Techniques for assessing Cormorant diet and food intake:
towards a consensus view

by

David N. Carss
&
The Diet Assessment and Food Intake Working Group

published in

Supplementi di Ricerche Biologia Selvaggina, XXVI (1997): 197-230
Proceeding 4th European Conference on Cormorants
(1-3 November 1995, Bologna, Italy)

INTRODUCTION - PELLET ANALYSIS - STOMACH CONTENTS ANALYSIS - FEEDING OBSERVATIONS - ESTIMATES DAILY FOOD INTAKE - OTHER NOVEL TECHNIQUES, AND SOME WORDS OF CAUTION - CONCLUDING REMARKS AND ACKNOWLEDGEMENTS - SUMMARY - RIASSUNTO - REFERENCES - APPENDIX 1

INTRODUCTION

General remarks

As with many applied ecological problems (e.g. see van Blaricom & Estes, 1988), the impetus for this paper comes from two sources. The first is scientific: Cormorants Phalacrocorax carbo are interesting birds to study because they are widespread, are at the top of many aquatic food-chains and therefore influence, and are influenced by, the structure and dynamics of fish communities. The second is political: Cormorants are perceived to damage fisheries throughout Europe, and elsewhere. In many countries, commercial and/or recreational fishermen may form a large, vocal body capable of considerable political influence. For instance the German Government (Bavarian State Ministries) has funded a three-year investigation into the impacts of Cormorants on fish stocks in Bavaria (see Keller et al., 1997). In The Netherlands, the Ministry of Agriculture, Nature Management and Fisheries established a working group on Cormorants and fisheries and then, along with the Ministry of Transport, Public Works and Water Management, the Organisation for the Improvement of Inland Fisheries, and the Product Board for Fish, provided funding for work which lasted about four years (van Dam et al., 1995). The Scottish Office has part-Funded a six-year investigation into sawbill duck (Mergus spp.) and cormorant predation on fisheries throughout Scotland (see Marquiss et al., 1991; Marquiss & Carss, 1996). More recently the UK Government, through the Ministry of Agriculture, Fisheries and Food, and the Department of the Environment, has commissioned a number of studies into bird predation at fisheries in England and Wales, at a reported cost of £ 1 M.

There are calls to devise a pan-European management plan for Cormorants and to "control" birds by culling, despite the scarcity of scientific evidence for any detrimental effects of fish predation on natural water bodies. Attempts to provide such evidence, for any fishery type, in any country necessarily require rigorous estimates of diet, including such things as fish species composition and size distribution. Moreover, knowledge of foraging ecology is also essential when assessing possible impact, particularly information on daily food (and energy) requirements, diet shifts, prey selection, the influence of foraging habitat, and any seasonal and annual variation. Thus, knowledge of cormorant diet has both scientific and political value.

There has been little attempt to synthesise the results of the relatively large number of studies investigating cormorant diet, and, in many cases, groups and individuals are working relatively independently Indeed, differences in methodology often make it impossible to even directly compare such studies. Most workers acknowledge methodological problems but often tackle them in, different ways. As the interest in cormorant impact on fisheries continues to grow, it was felt appropriate to make attempts to "get our house in order" by discussing the various methods of diet and food intake assessment at the 4th European Cormorant Conference, Bologna, Italy, November 1995, at a half-day workshop and a subsequent smaller meeting attended by all members of the Group (see Appendix).

Apart from the individual workshop contributions (see elsewhere in this volume) this paper is offered as a tangible result of our deliberations. It is not intended to be a comprehensive review of the scientific literature within our field but a discussion of methods, associated sources of error attempts to quantify them, areas of consensus, and gaps in current knowledge. Such gaps in knowledge often prevent important advances being made. Our ultimate aim is to standardise methodologies wherever possible so that biases are minimised, and always quantifiable, so that dietary studies are valid, use the appropriate methodology, and are comparable.

Research priorities

On a political level it appears that (a) quantifying cormorant consumption of commercially important fish is a key objective. However following on from this is the more problematic issue of (b) whether such consumption constitutes "damage", or "serious damage, to the fish stock in question. This paper is concerned with (a), for. which data are required on (i) cormorant "population" size, (ii) daily intake of different components of the population, and (iii) composition of the diet.

Data for (i) are probably achievable, though may have wide confidence intervals. Information for (ii) can theoretically be obtained by doubly labelled water techniques, nest balances, assimilation rate estimates, energy budgets derived from activity budgets and estimates of activity-specific costs (and possibly from pellets). It is debatable how precise these estimates have to be, and likely that the mean estimated daily intake of the population will have wide confidence intervals. Sampling the diet (iii) is a major challenge; particularly if a population estimate is required, as sample size has to be fairly large.

This paper is structured (1) diet assessment methods - analyses of pellets, stomach contents, and regurgitations, (2) direct observations, and (3) foraging- ecology - food intake and daily energy expenditure. Although this represents a progression of increasing complexity, it is not meant to demean the value of the more "hands-on' approaches of pellet and stomach contents analyses. These methods are relatively quick, and therefore cheap, and there are considerable pressures to use: them, but they must be used with care.

PELLET ANALYSIS

Introductory remarks

Pellets are attractive sampling units. They are relatively .easy to collect and, in some situations, are the only available means of assessing the food of cormorants. Pellet analysis is a useful method of obtaining a rough index of cormorant diet in qualitative terms but there is serious doubt as to whether it can be used to derive quantitative information on, for example, the species composition or size-range of fishes taken. Pellet analysis should not be used to estimate the daily food intake of cormorants. With reservations (in particular relating to the under-recovery of small fish), pellets may be used to investigate spatial or temporal variation in the relative frequencies of particular items in. samples of varying provenance. However, great care must be taken when interpreting the results of such studies as there are serious potential biases.

Advantages

Pellet analysis is a relatively cheap method of obtaining dietary information. Large samples can be collected relatively quickly, remembering that the sampling unit is an individual pellet, not the number of fish it contains. Collection involves little, or no, disturbance to birds, and analysis requires minimal laboratory facilities, and is fairly easy. In many situations, for example large, expansive lakes where observing foraging birds is difficult, areas where birds cannot be shot, or during the non-breeding period, pellets may be the only means of assessing the food of cormorants.

Studies involving the technique

Pellet analysis is a commonly used method. For example, of eight papers in the recent ARDEA Special Issue (83.1, 1995) which dealt with aspects of cormorant diet, six involved the analysis of pellets. One study used the stomach contents of drowned birds, another relied on direct foraging observations, but the remainder used pellets to calculate a wide variety of parameters. At its simplest, analysis was used to determine the prey species composition by fish number (in % of the diet) of particular birds (i.e. non-breeders on the coast, Warke & Day, 1995). van Eerden & Voslamber (1995) calculated the average length of prey fishes, and by implication, their maximum swimming speed while Keller (1995) calculated the length (and hence, mass) of fishes eaten, and the estimated fish mass per pellet (considered to be an individual bird's daily food intake) which was then used as a basis for speculation about seasonal changes in daily food intake. Similarly, Platteeuw & van Eerden (1995) assumed pellet contents accurately reflected daily food intake and compared calculated values with commuting distance from colony to forging site. Finally, Dirksen et al. (1995) calculated fish lengths from remains in pellets, converted these to fresh weights, calculated daily food intake from estimated fish mass per pellet, compared intake on a seasonal basis, and finally estimated fish consumption (kg/ha) for two adjacent lakes (see also Veldkamp, 1995).

Although some authors acknowledged potential biases in the results of their pellet analyses (i.e. Keller, 1995; Platteeuw & van Eerden, 1995; Dirksen et al., 1995; Veldkamp, 1995), only in the two latter papers were attempts made to quantify them and no author produced confidence limits for their final calculations of such things as total fish consumption.

Known sources of error and some words of caution

Several attempts have been made to determine the accuracy of pellet analysis by feeding fish of known length to captive Cormorants (Cape cormorant P. capensis: Duffy & Laurenson, 1983; shag P. aristotelis: Johnstone et al., 1990, great cormorant P. c. sinensis: Zijlstra & van Eerden, 1995). Each has shown that some key hard remains (most workers record otoliths) are partially or entirely digested, making it impossible to estimate the original length or number of fish ingested per meal or per day. At least five probable sources of error may occur:

1) Most "lost" or eroded otoliths are from the smallest fish (Johnstone et al:, 1990) and so recovery is size related (see also Carss & Elston, 1996 for bones in otter Lutra lutra faeces). If some of the diet comprises small individuals or small species of fish, by implication, their remains will be underrepresented in pellets. There is thus also the possibility of species-related differential recovery of remains in pellets.
2) As erosion of hard parts leads to inaccuracies in size estimation, such errors are compounded if estimated fish lengths are then expressed in terms of estimated fresh weights.
3) For some Cormorants (e.g. shags), the assumption that, a pellet's contents reflect the remains of food taken during the preceding 24 hr period is invalid and this has important consequences for the use of pellets to estimate daily food intake, and hence, energy requirements (Russell et al., 1995).
4) There may be seasonal variation in pellet production.
5) Hard parts derived from the stomachs of piscivorous fish may be recorded in pellets and assumed erroneously to be cormorant prey, as demonstrated for double-crested cormorant P. auritus regurgitations (Blackwell & Sinclair, 1995).

As a consequence of such demonstrable biases, the results of some of the studies discussed in the preceding section must be viewed as tentative, or as stated by Dirksen et al. (1995) in relation to Basal Metabolic Rate (BMR) estimates from pellets, they merely "illustrate that the pellet data are in the right order of magnitude". Zijlstra & van Eerden (1995) suggested that pellets produced in captivity, by potentially stressed birds, differ strongly from those in the field. Moreover, in some studies there is an implicit suggestion that some of the biases discussed above may be less applicable to the pellets produced by free-living cormorants, arguing (e.g. Dirksen et al., 1995) that the results from such analyses do not seem to disagree with theoretical values (e.g. of daily energy demand) and highlighting the fact that studies based on "wild" pellets produce comparable figures. Nevertheless, the fact that hard parts recovered from pellets are eroded means that biases caused by digestion must be taken seriously.

Gaps in current knowledge, and future work

Attempts have been made to reduce the biases associated with pellet analysis. For instance, Suter & Morel (1996) ignored visibly eroded grayling otoliths in pellets and, by using seemingly uneroded ones, found good correspondence between length frequency estimates from pellets collected at a colony and from fish samples taken nearby. Nevertheless, it is obvious that, within the recent cormorant diet literature at least, a consensus has yet to be reached on how best to analyse pellets, or the validity of basing large-scale calculations on such analyses. In fact, the consequences of potential biases are currently unknown but indications are that they could be substantial. Some of these biases could however be quantified by large-scale, well-designed experiments. Given the level of interest in pellet analysis, such feeding trials seem essential.

To this end, feeding trials should be undertaken, primarily to assess biases and attempt to determine realistic "correction factors". Trials could be used to determine the most appropriate (species-specific) hard parts ("key bones"), as has been done for salmonids in the gut contents of sawbill ducks (Feltham & Marquiss, 1989) and for salmonid and eel remains in the faeces of otters (Carss & Elston, 1996), see also Veldkamp (1995) for cyprinid remains. Then, attempts could be made to quantify size- (or species-) related differential recovery. Such trials should involve feeding captive birds on fishes of known species, length, and mass and should take into account the possible influences of stress (e.g. see Trauttmansdorf & Wassermann, 1995; Zijlstra & van Eerden, 1995 and also Cherubini & Mantovani, 1997), deliberate disturbance (on pellet content), and activity (e.g. Zijlstra & van Eerden, 1995; Carss et al., in press, for otter). Such trials may also produce much needed information on pellet formation, e.g. the influence of meal size and inter-meal interval, and the relationship between degree of digestion and quality (e.g. fat content) of fish.

The following were considered to be the most "important" potential trial fishes; (i) Cyprinids; roach Rutilus rutilus, rudd Scardinius a erythropthalmus, bream Abramis brama, carp Cyprinus carpio, chub Leuciscus cephalus, (ii) perches; perch Perca fluviatilis, pike-perch Stizostedion Lucioperca, ruffe Gymnochephalus cernuus, (iii) Salmonids; Salmo and Oncorhynchus spp, grayling Thymallus thymallus, whitefishes Coregonus spp, (iv) Other freshwater fishes; eel Anguilla anguilla, pike Esox lucius, pumpkinseed Lepomis gibbosus, catfishes Ictalurus spp, (v) Marine species; sea bass Dicentrarchus labrax, gilt-head bream Sparus auratus, mullets Chelon/Liza/Mugil spp., flatfishes Pleuronectes spp., eelpout Zoarces viviparus. It is unlikely that trials involving all these species can be undertaken but experiments using representatives of each group would be a useful first step.

Other areas identified as requiring further work were:

1) Development of a standard method for collecting pellet samples in the field. Such a method could include a standard period of time spent in each study site and collection of pellets on a standard transect route. Sampling should be done by experienced à collectors, and ideally the some individual, and should include searches for small pellets. Sampling should be done as early in the morning as possible to reduce the possibility of pellets being scavenged by birds and/or mammals (this of course assumes that pellets are only produced at night, which may not always be the case).

2) The effect of sample size on estimates of diet from pellets is currently unknown. Although not stated directly, most workers appear to collect as many fresh pellets as possible,: this must be the best procedure although more time is required to collect! large samples. Bootstrapping on a database produced from a large number; of pellets would allow the effect of sample size to be determined (cf. Marquiss & Carss, 1997, for stomach contents, Carss & Parkinson, 1996 for otter faeces).

3) Perhaps less crucial at this stage is an investigation into the effects of sampling interval on analysis (see Carss & Parkinson, 1996 for otter faeces). For example, pellet decomposition is almost certainly weather-dependent;-moreover some pellets" contents may decompose quicker than others (e.g. smaller pellets, or smaller remains within them, may be lost differentially). This could be examined in the field by marking individual fresh pellets (cf. Jenkins & Burrows, 1980 for otter faeces) or under more controlled conditions elsewhere. Sampling intervals should also be chosen to~allo4v detection of any temporal variation in diet, bootstrapping large data sets (see 2 above) may help to elucidate this.

4) Development of a standard method for analysing pellets. At present many workers fail to document their laboratory methods but many of those that do cite Marteijn & Dirksen (1991). Following these authors, pellets should be stored at -20 °C, and later thawed and soaked individually in lukewarm water to remove any mucus, remains can then be rinsed with cold water through a fine (e.g. 0.3 mm) sieve and allowed to dry at room temperature. Fish remains (e.g. otoliths, jaws, pharyngeal bones, operculae, cleithra) can then be extracted carefully and identified using reference bones. The number of individuals represented in a pellet has been defined (Marteijn & Dirksen, 1991) as the highest total of any of these key bones, taking right and left parts separately Bones can then be measured and original fish lengths calculated from a series of allometric equations. Precise errors associated with such a technique are currently unknown, but could be quantified with feeding trials.

5) Standardisation of (i) fish length measurements (i.e. fork or total length), allometric equations relating (ii) key bone size to fish length, and (iii) fish length to fresh weight, is also necessary.

Concluding remarks

There is a tendency for views on pellet analysis to be polarised; some believing the technique to be problematic, others confident that, for their studies-at least, it is robust. It seems clear that pellets could be an extremely useful sampling technique because they potentially allow large numbers of individuals' to be sampled through time. Large data sets, derived from the analysis of many thousands of pellets, currently exist. Experimental feeding trials should not be seen as an attempt to discredit the method, on the contrary they can only enhance the value of these data, quantifying the associated biases, and increasing confidence in the technique.

STOMACH CONTENTS ANALYSIS

These methods apply to the stomach contents of dead birds but also to regurgitation and samples obtained by stomach flushing. Daily food intake cannot be determined from stomach contents analyses because we do not know whether a bird has stopped feeding for the day, and so cannot get average values. Maximum values are overly influenced by subjective judgement of what is a full stomach, and might overestimate daily intake by ignoring lower values.

Advantages

Stomachs often contain relatively fresh material but there are also some established methods for dealing with well-digested prey. Moreover, differences between relatively fresh fish and the retained items at the bottom of the stomach are clear. Thus, some of the more serious errors associated with such well-digested items (see pellet analysis section) can be avoided. In some situations, stomach contents may be the only means of assessing cormorant diet, for instance where pellets cannot be collected and where direct observations are difficult. Stomach contents samples can be accompanied by age, sex, and parasite infestation information for each bird. They are also site- specific (i.e. foraging grounds are known) or can be implied from the location at which a bird sample was collected.

Disadvantages

There are several disadvantages to stomach contents analysis, the most obvious being the necessity of killing birds. Licences are required to kill Cormorants in European countries and so in most cases, the samples available for stomach contents analysis are small. Sample sizes (the sampling unit is an individual stomach, not the number of fish it contains) may be further reduced if some stomachs are empty. The likelihood of obtaining empty stomachs can be reduced by shooting birds later in the day, after they have had a chance to feed, but this adds à bias if some prey are digested quickly or if there are diurnal variations in the prey selected. Overall, patch foraging (i.e. 'repeated foraging in a localised area/habitat type) may lead to a non-random distribution of prey amongst stomachs (e.g. Marquiss & Leitch, 1989, for grey heron) which could influence assessments of diet, particularly from small samples (e.g. Carss & Marquiss, 1997). Moreover, foraging behaviour may influence vulnerability to shooting and again lead to bias in the resulting assessments of diet. For example, single birds (in general thought to take larger fish) are thought to be more vulnerable to shooting than flock-feeding birds (thought to take smaller fish). In theory, the more digested the contents the greater the potential for bias, as some items will be more resistant to digestion than others. Some workers have attempted to resolve this problem by examining only intact material (e.g. Mills, 1962, 1965). However, Marquiss & Carss (1997) have shown that, for goosander stomach contents, smaller fish were under represented by this method. This was probably because the digestion of fish is so rapid that small items can disintegrate while larger ones resist disintegration for longer Studies estimating diet From intact items alone will underestimate the proportion of small fish species and overestimate the mean size For some larger ones. Although not often quantified, regurgitations and stomach flushing are usually assumed to provide "complete" samples. However, one study of shags (Wanless et al., 1993a) has demonstrated that some food must have been digested before birds returned to the nest. Furthermore, the fact that adult birds swallow fish head-first means that when they regurgitate partly-digested food for nestlings, the Fish heads will be the most digested material and some otoliths within them may not be passed in the regurgitation. Nevertheless, the method is thought to provide necessary data on the size and species of prey taken which are needed in balance studies.

For nestling regurgitations, biases are associated with the fact that adults may eat low, quality food themselves and feed higher quality food to nestlings (as suggested by Harris & Wanless, 1993). So, nestling diet may not reflect that of the adult population. For grey herons, Moser (1986) suggested this idea but Marquiss & Leitch (1989) could find no evidence for it.

Proposed standard method

As with pellet analysis, a standardised technique for analysing stomach contents would allow direct comparison between studies. In the absence of a widely-accepted method of analysing stomach contents, the following is offered as a standard technique.

Many seabird researchers inject formalin into stomachs to stop the digestion process and preserve material. Attempts to do this for Cormorants were counter productive (Ekins & Carss, unpublished) as the resulting fish flesh could not later be digested in vitro. Birds should therefore be stored frozen at -20 °C as soon as possible after death. After thawing, biometric measurements should be taken and the body cavity then opened with scissors. To open the body cavity, an incision should be made in the flank below the sternum which passes through the sternum and clavicle. Next the skin of the throat is opened by cutting across the gullet just below the head and down the length of the neck to join the top of the initial body cavity incision. Finally the body cavity incision is continued down the remaining length of the body to the vent. The trachaea and various internal organs are then separated from the stomach, which is then lifted out of the body cavity in its entirety and severed from the hindgut (small intestine and colon, containing only extremely well-digested liquid material) which is discarded. The hind-gut is then removed to allow access to the ovaries or testes (for sex determination). The stomach it self i.e. the foregut, including the proventriculus (or the gizzard of ducks) should be opened from top (neck) to bottom. Whole fish should be removed carefully, identified and measured. Partially-digested material should be flushed out, using a water bottle, into a storage container.

Biological washing powder (e.g. "Biotex") and further water should be added to the container to form a saturated solution which completely covers the stomach contents. The washing powder digests all remaining Fish flesh from partially-digested fish after about 4/6 days. The process can be speeded up by placing samples in an oven at about 37 °C (higher temperatures may denature the enzymes in the washing powder) and stirring occasionally. The stage of flesh-removal may be checked by pouring the contents into a fine (e.g. 0.25 mm) sieve for examination, if further in vitro digestion is required, simply recant and add more saturated washing powder solution. Once samples are ready they should be poured into a fine sieve and rinsed well with cold water. Increasing the water pressure during rinsing can help dislodge, stubborn pieces of flesh or cartilage, and to disarticulate skeletons or skulls, but care must be taken not to lose any material over the rim of the sieve.

Once thoroughly rinsed, bones should be guided to the edge of the sieve by a gentle flow of water and transferred to filter paper for drying. Care should be taken to ensure that all bones are collected as fine vertebral spines . and very small pharyngeal teeth sometimes become lodged in fine-meshed sieves. Finally, samples should be .air dried at room temperature for 1/3 days before examination, alternatively they can be placed in paper packets after drying for storage and later analysis. There is no evidence that biological washing powder causes damage to bones or that air drying leads to significant shrinkage of material over the time periods specified above.

Analysis of the dried remains Involves placing them in a petri. dish for examination under a low power binocular microscope. Key bones, including atlas, thoracic, and caudal vertebrae, pharyngeal teeth, pelvic girdles, opercular bones, cleithra, lower jaws (dentaries), and cyclostome; teeth (from Lampetra spp.) can be extracted. The appearance of such key bones can be used for identification, and their size to estimate fish lengths and hence fresh weights from a series of regression relationships. (e.g. see references in Marquiss & Carss, 1997; see also Watt et al., in press). To provide the most accurate assessments of diet, key bones must be robust and resistant to digestion, relatively easy to identify, and diagnostic. The only published examples are salmonid atlas vertebrae (in sawbill duck stomach contents, i.e. Feltham & Marquiss, 1989; Feltham, 1990) and salmonid atlases and eel thoracic vertebrae (in otter faeces, i.e. Carss & Elston, 1996). It is recommended that a similar approach is taken for other potential key bones. It should be noted that Feltham (1990) rejected the use of otoliths, in the stomach contents of red-breasted mergansers Mergus merganser, because their presence did not correspond well with the presence (or size) of other bones from the same fish species (see also van Heezik & Seddon, 1989).

Sample size is important, as it may affect the accuracy of diet assessments (see discussion in Marquiss & Carss, 1997). These authors quantified the effect of sample size on the accuracy of estimates of both cormorant and goosander diet by re-iterative sub-sampling, and concluded that "adequate" estimates were possible from samples of 12-15 stomachs containing food but more analysis is required from large samples of stomachs containing diverse fishes.

Gaps in current knowledge and future work

Opportunities for obtaining samples of shot birds may be limited, so consideration should be given to the best strategy for collecting samples i.e. how best to utilise the opportunity of a small sample from a specific area. Data suggest that a sample from one localised area is more informative than one from a whole catchment. This is because Carss & Marquiss (1997) and Marquiss & Carss (1996) have found variation in the gut contents of cormorants, goosanders and red-breasted mergansers from Scotland, associated with location, habitat, and time of year Thus, homogeneity of samples cannot be assumed, for example within a single watershed, irrespective of season. This means that in practice, small samples from a restricted time and place are more useful than if spread out over a whole watershed, or year Further work is also required to determine the most appropriate key bones (for species identification and length estimation) for fishes other than salmonids and eel. There are some quite large data sets on stomach contents analysis (e.g. ITE data for Scotland) so it should be possible to explore the influence of sample size, and for some other fish species, the most appropriate key bones.

FEEDING OBSERVATIONS

Introduction

Feeding observations have long been used as a means of assessing the diet of birds (e.g. Hartley, 1948) but they have been little used in studies of cormorant diet, particularly so for the "continental sub-species" P. c. sinensis. Many European studies have been undertaken on large still-water bodies where such observations may b~ impracticable (e.g. Veldkamp, 1995) and observational dietary studies have usually been restricted to smaller waterbodies and rivers (e.g. Schaffer, 1982; Davies & Feltham, 1996; Doherty & McCarthy, 1997; Stickley et al., 1992 for double-crested cormorant, Carss, 1993a for shag). Where appropriate, direct observation can be a valuable method of obtaining useful information on cormorant diet.

Advantages

Direct field observation of foraging birds to record the fish they consume has several advantages. It is non-destructive and can provide large amounts of data with little disturbance to birds. A major advantage over other methods (e.g. pellet analysis, and to a lesser extent, stomach contents analysis) is that spatial and temporal variation in diet might be assessed with some accuracy (Davies & Feltham, 1996), with data ox: prey and precise feeding locations being collected simultaneously. Such observations may be valuable on rivers where cormorant "populations" are often extremely dynamic - observations allow this spatial variability to be taken into consideration, i.e. continual observations "focal" birds. In addition to dietary information, observations can also provide data on diving behaviour and an index of foraging performance (see late ).

Disadvantages

It may be difficult to get close enough to birds to enable the identification of all prey items caught during a feeding bout, thus data from birds close inshore may be less biased: than those from birds farther away. Biases will be further compounded if birds are feeding differently (e.g. taking different species/sizes of fish) in the two situations. Moreover, observations may be biased because particular prey species may be more, or less, easy to recognise. This problem may be particularly acute at fisheries with a diverse fish fauna. It may be possible to reduce such potential biases by categorising fish on their type (e.g. cyprinid/flatfish, Davies & Feltham, 1996) or body shape (e.g. Carss and Godfrey, 1996, for osprey Pandion haliaetus). Errors could also occur when attempting to estimate the size of fish caught, especially when judging the lengths of smaller prey items, which generally have shorter handling times. The sizes of fish caught by birds during observations are usually estimated in relation to bill or head length (e.g. Schaffer, 1982; Davies & Feltham, ,1996; Ulenaers et al., 1992 for great-crested grebe Podiceps cristatus, Carss 1993a, for shag, 1993b For grey heron Ardea cinerea). Despite the common use of this technique, there have been few attempts to quantify observer bias under experimental conditions (e.g. see Davies, 1996; Bayer 1985 for herons Ardea spp. and egrets Egretta spp., Carss & Godfrey, 1996 for osprey).

It is often assumed that Cormorants bring most prey to the surface to swallow. In some cases this may include small fish, often. taken at a high rate (e.g. 6 or 7 per minute, S. Volponi, pers. comm.), but the possibility of birds swallowing prey items underwater must be considered. Sometimes birds surface without a fish but may shake their heads, gape, exhibit throat fluttering or take a drink, implying that they have just swallowed a fish underwater Wanless et al. (1993b) recorded that shags feeding on sandeels Ammodytes spp. caught and swallowed about 6-7 fish per dive. In another study, Carss (1993a) observed shags feeding at cage fish farms in Scotland and showed that the calculated weight distribution of fishes eaten, based on visual estimates of their lengths when brought to the surface, differed to that expected from the stomach contents of birds shot in the same area. Greater numbers of small fish were found in the stomachs of shags than were recorded during observation's, indicating that about 50% of the small fish that shags caught were swallowed underwater The frequency with which Cormorants may swallow fish underwater has yet to be quantified, thus at certain fisheries, licensed shooting of birds may be necessary to determine the accuracy of estimates of diet from direct observations.

Feeding observations cannot be used to estimate the daily food intake (DFI, see next section) of cormorants. This can usually only be assessed by following individuals continuously all day and relies on both accurate and unbiased time budgets. Cormorants often move between roosts and several foraging sites and it is, therefore, not known whether birds have fed to satiation during the observation period, or whether they will forage elsewhere after observations have terminated.

Foraging behaviour

Diving behaviour

Cormorants capture their prey by pursuit diving, surfacing between dives in order to breathe. Many studies have examined the relationship between these dive and pause times (i.e. the inter-dive interval spent on the surface) and this may provide a measure of diving performance (Dewar, 1924; Cooper, 1986; Wanless et al, 1993b for shag). Cormorants' ability to dive, both in terms of duration and depth, may be a consequence of their body size (see Cooper 1986 and references therein). It was unclear whether the larger cormorant species had longer dive duration because they chose to forage in areas with deeper water or whether their physiology enabled them to stay down longer which, by analogy with diving ducks (Butler & Jones, 1982), may be the case.

Diving behaviour may also be influenced by foraging habitat and other environmental factors (Wilson & Wilson, 1988). Many studies have demonstrated that dive duration is related to the depth of water in which birds are foraging (e.g. Ross, 1 974; Cooper, 1986; Wilson & Wilson, 1988; Trayler et al., 1989; Hustler 1992). For instance, Trayler et al. (1989) found that both dive duration and inter dive interval differed significantly with water depth. Both Cooper (1986) and Wilson & Wilson (1988) suggested that Cormorants diving in shallow water or feeding on surface-shoaling fish, tended to have relatively short dive periods.

Both dive duration and inter-dive interval may also be related to foraging success. For great-crested grebes, Ulenaers et al. (1992) found that the time spent on the surface was related to the size of fish caught, and that dives which resulted in the capture of a prey item were significantly longer than those which did not. In addition, the duration of a foraging bout may also have an influence on dive/pause ratios.

Foraging success

While not forgetting that some, perhaps the smallest, prey items may be swallowed underwater and not recorded by direct observation (see above), an index of foraging performance may also be obtained through direct observations. A widely-used measure of foraging success is the proportion of successful dives that a cormorant makes during a foraging bout, i.e. the proportion of dives which result in a prey item being brought to the surface and swallowed. This might provide a good index of foraging performance, as. it includes the time costs involved in foraging (dive and rest budgets.

By measuring the diet of birds (either by analysis of stomach contents or observations) and the number and size of fish eaten per unit time, intake rate (e.g. g min^-1 food eaten) may also be assessed. Estimates of the weight of fish consumed are calculated from published length - weight relationships (see Schaffer 1982; Carss & Marquiss, 19Q7; Ulenaers et al., 1992 for great crested grebe, Carss, 1993a for shag). Furthermore, energy intake during foraging (i.e. kJ min^-1) can also be estimated from food intake observations by using published energetic values (i.e. kJ g^-1). For different fish species (e.g. Scherz & Senser, 1989), thus enabling estimates Of the length of foraging time required to meet daily energy expenditure (DEE). Such an approach has indicated that the duration of Feeding bouts tends to be shorter when intake rates are high (Morrison et al., 1978; Schaifer 1982). More recently the method has been used to assess the relative profitability Of particular diets and foraging areas (Davies & Feltham, 1996).

By examining both the proportion of successful dives and the intake rate, valuable information on the relative profitability of prey species, or types, can be collected. For instance, on the lower R. Ribble, north-west England, the proportion of cyprinid fish seen to be caught by Cormorants was negatively correlated with the proportion of successful dives in a foraging bout (Davies, 1996). Those Cormorants which fed on large numbers of flatfish tended to catch these items more frequently than did those feeding on cyprinids. The mean intake rate of birds was positively correlated with the proportions of cyprinids in the diet. This was a result of cyprinids having both a larger mean length, and higher calorific content, than flatfish. When birds fed mostly on cyprinids, they were thus able to meet their DEE by foraging for about an hour compared to 1.2 - 2.25 hours at other times. It appeared that the less-Frequent capture of cyprinids was compensated by higher energy returns.

As with diving behaviour, foraging success may also be related to environmental factors. Water conditions, for instance, may play a significant role in determining the foraging success of cormorants. Tide-related differences in foraging patterns (e.g. distribution and foraging success) are known for many piscivorous birds. For cormorants, Morrison et al: (1978) found that olivaceous (neotropic) Cormorants P. olivaceous (brazilianus) feeding in tidal areas were more successful during periods of-low tide when fish appeared to become trapped and concentrated in pools. Furthermore, Richner (1995) found that the proportion of foraging Cormorants observed on a Scottish estuary was highest during the ebbing tide, presumably as a result of variation in the availability and movements of the main prey flounder Platichthys flesus. The effects of water turbidity on foraging success may also be important (e.g. Eriksson, 1985 for osprey and divers Gavia spp.). The development of social fishing by Cormorants on large, eutrophic lakes in Europe is thought to have resulted in increased Foraging success in these turbid waters (see van Eerden & Voslamber 1995).

Conclusions

In areas where pellets cannot be collected or where shooting Cormorants is impracticable, observations may be the only method of collecting large amounts of dietary information. Variation in diet, both seasonally and spatially, has been shown to vary markedly, and in some cases observations can demonstrate these differences with some accuracy. Feeding observations may be an extremely useful method Of assessing cormorant diet at certain fisheries, particularly those where Cormorants are taking only large (potentially commercially valuable) fish. In such circumstances most, if not all, prey items will be brought to the surface and the main bias will be caused by inaccuracies in length estimation (and hence weight, and energy content) of prey fishes. Further biases will be encountered at Fisheries where Cormorants also take small Fishes. Here, an unknown proportion of these fish may be swallowed underwater, and this may have implications for assessing the potential impact at fisheries by leading to an overestimate of the "importance" of large prey items. Attempts to quantify the proportion of fish swallowed underwater should be made, using similar methods to those developed by Wanless et al. (1993) working on shags. There may also be biases associated with birds feeding differently (e.g. taking different species/sizes of fish) close inshore, where they may be more easily observed, and farther out, where they may be harder to watch. Extrapolating from the "easily-visible" sample to the local "population" would, in this case, not be justified.

Information on diving behaviour and foraging success collected as part of feeding observations 'can suggest the relative profitability of foraging areas. These data should, however, be collected in conjunction with information on the relative abundance of fish species, thus providing greater understanding of the reasons for variation in foraging success and, perhaps, explanations of prey choice at particular feeding sites.

ESTIMATES DAILY FOOD INTAKE

Introduction

Estimates of the potential damage to fisheries caused by cormorants, have typically been derived from data on the number of birds feeding at a fishery their diet, the availability or abundance of their prey and the bird's daily food intake (e.g. Barlow & Bock, 1984; Im & Hafner 1984; Moerbeek et al., 1987; Kennedy & Greer, 1988; Marteijn & Dirksen, 1991; Osieck, 1991; Linn & Campbell, 1992; Feltham & Davies, 1995; Davies, 1996). Most of these data are clearly fishery-specific and have been collected locally for the purpose of estimating losses to that particular fishery Estimates of daily food intake (DFI) have however been derived in many different ways, often with serious associated biases, thus making meaningful comparisons between studies extremely difficult (see Feltham & Davies, 1997). In addition, much of the information required for such comparisons is frequently missing from published studies and there has not, until now, been any serious attempt to standardise the methods used by cormorant researchers. The discussion below details the consensus reached on how estimates of DFI should be derived in future studies. Essentially it represents a set of agreed methods that should (i) reduce previous biases (ii) allow comparisons of potential damage at different fisheries and (iii) help to identify what further work needs to be carried out to produce the robust and definitive estimates of DFI that we urgently need.

What we should and shouldn't do

Pellets, the stomach contents of shot birds, and regurgitations cannot be used to derive good estimates of DFI because of the associated biases in estimating diet mentioned previously (see early section). Similarly, estimates of DFI derived from feeding captive adult Cormorants or from the food requirements of well-grown chicks almost invariably underestimate DFI as neither the flying nor swimming components of the bird's energy budget are included in such calculations. The solution to these problems is to base future estimates of DFI on considerations of the energy requirements of wild birds. One method that is widely used in the field to estimate daily energy expenditure (DEE) is the doubly-labelled water (DLW) technique (see e.g. Feltham, 1995; Keller, 1997) and some workers have begun to use heart rates as an estimator (e.g. Bevan et al., 1995). These techniques can however be very expensive. Another method for estimating DEE is through the construction of time-energy budgets, a technique where time-budget data are collected and energy costs assigned to the different behaviours. It should be stressed however that none of these methods will provide an absolute measure of DFI that can be applied to all studies. This is because food intake will not be constant but will vary in response to many factors e.g. season (breeding/nonbreeding), general activity levels, foraging distances travelled, prey availability etc. It will therefore only be possible to produce a range of likely DFI values for any study It is crucial therefore, in order to maximise the usefulness of existing data and future studies, that all workers use the same figures in their calculations. This will enable us to produce upper and lower estimates of potential damage to any particular fishery based on the highest and lowest estimates of DFI derived from energetics studies. A further improvement in damage estimates might be obtained by using the Monte Carlo simulation approach adopted recently by some British workers (e.g. Davies, 1996).

Estimating Daily Energy Expenditure from time-budgets

In order to standardise methods, we recommend the use of focal animal sampling when collecting these data in preference to scan sampling, as the former allows continuous sampling of the bird's behaviour and provides greater detail. For similar reasons we do not consider instantaneous sampling or one-zero sampling as ideal recording methods. This is because accurate estimates of the frequency and duration of behavioural states (particularly swimming and diving) are important when constructing time budgets to which energy costs will be applied. The absolute length of sampling period, rules for choosing which individuals to observe, etc. will depend on the number of birds available for study, their mobility and whether individuals are marked. As no single sampling regime is appropriate for all situations, the methods used need to be clearly stated if meaningful comparisons between studies are to be made. If time-budget data are to be collected by more than one observer we suggest that a simple index of concordance is used to quantify inter-observer reliability Although there are several indices to choose from, the Kappa coefficient (k) is simple of calculate and is, therefore, probably the most convenient. Finally, relevant meteorological data should be collected during observations on birds, the most important of which are those relating to temperature, wind and precipitation (see e.g. Gremillet et al., 1995; Davies, 1996).

Once time-budgets have been constructed it is important that the energy costs used to convert them to estimated energy budgets are consistent in the future. It was agreed that until more empirical data are available, the most sensible data to use for this purpose are those of Gremillet et al. (1995). Full details of the assumptions made are given in their paper:

It is important to stress however that only Resting Metabolic Rate and Swimming costs were derived empirically in this study. It is also important to note that as these data were generated from studies on sinensis and assumed a body mass of 2230g (Dif, 1982), appropriate conversations need to be carried out to take into account the larger mass of carbo. Note also that resting metabolic rates of the other activities probably scale to different factors but again; for the sake of standardisation, it is probably best to use mass^0.72 until empirical data have been obtained.

The use of the above conversions is intended to facilitate comparisons between future studies by standardising many of the common assumptions made in time/energy budget studies. They are not intended as the definitive estimates of energy costs but are perhaps the most comprehensive we have at the moment: However each study will inevitably require its own assumptions (in additional to those above) and the extent to which these need to be detailed cannot be understated. In this respect we could do worse than follow David Gremillet's lead in the paper cited above.

Estimating Daily Energy Expenditure from doubly-labelled water

The use of stable isotopes to measure CO₂ production in free-ranging birds provides' a more direct way to estimate DEE, and hence daily food intake, than any of the previously mentioned methods. The double-labelled water (DLW) technique (Lifson et al., 1955; Lifson & McClintock, 1966) is now becoming an increasingly common tool in studying energy expenditure of free-living animals. There have been at least forty validation studies, in species ranging in size from scorpions to humans, and most have show that the method, applied to vertebrates, typically provides estimates Of energy expenditure accurate to within 5% (e.g. Speakman & Racey, 1988). However, most validation studies have been carried out on resting animals and more recent work suggests that there may be a reduction in the accuracy of the technique in very active (e.g. diving, as opposed to anything more active than sleeping) animals (Bevan et al., 1994; Bevan et al., 1995; Boyd et al., 1995).

When using the doubly-labelled water method it must be remembered that all the assumptions for applying the technique have to be met (otherwise inaccuracies will be incorporated into the Final calculation). These are fairly straightforward and are described in full in Tatner & Bryant (1989) together with details of the methodology. One part of the calculation of DEE from DLW studies that requires particular mention is the choice of an appropriate respiratory quotient (RQ) for converting the CO₂ production (i.e. what the DWL technique actually estimates) to DEE in kJ d^-1 from which an ultimate daily food intake figure (g d^-1) is derived. RQ (the ratio of CO₂ production to O₂ consumption) can change for a variety of reasons such as the state of the animal (e.g. gaining/losing mass) and the substrate being metabolised (i.e. fat, protein or carbohydrate). For this reason it is preferable, wherever possible, to derive empirical estimates of RQ during the course of a DLW study from, for example, respirometry measurements. Where this is not possible it is preferable, in the interests of standardisation, to assume a mean RQ of 0.81.

Data from DLW studies have b en used to generate allometric equations from which estimates of the DEE of Cormorants have been derived (see Feltham & Davies, 1995 and 1997). However it is important to remember that these too are based on a variety of assumptions. One particular point to make is that DEE estimates from 'such equations are based solely on body mass and do not therefore always reflect the different lifestyles of the birds being considered. This problem can of course be overcome to some extent by basing predictive equations on those data from-birds with the :most similar lifestyles to cormorants. At the moment, however, this means simply restricting the data set to a variety of cold-water seabirds that use flapping flight. The only way such problems can be entirely avoided is to measure directly the energy expenditure of cormorants. Unfortunately, there has been only one empirical study of this kind, Thomas Keller's work on sinensis (1997). The predictive equation mentioned above and Keller's empirical measurements suggest that, for adult sinensis, the likely range in DFI lies between 522-707g (see below), estimates somewhat higher than those previously derived from more gravimetric methods. It must be stressed, however, that at present we know absolutely nothing about the spatial and temporal variation in daily energy expenditure of sinensis, and have no empirical data for carbo at all. Whilst DLW studies on Cormorants will ultimately provide the much needed information on their energy expenditure, a great deal of work still needs to be done and future collaborative studies are already being discussed.

As future DLW studies are so important, it was considered prudent at the outset to try as far as possible, to standardise these methods too. The following suggestions were therefore made:

i) A minimum of 8-10 recaptured wild birds should be used in such studies.
ii) Handling time should be minimised (see next section).
iií) Blood samples for estimating background isotope levels should be taken from the study bird itself prior to isotope administration.
(iv) Injections of isotope should be intraperitoneal in preference to intramuscular . as injections in the pectoral muscles of penguins have been shown to impede their swimming activity (Wilson & Culik, 1996).
(v) Birds should be recaptured 2-6 days (i.e. 1-2 biological half lives of the O₁₈ isotope) after they were injected. It is however preferable that birds are recaptured as late as possible to minimise any handling effects on the final estimate.
(vi) Tail mounted radio-transmitters should be used in conjunction with DLW work in preference to harnesses. Furthermore, radio-tracking data (Gremillet, unpubl.) show that birds often behave atypically during the first 24 h after release. This should be considered in subsequent analyses.
(vii) As the DLW method is initially intrusive, the potential effects of handling, injection and carrying a radio transmitter cannot be overlooked. It is clearly advisable, therefore, that some measure of the variance in the behaviour of birds within the population to be sampled is obtained before DLW measurements begin.

Converting DEE estimates from time-budgets and DLW studies to DFI estimates

Once DEE estimates have been obtained, the method by which these are converted to DF~I estimates must also be consistent in the future. Although the mathematics of this conversion is simple, the variation in the assumptions made by previous workers has led to significant differences in final estimates of DFI. By using the same conversion figures, this source of variance can easily be eliminated.

Estimates of DEE simply tell us how much energy an animal must assimilate in order to meet its daily needs, not how much it needs to consume. As no vertebrate has an assimilation efficiency of 100% (and Cormorants are no exception) this fact should to be taken into account when converting DEE estimates to those of DFI. It is generally assumed that Cormorants have a similar assimilation efficiency to most seabirds, somewhere between 70-80%. This may not seem a large range but it can (and does) produce quite different estimates of DFI. Recently, Brugger (1993) has shown that nitrogen-corrected assimilation efficiencies were more meaningful than uncorrected ones, and that failing to correct for nitrogen can lead to significant overestimates of assimilation efficiency A her study was carried out on P. auritus, we recommend that her assimilation efficiency of 77.65 /o should be used in . future conversions, at least until studies on captive P. carbo are available. Put simply DEE estimates need to be multiplied by 1.288 to produce an estimate of daily energy intake.

As estimates of DFI are usually expressed as g d^-1 (rather than kJ d^-1), the calorific content of the prey must also be known. Ideally this should be derived empirically by bomb calorimetry of fish sampled in the areas where cormorants feed. If this is not possible, published species-specific calorific values can be used, provided good estimates of diet are available. It should be noted however that many calorific content values in the literature originate from Nutrition Journals and. typically refer to fish flesh only, not whole fish. If neither of the above option are possible, an average figure of 5.4 kJ g^-1 wet mass (Feltham & Davies, 1995) would be a suitable standard figure to use for most cyprinid- and percid- dominated fisheries. Note, however that if eel is known to be a significantly dietary component, this figure is inappropriate and would seriously overestimate daily food intake because e ( s have a flesh calorific content 3-4 times g eater than most other "coarse i.e. non salmonid) species. Whichever method is used, the assumptions made should be clearly stated to permit comparisons. This has not always been the case to date in published studies on cormorants.

Conclusions

Estimating DFI from considerations of the bird's energy requirements derived either from time-budget studies or DLW studies should provide us with the most reasonable estimates of intake. In this respect it is crucial that we try to be consistent in the methods we adopt and the calculations we carry out. Hopefully, the suggestions made above will go some way towards achieving this. Until more studies of the kind mentioned above are completed, we are left with rather few reasonable estimates of DFI, despite the dozens that have appeared in the literature. The list below shows those estimates currently available which have been derived by the methods advocated in this short review. All have been standardised using Brugger's assimilation efficiency and have used the authors" own assumptions regarding the calorific content of prey. Mean body masses (BM), assumed for predictive equations or measured empirically, are also given below together with sources.

¹ Feltham & Davies, .1997 (BM: Adult sinensis 2275 g, Imm. sinensis 2079 g, Adult carbo 2901 g, Imm. Carbo = 2657g); body mass data for P. c. sinensis are from Cramp & Simmons (1977) and for P. c. carbo from British birds shot under licence (Feltham, unpublished);
² Gremillet et al., 1995 (BM = 2230 g);
³ Sato et al., 1988 (BM = 1915 g);
⁴ Keller, 1997 (BM = 2122 g);
⁵ Gremillet et al., in press (BM = 2870g). Note: these values are not directly comparable to others in the table as they refer to the estimated DFI of both adults and nestlings.

OTHER NOVEL TECHNIQUES, AND SOME WORDS OF CAUTION

Stomach temperature loggers

The deployment of stomach temperature loggers has been restricted to breeding adult cormorants. Nevertheless, stomach temperature data are extremely valuable for the detailed analysis of such things as the foraging strategies of birds at sea because they allow the calculation of the immediate catch per unit effort, mean prey size and giving-up time. Stomach temperature records give a unique opportunity to determine the mass of single prey items as well as the time at which they were captured (Wilson et al., 1992; Grémillet & Plos, 1994). There are four essential phases in this kind of study. (1) Device deployment: the logger may be voluntarily taken by the birds if it can be hidden in Fish, or force-fed to captured birds, if not. (2) Device retention: the probes are usually rejected by Cormorants in less than 24h. This results in short-term measurements and in a high rejection rate Of the loggers at sea. This problem can been solved by the attachment of a special spring crown to one end of the logger (see Wilson et al., in press.). Using this, loggers have remained in stomachs for at least 48 hrs (Gremillet & Argentin, in press.). (3) Device recovery: this u critical for standard loggers which may be rejected at sea or far from the nest. For loggers equipped with springs, birds must be recaptured and devices removed with a special tool (Wilson et al., in press.). (4) Data analysis: the accuracy measurements may be influenced by the activity level of the bird, stomach contents, stomach blood perfusion, stomach "churning", body orientation, as well as the size and buoyancy of the device itself. We therefore recommend that the guidelines given in Wilson et al. (1995) are followed so that errors are kept as small as possible.

Automatic nest balances

Automatic weighing units are presently the most powerful and accurate tools for the determination of the daily food intake in breeding cormorants. This method does not require the 'capture of the birds, and the food loads brought back to the nest by breeding adults can be measured over extended periods with an accuracy of 0.10 g (Gremillet et al. , in press)..

Using additional information bout the time-budget of the birds at sea (as recorded by radio-tracking), and their digestion rate, food loads can be corrected to calculate the total amount of food taken at sea Gremillet et al., in press.). However, one critical point concerns the digestion rate of free-ranging individuals which is believed o differ from the that recorded in captive birds (Gremillet et al., in. press.). A measurement of the digestion rate in free-living birds through recording of the stomach pH (see e.g. Peters, 1995 would enable correction of the food loads recorded on the nest balance.

Time-energy budgets

Time-energy-budgets can allow the calculation of the food requirements of a cormorant "population" over a complete breeding season or the whole year (e.g. Gremillet et al., 1995). They are thus extremely attractive compared to data collected from stomach temperature loggers or nest-balances, the deployment of which are restricted to particular phases of the breeding season. However, this generalization also contains an important reduction in the accuracy of any Final result: its variance is the sum of the variance of each single step in the model. In order to keep this error as small as possible, the following points may be considered when studying the energetics of both adults and chicks:

(i) Time budgets may be recorded by direct observation at breeding sites and by radio-tracking or "dead-reckoning" for foraging birds. The influence of breeding phase, feeding area and meteorological conditions on time budgets may be analysed separately for sub-samples of birds.

(ii) Considering that Cormorants have a special plumage and thus particular thermoregulatory costs, energetic costs at rest should not be calculated from allometric relationships, but measured directly via gas respirometry or heart rate measurements. Again, the influence of weather conditions on thermoregulation should be considered.

(iii) Annual, or seasonal, variations in the mean energy content of the food should be assessed.

(iv) Sensitivity analysis is recommended, in order to identify those parameters which have the greatest effect and to point out critical steps or assumptions (such as the calculation of flight costs).

Stress

Apart from methods such as pellet analysis and direct observations, most current/proposed techniques will involve the capture and handling of birds e.g. for stomach flushing, radio-tagging, DLW-injection, or attachment of data loggers. Recent stomach temperature data (Plotz et al., in press.) showed that in seabirds, stomach temperature may increase rapidly when birds are caught or disturbed. This demonstrates again that birds are generally extremely sensitive to stress. The consequences of disturbance or capture may vary from anything between zero effect or a higher subsequent food intake (Gremillet et al., 1995) to death. Moreover heart rate measurements in penguins (Culik et al., 1990) and stomach temperature records in Cormorants (Gremillet & Argentin, in press.) have shown that stress levels are determined more by retention and length of handling time than by the actual type of handling. Thus it may be less critical for a cormorant to swallow a stomach temperature logger within 30 seconds than to be ringed and weighed for 5 minutes. We consequently recommend that handling time is kept as short as possible, even for routine operations.

CONCLUDING REMARKS AND ACKNOWLEDGEMENTS

This paper is a starting point for standardising techniques to assess cormorant diet and Food intake. It is not perfect, and producing it was not always easy. Members of the Group have debated some of the finer points keenly in the months which have elapsed since our original discussions. The paper should not be considered as the final statement on the subject, merely a consensus, from which cormorant researchers can work. We would like to think that our ideas can be updated as new data become available. There is also the potential For developing new techniques, such as fish species-specific molecular markers; activity studies have developed new techniques, dietary studies must too.

Two further points arose .from our discussions which we would like to emphasise. First, policy makers must be aware of the necessarily wide confidence limits which apply to some of the data we can produce (e.g. estimates of daily food intake). Such wide variation obviously has important implications for cormorant management plans and policies, at both regional and Governmental levels, and must always be considered. Second, some of the biases and sources of error we have discussed are actually relatively small. Our attempts to fine-tune particular aspects of our cormorant diet/food intake methods must be seen against a background of the huge variability often associated with current fisheries assessments. We would welcome more discussions with fisheries scientists on all aspects of cormorant/fisheries interactions.

Many people worked very hard to provide their various contributions to this paper on time, and it would not have been produced without their enthusiasm and commitment. The Group would like to thank Jeff Kirby and Mennobart van Eerden for their encouragement in getting the ball rolling in the first place, and Nicola Baccetti and GC who provided us with a stimulating environment within which to work while in Bologna. We would also like to thank everyone who commented on earlier drafts of the manuscript.

All members of the Working Group contributed to the draft outline of the present paper Thereafter specific sections were written by particular individuals, or smaller groups, 'as follows. The pellet section was drafted by DNC, incorporating comments from GC, DG, TK, and WS. Additional comments were received from Ronnie Veldkamp, Paul Hald-Mortensen, and Mennobart van Eerden and this section was refereed by Sarah Wanless (ITE, Banchory). The stomach contents section was drafted by DNC, incorporating comments from Mick Marquiss, and was refereed by Paul Thompson (University of Aberdeen). The direct observations section was drafted by JD, incorporating comments from DNC, WS, and SV, and was refereed by MM (ITE, Banchory). The DFI section was drafted by MJF, incorporating comments from RB, DG, and TK, while DG drafted section, these sections were refereed by David Bryant (University of Stirling). The introduction and concluding remarks were drafted by DNC. Finally, both Mike Harris (ITE Banchory), and especially Sarah Wanless, refereed the entire manuscript and their numerous comments greatly improved the final version of this paper.

The Diet Assessment & Food Intake Working Group is an umbrella title, designed to cover the widest possible spectrum of interested parties. "Membership" is free and open to all who have an interest in the subject and we would welcome comments on any aspect of our work. Correspondence should be addressed, in the first instance, to the Chairman.

SUMMARY

In many studies quantifying cormorant consumption of commercially important fishes is a key aim. However following on from this, is the issue of whether such consumption constitutes serious damage to the fish stock in question. This ultimate objective is even more problematic and, although outside the scope of the present review, should always be borne in mind when considering bird/fish interactions. Quantifying consumption by Cormorants requires information on (i) bird numbers, (ii) daily food intake of various components of the "population" and (iii) diet composition. Data for (i) and (ii) are probably achievable but (iii) is a major challenge, particularly if a population estimate is required, as sample size has to be fairly large. There has been little attempt to synthesis the results of the relatively large number of studies investigating cormorant diet, and many are undertaken in relative isolation. Indeed, differences in methodology often make it impossible to even directly compare these studies. Most workers acknowledge methodological problems but tackle them in different ways. This paper is not intended to be a comprehensive review of the scientific literature but a discussion of methods, associated sources of error attempts to quantify them, areas of consensus and gaps in current knowledge. It is structured in three parts: (1) direct assessment methods - pellet analysis, stomach contents and regurgitations (2) feeding observations, and (3) foraging ecology - food intake, daily energy expenditure and novel techniques. Our ultimate aim is to standardise methodologies wherever possible so that biases are minimised, and always quantifiable, so that dietary studies are valid, use appropriate methodology and are comparable. Two other important points arose from our discussions: (i) that policy makers must be aware of the necessarily wide confidence limits which apply to some of the data we can produce and (ii) that our attempts to fine-tune certain aspects of our work must be seen against a background of the huge variability often associated with current fisheries assessments. We would encourage constructive dialogue with fisheries scientists on all aspects of cormorant/fisheries interactions. This paper is not the final statement on the subject, merely a consensus from which cormorant researchers can work. We would like to think that our ideas will be updated as new data become available.

RIASSUNTO

Tecniche di determinazione della dieta e del prelievo trofico del Cormorano: verso una visione concorde

Esistono molti studi che hanno come scopo principale quello di quantificare il prelievo di pesce di interesse commerciale effettuato dai cormorani. Passo successivo è, comunque, quello di stabilire se tale prelievo costituisca un danno serio per gli stock ittici in questione. Questo obiettivo principale è ancora più problematico e, sebbene al di fuori dagli scopi del presente lavoro, dovrebbe essere sempre tenuto presente ogni qual volta si considerino le interazioni tra uccelli ittiofagi e pesci. La quantificazione del consumo operato dai cormorani richiede informazioni su: (i) numero degli uccelli, (ii) prelievo trofico giornaliero delle varie componenti della "popolazione', e (iii) composizione della dieta. Mentre i dati relativi ai primi due aspetti sono probabilmente Facili da reperire, il punto (iii) costituisce il problema principale, in particolare se si intende ottenere una stima riferita alla popolazione, dato che le dimensioni del campione devono essere abbastanza grandi. Ci sono stati pochi tentativi di sintetizzare i risultati del numero relativamente ampio di studi rivolti alla dieta del cormorano, che per la maggior parte sono stati fatti in modo piuttosto isolato realtà, le differenze metodologiche rendono spesso impossibile perfino il confronto diretto di questi studi. La maggior parte dei ricercatori riconosce I'esistenza di problemi metodologici, ma li risolve in modo diverso. II presente contributo non intende essere una revisione complessiva della letteratura esistente sull'argomento, ma piuttosto una discussione sui metodi, sulle fon i di errore ad essi associate, un tentativo di quantificare tali fonti di errore, un inquadramento dei punti di accordo esistenti e delle lacune conoscitive. È strutturato in tre parti: (1) diretta definizione dei metodi di lavoro su borre, contenuti stomacali e materiali rigurgitati, (2) osservazioni su soggetti in alimentazione e (3) sull'ecologia trofica: prelievo alimentare, costo energetico giornaliero e moderne tecniche di indagine al riguardo. Il nostro scopo ultimo è quello di standardizzare le metodologie in ogni possibile modo, allo scopo di ridurre e rendere comunque quantificabili i dati viziati, in maniera tale da rendere validi gli studi sulla dieta, da de6 re le metodologie appropriate e da renderli comparabili. Dalle discussioni effettuate in occasi e del convegno sono emersi due principali argomenti: (i) che le Amministrazioni interessate al problema vengano rese edotte dei limiti fiduciali necessariamente ampi che caratterizzano i dati che i ricercatori possono produrre, e (ii) che i nostri tentativi di modulare certi dettagli del nostro lavoro devono essere letti sullo sfondo della enorme variabilità spesso associata alle quantificazioni degli stock ittici. È da incoraggiare il dialogo costruttivo con specialisti del settore (ittiologi) su tutti gli aspetti dei rapporti Cormorano - pesca. Questo lavoro non è un punto d'arrivo sull'argomento, ma solo una base di consenso a partire da cui i ricercatori interessati ai cormorani possono impostare il proprio lavoro. Vogliamo credere che le nostre idee sull'argomento verranno periodicamente aggiornate, man mano che nuovi dati saranno disponibili.

REFERENCES

Barlow C. G., K. Bock, 1984 - Predation of fish in farm dams by cormorants Phalacrocorax spp. Australian Wildlife Research, 11: 559 566.

Bayer R. D., 1955 - Bill length of herons and egrets as an estimator of prey size. Colonial Waterbirds, 8: 104-109.

Bevan R. M., A. J. Woakes, P, J. Butler, I. L. Boyd, 1994 - The use of heart rate to estimate oxygen consumption of free-ranging black-browned albatrosses Diomedea melanophrys. Journal of Experimental Biology 193: 119-137.

Bevan R. M., A. J. Woakes, P, J. Butler, J. Croxall, 1995 - Heart rate and oxygen consumption of exercising gentoo penguins. Physiological Zoology, 68 (5): 855-877.

Blackwell B. F, J. A. Sinclair, 1995 - Evidence of secondary consumption of fish by double-crested cormorants. Marine Ecology Progress Series, 123 (1-3): 1-4.

Boyd I. L., A. J. Woakes, P J. Butler, R. W. Davis, T M. Williams, 1995 - Validation of heart rate and doubly-labelled water as measures of metabolic rate during swimming in California sea lions. Functional Ecology 9: 151-160.

APPENDIX 1

Contributors, after Chairman, in alphabetical order:

Carss, D. N. (Chairman) Institute of Terrestrial Ecology, Banchory, Kincardineshire, AB31 4BY Scotland, UK.
Bevan, R. M. The Wildfowl & Wetlands Trnst, Slimbridge, Gloucestershire, GL2 7BT, England, UK.
Bonetti, A. Via Concordia 10, 6900 Lugano, Switzerland.
Cherubini, G. Instituto Nazionale per la Fauna Selvatica, Via Ca' Fornacetta 9, 40064 Ozzano Emilia, Italy.
Davies, J. Liverpool John Moores University, School of Biological and Earth Sciences, Byrom Street, Liverpool, L33 AF, England, UK.
Doherty, D. University College Galway, Zoology Department. Galway, Ireland.
El Hili, A. Les Amis des Oiseaux, Faculte des Sciences, Campus Belvedere, 1060 Tunis, Tunisia.
Feltham, M. J. Liverpool John Moores University, School of Biological and Earth Sciences, Byrom Street, Liverpool, L33 AF, England, UK.
Grade, N. Institut Nature Conservation, Rua Filipe Folque 46, 30-1050, Lisbon, Portugal.
Granadeiro, J. P. Institut Nature Conservation, Rua Filipe Folque 46, 30-1050, Lisbon, Portugal. Gremillet, D. Institut fur Neereskunde, Dustenbrooker Weg 20. 24195 Kiel, Germany.
Gromadzka, J. Ornithological Station ul. Nadwislandska 108, 80-680 Gdansk 40, Poland.
Harari, Y N. R. A., Kibutz Naot Mordechai, 12120, Israel.
Holden, T. Liverpool John Moores University, School of Biological and Earth Sciences, Byrom Street, Liverpool, L33 AF, England, UK.
Keller, T. Technische Universitat Munchen, Institut fur Tierwissenschaften, Angewandte Zoologie, Block III, D-85350 Freising, Weihenstephan 12, Munich, Germany.
Lariccia, G., Dipartimento di Biologia Animale e dell'Uomo Viale dell'Università 32, 00185, Rome, Italy.
Mantovani, R. Instituto Nazionale per la Fauna Selvatica, Via Ca' Fornacetta 9, 40064 Ozzano Emilia, Italy.
McCarthy, T. K. University College Galway Zoology Department, Galway, Ireland.
Mellin, M. Eagle Conservation Committee, Otop Polish Society for the Protection of Birds, Phsudskieyo 7/9, 10-959 Olsztyn, Poland.
Menke, T. Kehvuriecles 6, D-2414B, Kiel, Germany.
Mirowska-Ibron, I. Eagle Conservation Committee, Otop Polish Society for the Protection of Birds, Phsudskieyo 7/9, 10-959 Olsztyn, Poland.
Muller, W. SVS/ASPU, PO Box 8521, CH-8036, Zurich, Switzerland.
Musil, P. Institute for Applied Ecology, C2-28163 Lostelec n.Cl., Czech Republic.
Nazirides, T. Vironia, 62043 Neo Petritsi, Greece.
Suter, W. Swiss Federal Institute of Technology Zurich, c/o Institute of Forest, Snow and Landscape Research, Zurcherstrasse 111, 8903 Birmensdorf, Switzerland.
Trauttmansdorff, J. F. G. Wilhelminenberg, O. Koenig Institut ABT. Donau, A-2000 Stockerau, Austria.
Volponi, S. Dipartimento di Biologia, Universita di Ferrara, Via Borsari 46, 44100 Ferrara, Italy.
Wilson, B. Liverpool John Moores University, School of Biological and Earth Sciences, Byrom Street, Liverpool, L33 AF, England, UK.