The impact of American mink Mustela vison on water birds in the upper Thames
*Present address and correspondence: Dr P. Ferreras, Department of Applied Biology, Estación Biológica de Doñana – CSIC, Apdo. 1056, 41080, Sevilla, SPAIN. Fax: 34 54621125; E-mail: firstname.lastname@example.org
1. The effect of mink predation on water birds during the breeding season was studied between March and September 1996 in a 33-km long stretch of the upper Thames river, England.
2. Mink presence significantly affected the density of breeding coots and the number of chicks hatched per pair of coots, as well as the average number of nests per pair of moorhens and the percentage of moorhen clutches hatched.
3. Mink diet during the birds’ breeding season (March–September) was studied through scat analysis. Ralliformes (coots or moorhens) represented 10% of the ingested biomass and were the fourth prey in importance after rabbits (45%), fish (25%) and small mammals (14%). Mink obtained 11% of their energy requirements from coots and moorhens.
4. Impact of predation by mink during the bird breeding season was moderate to high for moorhens (16–27% of adults and 46–79% of broods) and high for coots (30–51% of adults and 50–86% of broods).
5. Although moorhens seem well adapted to withstand predation by mink, nesting behaviour by coots make them very vulnerable to mink predation. We hypothesize that the persistence of coot populations in areas with high mink density requires immigration from surrounding populations with lower mink impact.
Introduced by accidental escapes from fur farms, the American mink Mustela vison Schreb. was first recorded to breed in the British countryside in the late 1950s, since when it has rapidly colonized most of the waterways of Britain (Dunstone 1993). The spread of mink was probably favoured by coincidental declines in potential competitors: the otter Lutra lutra (L.), and the polecat Mustela putorius L. The relationship of these semi-aquatic mustelids with their riparian prey is relevant to both predator and prey conservation (Kyne, Smal & Fairley 1989; Lodé 1993; Clode & Macdonald 1995). The impact of mink is thought to be a major factor in the recent decline of the water vole Arvicola terrestris (L.) in Britain (Woodroffe, Lawton & Davidson 1990; Strachan & Jefferies 1993; Barreto et al. 1998; Barreto, Macdonald & Strachan 1998). Less clear is the effect of mink predation on riparian bird species, a topic on which early studies led to contradictory conclusions (Linn & Chanin 1978; Lever 1978). As mink colonized different parts of Britain, considerable concern was expressed about the possible disappearance of moorhens. Smith (1988) reported that mink were associated with reductions in moorhen Gallinula chloropus (L.), coot Fulica atra L. and little grebe Tachybaptus ruficollis (Pallas) populations. Recently, Halliwell & Macdonald (1996) found no significant correlations between numbers of mink and moorhen censused along sections of four lowland British rivers. The impact of feral mink on native prey species is a problem with statutory implications in European Union states where the Habitats Directive (Council Directive 92/43/EEC of 21 May 1992, European Commission 1992) requires action to mitigate any such impact. Our study provides a scientific assessment of the widespread claims that American mink threaten populations of coots and moorhens in Britain.
Our main objective was to evaluate the role of mink predation during the breeding season on populations of moorhens and coots in a riverine ecosystem of southern Britain. First, we sought relationships between mink presence and either bird abundance or bird breeding success. Secondly, we assessed the predation impact of mink on coot and moorhen populations, based on mink energetic requirements, proportion of birds in the mink's diet, and the abundance and production of birds in the study area.
The study area comprised 33 km of the Upper Thames river, southern England (between Lechlade, Gloucestershire, 1°42′W 51°41′N, and Northmoor Lock, Oxfordshire, 1°23′W 51°43′N). River width varied between 5 and 20 m and river depth was generally > 1 m. The river was fringed with trees such as willow Salix fragilis, alder Alnus glutinosa, and ash Fraxinus excelsior, and vegetation such as nettles Urtica dioica, rank grassland, willowherb Epilobium spp., bramble Rubus fruticosus, blackthorn Prunus spinosa, and hawthorn Crataegus monogyna. The band of vegetation emerging from the water consisted of species such as bur-reed Sparganium erectum, common reed Phragmites australis, and reedsweet grass Glyceria maxima (for the scientific authorities for the European flora see Tutin et al. 1980). Adjacent land was mainly pasture, arable and woodland. Presence of mink was known from previous studies (Halliwell & Macdonald 1996) being mainly restricted to the 23 km downstream portion of the river stretch. The remaining upstream 10 km were mainly mink free; in the upstream stretch of river, lock-keepers occasionally killed transient mink, and reports of mink had been very rare during the 3 years prior to the study.
Bird abundance and breeding success
Between March and September 1996 water bird populations were censused and their breeding activity was monitored along the study area. For this purpose the whole river stretch was walked, recording every water bird sighting, as well as nest locations and status, such as nest-building, incubation, number of eggs, predated eggs, number of chicks or nest abandonment. Typically, 2 weeks (average 13·3 ± 8·3 days) elapsed between successive visits to the same nest. Courting and territorial behaviour were used to define territorial borders and to record numbers of breeding pairs, following the guidelines given by the British Trust of Ornithology (Taylor 1984; Carter 1989). These consist, basically, of a minimum of six visits to the area, recording the birds seen and their territorial behaviour on detailed maps. In our case, a minimum of 10 visits to each area was carried out. The information for each species from all the visits is transferred to species sheets and used for delineating the boundaries of territories based on territorial contests, location of nests and repeated sightings of adults. For statistical analysis, the study area was divided into contiguous 500-m sections, where habitat characteristics, mink abundance and water bird (coots and moorhens) abundance were quantified. Bird abundance refers to the number of territories of a given species per sector, rather than the actual number of birds, since birds were not individually marked and we could not detect changes in the holders of a territory. The fate of every nest was recorded on successive visits and used to quantify breeding success for each territory. Several parameters were used to evaluate breeding performance: number of nest attempts, number of clutches laid, proportion of successful clutches (those with at least one hatched egg) and proportion of hatched eggs from the total that were laid. For statistical analysis, parameters of breeding performance were averaged for the territories contained within each 500 m sector. The identity of predators responsible for taking eggs could be determined only occasionally. To diagnose mink predation, we followed Craik (1995); canine marks were typically 1–2 mm wide and, if paired, ≈ 10 mm apart on the shell, and nests were little damaged.
A detailed habitat survey was obtained in the field by mapping on 1:2000 enlargements from 1:10 000 Ordnance Survey maps all habitat features found on the banks up to 100 m from the river border, as well as fringe, and other physiographic characteristics (channel width, presence of locks). Habitat features found on the banks were: conifer woodland, deciduous woodland, lines of trees, scattered trees, hedges, lines of scrubs, tall herbs, grazed grass, hay meadow grass, rough grass and pollarded trees. These variables were chosen because they had previously proven useful in models of the abundance of moorhen, coots and 27 other species of British riverine birds (Rushton, Hill & Carter 1994). Values for total overhanging vegetation were obtained by adding lines of trees and lines of scrubs. This information was converted to scores for every 500-m sector of the river, as the proportion of each habitat feature covering the 1000 m (= 500 m × 2 banks) of each section, except for scattered trees and pollarded trees, for which stem counts were recorded. Locks were recorded as presence or absence, and channel width as one of three categories: < 10 m, 10–20 m and > 20 m.
Data on mink presence and abundance were gathered simultaneously by trapping and radio-tracking in the lower 23 km of the river (N. Yamaguchi, personal communication). Total and resident mink densities were assigned to each 500-m sector. Field signs and frequent searches confirmed lock-keepers’ reports that mink were absent from the remaining 10 km upstream throughout the study, except in a localized 1500 m area. This small area (three sectors) from the widely mink-free 10 km was included together with those sectors inhabited by mink from the downstream 23 km in the analyses of mink–bird relationships.
Fresh mink scats were collected between March and September 1996 by searching the bank, ledges under bridges, the bases and the crown of pollarded trees, burrows and hollow trees (Dunstone 1993; Halliwell & Macdonald 1996). Individual scats were stored in paper bags and oven-dried at 60 °C for 48 h, then weighed to the nearest 0·01 g. Residues were examined under a binocular microscope using 10× magnification (Strachan & Jefferies 1996).
Remains of avian prey were identified to Order, using features of downy barbules (Day 1966). Mammalian prey were identified from hair and tooth characteristics (Teerinck 1991). Fish remains were identified to family using keys based on scales and vertebrae (Conroy et al. 1993).
The relative contribution of prey types to mink diet, was calculated using three techniques: (i) percentage of occurrence; (ii) percentage of dry weight; and (iii) percentage of biomass ingested, estimated by transforming the dry weight of scats to biomass consumed. The percentages of occurrence are provided since they allow comparison with other studies despite flaws in this measure, which over-represents minor items and under-represents major ones (Wise, Linn & Kennedy 1981). For the percentage of biomass ingested, we used correction factors calculated for mink (Fairley, Ward & Smal 1987), except for insects, for which we used the factor for Martes martes (Lockie 1961), and for bird's eggs, for which we used the factor for Herpestes ichneumon (Palomares & Delibes 1990).
Impact on water bird populations
We centred our study on two species of water birds, coots and moorhens, although some others were observed nesting in the river vicinity, such as mallards Anas platyrhynchos L., mute swans Cygnus olor (Gmelin) and great crested grebes Podiceps cristatus (L.). Coots and moorhens were selected since their nesting behaviour, amongst the bankside fringing vegetation or over the water, may make them particularly vulnerable to predation by mink. Data on swans and great crested grebes could not be analysed statistically due to their scarcity in the area (under 10 pairs each), nor mallards because of their nesting behaviour inland which made the detection of nests more difficult. In order to estimate the impact of mink on coot and moorhen populations, we compared the estimated consumed biomass with the total population and production of birds during the study. The proportions of prey biomass were transformed to proportions of ingested energy according to the energy content of prey types given by Dunstone (1993) except for eggs, where the value given by Arnold, Alisauskas & Ankney (1991) was used. We calculated the total energy requirements of the mink population, based on its composition (number of mink, sex and age) and the individual energy requirements (Dunstone 1993). Total numbers of coots and moorhens preyed upon by mink during the study were calculated from the total energy requirements of the mink population, the proportion of energy acquired from Ralliformes, the energy content of this food and the weight and proportions of the different age categories of birds taken according to the scat analysis. The impact of predation was then estimated by comparing the number of individuals taken with the total population of birds estimated from field censuses in the downstream 23 km, where mink were present.
Our primary goal was to assess relationships between mink abundance and number of water bird territories and breeding success. In an initial analysis we used univariate statistical tests (Student's t-test or Mann–Whitney U-test) to check for the effect of mink presence on numbers of bird pairs or average breeding performance in 500-m sectors. However, habitat selection by mink and water birds could confound this analysis, for example if they select different habitats, thereby giving an illusion of avoiding each other. Therefore, we controlled for habitat variation before relating mink and bird variables. Principal Components Analysis (PCA) was used to ordinate the habitat data. We then used the component scores of the most meaningful PCA, together with mink abundance variables as independent model predictors and either the number of territorial pairs per sector or each of the breeding success parameters (averaged by sector) as dependent variables. Stepwise multiple parametric regression was used to obtain the best model relating each of the breeding success variables (percentages were first arc-sin transformed) to habitat PCA scores and mink abundance variables. Since the numbers of bird territories per sector were discrete low numbers (0–5), the errors associated with modelling such data are likely to follow a Poisson rather than Normal distribution. Therefore, we used Generalized Linear Models (McCullagh & Nelder 1983) to relate such variables to habitat PCAs and mink abundance. We assumed that the residual error would follow a Poisson distribution (following the GLIM methodology as in Rushton, Hill & Carter 1994; see also Jongman, ter Braak & van Tongeren 1987). The goodness-of-fit of each model to the data was evaluated using the deviance test, and the ‘Wald’ test (Aitken et al. 1989) was used to test whether the regression coefficients were statistically different from zero. Models with the greatest reduction in deviance and minimum number of predictor variables were selected.
The four first Principal Components (PCs) were identified as an effective statistical description of the habitat structure of the river sections, according to the ‘scree test’ (Tabachnick & Fidell 1996). Their eigenvalues were 2·35, 2·25, 1·74, 1·64, respectively, and amounted to 47% of the total inertia of the data matrix. The most obvious feature of the ordination (first PC) is the amount of deciduous woodland and hay meadows. The second axis reflected a trend marked by the amount of scrub, total overhanging vegetation and absence of tree lines. The third PC was defined by a high score for arable land and tall herbs and a low proportion of grazed land. The fourth axis distinguished abundant reeds/rushes and rough grass on the bank from sectors where banks were typified by gardens or recreational areas.
Birds and mink abundance
A total of 47 coot and 103 moorhen territories was identified, 18 and 70 of them, respectively, occupying the downstream 23 km mainly inhabited by mink. Without any habitat considerations, mink presence (either resident or transient) was associated with a significant reduction in the number of coot pairs per sector (1·0 ± 0·2, 0·6 ± 0·1, no mink vs. mink; U = 1145·0, P < 0·01), and presence of resident mink was associated with a reduction in the number of coot chicks hatched per pair (2·6 ± 0·5, 0·9 ± 0·4, t = 2·81, P < 0·01; Table 1). Mink presence (either resident or transient) was also associated with a significant reduction in the average number of nests per pair of moorhen (2·5 ± 0·2, 1·7 ± 0·2, no mink vs. mink; t = 2·49, P = 0·02) and with an increase in the percentage of moorhen clutches that hatched (64·0 ± 6·2, 82·2 ± 6·5, no mink vs. mink; t = 2·01, P = 0·05). The number of moorhen pairs per sector was not significantly affected by mink presence (U = 925, P = 0·87; Table 1).
Average number, standard error (SE), minimum and maximum birds pairs per 500-m sector (n
= number of sectors, excluding those sectors with either no pairs or missing values). Statistics are also shown for nesting success parameters (averaged per 500-m sector). Results of univariate tests comparing mink and no-mink areas are shown for the bird variables (NS = not significant; *P
When habitat is included in the analysis, the most parsimonious model for coot abundance included the total number of mink (residents + transients) present during the study and a great part of total habitat variation (PC1, PC2 and PC3). This model reduced significantly the original deviance (D2 = 20·1%, P < 0·01) and included two significant terms, the total number of mink (negative association, t = – 2·65, P < 0·05), and the amount of deciduous woodland and hay meadows (positive association, t = 2·05, P < 0·05). The best model for moorhen abundance included the number of resident mink and the same habitat variables as coot abundance. Although the model produced a significant reduction in deviance (D2 = 15·9%, P < 0·05), none of the parameter estimates for each independent variable produced a significant t-value (P > 0·05).
Multiple regression revealed no significant relationship between moorhen breeding success and either mink abundance or habitat. Coot breeding success was related to habitat features and abundance of resident mink. Number of hatched chicks per coot pair was significantly related to the amount of scrub and overhanging vegetation (PC2; F = 4·94, P = 0·04, r = – 0·41) and the proportion of hatched eggs was related to the total number of resident mink (F = 4·56, P = 0·04, r = – 0·40).
Mink diet and impact of predation on coots and moorhens
All the scats (115) but one were collected in the 23-km downstream area to which mink were known to be restricted (Halliwell & Macdonald 1996). Rabbit Oryctolagus cuniculus L. was the predominant prey (28% occurrences in scats, 33% dry weight, 45% ingested biomass; see Table 2). The second prey in importance was fish (mainly Percidae and Cyprinidae) as ingested biomass (25%), or small mammals (mainly Apodemus sylvaticus (L.), Microtus agrestis (L.), and Sorex spp.) as percentage of both occurrences (24%) and scat dry weight (22%). Ralliformes (coots and moorhens, Rallidae family) were fourth in importance, representing 10% of the ingested biomass (Table 2). We could identify either the species or the age (chick or adult) of the remains of Ralliformes in only eight scats: two were adult moorhens, one an adult coot and five unidentified chicks. These proportions were extrapolated to the whole sample in calculations of impact of predation. Moorhen eggshell remains were found in only two scats and no coot eggshell remains were found.
Table 2. Diet composition estimated from mink scats (n = 115) collected in the upper Thames in 1996
|Passerines|| 4·0|| 4·5|| 2·1||7·29|| 2·3|
|Crustacean|| 4·0|| 0·5|| 0·2||3·00||<0·1|
|Anatidae|| 3·9|| 3·7|| 2·4||7·35|| 2·7|
|Undetermined bird|| 2·8|| 1·9|| 0·9||7·29|| 1·0|
|Insects|| 2·3|| 0·1||<0·1||3·00||<0·1|
According to a simultaneous survey of the mink population (N. Yamaguchi, personal communication) between three and eight adult male mink occurred in the area at different stages of the study; however, when the maximum number of mink was present, the ranges of five extended beyond the study area. Hence, we estimated the effective number of males hunting in the area to be between three and six. Two breeding adult females and between one and three non-breeding females were present in the area during the study. Thus, we assumed minimum and maximum mink populations of 3–6 males, 2 breeding females and their dependent kittens, and 1–3 non-breeding females. Energy requirements of captive adult male and female mink were given by Dunstone (1993; p.97) as 1500 kJ day–1 and 1000 kJ day–1, respectively. Total energy requirements during our study (169 days, from 19 March to 4 September 1996) for each adult male and each non-breeding adult female, would be 253 500 kJ and 169 000 kJ, respectively. Females would require 6% more energy during gestation, and up to 100% more before weaning the kits (Dunstone 1993; p. 99). Total requirements for a breeding female during the study (including those of her kittens) was estimated as 338 721 kJ. For this calculation we assumed: 50 days of gestation (Joergensen 1985); parturition occurring in the first week of May (Dunstone 1993; p. 146); 7 weeks from birth to weaning (Dunstone 1993; p. 146); and two kittens (one male and one female) reaching weaning age per female. According to these figures, energy requirements of the mink population during the study was between 1606 942 kJ (minimum) and 2705 442 kJ (maximum). Considering the energy value of the different prey categories and the ingested biomass in the mink diet, we estimated that 11·1% of this energy requirement, i.e. 24·2–40·7 kg of fresh biomass, came from coots and moorhens. Considering the proportions of the species and age classes identified in the scats, and their average weights (326 g, 731 g, 140 g and 184 g for adult moorhen, adult coot, and whole hatched clutches of moorhen and coot, respectively; Perrins 1987), we estimated that the mink present in the area during the study preyed upon 23–38 adult moorhens, 11–19 adult coots, 45–77 moorhen broods and 11–19 coot broods. Comparing these figures with the total bird populations and their productivity in the area inhabited by mink (the 23 km downstream: 140 adult moorhens, 37 adult coots, 97 moorhen clutches, and 22 coot clutches), 16–27% of the adult moorhens, 30–51% of the adult coots, 46–79% of the hatched moorhen clutches and 50–86% of the coot clutches would be predated by mink.
This is the first attempt to analyse quantitatively the effects of introduced American mink on a population of native British birds. Both the relationship found between mink and bird abundance and the impact of predation from diet analysis, indicate that mink do influence coot populations but this is less conclusive for moorhen populations.
Coot numbers were clearly affected by mink presence according to both statistical approaches. This negative relationship might have arisen either because the birds avoid areas with high mink density, or by a constant reduction of bird populations by mink. In both cases, we hypothesize that the mechanism maintaining coot populations is a source–sink system whereby surplus birds from nearby mink-free populations fill the vacancies created by the activities of mink. Moreover, a clear relationship seems to exist between coot breeding success and mink presence; fewer chicks were born per pair in the area occupied by mink, and the proportion of hatched eggs was lower in the presence of mink.
Moorhen abundance was negatively, but not significantly, associated with number of resident mink. Although some parameters of moorhen breeding performance appeared to be influenced by mink presence (fewer nests were built per pair and more clutches hatched, paradoxically, in the sectors with mink), these relationships did not hold when controlling for habitat parameters. The positive relationship between mink presence and percentage of moorhen clutches that hatched could be explained by an independent effect of habitat on mink presence and moorhen breeding success, i.e. the nests which are safe from mink in an area inhabited by this predator, could also be safe from other predators.
Our results should be interpreted with caution. First, we assumed that scat analysis is a valid indicator of mink diet and the variance attached to the relationship between input (ingested biomass) and output (matter found in faeces) does not bias our calculations of population impact. Despite the criticism that ‘correction factors’ may not accommodate wide variation in the amount of indigestible remains (Dunstone 1993; Strachan & Jefferies 1996), we employed such factors as the only way of estimating prey intake from scat composition. Secondly, we calculated the number of each prey taken on the assumption that mink consume the whole prey. However, Birks & Dunstone (1984) note that mink dens contain remains of prey species, and probably some parts of the prey, such as large bones, large feathers or hard skin are not usually ingested. In this respect, our estimate of mink impact will be conservative. Such a bias could be somewhat compensated to the extent that some birds were consumed as carrion (Dunstone 1993). A further bias in our estimates could be expected from a likely underestimation of eggs in the mink diet. In fact, we found moorhen eggshell remains in only two scats and no scat contained coot eggshell pieces, although we have field evidence of mink predation on eggs from both species. Such under-estimation may not have a great effect on the birds’ demography, since many eggs are destroyed at once whereas only one adult is killed at a time, and the reproductive value of a whole clutch (8–10 eggs) can be close to that of one adult. Another potential bias is the small sample from which we estimated the proportional representation of coots and moorhens in the diet. The small size of our sample of scats may introduce random sampling error to the estimated proportion of each prey in the mink diet but we have been cautious and have used only conservative estimates.
Coots and moorhens accounted for 11% of the minks’ energy, whereas other water birds (Anatidae) represented 4% of dry weight in the mink scats, 2% of ingested biomass, and 3% of the energy intake. These differences may reflect the different availabilities of the Ralliformes and other waterfowl. In general, higher proportions of waterfowl in mink diet have been related to high waterfowl densities and to the scarcity of other prey such as fish and crayfish (Eberhardt & Sargeant 1977; Chanin & Linn 1980). Moreover, predation on waterfowl is typically more intense during the breeding season (Gerell 1967; Eberhardt & Sargeant 1977) when nesting hens are more vulnerable.
Vulnerability may also explain the high proportion of Ralliformes compared to mallards in mink diet. Coots and moorhens are more easily killed by mink than are the larger ducks because they are smaller, they spend more time in reed beds where they are accessible to ambush by mink, they nest on the ground, and they have poorer ability to fly (Chanin 1981; Wise, Linn & Kennedy 1981). Vocalizations of coot hatchlings and the tendency for parents to leave juvenile coots unattended may also increase their vulnerability (Eberhardt & Sargeant 1977). Ducklings appear less susceptible to mink predation than juvenile coots, perhaps because of maternal defence (Sargeant, Swanson & Doty 1973) and lower use of river fringes (Gerell 1970) by ducks compared to coots and moorhens.
The impact of predation by mink on waterfowl in our study is higher than that in some areas of North America. In North Dakota one individual mink family (female with kits) ate three of the estimated 36 adult coots (8%) and 76 of their young (52%) during the nesting season (Eberhardt & Sargeant 1977). In Manitoba, Arnold & Fritzell (1987) concluded that male mink predation on breeding waterfowl (including American coots Fulica americana, ducks and grebes) had little impact on their populations. However, these authors ignored different digestibility of prey types in their calculations, which probably underestimated the amount of avian prey consumed. In our area, mink did seem to take a significant part of the coot population. Some predated pairs were rapidly replaced by other coots that nested in the same territories. We propose as a hypothesis to be tested in further studies that these coots could be immigrants from other populations existing in nearby lakes, ponds and a reservoir, where circumstantial evidence indicated that mink predation was lower. However, the local conditions, and especially the availability of major prey such as rabbits and fish, doubtless play an important role in determining the impact of predation by an opportunistic species such as the mink (Gerell 1967; Dunstone & Birks 1987). How many coot and moorhen are taken by the American mink may depend not only on their abundance, but also on the availability of other prey. For example, if rabbit numbers declined due to disease, then the impact of predation on the remaining prey could increase. At least in our study area, mink presence is evidently compatible with apparently sustainable populations of coots and moorhens.
We gratefully acknowledge the funding of The Environment Agency, People's Trust for Endangered Species and Tusk Force. Rob Strachan, Nobuyuki Yamaguchi, Guillermo Barreto, Chris Strachan and Adam Grogan kindly helped in different stages of the study. Paul Johnson provided invaluable statistical support. We thank Miguel Delibes, Hans Kruuk and two anonymous referees for their helpful comments on the manuscript. The Spanish Ministry of Education provided a postdoctoral fellowship to the senior author during the study.