The impact of great skua predation on seabird populations at St Kilda: a bioenergetics model

Abstract

1. The number of great skuas Catharacta skua Brünnich breeding at St Kilda, Outer Hebrides, has increased rapidly in recent years, making it currently the second largest colony outside Shetland and the fastest growing in the UK. In comparison with Shetland, where the diet consists mostly of sandeel Ammodytes marinus Raitt and discarded gadoid fish, and very rarely of birds, great skuas at St Kilda feed far more extensively upon seabirds. This paper incorporates metabolic, dietary and demographic data to estimate the total mass of different prey consumed at St Kilda and to assess the potential impact of this predation on other seabird populations.

2. On the basis of a bioenergetics model incorporating fundamental life-history parameters, the great skua population at St Kilda was estimated to require 141 × 10⁶ kJ of energy per season, most of which (88·0%) was necessary for the maintenance and activity of breeding adults. Energy demand was considerably lower for non-breeders (2·5% of the total), and for chicks and fledglings (9·2%).

3. In addition to seabirds, great skua diet at St Kilda also included a considerable proportion of fish and goose barnacles Lepas sp. However, because of differences in mean meal mass and caloric density, meals of larger seabird prey were more important in terms of their energetic contribution in the diet than in terms of their relative abundance.

4. Combining the bioenergetics and prey consumption models, it was estimated that a total mass of 12·2 tonnes of fish, 1·6 tonnes of goose barnacles and 8·8 tonnes of seabirds was consumed by the great skua population at St Kilda to fulfil its total energy requirement in 1996. Overall seabird consumption was estimated to be 40,800 seabirds of seven different species. Although a proportion of birds killed are likely to be visiting non-breeders, the magnitude of this figure nonetheless suggests that great skuas may have a considerable impact on the internationally important populations of some seabirds at St Kilda.

Introduction

Considerable attention has been focused in recent years on the dynamics of the relationship between raptors and their avian or mammalian prey in terrestrial ecosystems (e.g. Cresswell & Whitfield 1994; Korpimäki 1994). Much less is known about the interactions between their ecological counterparts in marine ecosystems, the predatory gulls (family Laridae), skuas (family Stercorariidae) and giant petrels Macronectes spp. and the impact they may have on seabird populations (but see Furness 1981; Watanuki 1986; Young 1994).

In the southern hemisphere, brown, Tristan and Falkland skuas Catharacta antarctica (Lesson), and south polar skuas C. maccormicki (Saunders), feed mostly by predation on seabirds, with mammals and scavenged prey important only at a few specific sites (Jones & Skira 1979; Schramm 1983; Green 1986; Furness 1987; Norman & Ward 1990; Ryan & Moloney 1991; Young 1994; Mund & Miller 1995). The only exceptions to this are south polar skuas, which have been recorded feeding at sea on fish and krill, usually in areas of sympatry with brown skuas C. antarctica lonnbergi (Young 1963; Trivelpiece & Volkman 1982; Hemmings 1984; Pietz 1987).

The UK holds ≈ 60% of the world population of great skuas Catharacta skua Brünnich (Furness 1987). For most of the current century, comparatively few great skuas bred outside the main strongholds of Orkney and Shetland, although numbers appeared to be increasing in the Outer Hebrides during the 1980s (Furness 1987; Rennie 1988). At the large Shetland colonies, great skuas appeared to be dependent primarily on sandeels Ammodytes marinus Raitt and discarded whitefish from trawling activities (Furness & Hislop 1981; Hamer, Furness & Caldow 1991). Except in localized instances of dietary specialization, seabirds tended to be of relatively minor importance, at least up until the late 1980s when sandeel availability in Shetland waters declined dramatically (Hamer, Furness & Caldow 1991). However, a recent study of diet and breeding ecology at St Kilda, Outer Hebrides, has found that adults at that site feed far more extensively upon seabirds yet maintain high levels of chick growth and annual productivity (Phillips et al. 1997).

Great skuas were first recorded breeding on Hirta, the main island in the St Kilda archipelago (57°49′N 08°35′W), in 1963, and the size of the breeding population increased steadily up until at least 1990 (see Phillips et al. in press). In recent years there has been a rapid expansion to 213 pairs on Hirta and 16 pairs on the other islands in 1996, which makes Hirta currently the second largest colony outside Shetland and also the fastest growing colony in the UK (Phillips et al. in press). The birds at St Kilda represent > 1% of the world population of this species.

Seabird populations at St Kilda are of major international importance, and for this reason the archipelago has been designated as a UK National Nature Reserve, EC Special Protection Area and World Heritage Site. The archipelago holds ≈ 20% of the North Atlantic population of northern gannets Morus bassanus (L.), ≈ 20% of the world population of the grabae race of Atlantic puffins Fratercula arctica (L.), the largest colony of northern fulmars Fulmarus glacialis (L.) in western Europe, the largest colony of Leach's petrel Oceanodroma leucorhoa (Viellot) in the eastern Atlantic and the largest colony of British storm petrels Hydrobates pelagicus (L.) in the UK (Tasker, Moore & Schofield 1988). The impact predation by great skuas may be having on these species is unknown.

In this paper we develop a model of great skua energy requirements at St Kilda using fundamental life-history parameters. By incorporating information on the meal size and caloric value of the main food items, and the proportion of the diet of breeding adults, non-breeders and chicks made up by different prey, it is possible to estimate their relative contribution to the overall population energy demand. Since the species composition of avian prey is also known (Phillips et al. 1997), we can then estimate approximately how many seabirds are consumed by great skuas at this site during the breeding season, and hence assess the potential impact this predation may have on prey populations.

Methods

Bioenergetics model of population energy requirements

The St Kilda archipelago consists of four main islands (Hirta, Dùn, Soay and Boreray) and several large sea stacks. The majority of the data included in this paper were collected during 1996 on the largest island, Hirta, which holds > 90% of the breeding adults. A bioenergetics model was constructed, incorporating and developing elements from a number of previous studies (Wiens & Scott 1975; Furness 1978; Wiens 1984; Cairns et al. 1990).

A complete census of great skuas was undertaken in 1996 by marking all nests on Hirta and Dùn, and counting apparently occupied territories on the two other islands (see Phillips et al. in press). The mean number of non-breeders attending the only club site at St Kilda, on Hirta, was determined by direct counts at intervals throughout the breeding season.

Energy requirements were estimated separately for breeding adults, non-breeders and chicks throughout the entire time spent at the colony, including the pre-incubation and post-fledging periods (see Table 1 for details). The major components in the model are illustrated in Fig. 1. All final estimates of energy requirements, including egg production costs of breeders, took assimilation efficiency into account. This was assumed to be 0·76 for all food types, a value obtained during controlled feeding trials of captive great skuas fed whole sandeels and whiting Merlangius merlangus (L.) (G. Hilton, unpublished data).

Table 1.  Parameters used in the great skua bioenergetics model

Parameter	Value	Source
Total breeding population in 1996	229 pairs	Phillips et al. (in press)
Mean non-breeding population	≈13 birds	This study
Pre-breeding period	≈30 days	P. Catry, personal communication
Incubation period	29 days	Furness 1978
Chick-rearing period	≈47 days	This study
Post-fledging period at colony	≈18 days	R.W. Furness, personal communication
Adult BMR	538 kJ day^−1	Bryant & Furness (1995)
Adult FMR:BMR ratio	3·5	See Methods
Mean clutch size	1·91	Hamer, Furness & Caldow (1991)
Mean fresh egg mass	96 g	Furness (1978)
Mean egg energy density	6·45 kJ g^−1	Meathrel et al. (1987); Meathrel, Ryder & Termaat (1987)
Egg synthesis efficiency	0·75	Ricklefs (1974, 1983)
Food assimilation efficiency	0·76	G. Hilton, unpublished data
Mean brood size at 20 days	1·06	This study
Mean brood size at fledging	0·84	Phillips, Thompson & Hamer (1997)
Mean chick fledging mass	1170 g	Phillips, Thompson & Hamer (1997)
Energy requirement of fledglings	1025 kJ day^−1	See Methods

Figure 1. Structure of the bioenergetic model of population energy requirements. *Calculation already incorporates assimilation efficiency.

The energy requirements for the maintenance and activity of individual adults were calculated assuming an overall field metabolic rate (FMR) throughout the season equivalent to 3·5 times basal metabolic rate (BMR). This FMR:BMR ratio was the average estimated by Caldow (1988) for great skuas breeding on Foula, Shetland, in 1973–76 on the basis of time-budgets and incorporating multiplicands of BMR for specific activities, and lies in the middle of the range recorded in most seabird studies to date, with one or two exceptions (see Discussion). Non-breeders do not have to produce a clutch or provision chicks, but are probably not as efficient at finding food as breeding adults and hence were assumed to have approximately equivalent foraging and maintenance costs (following Cairns et al. 1990). Seasonal energy requirements were calculated assuming a constant number of birds was present throughout the breeding period, although not necessarily the same individuals in the case of non-breeders.

The energy cost of clutch formation (E_clf in kJ) for each breeding pair was estimated according to the equation E_clf = Cls × M_egg × Cal_egg × (1/E_syn), where Cls is mean clutch size, M_egg is mean fresh egg mass in g, Cal_egg is average energy density of larid eggs in kJ and E_syn is egg tissue synthesis efficiency (for references see Table 1).

Total metabolic energy requirement (ME in kJ) from hatching to fledging of each chick was estimated from the equation ME = 35·15 × M

• 

where M_c is chick mass in g (Drent, Klaassen & Zwaan 1992). Energy demand of each fledgling during the short period before departure was estimated from a regression equation fitted to daily energy requirement (in kJ) at fledging (from Figs 3 in Drent, Klaassen & Zwaan 1992) vs. fledgling mass (Tables 1 in Drent, Klaassen & Zwaan 1992) for four larid species. Great skuas fledge weighing c. 1170 g, only slightly heavier than herring gull Larus argentatus Pontoppidan fledglings, the largest species in that study. Both of these equations take account of assimilation efficiency.

A sensitivity analysis of the model was performed in two ways: by increasing the value for each input parameter by 1% (following the approach of Furness 1978), and also by raising or lowering the value for each input parameter to what were considered to be reasonable extremes and noting the effect this had on the output of the model (following the approach of Diamond, Gaston & Brown 1993). Likely extremes for many parameters, particularly those measured directly for the St Kilda population in 1996, and colleagues that exhibit little variation in published studies, were taken to be ± 10% of the starting value. Slightly narrower limits (± 8%) were set for assimilation efficiency, as upper and lower values for this variable are between 0·72 and 0·82 even across a wide range of bird species (Gabrielsen et al. 1991; Drent, Klaassen & Zwaan 1992; Mehlum, Gabrielsen & Nagy 1993). Wider potential extremes were considered appropriate for mean duration of the pre-breeding and post-fledging periods, fledgling energy requirements and FMR:BMR ratios of breeding adults and non-breeders, as these were not measured directly. The average number of non-breeders foraging around the colony was also given a high upper, but not lower boundary, for reasons outlined in the Discussion.

Diet composition

Diet composition of breeding adults and non-breeders at St Kilda in 1996 from June onwards, i.e. shortly before the start of the chick-rearing period, was known from analysis of regurgitated pellets (see Phillips, Thompson & Hamer 1997; Phillips et al. 1997 for details). Diet prior to this was assessed from regurgitates of adults nest-trapped during incubation, on the assumption that these accurately reflected the proportion of meals in the three main prey categories – seabirds, fish and goose barnacles Lepas sp. – and using the ratio of different bird species in pellets collected in June as an index of species composition (see below). A similar approach was adopted to determine the diet composition of the young, on the basis of the relative proportion of the three main prey types in regurgitates obtained from chicks during routine ringing and measuring, and using the ratio of bird species in pellets of adults collected throughout chick-rearing as an index of species composition.

Different bird species in the diet were identified from feather characteristics and morphology of legs, feet, bill and wings found in pellets, and different fish species from the presence of vertebrae and sagittal otoliths (Phillips et al. 1997). Pellets containing bird remains were identified to the appropriate species with the exception of a proportion of pellets of auk (Alcidae) and storm petrel (Hydrobatidae). Different species were assumed to occur in a similar proportion among these pellets as those positively identified. A small number of pellets of other food types (squid, decapod crustacea, and bird eggs) and single pellets of uncommon bird species were recorded in 1996, but because of their relatively low incidence these were not considered further.

In the great majority of cases, pellets contained the remains of only one type of prey and can therefore be assumed to represent individual meals (Furness & Hislop 1981). The exceptions to this are pellets resulting from predation of Leach's petrel and British storm petrel. These small species are swallowed whole, and on the evidence of groups of pellets found together on breeding territories clearly consisting of combinations of wings, whole legs or body feathers, it was estimated that at least three pellets result from a meal of a single individual. A correction factor, that three pellets of these species represented a single meal, was therefore applied throughout. This was not the case for the larger seabirds, which are not eaten whole. Pellets of these species were much more likely to occur singly on territories and did not contain wings or legs. The extent to which pellet ratios adequately reflect the proportion of meals of different prey is considered in the Discussion. In our analyses, we considered a single meal to be the quantity of food present in a bird's proventriculus on its return from a foraging trip. Information on relative prey frequencies during different stages of the season was integrated, on the basis of the relative duration of each stage, into an index for the overall proportion of meals of different types consumed during the entire period that breeding adults, non-breeders and chicks spend at the colony.

Meal mass and caloric density

By weighing chicks before and after a feed, Furness & Hislop (1981) found that a mean of 91 g (range 50–125 g, n = 18) of discarded gadoid fish or bird meat was fed to great skua broods at each meal. We therefore assumed a mean meal mass for great skua adults of 100 g (following Furness & Hislop 1981) for meals of fish or larger seabird prey. This may be a more or less universal value for Catharacta skuas (see Discussion). In order to simplify the analyses, we assumed the indigestible portion of the meal, which is later regurgitated as the pellet, was of negligible mass. In contrast to the larger seabird prey species, Leach's petrel and British storm petrel are small enough to be swallowed whole. Consequently, meals of this type were assumed to be equivalent to adult mass of the two species at St Kilda (Cramp & Simmons 1977). Meals of goose barnacle were assumed to weigh 40 g, for reasons outlined below.

Over 90% of otoliths recorded in great skua pellets at St Kilda in 1996 were Norway pout, Trispoterus esmarki (Nilsson), or poor cod T. minutus (L.), whiting and haddock Melanogrammus aeglefinus (L.) (Thompson, Hamer & Phillips 1998). Based on the length of intact otoliths and regression equations (Härkönen 1986), the whiting and haddock appeared to be large fish (over 100 g) and were probably not taken whole. The Norway pout and poor cod in the diet will have weighed considerably less. However, most pellets of these last species contained otoliths from several fish, and so the average meal size of 100 g suggested by Furness & Hislop (1981) seems appropriate. Fisheries samples of whiting caught in June and August had a mean caloric value of 5·2 kJ g–1 fresh mass, which appeared to be fairly typical of gadoid fish during the summer (Hislop, Harris & Smith 1991).

The edible portion of an individual seabird consists primarily of the skin, pectoral and leg muscles, viscera (kidney, liver, gut, heart), fat deposits and blood. No single study has addressed the question of what proportion of fresh body mass these components comprise. Pectorals, leg muscles and viscera together constituted a mean of 33·9% (range 29·5–38·9%) of fresh carcass mass of fulmars, kittiwakes Rissa tridactyla (L.), common guillemots Uria aalge (Pontoppidan), and puffins (G. Hilton, unpublished data). Blood volume of birds is between 6 and 13% of body mass (Dein 1986). Approximately 8% of total body mass of breeding guillemots and, assuming samples of adults collected at St Kilda shortly after egg-laying were representative, ≈ 6%, ≈ 14% and ≈ 15% of body mass of puffins, fulmars and Manx shearwaters Puffinus puffinus (Brünnich), respectively, was lipid (Osborn & Harris 1984; Furness, Thompson & Harrison 1994). Most, if not all of this fat will be in accessible tissues, although some will already have been accounted for in muscles and visceral components. For the purposes of analysis, it was assumed that, overall, ≈ 65% of fresh body mass from carcasses of the larger seabirds (kittiwakes, puffins, fulmars, guillemots and Manx shearwaters) would be consumed. These prey therefore provide several meals per carcass if we assume that great skuas carry 100 g of food, on average, back to the colony. Several great skuas were frequently observed feeding from single corpses below the cliffs around the colony, so this assumption seems to be reasonable. If less than 65% of each carcass is consumed, the final figure for total seabird consumption will be underestimated to a greater or lesser extent (see sensitivity analysis).

There are published values for mean mass of adults of the relevant prey species at various UK colonies, and in some cases St Kilda itself (Cramp & Simmons 1977, 1983; Galbraith 1983). Caloric values of whole birds show some degree of interspecific variation, ranging from 7·5 to 10·9 kJ g–1 fresh mass in adults and fully grown chicks (Brisbin 1968; Dunn 1975; and references therein; Adams et al. 1991). This appears to depend largely on relative fat content, which is high in seabirds. The edible body components extracted from carcasses of the larger prey species by great skuas will be the most energy-dense tissues, and so the maximum value of 10·9 kJ g–1 was used in analyses. Leach's petrels and British storm petrels are swallowed whole, and consequently a greater proportion of the prey ingested will be indigestible or of low caloric value. Mean energy content of meals of these two species was therefore adjusted downwards, by assuming that only 65% of each petrel consists of digestible material of high calorific value, as above.

Although 100 g is probably an appropriate mass for meals of bird meat and fish, meals of goose barnacles weigh considerably less. The total mass of seven large undigested goose barnacles regurgitated by a nest-trapped adult on Hirta in 1996 was 35 g. Counts of large fragments and complete shell plates found in pellets suggest this is towards the upper end of the range in terms of the number consumed per meal. Even allowing for a degree of water absorption since ingestion in the sample collected, a single meal of goose barnacle is therefore unlikely to weigh more than 40 g on average, and this value was used in further analysis. As a published value was not available, mean caloric density of goose barnacles from the regurgitate was determined by drying to constant mass, followed by homogenization in an electric mill and direct combustion of weighed samples in a ballistic bomb calorimeter (CB-370, Gallenkamp, London, UK) calibrated with benzoic acid.

Prey consumption model

Using the above values for mean meal mass and caloric value per gram fresh mass, it is possible to estimate the energy content of meals of a particular type. The relative importance of different prey in terms of the overall dietary energy requirement was then calculated by combining information on relative energy content with data on the proportion of meals of each type in the diet of breeding adults, non-breeders and chicks. Since energy requirements in kJ have been estimated using the bioenergetics model (see Table 2), the total amount of energy provided by each type of prey can be quantified, and from this figure it is straightforward on the basis of caloric densities to back-calculate the total mass of each food type necessary to satisfy that requirement. By incorporating data on mean mass of the prey species, the numbers of birds eaten can then be calculated on the assumption that all the edible components (i.e. 65%) of carcasses of the larger seabird prey species are consumed.

Table 2.  Energy requirements of great skuas at St Kilda in 1996

Individual/pair basis		Entire colony
	Daily	Per season	Daily	Per season
Breeding adults (229 pairs)
Maintenance and activity	2480 kJ bird^−1	307×10^3 kJ bird^−1	1135×10^3 kJ	141×10^6 kJ
Egg production costs	–	2070 kJ pair^−1	–	475×10^3 kJ
Non-breeders (13 individuals)
Maintenance and activity	2480 kJ bird^−1	307×10^3 kJ bird^−1	32·2×10^3 kJ	3·99×10^6 kJ
Chicks
Hatching to fledging	–	45·7×10^3 kJ chick^−1	–	11·1×10^6 kJ
(1·06 chicks per pair)
Fledging to departure	1025 kJ fledgling^−1	18·5×10^3 kJ fledgling^−1	197×10^3 kJ	3·55×10 ^6 kJ
(0·84 fledglings per pair)
Total colony energy requirement	–	–	–	160×10^6 kJ

Sensitivity analysis

The relative importance of variation in input values for each element in the prey consumption model was considered in a sensitivity analysis, as above, by increasing the value for each input parameter by 1%, and also by raising or lowering the value for each input parameter to what were considered reasonable extremes. Likely extremes for mean caloric density were taken to be ± 25%, as this encompasses most variation in published figures for bird meat, crustacea, and gadid fish species caught during the summer (Brisbin 1968; Hunt 1972; Adams et al. 1991; Hislop, Harris & Smith 1991). Probable lower and upper limits to mean meal mass for fish and the large seabirds were considered to be 70 g and 130 g, i.e. ± 30% (Young 1963; Furness & Hislop 1981). Mean mass of meals of goose barnacles was given a higher upper boundary (+ 50%) as more assumptions were involved in deriving the original estimate. The data on mass of adult Leach's petrel and British storm petrel were collected at St Kilda and the mean was therefore considered to be within ± 10% of that of birds consumed in 1996. Wider likely extremes (± 50%) were applied for the proportions of different prey eaten by great skuas, because of possible problems in the use of pellets as indicators of diet (see Discussion).

Results

Bioenergetics model

Energy requirements

Energy requirements of each section of the great skua population are indicated in Table 2. As a proportion of the seasonal total for the colony, by far the largest component (88·0%) was for the maintenance and activity of breeding adults. Energy demand was much lower for chicks and fledglings (9·2% of seasonal totals), and for the small number of non-breeders (2·5%). Egg production costs were low, at 0·3% of the seasonal total for the colony, or less than 0·4% of adult activity and maintenance requirements.

Sensitivity analysis

Results from the sensitivity analysis are indicated in Table 3. Potential variation in input values for the number of non-breeders foraging around the colony, and their FMR:BMR ratio, had a comparatively minor effect on the overall estimate for energy demand of the entire great skua population. Similarly, variables associated with energy requirements of chicks and fledglings for growth and maintenance, and those of breeding adults for clutch formation, tended to have only a small influence. More important determinants of the overall population energy demand were the durations of the pre-breeding and post-fledging periods, primarily because insufficient information was available for St Kilda and wide potential extremes were deliberately set. Not surprisingly, as the largest element in the energy budget for the colony was the activity and maintenance energy requirements of breeding adults, possible variation in their BMR, FMR:BMR ratio and, to a lesser extent, assimilation efficiency, have a major influence on the final figure.

Table 3.  Sensitivity tests for the bioenergetic model. Shown in the table are the percentage changes in output (population energy requirement) resulting from a 1% increase in the value of an input parameter, and the percentage change in output at probable maximum and minimum extremes in input parameters

Parameter	% change in output for 1% increase	Likely extremes (%)	% change in output at extremes
Total breeding population	0·98	+10	+9·75
Non-breeding population	0·02	+200, −30	+5·00, −0·75
Pre-breeding period	0·22	±50	±10·95
Incubation period	0·21	±10	±2·12
Chick-rearing period	0·34	±10	±3·43
Post-fledging period at colony	0·15	±50	±7·68
Adult BMR	0·91	±10	±9·05
FMR:BMR ratio of breeders	0·88	±15	±13·21
FMR:BMR ratio of non-breeders	0·02	±15	±0·37
Clutch size	<0·01	±10	±0·03
Fresh egg mass	<0·01	±10	±0·03
Egg energy content	<0·01	±10	±0·03
Egg synthesis efficiency	<0·01	+10, −10	−0·03, +0·03
Food assimilation efficiency	−0·90	+8, −8	−6·73, +7·90
Brood size at 20 days	0·07	±10	±0·69
Brood size at fledging	0·02	±10	±0·22
Chick fledging mass	0·07	±10	±0·71
Energy requirement of fledglings	0·02	±20	±0·44

Prey consumption model

Relative energy content of different meals

On the assumption that meals of these two types were of the same mean mass (100 g), the estimated energy content (in kJ) of a single meal of bird meat from one of the larger seabirds was just over double that from fish, because of differences in caloric value per gram of fresh tissue (Table 4). Measured caloric value of goose barnacles was low. The mean value ± SD for goose barnacle tissue itself was 27·02 ± 1·81 kJg–1 dry mass (n = 7), with an overall energy density equivalent to 1·9 kJ g–1 fresh mass for the whole sample including shell plates. Because meals of goose barnacles were small (40 g) in addition to being poor energetically, their total energy content was under 15% of a fish meal and under 7% of a meal of the larger bird species. Estimated energy content of Leach's petrel and British storm petrel meals were also low, simply because of the small size of adults of these species.

Table 4.  Mean meal mass, energy content and frequency of different prey types at St Kilda in 1996 and their relative contribution to total energy requirements. For estimation of mean meal mass see Methods

				Breeding adults		Chicks and fledglings		Non-breeders
test	Caloric value (kJ g^−1)	Meal wet mass (g)	Meal energy content (kJ)	Percentage of all meals	Percentage of energy consumed	Percentage of all meals	Percentage of energy consumed	Percentage of all meals	Percentage of energy consumed
Fish	5·2	100	520	45·4	39·4	52·6	38·7	51·3	57·2
Goose barnacle	1·9	40	76	16·3	2·1	0	0	24·3	3·9
Leach's petrel	7·1	45	318·8	5·5	2·9	6·8	3·0	7·4	5·0
British storm petrel	7·1	25	177·1	2·8	0·8	3·4	0·9	3·0	1·2
Kittiwake	10·9	100	1090	2·2	4·1	2·8	4·3	0·8	1·8
Puffin	10·9	100	1090	3·8	6·9	4·7	7·2	7·1	16·6
Guillemot	10·9	100	1090	18·8	34·2	23·2	35·8	0	0
Fulmar	10·9	100	1090	4·8	8·8	6·0	9·2	5·6	13·1
Manx shearwater	10·9	100	1090	0·5	0·8	0·6	0·9	0·5	1·2

Variation in meal energy content had a very strong effect on the relative contribution that different prey made to the overall energy demand of the colony (Table 4). Most importantly, although goose barnacles made up 16% of meals of breeding adults and 24% of those of non-breeders, they accounted for around 2% and 4%, respectively, of the energy intake of these two groups. Meals of Leach's petrel and British storm petrel were also much less important energetically than their relative number would suggest. In contrast, meals of larger seabirds (kittiwakes, puffins, fulmars, guillemots and Manx shearwaters) were considerably more important in terms of their energetic contribution in the diet than in terms of their relative abundance.

Mass of prey consumed

The total energy requirement of breeding adults, non-breeders, chicks and fledglings (from Table 2) was apportioned according to the relative contribution of each prey type from Table 4. These figures were then converted to a value for the mass of prey consumed according to differences in prey caloric density. In total, an estimated 12·2 tonnes of fish, 1·6 tonnes of goose barnacles and 8·8 tonnes of seabirds were consumed in 1996 by great skuas at St Kilda (Table 5). Maintenance and activity requirements of breeding adults accounted for over 85% of the total mass of each prey type consumed by the colony.

Table 5.  Total fish, goose barnacle and seabird consumption by great skuas at St Kilda in 1996

Total mass consumed (kg) by
	Breeding adults	Chicks and fledglings	Non- breeders	Colony total
Fish	10 660	1090	439	12 190
Goose barnacle	1535	0	83·3	1620
Leach's petrel	580	62·9	28·4	670
British storm petrel	160	17·7	6·5	185
Kittiwake	525	57·4	6·5	590
Puffin	890	97·1	60·9	1050
Guillemot	4415	481	0	4895
Fulmar	1135	124	47·8	1305
Manx shearwater	110	11·8	4·4	125

Number of seabirds consumed

On the basis of published values for adult mass (Cramp & Simmons 1977, 1983; Galbraith 1983), and assuming that 65% of each carcass of the larger species was consumed, the total numbers of seabirds eaten at St Kilda can be calculated (Table 6). The most important species in terms of overall numbers were Leach's petrel, British storm petrel and guillemot. In total, an estimated 40 800 seabirds were eaten by the great skua population during the 124-day breeding season.

Table 6.  Total numbers of seabirds consumed by great skuas at St Kilda in 1996

Total numbers consumed by
	Prey mass (g)	Breeding adults	Chicks and fledglings	Non- breeders	Entire colony
Leach's petrel	45	12 825	1400	630	14 850
British storm petrel	25	6480	705	260	7450
Kittiwake	375	2160	235	27	2420
Puffin	370	3700	405	250	4360
Guillemot	860	7900	860	0	8760
Fulmar	800	2180	240	90	2510
Manx shearwater	420	395	45	16	455

Sensitivity analysis

Changes in caloric density by ± 25%, or in mean mass of meals of fish or the larger bird species by ± 30%, generally resulted in a 10–20% change in the estimated mass of fish and goose barnacles, and estimated total numbers of the larger birds consumed during the 1996 season (Appendix). Changes to the value for goose barnacle caloric density had very little influence, whereas a potential increase (by 50%) or decrease (by 25%) in the mean meal mass had a substantial effect on total consumption of this item, but trivial consequences for the estimated quantities of other prey consumed. Altering the mass of the two small petrels by ± 10% had very little effect on the output of the model. The total number of seabirds consumed (and particularly of the five larger species) was more sensitive to a change in the percentage utilization of each carcass. An increase to 125% of the original value resulted in a 10% decline in overall numbers eaten, whereas a decrease to 75% of the original value resulted in a 16% increase in seabird consumption.

Altering the proportion of a particular prey in the diet of non-breeders or young to the proposed upper and lower extremes of ± 50% generally had a very minor influence on the output of the model. This was not the case for breeding adults. An increase to the proposed limit of 150% of the observed proportion of fish in the diet of breeders resulted in a considerable rise in the estimate of total fish consumption, and a substantial decline in goose barnacle and seabird consumption. A 50% decrease, not surprisingly, had the opposite effect. Raising or lowering the proportion of goose barnacle in the diet of breeding adults resulted in very large changes in estimated consumption of barnacles, but not of any other prey type. Adjusting the observed proportion of each bird species in the diet of breeders to the ± 50% extremes had dramatic consequences for that particular species, but this was balanced by smaller changes in the opposite direction for the other species. Consequently, at most there was a 15% change in the overall figure when the proportion of Leach's petrel changed, but generally only a 10% or smaller change occurred when the the proportion of one of the other six species was altered.

Discussion

As in all bioenergetics studies of this nature, there are a number of potential sources of error in the estimation of values for the different parameters used in the model. The results of the sensitivity analysis highlighted the importance of several variables, in particular those associated with total number and daily energy expenditure of breeders. One of the most crucial parameters was clearly the FMR:BMR ratio of breeding adults, because the overall energy requirement of this group was very high compared to that of non-breeders or young (chicks and fledglings). The use of a seasonal average of 3·5 times BMR, however, seems a reasonable approach, because this is in the mid-range of published values for seabirds, which tend to average between three and four times BMR across the season as a whole (Gabrielsen, Mehlum & Nagy 1987; Birt-Friesen et al. 1989; Gabrielsen et al. 1991; Montevecchi, Birt-Friesen & Cairns 1992; Mehlum, Gabrielsen & Nagy 1993; Furness & Bryant 1996). A few species, including Cape gannets Morus capensis (Lichtenstein) and northern gannets, which use sustained flapping flight and plunge-dive during foraging, and common guillemots, which are wing-propelled pursuit divers, have an FMR:BMR ratio in excess of 4 : 1 (Birt-Friesen et al. 1989; Cairns et al. 1990; Adams et al. 1991). However, the foraging modes of these species are likely to be energetically expensive in comparison with great skuas.

The equations employed in this study to estimate total energy demand of growing chicks and daily energy requirements of fledglings were based on detailed work on four closely related semi-precocial species which exhibited very similar patterns of total metabolizable energy intake during development, and as such are likely to be appropriate for great skuas. Recent work (e.g. Bolton, Houston & Monaghan 1992) suggests that egg production may be limited by the availability of specific nutrients at the time of clutch formation rather than energetic constraints. If this is the case, adult foraging effort may be comparatively high over a short period prior to egg-laying, but this would not be incorporated in the component for egg production using the partitioning method adopted here. However, adult activity costs during the season as a whole are taken account of elsewhere in the model, and, assuming any increase in activity occurred for a reasonably brief period during egg formation, the effect on overall energy requirements is likely to be slight.

One other possible error is the estimate of the number of non-breeders present at the colony, hence the comparatively high upper boundary proposed for this parameter in the sensitivity analysis. Counts of birds at the club site at St Kilda remained fairly uniform during the course of the season. However, turnover rates of non-breeding skuas are usually high, with individuals at Foula generally in attendance for 16–21 days at the colony, although ringing recoveries suggest that non-breeding great skuas remain in northern latitudes throughout the summer (Klomp & Furness 1992). Whether these birds forage in close proximity to breeding colonies when not actively in attendance at a club is unknown. If they do feed near the colony, then the predicted energy requirement for non-breeding skuas at St Kilda will be underestimated in this study to an uncertain extent.

Sensitivity tests on related bioenergetics models generally found that variation in the exponents and multiplicands of metabolic equations had the greatest effect on output (Furness 1978; Diamond, Gaston & Brown 1993). This model did not require the use of these very general equations (those used in this study to calculate chick and fledgling energy demands were derived from closely related species) and so avoided many of the potential complications. Therefore, while bearing the results of the sensitivity analysis and the above considerations in mind, we can be fairly confident that the final figures for population energy demand are reasonably accurate.

Limitations of prey consumption model

Determination of the total mass of different prey necessary to fulfil the population energy demands of great skuas relied heavily on adequate estimation of mean meal mass and on the use of pellets to assess diet composition. The sensitivity analysis indicated that large increases or decreases in the proportion of different prey in the diet of non-breeders or chicks had comparatively little effect on the estimation of total food consumption, whereas any change in the prey ratio of breeding adults was much more influential. This underlines the importance of accurate characterization of diet, particularly of breeding birds.

Many previous studies of seabirds have incorporated the analysis of prey remains in pellets as a means of assessing diet (e.g. Duffy & Jackson 1986; Hamer, Furness & Caldow 1991; Mund & Miller 1995). Pellet analysis does have several inherent limitations. Soft-bodied prey, e.g. fish offal, are likely to leave few, if any, remains. However, this was unlikely to be a problem in this study because great skuas obtain very little discarded offal from fishing vessels, as fulmars compete much more effectively for this resource (Hudson & Furness 1989). Foraging ranges of great skuas feeding on goose barnacles or discards might be greater and, in theory, pellets sampled at the colony could be biased towards seabird remains if the latter are consumed closer inshore. However, great skuas regurgitate freshly obtained prey to their mate or chicks after returning from almost all absences from the territory (Furness 1987; Hamer, Furness & Caldow 1991). In addition, attendance of great skuas at St Kilda is very high, indicating that foraging trips are relatively short (Phillips et al. 1997). It is therefore unlikely that many foraging trips are of sufficient duration to allow for digestion of prey and the production of pellets away from the colony.

Pellets persist for differing time periods after production depending on their composition, potentially biasing analyses towards more resilient pellet types, particularly those of feathers (Furness & Hislop 1981). Restricting the comparison to the chick-rearing period, when the diet of adults was assessed solely by pellets, and that of chicks by regurgitates, an estimated 57% of meals of breeding great skuas, and 30% of those of non-breeding skuas were of birds. If birds were being grossly overestimated, this should be detectable in a comparison with their relative incidence in chick regurgitates, which can reasonably be considered an unbiased representation of diet. Bird meat did occur in a lower proportion of chick regurgitates (47%, n = 19) compared with breeders’ (although not nonbreeders′) diet during this period, but the difference was only moderate (10%). In addition, Watanuki (1989) found a high positive correlation (r = 0·82) between the percentage of seabirds in pellets or food remains and their occurrence in food loads delivered to slaty-backed gull Larus schistisagus Stejneger chicks. Annett & Pierotti (1989) also concluded, in a comparison of pellets, stomach contents, and chick and mate-feedings of western gull Larus occidentalis Audubon, that pellets reflected dietary composition with reasonable accuracy.

As far as estimating meal mass is concerned, the assumption that meals of fish and seabirds (other than the two smallest species) had a similar mean mass is probably valid. Because of wing-loading considerations, it is likely that a maximum payload exists just below that which would jeopardize flight performance. The maximum mass of food loads delivered by adult Catharacta skuas to their mate or brood is surprisingly consistent, at 119–140 g (Young 1963; Furness & Hislop 1981; Reinhardt 1997). Given that there is bound to be some variance in payload mass somewhere below this level, an assumed mean meal mass of 100 g seems appropriate. Bearing this and the above considerations in mind, it is therefore reasonable to conclude that the final estimates of total mass of prey and numbers of seabirds consumed are at least broadly applicable, and certainly in the right orders of magnitude.

Relative energetic contribution of different prey

Goose barnacles, although constituting 16% and 24% of meals of breeding adults and non-breeders, respectively, contributed comparatively little (under 4%) to total energy consumption. Goose barnacle tissue has a caloric value of 27·0 kJ g–1 dry mass, which is higher than, for example, whole pelagic amphipods and krill, or shrimp (Massias & Becker 1990; Mehlum, Gabrielsen & Nagy 1993). However, because of the large shell plates, the overall caloric density is very poor (1·9 kJ g–1 fresh mass), just under 75% of the value for non-gravid, ‘shelled’ benthic invertebrates such as starfish, crabs and urchins measured by Hunt (1972). Crustacea appear to be a suboptimal food choice in general. An adult great skua feeding solely on goose barnacles would require a total of 162 kg per season for activity and maintenance alone, which at 40 g a meal would require 33 meals a day, representing a very large number of foraging trips and/or long periods spent away from the colony leaving chicks unguarded and open to predation. This compares with a total of 59 kg of fish or 28 kg of birds, which would require around five meals and two meals (of the larger species) a day, respectively.

Impact on seabird populations

One possibility to consider before discussing the potential impact of predation on other species is that great skuas might obtain a proportion of seabird prey by scavenging rather than by active hunting. Studies of several species, including guillemots and kittiwakes at sites where presumably predation pressure is low, conclude that survival of breeding adults is very high during the breeding season (Birkhead 1974; Aebischer & Coulson 1990), although of course, at very large colonies, this might still mean a fair number of carcasses could be available. In addition, non-breeder mortality may be greater, and therefore more corpses might be available from this source.

A critical determinant of whether predation might have an impact on prey populations is whether hunting effort is directed mostly at breeding adults, visiting non-breeders, chicks or fledglings. A few pellets of kittiwake, fulmar and puffin, and many pellets containing guillemot remains found at St Kilda in 1996, were from chicks or fledglings (Phillips, Thompson & Hamer 1997; Phillips et al. 1997). Unfortunately it is difficult to determine the exact proportion because, excepting those that include an entire skull, pellets from adult prey rarely have a distinguishing characteristic. Positive identification will therefore often be slightly biased towards chicks. These are swallowed whole and tend to leave more obvious remains, particularly legs in the case of guillemot fledglings.

A large percentage of great skua predation might be directed at visiting non-breeders. In a study on Dùn, St Kilda, during the 1970s, of 2600 puffins killed by great black-backed gulls Larus marinus, 43% were non-breeders (Harris 1980). It is extremely difficult to assess relative number of immatures compared to breeding birds attending colonies, and usually impossible to determine the breeding status of prey. Transient non-breeding Leach's petrels and British storm petrels visit colonies in huge numbers (Fowler, Okill & Marshall 1982). In addition, detailed demographic and ringing studies suggested that numbers of non-breeding Manx shearwaters and fulmars were equivalent to 75% and 200–500%, respectively, of breeding population sizes (de Brooke 1990; Dunnet 1991). Similarly, immature auks also visit colonies in large numbers. Up to 50% of puffins trapped on the Isle of May (Scotland) during the latter part of the season were probably too young to breed, and an analysis of common guillemot life-table data indicated that non-breeder numbers were equivalent to ≈ 30% of the breeding population (Harris 1984; Cairns et al. 1990). This potential availability of very large numbers of immatures or non-breeders should be borne in mind, because if a large proportion of skua attacks are directed at this group, the impact on prey species’ population dynamics will be considerably lower than would otherwise be expected.

Surveys at St Kilda in 1987 by Tasker, Moore & Schofield (1988) estimated that there were 62 800 apparently occupied fulmar nest sites, 7800 apparently occupied kittiwake nest sites, 230 500 puffin burrows and 22 700 individual guillemots. Assuming, for illustrative purposes only, that great skua predation was directed solely at breeding adults, and excluding guillemots because we know that a very substantial proportion of birds consumed are chicks or fledglings, the seasonal total for each species eaten by great skuas in Table 6 would be equivalent to 2·0% of breeding fulmars, 15·5% of breeding kittiwakes and 0·9% of breeding puffins. In fact, the impact on the population is likely to be less than this, as an unknown proportion of birds eaten will be non-breeders, chicks or fledglings, even for these species. It is currently impossible to assess the impact of predation on Leach's petrel, British storm petrel and Manx shearwater by great skuas. Even the approximate sizes of their breeding populations are unknown due to the heterogeneous nature of the nesting habitat (which includes grass slopes, stone walls, crevices in boulder scree, etc.) and the inaccessibility of two islands, Soay and Boreray. Nonetheless, it is clear that very large numbers of these petrels, whether breeding adults or non-breeders, are being killed.

Other studies have concluded that substantial mortality occurs as a consequence of predation by skuas or gulls. During the 1970s, great black-backed gulls on Dùn were estimated to consume ≈ 1·5% of the breeding population of puffins per year (Harris 1980). A recent study at a Newfoundland colony also found a very high rate of predation by great black-backed gulls, with 2·9 puffins found dead for every 100 breeding pairs of puffins present (Russell & Montevecchi 1996). Annual predation rate of Leach's petrels by a colony of 3500 pairs of slaty-backed gulls on Daikoku Island, Japan, was apparently even higher, with up to 49 000 petrels killed in a single month and an estimated 13·2% of all adult petrels attending the colony killed annually (Watanuki 1986). This last study demonstrates that a large colony of gulls can have a very severe impact on a seabird population.

The great skua colony on Hirta has grown extremely rapidly in the last few years, the population having almost doubled since 1993 (Phillips et al. in press). Extensive surveying of St Kildan seabird colonies in 1987 by Tasker, Moore & Schofield (1988) did not provide any evidence that fulmar, kittiwake, puffin or guillemot populations had declined since great skua colonization in 1963, although that survey was undertaken before the recent expansion in skua numbers. More recent monitoring of subcolonies visible from land indicates no change in the number of apparently occupied fulmar nest sites, an increase in guillemot numbers, but a substantial decline in occupied kittiwake nest sites at St Kilda in 1996 compared with 1993 (Thompson & Walsh 1997). It is possibly no coincidence that of these species, the kittiwake is the one on which we might predict that great skua predation would have the greatest impact. In terms of available space, the great skua population at St Kilda could continue increasing for a number of years (Phillips et al. in press). Given that St Kilda is considered to hold amongst the largest European populations of puffins, fulmars, Leach's petrel and British storm petrel, the exact numbers of which are unknown for the last two species, close monitoring could be deemed a judicious approach in this situation.

Acknowledgements

Fieldwork in 1996 on Hirta was supported by a grant from Scottish Natural Heritage. For help with fieldwork or logistics at St Kilda we thank John Love, Stuart Murray, Gail Churchill and Phil Sharkey. We are extremely grateful to Geoff Hilton for allowing us access to unpublished data on assimilation efficiency and body composition of seabirds. Robert Furness, Kate Thompson and an anonymous referee made many helpful comments on the manuscript.

Received 4 August 1997; revision received 20 January 1999

Appendix

Table 7. Sensitivity tests for the prey consumption model. Shown in the table are the percentage changes in output estimates for kilos of fish, kilos of goose barnacles, and numbers of birds consumed resulting from a 1% increase in the value of an input parameter, and probable maximum and minimum extremes in input parameters (see Methods). Note that goose barnacles and guillemots are absent from the diet of chicks and non-breeders, respectively. Values in parentheses are means for ¹ Leach's petrel, ² British storm petrel, ³ kittiwake, ⁴ puffin, ⁵ guillemot, ⁶ fulmar, ⁷ Manx shearwater. A zero value indicates a negligible change (±<0·005 for a 1% change in an input parameter, and ±<0·05% for likely extremes).

Parameter	% change in input parameter	% change fish mass (kg)	% change goose barnacle mass (kg)	% change bird numbers
Fish caloric density	+1	−0·40	−0·40	−0·40
	+25	−9·1	−9·2	−9·1
	−25	+11·1	+11·2	+11·1
Goose barnacle caloric density	+1	−0·02	−0·02	−0·02
	+25	−0·5	−0·5	−0·5
	−25	+0·5	+0·5	+0·5
Bird caloric density	+1	−0·58	−0·57	−0·58
	+25	−12·7	−12·6	−12·7
	−25	+17·0	+16·8	+17·0
Fish meal mass	+1	+0·60	−0·40	−0·40
	+30	+16·1	−10·8	−10·7
	−30	−20·4	+13·8	+13·6
Goose barnacle meal mass	+1	−0·02	+0·98	−0·02
	+50	−0·9	+48·5	−0·9
	−25	+0·5	−24·6	+0·5
Large bird meal mass	+1	−0·54	−0·53	−0·09 (−0·54^1,2, +0·45^3–7)
	+30	−14·0	−13·9	−2·3 (–14·0^1,2, +11·8^3–7)
	−30	+19·5	+19·2	+3·3 (+19·4^1,2, −16·4^3–7)
Leach's petrel mass	+1	−0·03	−0·03	+0·33 (+0·97^1, −0·03^2–7)
	+10	−0·3	−0·3	+3·3 (+9·7^1, −0·3^2–7)
	−10	+0·3	+0·3	−3·3 (−9·7^1, +0·3^2–7)
British storm petrel mass	+1	−0·01	−0·01	+0·17 (+0·99^2, −0·01^1,3–7)
	+10	−0·1	−0·1	+1·7 (+9·9^2, −0·1^1,3–7)
	−10	+0·1	+0·1	−1·7 (–9·9^2, +0·1^1,3–7)
Percentage carcass utilization	+1	−0·04	−0·04	−0·49 (−0·04^1,2, −1·03^3–7)
	+25	−0·1	−0·1	−9·9 (−0·9^1,2, −20·8^3–7)
	−25	+0·1	+0·1	+16·2 (+1·0^1,2, +34·6^3–7)
Proportion fish
Non-breeder diet	+1	+0·03	−0·06	−0·04
	+50	+1·5	−2·8	−1·7
	−50	−1·7	+3·2	+2·0
Breeder diet	+1	+0·97	−0·68	−0·63
	+50	+51·3	−36·2	−33·3
	−50	−46·0	+32·4	+29·9
Chick diet	+1	+0·12	−	−0·08
	+50	+6·8	−	−4·6
	−50	−5·0	−	+3·4
Proportion goose barnacle
Non-breeder diet	+1	0	+0·07	0
	+50	−0·1	+3·8	−0·1
	−50	+0·1	−2·9	+0·1
Breeder diet	+1	−0·02	+1·11	−0·02
	+50	−1·2	+60·7	−1·2
	−50	+1·0	−51·2	+1·0
Proportion Leach's petrel
Non-breeder diet	+1	0	0	+0·01 (+0·04^1, 0^2–4,6,7)
	+50	−0·1	−0·1	+0·8 (+2·2^1, −0·1^2–4,6,7)
	−50	+0·1	+0·1	−0·7 (−2·1^1, −0·1^2–4,6,7)
Breeder diet	+1	−0·03	−0·03	+0·31 (+0·89^1, −0·03^2–7)
	+50	−1·4	−1·5	+15·5 (+44·9^1, −1·1^2–7)
	−50	+1·3	+1·4	−15·1 (−43·7^1, +1·3^2–7)
Chick diet	+1	0	−	+0·03 (+0·01^1, 0^2–7)
	+50	−0·1	−	+1·7 (+5·0^1, −0·2^2–7)
	−50	+0·1	−	−1·7 (−4·8^1, +0·2^2–7)
Proportion British storm petrel
Non-breeder diet	+1	0	0	+0·01 (+0·04^2, 0^1,3–4,6,7)
	+50	0	0	+0·3 (1·8^2, 0^1,3–4,6,7)
	−50	0	0	−0·3 (–1·8^2, 0^1,3–4,6,7)
Breeder diet	+1	−0·01	−0·01	+0·16 (+0·89^2, −0·01^1,3–7)
	+50	−0·4	−0·4	+7·9 (+44·8^2, −0·4^1,3–7)
	−50	+0·4	+0·4	−7·7 (−43·9^2, +0·4^1,3–7)
Chick diet	+1	0	−	+0·02 (+0·01^2, 0^1,3–7)
	+50	0	−	+0·9 (+4·9^2, 0^1,3–7)
	−50	0	−	−0·8 (−4·8^2, 0^1,3–7)
Proportion kittiwake
Non-breeder diet	+1	0	0	0 (+0·01^3, 0^1,2,4,6,7)
	+50	0	0	0 (+0·5^3, 0^1,2,4,6,7)
	−50	0	0	0 (–0·5^3, 0^1,2,4,6,7)
Breeder diet	+1	−0·04	−0·04	+0·02 (+0·88^3, −0·04^1,2,4–7)
	+50	−1·8	−2·0	+0·1 (+43·3^3, −1·8^1,2,4–7)
	−50	+1·8	+2·0	−0·1 (−44·2^3, +1·8^1,2,4–7)
Chick diet	+1	0	−	0 (+0·1^3, 0^1,2,4–7)
	+50	−0·2	−	+0·1 (+0·5^3, −0·2^1,2,4–7)
	−50	+0·2	−	−0·1 (−4·8^3, +0·2^1,2,4–7)
Proportion puffin
Non-breeder diet	+1	−0·01	−0·01	0 (+0·05^4, −0·01^1–3,6,7)
	+50	−0·3	−0·4	+0·1 (+2·5^4, −0·3^1–3,6,7)
	−50	+0·3	+0·5	−0·1 (−2·7^4, +0·3^1–3,6,7)
Breeder diet	+1	−0·06	−0·07	+0·03 (+0·8^2, −0·06^1–3,5–7)
	+50	−3·1	−3·3	+1·6 (+40·4^4, −3·1^1–3,5–7)
	−50	+3·2	+3·5	−1·6 (−41·8^4, +3·2^1–3,5–7)
Chick diet	+1	−0·01	−	0 (+0·09^4, −0·01^1–3,5–7)
	+50	−0·3	−	+0·2 (+4·4^4, −0·4^1–3,5–7)
	−50	+0·3	−	+0·2 (−4·6^4, +0·4^1–3,5–7)
Proportion guillemot
Breeder diet	+1	−0·37	−0·40	−1·3 (+0·73^5, −0·37^1–4,6,7)
	+50	−16·8	−18·2	−5·9 (+33·4^5, −16·8^1–4,6,7)
	−50	+20·3	+22·1	+7·2 (−40·4^5, +20·2^1–4,6,7)
Chick diet	+1	−0·04	−	−0·02 (+0·08^5, −0·04^1–4,6,7)
	+50	−1·9	−	−0·8 (+3·8^5, −2·0^1–4,6,7)
	−50	+2·3	−	+0·9 (−4·5^5, +2·4^1–4,6,7)
Proportion fulmar
Non-breeder diet	+1	−0·01	−0·01	0 (+0·03^6, −0·01^1–4,7)
	+50	−0·2	−0·3	−0·1 (+1·6^6, −0·3^1–4,7)
	−50	+0·3	+0·4	+0·1 (−1·8^6, +0·3^1–4,7)
Breeder diet	+1	−0·08	−0·09	−0·02 (+0·83^6, −0·08^1–5,7)
	+50	−4·0	−4·3	−1·2 (+40·8^6, −4·0^1–5,7)
	−50	+4·1	+4·5	+1·3 (−42·5^6, +4·1^1–5,7)
Chick diet	+1	−0·01	−	0 (+0·09^6, −0·01^1–5,7)
	+50	−0·4	−	−0·2 (+4·5^6, −0·5^1–5,7)
	−50	+0·4	−	+0·2 (−4·7^6, +0·5^1–5,7)
Proportion Manx shearwater
Non-breeder diet	+1	0	0	0 (+0·037, 0^1–4,6)
	+50	0	0	0 (+1·77, 0^1–4,6)
	−50	0	0	0 (–1·77, 0^1–4,6)
Breeder diet	+1	−0·01	−0·01	0 (+0·877, −0·01^1–6)
	+50	−0·4	−0·4	−0·1 (+43·37, −0·4^1–6)
	−50	+0·4	+0·4	+0·1 (−43·47, +0·4^1–6)
Chick diet	+1	0	−	0 (+0·097, 0^1–6)
	+50	0	−	0 (+4·77, 0^1–6)
	−50	0	−	0 (−4·77, 0^1–6)