Genetic population assignment reveals a long-distance incursion to an island by a stoat (Mustela erminea)
When new individuals from a pest species are detected after an eradication programme, it is important to determine if these individuals are survivors from the eradication attempt or reinvaders from another population, as this enables managers to adjust and improve the methodologies for future eradications and biosecurity. Rangitoto/Motutapu Islands in the Hauraki Gulf (New Zealand) had a multispecies mammalian pest eradication conducted in 2009. A year after this eradication a single stoat was trapped on the island. Using genetic population assignment we conclude that this individual was a reinvader, which probably swam a minimum distance of 3 km from the adjacent mainland. This swimming distance is greater than any previously known stoat incursions. Our results suggest that the original population on these islands was from natural dispersal rather than anthropogenic introduction and that it had some limited ongoing mixing with the mainland population. These findings highlight the invasion/reinvasion potential of stoats across large stretches of water, and will necessitate ongoing biosecurity indefinitely for these islands. The study also highlights the utility of genetic assignment techniques for assessing reinvasion, and emphasizes the need for pre-eradication genetic sampling of all pest species to enable such analyses to be carried out.
KeywordsConservation Eradication Island Microsatellite New Zealand Reinvasion
Eradicating invasive species and preventing reinvasion is a common goal for ecological restoration programs and this is often easier to achieve on islands because of their isolation (Courchamp et al. 2003; Towns and Broome 2003; Clout and Russell 2008). When pests reappear after an eradication attempt, determining whether they are survivors or re-invaders can have important implications for management: either quarantine or eradication procedures may need to be modified. When there is a long time between eradication and the reappearance of pests it can usually be assumed that the new individuals are re-invaders, but when pests reappear soon after “eradication” it may not be clear whether they are re-invaders or survivors of a failed eradication attempt (Abdelkrim et al. 2005; Abdelkrim et al. 2007; Russell et al. 2008).
Genetic techniques can provide vital information on the origin of invasive species populations, the underlying genetic population structure, and the levels of dispersal within and between populations (Robertson and Gemmell 2004; Rollins et al. 2006; Schwartz et al. 2007). Genetic population assignment provides a method for determining an individual’s source population, and is now a widely used tool in ecology (Davies et al. 1999). These assignment methods compute the probability of finding the genotype of a given individual within each of a sampled set of populations, and can be used to select or exclude populations as possible sources for the individual (Manel et al. 2005). In situations where pests have been detected post-eradication, these genetic techniques have been successfully used to determine the origin of these individuals (Kim et al. 2010; Abdelkrim et al. 2007; Russell et al. 2010).
Stoats (Mustela erminea) were introduced to New Zealand in 1884 (Thomson 1922) to control rabbits (Oryctolagus cuniculus). Since then stoats have had well-documented catastrophic effects on native New Zealand birds (King and Powell 2007). They are the primary nest predator in both alpine grasslands and beech forests (Smith et al. 2008) and they are a primary agent of decline for over half of all forest birds of New Zealand currently threatened or significantly declining (Innes et al. 2010). They pose one of the highest invasion risks to islands (Elliott et al. 2010; Veale et al. in press). There are numerous eye witness accounts of stoats swimming hundreds of meters from shore (Wodzicki and Bull 1951; Fitzgerald 1978; King and Moors 1979) and swimming is believed to be their primary island invasion pathway (Taylor and Tilley 1984; Veale et al. in press). The best-documented series of invasions have been on Maud Island 900 m offshore in the Marlborough Sounds where there have been three separate invasions by pregnant female stoats since 1982, resulting in a total of 18 stoats caught on the island. The first of these invasions extirpated the South Island Saddleback (Philesturnus caruncalatus) population on the island (King and Powell 2007). Due to the dramatic effects of stoat invasion, maintaining the stoat free status of islands is important for ecological restoration programs on New Zealand islands (Atkinson 2001; Parkes and Murphy 2003; Bellingham et al. 2010).
In July 2009 an attempt was made to remove the full suite of invasive mammals from the island including stoats. Standard aerial brodifacoum baiting methods were used for rodents, combined with subsequent trapping and shooting for predators that had not succumbed to secondary poisoning. After the poison operation, spotlighting, dog teams and trapping detected no live stoats, and only a single stoat carcass was found. No stoat sign was detected for the next 11 months, but in July 2010 a single adult male stoat was caught in a trap on Rangitoto almost a year after the poison drop. Given the dramatic potential effects of stoat reinvasion, it is important to determine if this stoat survived the poison drop and avoided trapping and detection for a year, or alternatively if it reinvaded from another source population.
Materials and methods
Stoat trapping was not undertaken on Rangitoto or Motutapu prior to the aerial poisoning operation as there was a perceived risk of creating trap-shy animals (Richard Griffiths pers. comm.). Non-invasive genetic sampling in the month leading up to the eradication obtained 260 hair samples from across Rangitoto/Motutapu using hair tubes (see Gleeson et al. 2010 for methods). Hair samples were sub-sampled for genotyping, selecting samples with large amounts of hair (to ensure ease of genotyping), and the greatest geographic spread (to maximize the number of stoats detected). Along with these hair samples, two pre-eradication tissue samples were obtained: one from a stoat caught incidentally in a rat trap a month before the poison bait drop, and another from the single stoat carcass found immediately after the poison operation.
There are two possible source populations for immigrant stoats to Rangitoto/Motutapu, the mainland and Waiheke Island (Fig. 1). Neighboring Rakino, Motukorea and Motuihe Islands are all stoat free, having had successful multispecies eradications. Stoats captured by local environmental groups from around the Auckland region were also obtained and frozen for genetic sampling. These trapping programs were located at Tawharanui, Ark in the Park, Kawakawa Bay and on Waiheke Island.
At least ten hairs per sample were isolated from each hair tube sample (samples were collected from distinct clumps to ensure only single individuals contributed). Tissue samples were dissected in the laboratory, where 50 mg of muscle tissue and caudal skin were removed. DNA was then isolated from both sample types, using a Bio-Rad AquaPure Genomic Tissue Kit (Cat# 732-6343) following the manufacturer’s protocol, and re-suspended in 100 μl of supplied buffer.
Variable microsatellite loci used for stoat genotyping
PCR amplifications were performed in 25-μl reactions containing 1 μl of DNA extract, 1× PCR buffer with MgCl2 (50 mM Tris/HCl, 10 mM KCl, 5 mM [NH4]2SO4, 2 mM MgCl2, pH 8.3), 200 μM of each dNTP, 10 μM of each primer, and 2 μl of FastStart Taq DNA Polymerase (Roche Diagnostics). Amplification conditions on a GeneAmp PCR System 9,700 thermocycler (Applied Biosystems) were: initial denaturation at 95°C for 4 min; 10 touchdown cycles of 20 s at 94°C, 20 s at 62–58°C, 20 s at 74°C; 40 cycles of 20 s at 94°C, 20 s at 58°C, 20 s at 74°C, and a final extension of 40 min at 72°C. The 5′-end of the forward primer of each pair was fluorescently labelled with either 6FAM, NED, or VIC dyes (Applied Biosystems) and amplification products were separated using capillary electrophoresis (ABI PRISM 310). Alleles were sized relative to an internal size standard (GS-350 ROX) using GENESCAN 3.1 (Applied Biosystems). The genotype profiles were analysed using the software GENOTYPER Version 2.5 (Applied Biosystems).
Genotyping for the hair samples was carried out using a step-wise protocol of exclusion that has been shown elsewhere to ensure rigorous and conservative determination of identity (Paetkau 2003; Weaver et al.2005). We required a perfect match between two PCR amplifications in order to accept that genotype in order to eliminate PCR errors resulting in either allelic drop-out or false alleles. Any samples that differed by one locus were checked for any potential errors as either scoring or amplification errors (Paetkau 2003). If these differences could not be explained by errors in scoring/typing, samples were then subjected to a further round of PCR and scoring (Poole et al. 2001; Mowat and Paetkau 2002). We used the software package Genalex 6.4 (Peakall and Smouse 2006) to estimate PID and PID-sib among full siblings as that provides an upper limit to the probability that pairs of individuals will share genotypes (Taberlet and Luikart 1999).
For the three populations we calculated allele frequencies, observed heterozygosities, and mean number of alleles per locus as summaries of genetic diversity. Our assignment methods rely on assumptions of Hardy–Weinberg proportions and no linkage disequilibrium, so we checked these in each population using Fisher’s exact tests. We also estimated FIS and FST for each population. All calculations were performed in Arlequin 3.5 (Excoffier and Lischer 2010).
For genetic assignment analysis, we used GeneClass2 (Piry et al. 2004), implementing the Bayesian assignment criterion of Rannala and Mountain (1997). This criterion was recommended by Cornuet et al. (1999) as the best of a set of assignment criteria, and they specifically demonstrated high assignment accuracy for samples as small as ours. For the post-eradication Rangitoto stoat, the criterion gives a posterior probability of finding this stoat’s genotype in the pre-eradication Rangitoto/Motutapu Island population (survivor hypothesis), and of finding the stoat’s genotype in either the mainland or Waiheke Island populations (reinvader hypotheses). We present the assignment results graphically following the methodology of Russell et al. (2010). The two-dimensional scatterplot created displays the clustering of samples, and the orders of magnitude spanning the posterior probabilities. This information is lost by compressing the two-dimensional points into the one one-dimensional score values provided by GeneClass2.
Genetic properties of the three sampled populations
No. of gene copies
No. of loci
No. of polymorphic loci
Mean no. of alleles/locus
F-statistics for the pre-eradication Rangitoto/Motutapu population and the two potential source populations
From the log posterior genotype probability plots, the post-eradication stoat caught on Rangitoto clearly grouped genetically with the mainland stoats, and did not group with the pre-eradication Rangitoto/Motutapu samples (Fig. 3) (Geneclass score >0.99). Indeed it is less similar to the pre-eradication samples than over half of the mainland samples, and had alleles not present in the pre-eradication population sample. It also clearly did not group with the Waiheke Island stoat population (Fig. 4), which was genetically distinct from both the pre-eradication Rangitoto/Motutapu population, and the mainland population.
Using genetic assignment methods we have demonstrated clear discrimination between mainland and island populations despite relatively small sample sizes. With this strong genetic signal, we were able to confidently assign the post-eradication Rangitoto stoat to the mainland population. This supports the hypothesis that this stoat arrived from the mainland, presumably by swimming a distance of at least 3 km. Our results show there is a risk of reinvasion of Rangitoto/Motutapu and many other islands around New Zealand that were previously thought to be outside the swimming range of stoats. Subsequent to this incursion on Rangitoto, stoats have been detected and caught on Kapiti Island (New Zealand), which is a distance of 5.2 km offshore. While it is not possible to know with certainty the invasion pathway, it is plausible that stoats arrived there by swimming, extending the risk to islands further offshore.
An alternative hypothesis as to how the reinvading stoat arrived on Rangitoto is that it was a stowaway. This hypothesis seems unlikely given that stoats have never been recorded on boats, nor have they been detected on islands that are clearly not reachable by swimming yet have high levels of boat activity, such as Great Barrier Island (15.7 km from the mainland). Stoats are good swimmers, and their distribution on islands is correlated only with two factors that affect swimming dispersal (distance offshore and island area), and is not correlated with any anthropogenic factors such as wharfs or human activity (Veale et al. in press).
While this apparent stoat re-invasion means it will be necessary to maintain ongoing surveillance, the fact that the stoat was caught suggests that the current trapping program is adequate. The stoat captured post eradication was caught in a coastal trap near a likely entry point, in a relatively sheltered area of Rangitoto, and near the closest point to the mainland. The fact that the stoat is likely to have come from a developed suburban area is somewhat surprising, although not unheard of. It has been previously found that domestic cats in suburban areas near forested parks in Auckland occasionally prey upon stoats (Gillies 1998), and stoats have been seen in the larger forested parks on Auckland’s North Shore (pers. comm. Mel Galbraith).
The genetic differentiation between the pre-eradication Rangitoto/Motutapu population and the mainland population suggests that these populations were relatively isolated before the eradication. Genetic population assignment would have been less successful if there had been more significant gene flow between them. This indicates either that incursions are relatively infrequent, or that incursions are rarely successful, with new individuals not likely to survive and reproduce because the island is already at carrying capacity (Granjon and Cheylan 1989). The genetic results do provide compelling evidence that the Rangitoto stoat population was founded by natural dispersal rather than a deliberate introduction, and that there has been some ongoing immigration from the mainland, given the relatively high genetic diversity observed. This supposition is supported by historical information, as during the period that stoats could have been introduced, the owners of Motutapu were introducing game birds, and rabbit levels were relatively low (Thomson 1922). Given these factors it seems highly unlikely that stoats would have been deliberately introduced.
Pre-eradication sampling was critical in enabling population assignment in this case. The hair tube sampling conducted immediately prior to the eradication was fortuitous as no pre-eradication sampling had been planned. We urge all pest managers undertaking eradication projects to include the collection of a spatially representative sample of all target species pre-eradication. The genetic lab-work need not be performed until there is a subsequent detection. The minor cost of pre-eradication sampling is outweighed by the potential cost of unfocused remedial action. In this case we have learnt that eradication of stoats from large islands using secondary poisoning via rodents is possible, but that ongoing trapping will be required to keep this particular island stoat-free.
We thank Rachel Fewster and Andrea Byrom for their assistance in the planning and analysis of this project. Thanks to Richard Griffiths (Department of Conservation) and his team on Rangitoto/Motutapu, and Ark in the Park, Weka Watch, Auckland Council and Rob Morton for stoat samples. Also thanks to the hair tube team: Adam Field, Amy Holliday, Anya Elgort, Chloe McLaren, Daniel Laird, Josh Guilbert, Katrina Parry, Louis Scarrold, Oliver Hannaford, Owen Lee and Stan Woodhouse.
- Bellingham PJ, Towns DR, Cameron EK et al (2010) New Zealand island restoration: seabirds, predators, and the importance of history. N Z J Ecol 34:115–136Google Scholar
- Courchamp F, Chapuis JL, Pascal M (2003) Mammal invaders on islands: impact, control and control impact. Biol Rev 78:347–383Google Scholar
- Fitzgerald BM (1978) A proposal for biological control. In: Dingwall PR, Atkinson IAE, Hay C (eds) The ecology and control of rodents in New Zealand reserves. Information series no. 4, pp 223–230Google Scholar
- Gillies C (1998) Aspects of the ecology and management of small mammalian predators in northern New Zealand. Department of Biology, University of Auckland, AucklandGoogle Scholar
- Gleeson DM, Byrom AE, Howitt RLJ (2010) Non-invasive methods for genotyping of stoats (Mustela erminea) in New Zealand: potential for field applications. N Z J Ecol 34:356–359Google Scholar
- Granjon L, Cheylan G (1989) The fate of black rats (Rattus rattus) introduced on an island, as revealed by radio-tracking. Comptes Rendus Acad Sci Ser III Sci Vie-Life Sci 309:571–575Google Scholar
- Huang CC, Lin RC, Li SH et al (2007) Characterization of polymorphic tetranucleotide microsatellite loci from the Siberian weasel (Mustela sibirica). Mol Ecol notes 7:483–485Google Scholar
- Innes J, Kelly D, Overton J et al (2010) Predation and other factors currently limiting New Zealand forest birds. N Z J Ecol 34:86–114Google Scholar
- Mckinlay B (1997) The likelihood of arrival of stoats on islands. National predator management workshop St Arnaud, Nelson Lakes, pp 51–53Google Scholar
- Mowat G, Paetkau D (2002) Estimating marten Martes Americana population size using hair capture and genetic tagging. Wildl Biol 8:201–209Google Scholar
- Poole KG, Mowat G, Fear DA (2001) DNA-based population estimate for grizzly bears Ursus arctos in northeastern British Columbia, Canada. Wildlife Biology 7:105–115Google Scholar
- Rollins LA, Woolnough AP, Sherwin WB (2006) Population genetic tools for pest management: a review. Wildl Res 33:251–261Google Scholar
- Russell JC, Towns DR, Clout MN (2008) Review of rat invasion biology: implications for island biosecurity. Sci Conserv 286:1–53Google Scholar
- Taylor RH, Tilley JAV (1984) Stoats (Mustela erminea) on Adele and Fisherman Islands, Abel Tasman National Park, and other offshore islands in New Zealand. N Z J Ecol 7:139–145Google Scholar
- Veale AJ, Hannaford OD, Russell JC et al (in press) Modeling the distribution of stoats on New Zealand offshore islands. N Z J EcolGoogle Scholar
- Wodzicki KA, Bull PC (1951) The small mammals of the Caswell and George sounds area, pp 62–69. In: Poole AL (compiler) Preliminary reports of the New Zealand-American Fiordland expedition: expedition investigations in Fiordland New Zealand, in 1949. NZ Department of Scientific and Industrial Research Bulletin, vol 103, p 99Google Scholar