Search
Author
Title
Vol.
Issue
Year
1st Page

Journal of Animal Science - Animal Genetics

Population viability analysis of American mink (Neovison vison) escaped from Danish mink farms

 

This article in JAS

  1. Vol. 91 No. 6, p. 2530-2541
     
    Received: Oct 29, 2012
    Accepted: Feb 15, 2013
    Published: November 25, 2014


    2 Corresponding author(s): biocp@nf.au.dk
 View
 Download
 Share

doi:10.2527/jas.2012-6039
  1. C. Pertoldi 21,
  2. S. Rødjajn,
  3. A. Zalewski§,
  4. D. Demontis,
  5. V. Loeschcke and
  6. A. Kjærsgaard‡#
  1. Aalborg University, Department 18/Section of Environmental Engineering Sohngårdsholmsvej 57, 9000 Aalborg, Denmark
    Aalborg Zoo, Mølleparkvej 63, 9000 Aalborg Denmark
    Department of Biosciences, Aarhus University, Ny Munkegade 114, 8000 Aarhus C, Denmark
    Mammal Research Institute, Polish Academy of Sciences, 17-230 Białowieża, Poland
    Institute of Evolutionary Biology and Environmental Studies, University of Zurich, Winterthurerstrasse 190, CH-8057 Zurich, Switzerland

Abstract

The American mink (Neovison vison) was introduced to Danish fur farms in the 1930s. An unknown number of mink have managed to escape these farms over the years. Today feral mink are found in the wild in most parts of Denmark. A population viability analysis (PVA) was performed using VORTEX, a stochastic population simulation software, to 1) predict the viability and potential population expansion from different sizes of founding populations of farm escapees, 2) investigate which parameters mostly affect the viability, 3) assess the effects of continuous escapes on the feral populations and how the feral populations are affected by management programs, and 4) discuss eradication strategies and their efficiency in management of the feral American mink population in Denmark. The simulations showed that juvenile mortality had the greatest effect on population viability followed by fecundity, adult mortality, and initial population size. Populations supplemented yearly by escapees all reached the carrying capacity and gained genetic variability over the years. Harvesting was modeled as the yearly number of mink caught in Denmark. Most of the simulated harvested populations crashed within few years after the first harvesting event. This indicates that the feral number of mink in Denmark is sustained due to supplements from mink farms and no true feral population exists. To manage the number of feral mink in Denmark it is essential to prevent escapees. The eradication effort would be most effective if focused on late summer and autumn when juvenile mink leave the maternal territory.



INTRODUCTION

Invasive alien species (IAS) are commonly regarded as a major threat for biological diversity at the global scale. In addition to the threats to biodiversity, the direct costs of IAS are immense. It is difficult to estimate precisely the economic losses caused by biological invasions; however, they include the increased costs of control, the management costs of reducing the alterations of protected areas, and the impact of introduced pathogens affecting wildlife. In consequence, the planning of more effective strategies to deal with biological invasions is a conservation priority on a global scale. An important applied question is therefore how to optimize alien species management through control and minimize economic costs. Modeling of population dynamics and their response on number control, based on population ecology theory, can inform on optimal strategies to achieve the best results in management program.

In contrast to feral mink from populations living in the wild 20 to 30 yr, farm mink are rather poorly adapted to the wild conditions and genetic adaptation is necessary to establish a population that can increase in size (Hammershøj et al., 2006). Farm mink are often inbred to maintain desired traits such as homogenous pigmentation, coloration, and large body size (Dunstone, 1993). Inbreeding can lead to inbreeding depression, in which the fitness of the inbred individual is less than the species average (Pertoldi et al., 2006; Charlesworth, and Willis, 2009). Demontis et al. (2011) found significant differences in fitness between color strains. The “wild-type” strain demonstrated significantly lower relatedness, defined as the proportion of alleles identical by descent in 2 related individuals, than the less common colors pearl and sapphire (relatedness = 0.03 to 0.26 vs. relatedness = 0.35 to 0.46).

The aims of this study were to 1) predict the viability and potential population expansion from different sizes of founding populations of farm escapees, 2) investigate which parameters mostly affect the viability, 3) assess the effects of continuous escapes on the feral population and how the feral population is affected by management programs, and 4) discuss eradication strategies and their efficiency in management of the feral mink population in Denmark. To evaluate the above described points we performed a population viability analysis (PVA) using the software VORTEX.


MATERIALS AND METHODS

The Study Organism

The American mink (Neovison vison) is a medium sized riparian mustelid endemic to North America (Kidd et al., 2009). The body length is approximately 30 cm and the BW is 0.7 to 1.5 kg, with the females half the size of the males. Farm mink are considerably larger and show less sexual dimorphism compared with the wild living mink (Belliveau et al., 1999; Asferg, 2010). The mink prefers a wide range of semiaquatic habitats such as riverbanks along streams, lakes, and near the coast (Gerell, 1970) but is also found in drier habitats (Birks and Dunstone, 1991). In these various habitats mink diet composition also greatly varies and they prey on small mammals, birds, amphibians, fish, and invertebrates (e.g., crayfish; Gerell, 1967; Sidorovich et al., 2010; Zalewski and Bartoszewicz, 2012). Mink have 1 litter per year and mating period starts in late winter to early spring. Due to a high degree of early kit mortality in the wild, fewer than 4 kits usually survive until the kits emerge from the den and as few as 1 on average is alive in late summer when the kits are weaned (Dunstone, 1993).

In the 1920s the mink was brought to Europe to be used in the fur-trading industry (Asferg, 2010). The present-day mink kept on farms are believed to originate from 3 subpopulations within the native range (Dunstone, 1993). Over the years, mink have managed to escape from farms, either by accident or intentional release (Bonesi and Palazon, 2007; Zalewski et al., 2011). At present, the feral mink is widely distributed across many European countries (Bonesi and Palazon, 2007; Hammershøj et al., 2007) and is an invasive species that is difficult or maybe even impossible to get rid of once established. Feral mink have a potential negative effect on the indigenous wildlife, due to both competition with other mustelid species and predation on birds and mammals (Hammershøj et al., 2005; Brzeziński et al., 2012). Feral mink can also cause economic loss for farmers due to surplus killing of housed poultry and fish (Bonesi and Palazon, 2007).

Population Viability Analysis

Population viability analysis is a tool for modeling the present and future probability of extinction in a species at population level. It incorporates known survival threats into a model of the extinction process (Lacy, 1993). Populations are considered genetically viable if they maintain ≥90% of the initial expected heterozygosity (HE; Foose, 1993) and critically endangered if the effective population size (NE) decreases by more than 80% over a period of 3 generations (IUCN, 2001). If a population for some reason ends up small and isolated, it becomes vulnerable to other forces that might drive it to extinction even if the factors that caused the initial decline are no longer present (Shaffer, 1987; Soulé and Mills, 1998). Stochastic processes are a particular threat to small populations. Stochastic processes that affect populations are categorized as catastrophes, genetic drift, environmental variation, and demographic stochasticity (Shaffer, 1987; Mills and Allendorf, 1996).

The usefulness of PVA has been discussed in the literature (Boyce, 1992; Shaffer et al., 2002). The criticism has mainly been on defining the appropriate time frames of the simulations (Armbruster et al., 1999), the difficulty in estimating the accurate input data (Boyce, 1992), and the lack of documentation that the model actually predicts the future (Coulson et al., 2001). Another weakness in PVA is the data quality (Beissinger and Westphal, 1998; Coulson et al., 2001) as the lack of relevant information makes it difficult to estimate many critical parameters. The main advantage of PVA is the possibility to assess the relative risks that influence the survival of a population (Beissinger and Westphal, 1998), which makes it possible to achieve an understanding of the risk factors threatening a particular population or species (Ralls and Taylor, 1997). Retrospective tests based on 21 long-term ecological studies have, however, also revealed highly accurate predictions of PVA. These authors even recommended its use when data availability is limited as the best available method for guiding management decisions.

Software

VORTEX (Miller and Lacy, 2005) is a software program that simulates the effect of deterministic forces, demographic and environmental stochastic events, and genetic drift in small populations. The software simulates the population size through annual events via random variables and user-specified distributions. It is designed to model the viability in small populations of long-lived, diploid sexually reproducing species with low fecundity, such as mammals, birds, and reptiles. The software uses variables such as reproduction, mortality, carrying capacity (K), and migration (Lacy, 1993) and has been used to evaluate management strategies in many species [e.g., the Puerto Rican parrot (Amazona vittata; Lacy et al., 1989), the domestic horse (Equus caballus; Thirstrup et al., 2009); the African Wild dog (Lycaon pictus; Bach et al., 2010)].

When modeling events such as reproduction, litter size, and determination of death, a pseudorandom number generator is used to model demographic stochasticity. The carrying capacity is specified by the user and whenever the population exceeds this upper limit, animals are randomly removed from the population (Miller and Lacy, 2005). VORTEX can incorporate annual fluctuations in birth and mortality rates due to EV by assigning a mean and a SD to the parameters used. These annual variations are important when determining the probability of extinction (Goodman, 1987) but field data are rarely available. VORTEX is most often used for exploring different management strategies for rescuing endangered populations and the consequences of these strategies. In this way unforeseen (negative) consequences might be discovered in the simulation and it is possible to choose the best management strategy. VORTEX also has an option for sensitivity testing, which is used to assess how much the different parameters and their interactions affect the probability of extinction (Lacy, 1993).

The program ideally requires considerable amounts of data (Lindenmayer et al., 1995), which is available for mink due to the many years of mink farming and consequent intensive research into their life history. To our knowledge, VORTEX has never been used on invasive species to predict population expansions or to investigate the power of current eradication strategies.

The program GENALEX v. 6.1 (http://biology.anu.edu.au/GenAlEx/Download.html; Peakall and Smouse, 2006) was used to estimate the pairwise relatedness (according to Queller and Goodnight, 1989) and the allele frequencies within and between the farm populations and the F1 hybrids from the 33 populations. The allelic frequencies were used as inputs for the VORTEX simulations.

The program Hybridlab 1.0 (Nielsen et al., 2006) was used for simulating a user-specified number of multilocus F1 hybrid pairs of the 33 mink populations from population samples of genetic markers (21 polymorphic microsatellites loci; Demontis et al., 2011). The output file of Hybridlab was used for the estimation of the allelic frequency of the F1 hybrids with GENALEX. A linear regression of fecundity with relatedness was performed to predict the mean fecundity of each of the 15 F1 hybrid populations.

Genetic Data

Allele frequencies of 21 loci genotyped in 33 breeding lines from 13 Danish mink farms and their corresponding mean fecundities were obtained from Demontis et al. (2011) and used in the present investigation. These color strains were analyzed: black (population 1 to 8), white (population 9 to 16), sapphire (population 17 to 19), wild-type (population 20 to 30), silver (population 31), mahogany (population 32), and pearl (population 33). First generation hybrids (derived from the cross between 2 different populations) were simulated from the 33 breeding lines and 15 hybrid populations were selected for the VORTEX simulations.

Estimation of the Input Parameters for VORTEX

Carrying Capacity and Demographic Parameters

The carrying capacity was estimated to be 30,000 individuals. The area modeled (the Jutland peninsula) is approximately 30,000 km2. It has been estimated that the population density of mink in Denmark is 1 to 2 mink per km2 (Hammershøj et al., 2006).

The size of the feral population in Denmark is not known. Whether or not the feral mink living on the Jutland peninsula belong to 1 population or different distinct populations is not known. In this study the feral population will be considered as 1 single population. Mink escape rates in Denmark have been estimated to 2% of the farm population with a mortality rate in the first 3 mo of 75% (Hammershøj et al., 2006). Basic simulations were made for initial population sizes of 10, 20, 50, and 100 individuals. In this study extinction was defined as when only 1 sex remained.

Both sexes of American mink reach maturity in the first spring after they are born (Dunstone, 1993). The female gives birth to an average of 4 to 6 kits per litter once a year (Hammershøj et al., 2006) and potentially for the rest of its life. Females in the breeding pool were set to be 100 and the males to be 33, to simulate the mink breeding strategy. Due to the territoriality of mink, the male only has limited access to females that it can defend throughout the mating period (Dunstone, 1993). The percentage of males in the breeding pool was estimated from the male–male and female–female territoriality and the breeding cycle and biology of the females.

The female mink is an induced ovulator (Sundqvist et al., 1988; Dunstone, 1993), which means that eggs are released as a reaction to the mating act. The male stays with the female until of the end of estrous period because the female can release eggs several times.

Every simulated population had a different fecundity calculated from the relatedness estimates of the F1 hybrid populations. It is assumed that lifespan in nature is around 5 yr (Bonesi et al., 2006). The maximum female reproduction age was therefore set to 5 yr. Death rates for the different age classes used in this study are based on estimates from Bonesi et al. (2006) and Hammershøj et al. (2006). The latter study estimated a rather high mortality rate, which was used in the sensitivity analysis of this investigation. The sex ratio was specified to be an equal number of males and females at birth and the age distribution was two-thirds 1 yr olds and one-third 2 yr olds, based on the typical distribution of sexes and ages on Danish mink farms (Hammershøj et al., 2006). The harvest and supplementation options were not used in basic simulations but were used in the sensitivity analysis, with yearly supplements of 10, 20, 30, or 40 individuals and harvesting of 4,000 individuals yearly from yr 50, 75, or 100 after founding, respectively.

Environmental Variation.

VORTEX provides an option of including density dependent reproduction in the simulation. But because this parameter has not been investigated for mink (to our knowledge) and there have been warnings about using a density dependent model in the literature (Boyce, 1992; Mills et al., 1996; Brook et al., 1997), no density dependent reproduction parameter was included. It was assumed that no catastrophes could affect the persistence of the population because the simulated area was so large and the population was assumed to be scattered around in this area, making catastrophes affecting the whole area unlikely. Variation in carrying capacity due to environmental variation (EV) was also impossible to estimate. Based on the literature the EV was set to be 10% of the carrying capacity, which is probably a rather conservative estimate (Song, 1996; Penn et al., 2000), but recommended choice by Miller and Lacy (2005).

Inbreeding and Genetic Management.

The impact of inbreeding depression on traits closely related to fitness is well documented both in captivity (Kristensen and Sørensen, 2005) and in the wild (Hedrick and Kalinowski, 2000). The level of relatedness and the inbreeding coefficient (FIS), which is defined as the probability that 2 alleles in a locus are identical by descent, are directly correlated, with the inbreeding coefficient in an individual being half of the relatedness between the parents (Queller and Goodnight, 1989). The inbreeding coefficients were calculated from 1 to observed heterogosity (HO; Lacy, 1993; Miller and Lacy, 2005). Both parameters would are useful to estimate the reproductive potential in mink populations, which affect population viability.

The default value in VORTEX of 3.14 lethal equivalents was used, which is also recommended by Miller and Lacy (2005). The starting allele frequencies were provided manually and the allele frequencies were specific for each F1 hybrid population.

Simulations

For all individual simulations we used 100 iterations, each spanning 200 yr (approximately 67 generations). We show the results of 3 major groups of simulations: 1) basic simulations in which only the number of founders is varied, 2) varying fecundity simulations to assess the impact of inbreeding as relatedness changes during the simulation of the population, and 3) sensitivity simulations where the effect of 1 parameter on population fitness is assessed in turn by keeping all other parameters constant.

The basic set of simulations was performed for each of 15 hybrid populations with starting population sizes of 10, 20, 50, and 100, respectively. The basic parameter settings used are shown in Table 1. The 15 selected F1 hybrid populations (P1 to P15) with varying relatedness estimates are shown in Table 2. The fecundity was set to be stationary, but it was estimated for every individual from the relatedness in the hybrid populations (see fecundity regression in Fig. 1a).


View Full Table | Close Full ViewTable 1.

Basic input parameters used for VORTEX simulations of the viability in 15 F1 hybrid populations

 
Variable
No. of years 200
No. of iterations 100
Extinction definitions Only 1 sex remains
Lethal equivalents 3.14
% due to recessive lethals 100 (due to limitations in VORTEX)
Reproductive system Polygynous
Age of first offspring 1
Maximum breeding age 5
Sex ratio 50:50
% adult females breeding 100%
Mortality rates 0 to 1 yr 78%
Mortality rates after 1 yr 50%
% males in breeding pool 33.3%
Density dependent reproduction No
Specified age distribution Yes
Initial population size 10 (20, 50, 100)
Carrying capacity 30,000 (±3,000)
Harvest 0
Supplemented 0

View Full Table | Close Full ViewTable 2.

Results from simulations of the 15 F1 hybrid populations using basic parameters (Table 1) and with an initial population size of 10 individuals1

 
Population Hybrid Fecundity Relatedness det-r Stoc-r (SD) PE n (SD) HE (SD) HO (SD) Fin MedianTE MeanTE
P1 5+8 (b+b) 4.964 0.344 0.031 –0.03 (0.255) 0.96 415.2 (2,581.97) 0.1046 (0.0981) 0.1044 (0.0981) 0.8956 9 13.7
P2 10+11 (w+w) 4.995 0.335 0.034 –0.002 (0.21) 0.89 2,157.22 (6,600) 0.1039 (0.0679) 0.1037 (0.0679) 0.8963 11 14.6
P3 25+29 (W+W) 5.181 0.278 0.055 0.012 (0.198) 0.88 3,293.2 (9,200.75) 0.206 (0.204) 0.1504 (0.1) 0.8496 9 12
P4 13+14 (w+w) 5.467 0.19 0.085 0.053 (0.171) 0.78 6,429.8 (12,239.43) 0.223 (0.1575) 0.191 (0.0861) 0.809 14 13.4
P5 3+5 (b+b) 5.494 0.182 0.088 0.062 (0.154) 0.74 7,526.16 (12,848.91) 0.2513 (0.1577) 0.2221 (0.1097) 0.7779 13 12.4
P6 4+8 (b+b) 5.885 0.073 0.125 0.102 (0.132) 0.6 12,095.91 (14,975.77) 0.305 (0.1988) 0.2363 (0.1186) 0.7637 15 10.7
P7 22+23 (W+W) 5.982 0.033 0.138 0.119 (0.11) 0.49 14,813.54 (14,727.54) 0.338 (0.1186) 0.3305 (0.1118) 0.6695 0 11.2
P8 1+11 (b+w) 5.985 0.032 0.139 0.116 (0.129) 0.6 11,747.58 (14,553.76) 0.3113 (0.1544) 0.2804 (0.1206) 0.7196 15 10.8
P9 23+26 (W+W) 6.004 0.026 0.14 0.123 (0.11) 0.44 16,316.39 (14,706.89) 0.355 (0.1344) 0.3279 (0.1117) 0.6721 0 13.1
P10 9+33 (w+p) 6.043 0.014 0.144 0.122 (0.121) 0.5 14,717.02 (14,904.45) 0.3623 (0.1477) 0.3091 (0.1027) 0.6909 44 13.6
P11 8+21 (b+W) 6.055 0.011 0.145 0.127 (0.107) 0.4 18,004.41 (14,898.6) 0.3063 (0.1431) 0.2709 (0.1007) 0.7291 0 14.8
P12 28+32 (W+m) 6.078 0.004 0.147 0.129 (0.109) 0.45 16,514.75 (15,139.93) 0.3858 (0.1516) 0.3283 (0.1056) 0.6717 0 12.3
P13 5+12 (b+w) 6.08 0.002 0.147 0.127 (0.116) 0.5 14,786.55 (14,996.93) 0.3395 (0.1425) 0.2959 (0.1039) 0.7041 44 13.5
P14 6+31 (b+s) 6.087 0.001 0.148 0.129 (0.114) 0.49 15,140.18 (15,042.41) 0.3345 (0.1302) 0.3271 (0.1208) 0.6729 0 11.4
P15 8+16 (b+w) 6.318 –0.07 0.17 0.154 (0.105) 0.38 18,321.92 (14,568.72) 0.3447 (0.109) 0.3394 (0.1025) 0.6606 0 11.7
1The individual hybrid populations are specified (Hybrid) by the parental population number and color morph (b = black, w = white, W = wildtype, p = pearl, s = silver, and m = mahogany) and the fecundity and initial relatedness are specified. Deterministic growth rate (det-r), mean stochastic growth rate (Stoc-r), probability of extinction (PE), mean final population size (n), mean final expected heterozygosity (HE), observed heterozygosity (HO). mean final inbreeding coefficient (Fin), median (MedianTE), and mean (MeanTE) time to extinction. Standard deviations are shown in parentheses.
Figure 1.
Figure 1.

a) Fecundity regression for the 33 farm mink populations simulated and b) the within-population (Pop) pairwise relatedness. U = upper limit of the 95% confidence interval. L = lower limit of the 95% confidence interval.

 

For the varying fecundity simulations incorporating effects of inbreeding, fecundity was modeled as a variable depending on the changes in relatedness during the 200 yr the simulations spanned, using the fecundity regression as the fecundity option.

In the sensitivity simulations all parameters were kept constant at the basic parameter settings (Table 1) except the one in focus (Jørgensen and Fath, 2011). The initial population size was 10. In addition to varying in turn some of the parameters included in the former simulations, we also analyzed the harvesting and yearly supplement the parameters (Table 3). Three representative populations were selected for these simulations (P2, P5, and P15 in Table 2; see below). A sensitivity analysis was performed to investigate which parameters were most sensitive to the survival of the populations. To compare the sensitivities of different parameters a sensitivity index was calculated. The sensitivity index of a parameter (Sx) is defined asin which ΔX is the change in the observed response variable (e.g., stochastic growth rate or final population size) and param is the examined parameter (e.g., fecundity or mortality rate).


View Full Table | Close Full ViewTable 3.

Results of the sensitivity analysis for 3 (low, medium and high initial relatedness) of the 15 F1 hybrid populations1

 
Sensitivity testing index 8.21 Stoc-r N 3.5 Stoc-r N 10.11 Stoc-r N
Juvenile mortality –10% 10.97 #5.77 30.48 22.87 nc 2.72
juvenile mortality –5% 12.60 8.54 34.19 34.29 nc 3.22
Juvenile mortality +5% 15.97 13.84 46.45 E nc E
Juvenile mortality +10% 17.99 E 37.9 E nc E
Adult mortality –10% 2.66 2.4 7.26 6.73 nc 0.78
Adult mortality –5% 2.99 4.79 9.35 15.9 nc 0.99
Adult mortality +5% 3.38 5.87 14.19 11.24 nc 0.6
Adult mortality +10% 2.92 3.22 6.94 4.09 nc E
Fecundity –10% 4.16 3.13 14.52 9.44 nc E
Fecundity –5% 4.29 1.96 20.97 14.41 nc 0.6
Fecundity +5% 3.77 2.93 11.29 12.73 nc 0.71
Fecundity +10% 3.77 2.21 9.35 8.06 nc 0.98
Lethal Equivalents –10% 0.19 0.89 0.32 2.37 nc 0.14
Lethal Equivalents –5% 0.26 0.55 0.32 3.58 nc 0.45
Lethal Equivalents +5% 0.26 2.33 3.55 5.4 nc 0.21
Lethal Equivalents +10% 0.06 0.17 0.65 1.42 nc 0.14
Yearly supplement 10 individuals 0.01 #0.06 0.05 #0.29 nc 0.04
Yearly supplement 20 individuals 0.01 #0.03 0.03 #0.14 nc 0.02
Yearly supplement 30 individuals 0.01 #0.02 0.02 #0.1 nc 0.01
Yearly supplement 40 individuals 0.00 #0.02 0.02 #0.07 nc 0.01
Harvest from yr 50 (4,000) A A a a a a
Harvest from yr 75 (4,000) A A a a a a
Harvest from yr 100 (4,000) A A a a a a
Initial population size = 20 0.00 0.2 0.06 0.72 nc 0
Initial population size = 50 0.01 #0.12 0.05 #0.55 nc 0.01
Initial population size = 100 0.01 #0.06 0.03 #0.29 nc 0.01
1For explanation of the calculation of the sensitivity index, see the Materials and Methods section. High index indicates that the parameter is sensitive to the viability. E = all population went extinct; a = could not be calculated due to complex parameters (see Fig. 4a to c); # = population reached carrying capacity; nc = could not be calculated. Stoc-r = stochastic growth rate; N = population size.


RESULTS

Genetic Analysis

Within the 33 populations the relatedness ranged from 0.026 to 0.513 (Fig. 1b) and the degree of genetic differentiation (FST) values within strains from 0.02 to 0.29 (Demontis et al., 2011). Fecundity ranged between 4.3 to 6.8 kits per litter. The correlation of fecundity with relatedness was r = –3.27R + 6.09 (Fig. 1a). The pairwise relatedness between populations ranged from 0.174 to 0.401 and FST from 0.00069 to 0.268. The fecundity in the F1 hybrid populations ranged from 4.778 to 6.657.

Simulations

In simulations with high mortality rates (juveniles 80% and adults 70%; Hammershøj et al., 2006) all populations went extinct (results not shown). The mortality rates were therefore decreased to the lower values seen in Table 1 for the basic and varying fecundity simulations and matching observed in other feral populations in Europe (Bonesi et al., 2006). The actual high Danish estimates of mortality were used in the sensitivity simulations. Simulations using the basic parameters (Table 1) and the allele frequencies from the hybrid populations but with fixed fecundity of 5 kits per litter showed no significant differences between populations (results not shown).

The results of the simulations with basic parameters for the 15 F1 hybrid populations (Table 1) showed that populations with low fecundity had small final population sizes (n < 15,000 for all starting population sizes). Carrying capacity was never reached in any of the populations with initial population sizes of 10 and 20 (Fig. 2c). For initial population sizes of 50 and 100, the final population size reached the carrying capacity for fecundities >5.5 (Fig. 2c).

Figure 2.
Figure 2.

a) Mean final population size and probability of extinction (PE), b) expected and observed heterozygosity (HE and HO, respectively) and inbreeding coefficient (FIS), and c) mean final population size with initial population size of 10, 20, 50, or 100 individuals for all 15 F1 hybrid populations.

 

The probability of extinction varied considerably with the fecundity in the simulated populations. In the simulations using basic parameters and fixed fecundity (calculated from the F1 hybrid relatedness), the probability of extinction ranged from 38 to 96% and decreased with increasing fecundity (Fig. 2a). All populations with an initial relatedness of 0.073 or more had a probability of more than 60% of extinction. The population with the greatest initial relatedness and accompanying low fecundity (P1; Table 1) had a probability of 96% of going extinct and in the last 4% population sizes did not exceed 500. Mean final HE and HO increased with increasing fecundity (Fig. 2b). The mean final inbreeding coefficients for the populations (basic parameters) ranged from 66 to 89%, with the smallest observed for populations with the lowest relatedness and greatest fecundity and vice versa (Fig. 2b). All hybrid populations lost more than 38% of their initial level of genetic variability during the 200 yr. The loss of genetic variability was calculated from the difference between the heterozygosity at yr 1 and yr 200. All populations with an initial population size of 20 lost at least 27% of their initial level of genetic variability. Seven of the 15 populations (fecundity > 6; relatedness < 0.03) with initial population size of 50 and 11 of the 15 populations (fecundity > 5.49; relatedness < 0.19) with initial population size of 100 lost less than 10% of their initial level of genetic variability.

Three representative populations (P2, P5, and P15) characterized by low, medium, and high fecundity were selected for detailed descriptions of the temporal patterns of the parameter dynamics (basic parameters, initial population size = 10). The population with low fecundity (P2; fecundity = 4.995), started a slow expansion after 120 yr and never reached a maximum (Fig. 3a). The population with medium fecundity (P5; fecundity = 5.494) had a slower population expansion, starting after 50 yr. In the next 50 yr it reached a maximum of about 8,000 individuals and remained at that level (Fig. 3a). The highly fecund population (fecundity = 6.318; P15) had a population expansion after about 30 yr. After roughly 20 yr it reached a maximum of 20,000 individuals and remained at that size (Fig. 3a). The stochastic growth rate (Fig. 3b) was low and sometimes negative for P2, medium for P5, and much greater for P15. All 3 populations had a drastic decrease of HE and HO (Fig. 3c and 3d) during the first 20 yr. However, after 20 yr it remained relatively steady during the rest of the simulation with final values of 0.1 (both HE and HO) for P2, 0.25 (HE) and 0.22 (HO) for P5, and 0.34 (both HE and HO) for P15 (Table 2).

Figure 3.
Figure 3.

a) Population size (n), b) stochastic growth rate (r), c) expected heterozygosity (HE), and d) observed heterozygosity (HO) for 3 (high, medium, and low relatedness) of the 15 F1 hybrid populations during the 200 yr.

 

In the simulations with basic parameters and fecundity varying over time due to changes in relatedness all 15 populations crashed, with mean time to extinction ranging between 8.5 to 10.8 yr (results not shown).

The sensitivity simulations of the same 3 selected populations mentioned above showed that the most vulnerable parameters were those associated with juvenile mortality (Table 3). Harvest of individuals (4,000 per yr from yr 50, 75 or 100) caused P2 and P5 (Fig. 4a to 4b) to go extinct even when the first year of harvest was at yr 100. Simulation (Fig. 4c) decreased in population size but did not go extinct. The somatotropin index of harvest from populations could not be calculated due to the complexity of the parameter (but see Fig. 4a to 4c).

Figure 4.
Figure 4.

Population size during the 200 yr with different initial year of harvesting for 3 of the 15 F1 hybrid populations for a) high, b) medium, and c) low relatedness.

 

When supplemented yearly all 3 populations reached carrying capacity and had greater mean stochastic growth rate and greater HE and HO compared with the basic simulations (Fig. 5a to 5d). The yearly supplement of 5, 10, 20, 30, or 40 unrelated individuals resulted in a final genetic diversity 7 to 11% greater than the initial genetic diversity (results not shown).

Figure 5.
Figure 5.

a) Mean final population (pop.) size (n), b) stochastic growth rate (r), c) expected heterozygosity (HE), and d) observed heterozygosity (HO) for 3 (high, medium, and low relatedness) of the 15 F1 hybrid populations with yearly supplements of 0, 10, 20, 30, or 40 unrelated mink.

 

Changes in the amount of lethal equivalents had no significant effect on the population viability (t test for stochastic growth rate and final population size could not reject the HO hypothesis of no difference for either of the F1 hybrid populations).


DISCUSSION

Denmark has the greatest production of mink pelts in the world, with about 1,500 mink farms and an average of 2,000 mink per farm (Hammershøj et al., 2006). Between 5,000 and 6,000 mink are caught every year in Denmark, and this number has been relative stable during the last decade (Asferg, 2010). In Denmark, however, it is questioned whether or not there is a true feral population because in an area with a high number of mink farms about 80% of the wild mink are recent farm escapees (Hammershøj et al., 2005). This leads to the view that no true feral population exists in Denmark. However, the closure of mink farms in some European countries has not caused a decline of the feral populations in those countries (Bonesi and Palazon, 2007). Therefore the mink populations in those countries must be true feral populations, which do not rely on continuous immigrants from the farms.

Populations are considered genetically viable if they maintain ≥90% of their initial HE (Foose, 1993) and critically endangered if they lose ≥80% of the NE over a time period of 3 generations (IUCN, 2001). None of the F1 hybrid populations simulated from basic parameters reached carrying capacity within the simulated 200 yr. The probability of extinction varied considerably depending on the fecundity with the greatest risk of extinction for populations with low fecundity. For all 15 F1 hybrid populations simulated with basic parameters, the population was not reduced to a critical level during the 200 yr. The probability of extinction was above 80% in 3 of the F1 hybrid populations simulated with basic parameters. These 3 populations had the lowest fecundities (fecundity = 4.964, 4.995, and 5.181) and greatest level of relatedness (relatedness = 0.344, 0.335, and 0.278). It is interesting that the F1 hybrid populations with the greatest probability of extinction were the ones with the fecundity closest to the mean fecundity estimated for wild populations (Hammershøj et al., 2007).

All populations simulated using basic parameters and an initial population size of 10 lost more than 38% of their genetic variability during the 200 yr. All populations with an initial population size of 20 lost at least 27% of the genetic variability. Seven of the 15 populations (fecundity > 6; relatedness < 0.03) with initial population size of 50 and 11 of the 15 populations (fecundity > 5.49; relatedness < 0.19) with initial population size of 100 could be considered genetically viable. This reflects the importance of high fecundity and low relatedness estimates for establishment of persistent populations when the number of founders is low. When populations were supplemented each year with 10, 20, 30, or 40 unrelated individuals, the final genetic diversity was 7 to 11% greater than the initial genetic diversity. There was only about 1% difference in the final genetic diversity between supplements of 10 and 20, 20, and 30, and 30 and 40. Based on these findings, it seems less important how many individuals are supplemented each year, with respect to the effect on the genetic diversity. As long as there was a yearly supplement, all populations remained genetically viable and all populations reached carrying capacity within the 200 yr. This is in line with Mills and Allendorf (1996) stating that 1 migrant per generation is considered enough to sustain genetic variability in a population.

The sensitivity simulations with harvesting incorporated (mink removal and culling) showed that with the number of mink caught every year in Denmark, the populations in the wild should decline or even go extinct. However, in Denmark no decline of the wild population has been observed, which could be due to the continuous yearly addition of farm escapees. A Danish study on mink caught outside of farms showed that about 4 out of 5 mink were born on farms and 60% had lived in the wild for less than 2 mo (Hammershøj et al., 2005). Even though the mink farming industry in Denmark is highly regulated, farms may act as a source for the mink population found in the Danish nature. Therefore there is possibly no true feral population in Denmark.

The sensitivity analysis further showed that the survival of populations is highly sensitive to juvenile mortality followed by fecundity and adult mortality. For simulations using basic parameters, but changing the initial population size, it was evident that the size of the founding population is also an important factor. In the basic scenario with a population of 10 or 20 founding individuals, no population reached the carrying capacity. On the other hand only 4 of the populations (P1 to P4) with 50 founding individuals and 2 (P1 to P2) with 100 founding individuals did not reach the carrying capacity. This indicates that the number of founders is a determining factor in the success of an introduced species. The results of the yearly supplement scenarios also indicate that the number of breeding individuals is a determining factor.

This is seen in simulations with low initial population size and low fecundity having a high probability of extinction. The fecundity found in wild mink is less than the 1 observed for farm mink (Hammershøj et al., 2007), which can be explained by the smaller body size in wild mink, early kit mortality in the wild, and scarcity of food leading to bad nutritional conditions for the females. If on average there are 5 kits per litter in feral mink, the model with harvesting of 4,000 individuals every year would lead to a population crash. This is, however, not seen in Denmark. The harvest model does not take into account that when there is a reduction in population size, there would also be a reduction in the number of caught mink. In reality this would cause a longer decreasing period and not the sudden population crash shortly after the harvest begins.

The American mink is an invasive species and is regulated, but it is traditionally thought impossible to eradicate completely (Bonesi and Palazon, 2007; Hammershøj et al., 2007). Free-ranging American mink have the potential of harming the native wildlife both due to competition with the native species for resources and due to preying on animals such as birds and small mammals, which do not have mink as a natural enemy (Hammershøj et al., 2005; Bifolchi et al., 2010; Brzezinski et al., 2012). Such an effect has not yet been observed in Denmark (Hammershøj et al., 2004). Mink in Denmark prey mostly on mammals and birds but also on amphibians and fish. This does not indicate competition for food with native Danish mustelids (Hammershøj et al., 2004). Feral mink can also cause economic damage to animal breeders when preying on domestic poultry and fish and should be controlled. Due to the mating strategy of mink, it might be possible to reduce the number of offspring. As mentioned in the introduction mink are induced ovulators (Sundqvist et al., 1988; Dunstone, 1993) releasing the eggs as a result of the act of mating. If a number of sterilized males are released just before the mating period, the number of fertilized eggs would be reduced and the number of offspring would decrease. This strategy is for example used in insect pest control (Klassen and Curtis, 2005). Combined with an organized culling in late summer or early autumn, when the juveniles leave the maternal territories, this could potentially reduce the number of feral mink in the Danish nature. Due to the relatively high costs of the above actions, priority should, however, be given to reduce the number of escapees. Our results indicate that the feral mink population in Denmark is sustained due to supplements from mink farms, but the closure of mink farms apparently does not always lead to declining numbers of free-living mink (Bonesi and Palazon, 2007). Farm mink are poorly adapted to the wild conditions and escapees might be the reason why there is no true feral mink population in Denmark, because the continuous flow of maladapted genotypes hinders adaptation.

 

References

Footnotes


Comments
Be the first to comment.



Please log in to post a comment.
*Society members, certified professionals, and authors are permitted to comment.