intervals (see Efron and Tibshirani 1993). This utilised resampling from the dataset with replacement to give 1,000 sets of randomly selected stomachs, which was sufficient resampling to give reproducible results. Differences were considered significant when there was no overlap in the 95% confidence intervals.

A Kruskal–Wallis test was used to test for differences in stomach fullness with time of day (all data pooled) and between years (age classes and regions pooled).

Condition

Icefish condition was investigated using analysis of covariance to indicate differences between year-specific regressions of: (a) length and total wet weight; and (b) length and gutted wet weight. Males and females were analysed separately. The analysis was carried out on 2+ and 3+ fish only, since 1+ fish could not be weighed with sufficient accuracy at sea and the numbers of larger (4+) fish sampled varied greatly between years. Length and weight were log10 transformed in order to meet assumptions of normality. The response variable was either log10(total wet weight) or log10(gutted wet weight); the covariate was log10(length) and year was a factor. In the model, if the interaction term between regressions was not significant this implied that their slopes were approximately equal. If this was the case, the interaction term was removed, producing an additive model. A significant effect of year within the additive model indicated a significant difference in icefish condition between years. These analyses were undertaken using Minitab Release 14.

Results

Distribution and abundance of Champsocephalus gunnari

Mackerel icefish were widely distributed across the South Georgia and Shag Rocks (SR) shelves. The largest catches were sampled in the SR area and to the N and NW of South Georgia (Fig. 1). Few trawls could be carried out to the south of mainland South Georgia as there was little ground that was suitable for bottom trawling. Catches were highest in 2006 and lowest in 2005 (Fig. 1). Annual biomass estimates and confidence intervals (CI) were estimated as: 47,802 tonnes in 2004 (95% CI: 18,959–84,615); 9,288 tonnes in 2005 (4110–13,237) and 86,578 tonnes in 2006 (28,148–169,947).

Some regional differences in size frequency distribution were evident (Fig. 1, see also North 2005). Notably, the catches in the SE tended to be dominated in number by smaller fish, although some of the largest fish were also caught in this region (Fig. 1). Catch-weighted length frequencies, with fish from all regions pooled for each survey year were used to indicate the following putative age classes (Fig. 2). Age 1+ fish were defined as <21 cm; Age 2+ as 21–26 cm; Age 3+ as 27–34 cm and Age 4+ as >34 cm. The length distribution of fish from which stomachs were sampled (Fig. 3) was generally representative of the length-distribution of the population, although larger fish tended to be sampled more often than small fish.
https://static-content.springer.com/image/art%3A10.1007%2Fs00300-008-0552-7/MediaObjects/300_2008_552_Fig3_HTML.gif
Fig. 3

Length frequency distribution (count) of individual Champsocephalus gunnari from which stomachs were sampled during each survey (2004–2006)

Stomach fullness

Stomach fullness showed no pattern with time of day for the pooled dataset. However there were significant differences in icefish stomach fullness between years (Kruskall–Wallis test: H = 595.92; df = 2; n = 4139; P < 0.001). The mean value for this test in 2004 (0.74) was considerably lower than in either 2005 (1.58) or 2006 (1.73).

Diet composition

Krill was the predominant prey of C. gunnari in all three summer seasons and in all regions for age classes 2+, 3+ and 4+ (Table 2, Figs. 4, 5, 6). Other species of euphausiid such as E. frigida, E. triacantha and E. vallentini were present in the diet in smaller proportions (Table 2). Euphausiids of the genus Thysanoessa were important to smaller (1+) fish caught mainly at SR and NW South Georgia (Figs. 5, 6). After krill, T. gaudichaudii was the next most important prey species in terms of numbers, mass and % IRI (Table 2). C. gunnari also took other amphipods, and various species of fish including notothenids (Chaenocephalus aceratus, C. gunnari, Pseudochaenichthys georgianus, Lepidonotothen larseni) and myctophids (Electrona antarctica, E. carlsbergi, Gymnoscopelus nicholsi, Krefftichthys anderssoni, Protomyctophum bolini and P. choriodon). The salp, Salpa thompsoni, was present in a small proportion (0.09%) of stomachs from 2006.
Table 2

All prey identified from C. gunnari stomach contents collected in 2004, 2005 and 2006

 

2004

2005

2006

Abr.

% M

% N

% F

% IRI

% M

% N

% F

% IRI

% M

% N

% F

% IRI

Crustacea Amphipoda

 Cyphocaris sp.

CRU

0.01

0.01

0.17

         

 Eurythenes obesus

CRU

0.01

0.01

0.17

         

 Eusirus perdentatus

CRU

 

0.01

0.17

         

 Hyperiella sp.

CRU

0.01

0.01

0.17

         

 Primno macropa

CRU

         

0.01

0.17

 

 Themisto gaudichaudii

THO

16.71

54.92

37.50

41.13

12.98

57.03

44.65

31.66

4.75

34.39

43.96

12.19

 Vibilia antarctica

CRU

 

0.01

0.17

         

 Amphipoda unid.

CRU

     

0.02

0.21

     

Copepoda

 Copepoda unid.

CRU

 

0.01

0.17

       

0.09

 

 Euchaeta sp.

CRU

 

0.02

0.17

         

 Rhincalanus gigas

CRU

     

0.02

0.21

     

Decapoda

 Decapod larvae

CRU

 

0.06

0.17

         

 Notocrangon antarcticus

CRU

0.21

0.06

0.68

     

0.13

0.04

0.60

 

Euphausiacea

 Euphausia frigida

CRU

0.15

1.00

2.23

0.04

    

0.23

0.90

1.28

0.01

 Euphausia superba

KRI

51.83

20.60

44.69

49.57

68.98

28.51

66.67

65.83

85.02

51.80

88.86

86.11

 Euphausia triacantha

CRU

 

0.01

0.17

     

0.05

0.10

1.11

 

 Euphausia vallentini

CRU

          

0.09

 

 Thysanoessa sp.

THY

1.18

6.30

7.19

0.82

0.72

6.13

11.93

0.83

0.63

3.45

11.56

0.33

 Euphausiid unid.

CRU

2.12

4.36

12.16

1.21

0.56

4.66

11.11

0.59

0.05

0.10

1.11

 

Mysidacea

 Antarctomysis sp.

MYS

7.37

9.45

23.80

6.13

1.02

1.65

4.32

0.12

6.16

8.79

12.67

1.34

Other Crustacea

 Crustacea unid.

CRU

0.80

0.73

6.34

0.15

0.33

0.48

5.56

0.05

0.01

0.01

0.17

 

Mollusca: Cephalopoda

 Cephalopod unid.

OTH

        

0.01

 

0.09

 

Chordata: Teleost fish

 Chaeocephalus aceratus

FSH

        

0.05

 

0.09

 

 Champsocephalus gunnari

FSH

0.21

0.05

0.34

     

0.98

0.03

0.26

 

 Electrona antarctica

FSH

0.08

0.07

0.34

         

 Electrona carlsbergi

FSH

0.21

0.03

0.34

 

0.26

0.02

0.21

  

0.01

0.09

 

 Gymnoscopelus nicholsi

FSH

0.97

0.01

0.17

 

0.09

0.02

0.21

     

 Harpagifer sp.

FSH

 

0.01

0.17

         

 Krefftichthys anderssoni

FSH

    

0.60

0.22

2.06

0.02

0.05

0.01

0.26

 

 Lepidonotothen larseni

FSH

0.21

0.03

0.51

 

0.10

0.02

0.21

 

1.02

0.16

0.94

0.01

 Myctophid unid.

FSH

5.81

0.87

2.91

0.30

0.05

0.02

0.21

 

0.02

0.01

0.26

 

 Notothenid unid.

FSH

0.02

0.07

0.17

 

0.01

0.02

0.21

 

0.02

0.01

0.17

 

 Notothenid larvae

FSH

1.08

0.59

3.42

0.09

0.20

0.30

0.82

   

0.09

 

 Patagonotothen guntheri

FSH

6.66

0.19

2.40

0.25

13.70

0.73

6.17

0.90

0.36

0.03

0.26

 

 Protomyctophum bolini

FSH

        

0.33

0.07

0.85

 

 Protomyctophum choriodon

FSH

0.38

0.06

0.86

0.01

0.17

0.02

0.21

 

0.05

0.01

0.17

 

 Pseudochaenichthys georgianus

FSH

        

0.01

 

0.09

 

 Trematomus hansoni

FSH

0.01

0.01

0.17

         

 Fish unid.

FSH

3.91

0.41

4.45

0.29

0.23

0.14

1.85

0.01

0.05

0.02

0.34

 

Urochordata: Salpa

 Salpa thompsoni

OTH

          

0.09

 

In the column denoted “Abr” the abbreviations that were used to denote the different diet categories applied when calculating the % IRIDC are indicated

KRI, E. superba; THY, Thysanoessa sp.; MYS, mysid (Antarctica maxima and A. ohlini); THO, Themisto gaudichaudii; CRU, other crustacea; FSH, fish

https://static-content.springer.com/image/art%3A10.1007%2Fs00300-008-0552-7/MediaObjects/300_2008_552_Fig4_HTML.gif
Fig. 4

Ontogenetic variability in the diet of Champsocephalus gunnari indicated by mean percent index of relative importance (% IRIDC) of prey categories with 95% confidence intervals (error bars) for fish in each putative age class. Data pooled from three seasons 2004–2006. KRI, E. superba; THY, Thysanoessa sp.; MYS, mysid (Antarctica maxima and A. ohlini); THO, Themisto gaudichaudii; CRU, other crustacea; FSH, fish

https://static-content.springer.com/image/art%3A10.1007%2Fs00300-008-0552-7/MediaObjects/300_2008_552_Fig5_HTML.gif
Fig. 5

Inter-annual variability in the diet of Champsocephalus gunnari illustrated by mean percent index of relative importance (% IRIDC) of prey categories with 95% confidence intervals (error bars) for fish in each putative age class. KRI, E. superba; THY, Thysanoessa sp.; MYS, mysid (Antarctica maxima and A. ohlini); THO, Themisto gaudichaudii; CRU, other crustacea; FSH, fish

https://static-content.springer.com/image/art%3A10.1007%2Fs00300-008-0552-7/MediaObjects/300_2008_552_Fig6_HTML.gif
Fig. 6

Regional variability in the diet of Champsocephalus gunnari indicated by mean percent index of relative importance (% IRIDC) of prey categories with 95% confidence intervals (error bars) for fish in each putative age class. Analysis was carried out on fish of putative ages 2+ and 3+ with samples from different years pooled. KRI, E. superba; THY, Thysanoessa sp.; MYS, mysid (Antarctica maxima and A. ohlini); THO, Themisto gaudichaudii; CRU, other crustacea; FSH, fish

Inter-annual and ontogenetic variation in the diet of icefish

There was clear evidence of both inter-annual and ontogenetic variation in icefish diet (Figs. 4, 5). When year-specific samples were pooled by age class (Fig. 4) an ontogenetic shift in diet from age 1+ to age 2+ was evident indicating that krill was generally less important in the diet of 1+ fish than for older fish. Sample size was small for the 1+ fish (Table 1) and there was inter-annual variability in the diet of this age class (Fig. 5). This produced wide 95% confidence intervals when the data set was resampled (Fig. 5). Also, although in all 3 years the 1+ fish had consumed appreciable quantities of the small euphausiid Thysanoessa, a considerable proportion of the diet could not be identified to the species level and so was included in the “other crustacea” category. This diet category consisted mainly of small euphausiids (predominantly Thysanoessa sp. and E. frigida) which were difficult to separate and identify when slightly digested.

The larger (older) fish were highly dependent on krill and to a lesser extent on T. gaudichaudii, with 60–80% of the diet (indicated by % IRIDC) made up from krill and the remainder being mostly T. gaudichaudii. The 4+ fish had consumed more mysids and fish than the younger fish (Fig. 4), but the importance of these diet categories was low (<10%) for all age classes (Fig. 5).

There were clear inter-annual differences in the diets of all four age classes. Across all year classes studied the amount of krill in the diet was highest in 2006 (Fig. 5) and the proportion of krill in the diet was generally much lower in 2004. This was most apparent for age 2+ and age 3+ fish, which took substantially less krill in 2004 than in either 2005 or 2006 (Fig. 5). Krill represented over 80% of the diet of age 3+ fish in 2006. The lack of krill in the diet in 2004 of age 2+ and age 3+ icefish was accompanied by an increase in T. gaudichaudii. These fish had also supplemented their diet with mysids (Antarctomysis maxima and A. ohlini) (Fig. 5). In the 4+ fish there was no significant difference between the amount of krill consumed in 2004 and 2005, but in both years krill was a less important component of diet than in 2006.

Regional variation in diet

This analysis was carried out on age 2+ and age 3+ fish pooled from all years (Table 1). There was little regional difference in the proportion of krill consumed. A greater proportion of krill was consumed in the SW when compared to the NE, SE and NW regions (Fig. 6). The proportion of T. gaudichaudii in the diet was lowest in the SW, where its consumption was lower than in the NE and NW. Most of the Thysanoessa sp. in fish stomachs was identified in fish caught at SR and NW South Georgia (Fig. 6). Most mysids were found in fish caught in the SE region of South Georgia (Fig. 6).

Predator size and prey size

There was no clear relationship between predator size and prey size (TL) when all prey were considered together (Fig. 7). Krill, T. gaudichaudii and Antarctomysis sp. all had a relatively constant mean size with increasing predator size. A range of sizes of fish had been preyed on by the icefish, particularly by larger individuals. The consumed krill had a modal size of 45 mm (Fig. 8; data pooled from 2005 and 2006 seasons). The mean size of Thysanoessa sp. found in icefish stomachs was 19 mm TL.
https://static-content.springer.com/image/art%3A10.1007%2Fs00300-008-0552-7/MediaObjects/300_2008_552_Fig7_HTML.gif
Fig. 7

Predator length versus mean prey length of seven different prey types identified from icefish stomachs (stomach data pooled from 2005 and 2006) based on the mean length of these prey in a subsample of 20% of the icefish stomachs

https://static-content.springer.com/image/art%3A10.1007%2Fs00300-008-0552-7/MediaObjects/300_2008_552_Fig8_HTML.gif
Fig. 8

Length frequency of all krill prey. Based on mean length measured from each stomach in the dataset (data pooled from 2005 and 2006)

Inter-annual variation in condition

Analysis of covariance of length-total wet weight and length-gutted wet weight regressions in male and female C. gunnari indicated some differences in condition between years. A significant interaction (P = 0.04) was found in the relationship between male length and total weight, so no further analysis of these data could be undertaken. There were significant differences between length-gutted weight regressions for males between years (Table 3) with condition in 2004 and 2005 significantly lower than in 2006. Comparison of both length-gutted weight and length-total weight regressions between years indicated a significant effect of year for females. Pairwise comparisons using Tukey tests indicated that there was a significant difference in fish condition between 2004 and 2006 for females using gutted weight. Using total weight, condition was significantly poorer in 2004 and 2005 than in 2006 (Table 3).
Table 3

Results from analysis of covariance, comparing the regressions of length and weight (either gutted wet weight or total weight) between years, to indicate changes in icefish condition

 

n

Interaction

Year

Tukey comparisons

F

P

F

P

Males GW

952

1.66

0.191 NS

4.45

4**

2004 < 2006

2005 < 2006

Males TW

1,209

3.22

0.040*

   

Females GW

1,016

1.27

0.283 NS

5.60

4**

2004 < 2006

Females TW

1,019

0.38

0.687 NS

23.97

<1**

2004 < 2006

2005 < 2006

GW gutted weight, TW total weight, NS not significant at P = 0.05 level

* Significant interaction of length made further comparison of regressions inapplicable

** Significant effect of year

Discussion

As with previous studies of mackerel icefish diet at South Georgia (Kompowski 1980; Kock et al. 1994; Barrera-Oro et al. 1998) the present study found that C. gunnari has a relatively low diversity diet dominated by krill and the hyperiid amphipod Themisto gaudichaudii. In addition, the present study showed inter-annual and ontogenetic patterns in C. gunnari diet that were supported by robust statistical comparisons of an index of relative importance of prey (% IRIDC) across three years of data. Some regional variation in diet was observed but was much less distinct.

Common measures of fish diet include the numbers, mass or frequency of occurrence of prey items (Hyslop 1980). Quantifying fish diets using single measures may produce a biased interpretation of diet and the use of several measures, aiming to give a more balanced interpretation, has been advocated (Hyslop 1980). An index of relative importance (IRI, Pinkas et al. 1971) is an attempt to reduce bias in the interpretation of dietary importance by combining three different dietary measures. Use of the % IRI was advocated by Cortes (1997) and this index has subsequently been widely used in studies of fish diet (Hahn et al. 2008) and diets of other taxa (Hart et al. 2002). However, there are problems associated with the use of % IRI. It has been criticised for lack of accuracy and precision (see Tirasin and Jorgensen 1999) and the multiplicative use of % frequency in the index means that increasing the taxonomic resolution of prey categories results in a reduction of the summed % IRI values (see Hansson 1998). Therefore, if several prey species are considered together in a category (as % IRIDC) the grouped % IRIDC value is not necessarily the sum of the individual species % IRI values. There is the potential for this to cause confusion (see Table 3 of Collins et al. 2008) and hence make comparisons between studies difficult. Although Cortes (1997) advocated the use of the % IRI as a standard index to facilitate dietary comparisons, he recommended that other measures of diet (numbers, mass and occurrence) also be reported. Liao et al. (2001) evaluated the performance of % IRI in comparison to the single measures of numbers, mass or frequency of occurrence, concluding that % IRI and % frequency of occurrence appeared to indicate levels of dietary importance with less bias in terms of the relative size of prey taxa than was observed when using either % N or % M.

Here, we used a bootstrapping technique to calculate confidence intervals for mean values of % IRIDC in order to facilitate statistical comparisons of the index for samples of C. gunnari grouped by age class, year or geographical region. This elucidated patterns in diet that provide fresh insight into ecological interactions of this species. These observations will be considered in relation to evidence of local variation in the availability of Antarctic krill during the period studied.

Although Antarctic krill was important to all age classes of C. gunnari, clear ontogenetic changes in diet indicated that icefish switched to larger prey as they grew. Smaller (younger) C. gunnari were less dependent on krill and took more of the smaller euphausiid Thysannoessa sp. than was taken by larger (older) fish. Krill at South Georgia tend to be of around adult size (>45 mm TL) having been advected to the region from recruitment at the Antarctic Peninsula (Atkinson et al. 2001; Murphy et al. 2007b). The modal size of krill in the present study was 45 mm TL (Fig. 8) and individuals of this size were probably too large to be taken by small icefish. The diet of larger (older) fish contained more mysids and fish, perhaps indicating a more benthic feeding habit (Frolkina 2002). Regional comparisons of mean % IRIDC indicated that T. gaudichaudii was a more important component in the diet of C. gunnari in the northeast and northwest than in other regions. To the southeast of South Georgia, mysids formed a greater proportion of C. gunnari diet (age 2+ and 3+ fish pooled) than in other regions. This is consistent with findings of Kock et al. (1994) and is probably a consequence of greater abundance of A. maxima in this area (Atkinson and Peck 1990). It seems likely that increased abundance of mysids in this region may provide adult C. gunnari with an alternative to E. superba and not necessarily that older fish actively seek mysids as prey.

The considerable decline (estimated at 80%) in the mackerel icefish stock, between January 2004 and January 2005, exceeds the episodic declines reported by Everson et al (1999). The three surveys considered here were conducted using the same vessel, with the same gear and fishing master. Hence, survey catchability was unlikely to have changed significantly between years. In addition, although fewer trawls were carried out in 2005 than in 2004 or 2006, licensed commercial fishing vessels also struggled to catch icefish during the 2004–2005 season (CCAMLR Online Statistical Bulletin; www.ccamlr.org) and fish were also not found closer inshore. Unreported catches are assumed to be negligible. Everson et al. (1999) presented a detailed analysis of the likely causes of inter-annual variability in icefish stocks, attributing declines to “krill poor” (low krill abundance, see Brierley et al. 2002) years. In such years, Everson et al. (1999) hypothesised that icefish struggle to find food and are also subjected to greater predation by the normally krill dependent Antarctic fur seals, which are abundant at South Georgia. Although icefish were abundant in January 2004, low stomach fullness, the low proportion of krill in the diet and poor condition all indicate that krill may have been in short supply at this time. This observation is consistent with acoustic surveys NW of South Georgia, which indicated that 2003–2004 was a relatively poor krill year (see Murphy et al. 2007a). Fur seals at South Georgia experienced high female mortality in 2003–2004 and poor breeding success in 2004–2005 (Forcada, BAS, unpublished data) which is also indicative of low proportion of krill in fur seal diet (Forcada et al. 2005) and hence of krill availability (Murphy et al. 2007b).

Of particular note is the indicated rapid recovery of the icefish stock from an estimated 9,000 tonnes in January 2005 to 87,000 tonnes in January 2006. The 2006 stock was dominated by putative 2+ fish (Fig. 2), which as 1+ fish in January 2005 were not fully selected by the bottom trawl survey (as a consequence of likely escapement through meshes and also because of the more pelagic existence of younger fish). Therefore, it seems that the mortality associated with icefish between January 2004 and January 2005 had less impact on the smaller fish. The proportion of fish in fur seal diet was high during the winter of 2004 (Forcada, BAS, unpublished) indicating some prey switching from krill to icefish. The impact of predation on icefish by both fur seals and gentoo penguins is likely to be spread across all age classes (Reid et al. 2005), yet the mortality between 2004 and 2005 shown here appears to have been limited to the older fish. Given that 2+ and 3+ fish are more krill dependent than 1+ fish, this would indicate that the mortality event was closely associated with the poor (krill) feeding conditions leading to starvation in winter and perhaps also to high post-spawning mortality. We were unable to assess the condition of 1+ fish in the present study because of small sample size. However, since small fish are less dependent than large fish on the variable supply of krill we consider that 1+ condition is less likely to vary between years. The condition of mackerel icefish has been found to influence spawning, with more fish spawning in years of higher krill availability (Everson et al. 1997, 2000). It is not clear if the conditions in 2003–2004 limited spawning, but the presence of 1+ fish during the 2006 survey suggests that some spawning had occurred.

Antarctic krill is a key component of the South Georgia ecosystem, with large populations of vertebrate species competing for this resource (Murphy et al. 2007b). Evidence of a long-term decline in krill (Atkinson et al. 2004) may result in this competition intensifying. It will therefore be important to understand the krill demands of species in order to facilitate sustainable fisheries management for both krill and icefish. Although detailed estimates of the daily feeding rate of icefish are currently not available, krill consumption can be roughly estimated as follows. If a daily feeding rate of 1.5% wet body weight per day is assumed (Flores et al. 2004) then multiplying this rate by the proportion of krill in C. gunnari diet (Table 2) and by the total mackerel icefish biomass in the study area would give annual krill consumption estimates of 130,000 tonnes in 2003–2004, 36,000 tonnes in 2004–2005 and 410,000 tonnes in 2005–2006. Values for icefish biomass given here are likely to be underestimates as the surveys were primarily conducted with a bottom trawl (headline height approx. 6 m) and a considerable portion of the icefish population is pelagic at any given time (Frolkina 2002). Kock (1985) estimated that icefish took 156,000–630,000 tonnes of krill per year, but that estimate was based on a summer feeding rate of 10% body weight per day. Other South Georgia shelf fish are also krill consumers (McKenna 1991; Collins et al. 2007, 2008; Reid et al. 2007; Clarke et al. 2008) but these species are less abundant than C. gunnari and they feed less exclusively on krill. Land-based predators are probably the greatest consumers of krill on the South Georgia shelf. These include fur seals and macaroni penguins, with published estimates of total annual krill consumption of 3.8 million tonnes (fur seals), and 8 million tonnes (macaroni penguins) (Boyd 2002). Note, however, that updated estimates of population size for these species suggest that the amount of krill consumed by fur seals is in fact greater than Boyd’s (2002) estimate and that macaroni penguins consume less (BAS, unpublished data).

The present study considered only the summer diet of C. gunnari. Further understanding of the ecosystem relationships of this species should be sought by studying diet in other seasons, particularly in winter, as this may be a key time when food is limited and natural mortality may be high. Feeding chronology was only briefly considered here because data were only available for daylight hours. A more comprehensive look at diurnal feeding patterns would be possible if data were collected from C. gunnari populations throughout the 24 h cycle. This is important because diet may vary diurnally and stomachs obtained during daylight (as in this study) may result in biased interpretation of prey importance. For example, too little importance may have been allocated to prey taken at night that were easily digested and hence present in less number and bulk during the daytime catches. Other potential sources of bias in the present study may have resulted from differential digestion rates, which may have under represented rapidly digested prey. This study focused on prey of macrofaunal size and above and may have therefore overlooked smaller prey such as copepods or other plankton. These may have been present particularly in the diet of small (age 1+) icefish.

In conclusion, although the % IRI has limitations, it is useful to report this measure of dietary importance, since it seems that the inclusion of several measures of diet may help to reduce the bias of its component single measures (Bigg and Perez 1985). Calculating confidence intervals for the % IRIDC has enabled robust statistical comparisons of fish diets in sample groups using this index. This demonstrated clear ontogenetic and inter-annual patterns in the summer diet of mackerel icefish at South Georgia and suggested that the decline in the icefish population following the 2004–2005 season was related to a reduction in food availability associated with a krill poor year. Future work should investigate seasonal patterns in the diet of C. gunnari to provide a full understanding of the ecological interactions of this species with its preferred prey, E. superba. In addition, understanding the effects of food availability on icefish condition may enable us to predict the likely success of cohorts. By quantifying feeding rates and monitoring condition, the uncertainty of natural mortality may be reduced. This could improve our management of key species.

Acknowledgments

We wish to thank Len Featherston (Master) and the crew and scientists on Dorada during the surveys in January 2004, 2005 and 2006. Thanks also to Sarah Clarke, Will Reid and Jamie Watts who assisted with some of the stomach contents analysis, and thanks to Pete Rothery for statistical advice. The study contributed to the BAS Discovery 2010 programme.

Copyright information

© Springer-Verlag 2008