 |
 |
 |
 |
|
|
|
|
| Fish Eating Birds and Salmonids in Scotland |
| 3. The diet of sawbill ducks and cormorants in Scotland |
| M Marquiss & D N Carss |
| SUMMARY |
| The diets of both sawbills and cormorants in Scotland were described using key bones (from 12700 fish collected from 1578 bird stomachs, 1990-96) to estimate the species and sizes of fish consumed at particular places and times of year. |
| The prey taken commonly and most widely by all 3 species, were brown trout, salmon parr, eel and minnow. Other species that sometimes predominated diet included 3-spined stickleback, stone loach, lamprey and flounder taken by red-breasted mergansers; lamprey, stone loach and frog taken by goosanders; grayling, roach and flounder taken by cormorants on rivers, and perch and rainbow trout taken by cormorants on stillwaters. |
| Diet was diverse but there were predictable patterns. Differences between bird species and between localities were generally greater than differences between times of year, and between years. The stomachs of larger bird species contained larger fish and fewer salmon than did those of smaller ones. In the north, the diet of all bird species was less diverse with a greater proportion of salmon than in the south. More salmon were consumed in March/April and in early winter, than at other times of year. |
| Using latitudinal trends in the proportion of salmon in the diet during winter and spring, red-breasted merganser diet averaged about 30% salmon (by mass) on southern rivers and 56% in the north. The equivalent figures for goosander diet were 9% in the south and 41% in the north, and for cormorant, <1% in the south and 18% in the north. Trout was a greater dietary component than was salmon in all 3 bird species, but its proportion showed no discernible latitudinal trend. |
| The median length of brown trout taken by cormorants from lochs was 239mm. On rivers, red-breasted mergansers took smaller brown trout (median 86 mm) than did goosanders (111 mm), and cormorants took the largest (163 mm). The average length of salmon taken by mergansers was 83 mm, less than that for goosanders (92 mm) and cormorants (101 mm). |
| In terms of the numbers of commercially important fish, the highest numbers (>3 fish per 100g food) of large juvenile salmon (>89 mm length) were found in mergansers from the Rivers North Esk, Spey and Findhorn, and in goosanders from the Beauly, Dee and Deveron. The highest numbers (>0.25 fish per 100g food) of large 'angleable' trout (>199 mm) were found in cormorants from the Rivers Laggan and Deveron, and Cobbinshaw Loch. |
| These dietary patterns for birds in Scotland, together with studies from elsewhere in Europe and North America, are consistent with the hypothesis that these fish-eating birds take the largest of prey that are most available to them. Future studies should concentrate on the factors predisposing fish to predation. A first step should be improved estimates of fish abundance in the fish communities of the lower, wider stretches of rivers where most birds forage. |
| INTRODUCTION |
| Both sawbills and the great cormorant are widely perceived to be damaging to salmonid fisheries on Scottish rivers, and elsewhere (Marquiss & Carss 1994, Russell et al. 1996). This chapter describes the diet of these fish-eating birds in terms of the sizes of salmon and trout consumed, their proportionate mass in the overall diet and thus, the numbers consumed per unit mass of fish intake. To derive the latter statistics we need the composition of the whole diet including the mass of other foods consumed. |
| Previous studies of fish-eating bird diet in Scotland |
| In the early 1960s, sawbill diet in Scotland was documented using the stomach contents of 147 goosanders and 148 mergansers, shot on a number of major rivers and a few lochs and estuaries (Mills 1962a). Juvenile salmon were the most commonly found prey, but goosanders also took (in decreasing order of frequency) perch, brown trout, eel, minnow, pike, gudgeon, and bullhead. As well as juvenile salmon, mergansers took brook lamprey, minnow, common goby, eel, stickleback, flounder, perch, brown trout, butterfish and saithe. More recently, Feltham (1990, 1995b) examined the stomach contents of 99 mergansers shot between 1987-90 on the rivers North and South Esk. Here juvenile salmon constituted the highest proportion of the diet by mass but birds also took brown trout, eel, minnow, stickleback, sandeel, stone loach, lamprey and flounder. Feltham (1995a) also documented the predominantly salmon diet of goosanders in summer on the River Dee. |
| Mills (1965) examined the stomachs of 129 cormorants from Scottish fresh waters during the early 1960s, mostly in the Conon and Ness catchments but also from other lochs, rivers, estuaries and firths. On rivers, cormorants took mostly stickleback, salmon and trout but some flounder, eel, perch, and pike were also eaten. On lochs, cormorants took mostly perch and salmon with some trout and eel. McIntosh (1978) described the stomach contents of 26 cormorants shot during the 1972/73 winter on the lower reaches of the River Tweed. Here, the diet was mainly roach, salmonids and flounders although grayling, eel, gudgeon, dace and brook lamprey were also recorded. |
| The usefulness of these studies for modelling impact on fisheries has been limited because, either they were essentially qualitative (e.g. Mills 1962a, 1965, McIntosh 1978, see below) or were restricted to a few river systems (e.g. Feltham 1990, 1995, McIntosh 1978). Detailed knowledge of bird diet provides information on whether or not fishes of commercial concern (such as juvenile salmon and large trout) are eaten, but to model impact it is also necessary to know how much such predation varies, both spatially and temporally. This requires samples from a variety of places, seasons and years. |
| We investigated bird diet using samples of birds shot under licence by fisheries managers where and when it was thought that the birds were causing serious damage to fisheries. Our results do not therefore document the overall diet of these birds in Scotland but are biased towards times and places where the birds were most likely to have taken fish of commercial interest. In this Chapter our aim is to provide estimates of the diet of these birds and the number of trout and salmon they consumed. |
| METHODS |
| Every method for estimating the diet of fish-eating birds has well-known associated biases and problems of interpretation (see extensive reviews: Duffy & Jackson 1986, Marquiss & Leitch 1990, Harris & Wanless 1993, Carss et al. 1997a). The previous Scottish studies by Mills (1962a, 1965) and McIntosh (1978) used only data from the intact fish in stomachs. This method is biased towards the largest items. The prevalence of smaller fish species is underestimated and the average size of juvenile salmon eaten overestimated by about 20% (Marquiss & Carss 1997). |
| Measurements of uneroded boney material in the foregut give more accurate estimates of diet. In the present study we separated out and measured hard 'key' bones that are resistant to rapid digestion (Feltham 1990, Carss & Marquiss 1997) and reconstituted the diet using a series of species-specific relationships between the size of these bones, fish length, and fish weight. |
| Laboratory procedure |
| Bird
carcases were cut open, the gut removed and the hindgut
(small intestine and colon, containing only well digested
liquid material) discarded. The foregut, including the
proventriculus (and the gizzards of ducks), was opened
and the contents examined. For simplicity we refer to
these as stomach contents. These stomach contents were
categorised according to their relative states of
digestion: (i) Very fresh fish, presumably swallowed immediately prior to the bird being shot, which showed no visible signs of digestion and had intact tail and fin rays. (ii) Intact fish with part-digested skin, fins and tail which could not be measured directly with any confidence. (iii) A section of more heavily digested material, where fish still had much of their flesh, some skin and fins were visible but it was impossible to measure individuals. There were also skeletons and skulls that were beginning to fragment and disarticulate. (iv) A distinct layer at the bottom of the stomach of loose, eroded bones, grit, pebbles and other non-digested material, which remained from previous meals. |
| Intact fish (i) were removed, identified (Wheeler 1978) and fork length (snout to the fork of the tail) measured to the nearest 1mm. Material (ii) to (iv) was rinsed out of the stomach and digested in vitro in a solution of biological washing powder to remove the flesh. The remaining hard parts were then rinsed, dried and examined with a binocular microscope. Bones that were discoloured and friable, visibly eroded or broken and worn smooth, were from the bottom layer (iv) and discarded. Keybones from recent meals were extracted from the remaining undamaged material. These included atlas, thoracic and caudal vertebrae, pharyngeal teeth, pelvic girdles, opercular bones, cleithrae, lower jaws (dentaries) and cyclostome teeth (from lampreys). Key bones were identified and measured. The measurement was used to estimate fish lengths and hence fresh weights, by a series of regression relationships (references detailed in Appendix of Carss & Marquiss 1997, additional regressions in Carss 1993a, Carss & Brockie 1994, Carss & Elston 1996, Carss & Marquiss 1996a). |
| Finally, the diet of birds from particular samples was reconstituted by adding together the total fish, by number and by mass, contained in their stomachs. Some stomachs were empty and a few contained only marine fish, presumably because the shot birds had newly arrived in fresh waters. These stomachs were not included in our estimates of diet. |
| Estimating the length of juvenile salmon from atlas vertebrae |
| There was a potential problem in estimating the length of young salmon using the width of the atlas vertebra (Feltham & Marquiss 1989). As a juvenile salmon changes from parr to smolt, its body elongates; for wild fish of any given atlas width, smolts are longer than parr by up to 11 mm (Armstrong & Stewart 1996). The Feltham & Marquiss (1989) equation was derived from an unknown mixture of parr and smolts. It is closer to Armstrong & Stewart's regression for parr than their one for wild smolts. If used to estimate the lengths of smolts, their average length might be underestimated and this would result in some small smolts being misclassified as parr on the basis of estimated length (e.g. Feltham 1995a, 1995b). |
| There is as
yet no way of classifying fish as smolts or parr from the
partially digested remains of most of the salmon in
stomachs so we could not apply the wild smolt and parr
equations of Armstrong & Stewart separately. However,
we could derive an overall equation for atlas width to
fork length that would be applicable to all juvenile
salmon atlases from our sample of stomachs, by using the
most pristine of whole fish we recovered from these same
stomach samples. The fork lengths of these fish were
measured, then the atlas vertebrae digested in vitro
and the width measured. Bird carcasses had been frozen so
we assumed that the pristine fish we recovered had shrunk
during freezing (Armstrong & Stewart 1997). We
estimated their original fork length (FLlive)
using a regression derived from a sample of freshly
killed juvenile salmon that had been measured fresh, then
frozen, thawed and remeasured (FLfrozen): FLlive = 1.980 +
1.032.Flfrozen |
| Estimated original fork length was then regressed on atlas width. The regression was similar to that of Feltham & Marquiss (1989) but differed markedly from that given for wild smolts by Armstrong & Stewart (1996) (Table 3.1). These authors' regressions for parr (b) and wild smolts (c) had lower gradients than the regression lines of Feltham & Marquiss (a), and the present study (d). This was consistent with the hypothesis that regressions (a & d) were derived from samples of juvenile salmon that were predominantly parr but contained a few large fish that were elongate, i.e. smolts. Equation (d) was used in the present study because it was directly appropriate to the material we examined. |
| Table 3.1 Regresson equations of fork length on atlas width. |
| Provenance | Sample |
n |
gradient |
intercept |
% variance explained |
| (a) Feltham & Marquiss (1989) | parr & smolts | 88 |
60.5 |
- 8.95 |
0.961 |
| (b) Armstrong & Stewart (1996) | parr | 29 |
57.1 |
-3.47 |
0.971 |
| (c) Armstrong & Stewart (1996) | wild smolts | 20 |
56.9 |
+7.96 |
0.755 |
| (d) Present study | parr & smolts from bird stomachs | 83 |
60.7 |
-6.95 |
0.943 |
| Data analysis |
| The accuracy
of diet estimates are heavily influenced by the number of
stomachs used, particularly where samples comprise less
than 10 stomachs (Marquiss & Carss 1997). Our sample
sizes varied for different places and times of year, and
were often <10. We therefore searched for significant
variation associated with bird species, location, year
and time of year, using General Linear Models (GLMs,
computed with Minitab statistical package 'release 11',
1996) comparing variation within and between samples,
weighted by the number of stomachs within a sample. The
variables compared included: (i) mean fresh mass of reconstituted stomach contents, (ii) mean mass of fish in a sample, (iii) number of fish per stomach, (iv) number of fish species in a sample, (v) proportion by mass of the most widely taken fish species (i.e. brown trout, salmon, eel and minnow), (vi) mean fork length of brown trout and salmon per stomach, (vii) number of salmon consumed per 100g fresh mass of stomach contents, (viii) number of large salmon (>89mm) consumed per 100g fresh mass of stomach contents, (ix) number of brown trout consumed per 100g fresh mass of stomach contents, (x) number of large trout (>199mm) consumed per 100g fresh mass of stomach contents. |
| By combining these last 4 variables (vii-x), along with published estimates of the daily food consumption of birds, we later calculate the numbers of these four categories of salmonids consumed per bird per day (Chapter 4). |
| The factors
entered into the statistical models categorised each
sample of stomachs as far as was possible from the data
provided with carcasses: (i) bird species (goosander, red-breasted merganser, cormorant), (ii) locality (21 rivers and 7 standing waters), (iii) year (1990-1996), (iv) time of year (2-monthly periods: Jan/Feb, Mar/Apr, May/Jun, etc). |
| Where necessary, highly skewed variables were transformed to near normality using an optimally calculated exponent (Lambda, Box-Cox transformation, Minitab "release 11", 1996). In each case the average values subsequently quoted in comparisons between samples were derived by back transformation from the mean transformed variable. |
| The number of prey species in a sample (iv) increased with sample size (number of stomachs, N) in a log linear relationship (number species = 1.79+3.65 logN; F 1, 183 = 204.6, p<0.001, r2 = 0.53) so in the analysis of variance for this variable, sample size (Log N) was used as a covariate. A preliminary analysis showed that much of the variation was explicable in terms of the provenance of samples with a less diverse, more salmon dominated diet, for carcases from northern river systems. We therefore used latitude as a covariate in the statistical models to test whether this explained more of the variance than did location alone. |
| Analyses proceeded from the full dataset, partitioning variance and describing the most important differences between samples by combining those where there was no statistical significance (p>0.05), or those where the variation was minor compared with the factor(s) that explained most. In doing so we were nevertheless cautious of the precision of estimates, as even combined samples were sometimes less than 10 stomachs. |
| RESULTS |
| Between 1990 and 1996, we received the carcasses of 1099 goosanders, 186 red-breasted mergansers and 293 cormorants, from which we recovered the identifiable remains of about 12700 food items, almost entirely (>99%) fish. Full dietary details of these birds, which formed 185 samples, are given in Appendices 4 & 5). As was expected from their large size (Chapter 4), the stomachs of cormorants contained the greatest weight of food and the largest fish, but on average fewer items than did those of sawbills (Table 3.2). Similarly goosanders (the larger of the two sawbill species) contained more food, comprising fewer, larger fish, than did red-breasted mergansers. |
| Table 3.2. Summary statistics for the contents of goosander, red-breasted merganser and cormorant stomachs. |
Goosander |
Red-breasted merganser |
Cormorant |
|
| Number of carcasses | 1099 |
186 |
391 |
| Proportion containing freshwater fish | 91% |
85% |
75% |
| Number of food items | 9386 |
1741 |
1575 |
| Mean number of fish/stomach | 5.78 |
8.31 |
3.01 |
| Mean fresh wt/stomach | 62g |
44g |
229g |
| Mean fresh weight/item | 9.9g |
4.9g |
63.6g |
| Mean number of prey species (adjusted to samples of 5 stomachs) | 4.3 |
4.8 |
3.1 |
| ANOVAs: fish/stomach, F2, 143 = 21.7; fresh wt/stomach, F2, 143 = 116.8; fresh wt/item, F2, 143 = 139.7; prey species, F2, 143 = 19.1; all p<0.001. |
| General patterns in diet |
| Although the contents of stomachs were diverse and included 17 species of freshwater fishes, the prey taken commonly and most widely, by all 3 bird species, were brown trout, salmon, eel and minnow. Other species that were taken in large amounts at some places and times of year included 3-spined stickleback, stone loach, brook lamprey and flounder (by red-breasted mergansers); river lamprey, stone loach and frog (goosanders); grayling, roach and flounder (cormorants on rivers) and perch and rainbow trout (cormorants on still waters). |
| Amongst the 185 samples of stomachs, there was highly significant variation in the proportions by mass of brown trout, salmon, eel and minnow, the four main prey species, as well as the total number of prey species present (Table 3.3). Most of this variation was attributable to location, somewhat less to bird species and least to differences between years and between time of year. The exception to these overall trends was the proportion of trout in stomachs, which varied more between years than it did between bird species. However this variation was still only a quarter of that attributable to differences between locations. Therefore, to best detail the main differences in the stomach contents of birds, we sorted them primarily by location and bird species but where sample size allowed, also by time of year and year. |
| Table 3.3. Partition of the variation amongst samples (n = 185) in the diversity (number of prey species) and composition (proportion by mass) of stomach contents, associated with bird species, location, time of year, and year. |
| Factor | Variable |
||||
Proportion trout |
Proportion salmon |
Proportion eel |
Proportion minnow |
Number of prey species |
|
| Bird spp. | 0.02 |
0.15 |
0.08 |
0.13 |
0.06 |
| Location | 0.31 |
0.27 |
0.23 |
0.34 |
0.19 |
| Time of year | 0.03 |
0.05 |
0.02 |
0.02 |
0.01 |
| Year | 0.08 |
0.05 |
0.04 |
0.03 |
0.01 |
| Multivariate ANOVA, bird spp (F10,290 = 26.4), location (F105,713 = 4.4), time of year (F30,582 = 4.4) and year (F30,582 = 2.8). All factors across all variates are highly significant (p<0.001). Within cells, the variation is statistically insignificant (p0.05) for values of 0.03 or less. |
| Geographic variation in diet |
| Most of the stomachs were from birds shot in March or April, so we combined data from these months over all years to detail variation in the spring diet for all 3 bird species in relation to location (Figure 3.1). |
|
|
|