Molecular techniques, wildlife management and the importance of genetic population structure and dispersal: a case study with feral pigs

Authors


Peter Spencer, School of Veterinary and Biomedical Sciences, Murdoch University, Western Australia 6150, Australia (fax + 61 89310 4144; e-mail P.Spencer@murdoch.edu.au).

Summary

  • 1Understanding the spatial structure of populations is important in developing effective management strategies for feral and invasive species, such as feral pigs Sus scrofa. World-wide, feral pigs can act as ‘triple threat’ pests, impacting upon biodiversity, agricultural production and public health; in Australia they are a significant vertebrate pest. We utilized a molecular approach to investigate the structure of populations of feral pigs in south-western Australia. These approaches have been underutilized in pest management.
  • 2Using 14 highly polymorphic microsatellite markers from 276 adult pigs, we identified eight inferred (K = 8) pig populations that would be difficult to define with standard ecological techniques. All populations had moderate heterozygosity (HE = 0·680) and moderate to high levels of differentiation (FST = 0·118; RST = 0·132) between populations.
  • 3The molecular approach identified feral pig groups that appeared to be acting as a source for reinvasion following control operations. It also identified populations where current control measures were less successful in reducing ‘effective population size’. Additionally, the data indicated that dispersal rates between, but not within, the inferred feral pig populations were relatively low.
  • 4The potential for the spread of directly transmitted wildlife diseases between the pig populations studied was low. However, under some circumstances, such as within major river catchments, the role of feral pigs in the transmission of endemic or exotic diseases is likely to be high.
  • 5Synthesis and applications. A molecular-based approach allowed us to determine the genetic structure and dispersal patterns of a cryptic, destructive and invasive vertebrate pest. Our results indicated that the feral pig populations studied were unlikely to be acting as closed populations and, importantly, it identified where movement between groups was likely to occur. This should lead to more informed decisions for managing the potential risk posed by feral species, such as pigs, in the transmission of wildlife diseases. The suggested technique could help in understanding the dynamics of many other free-ranging pest animal populations.

Introduction

Biological invasions by non-native species constitute one of the leading threats to agriculture, human health and ecosystem sustainability. The effects of invasive species are numerous and often irreversible (Edwards et al. 2004). Understanding the spatial structure, dispersal and population genetics of the species of concern is important if the impact of the species is to be reduced or reversed. This is particularly so if effective control programmes for feral or invasive species are to be developed, and if risk analyses are to be informative and reliable (Edwards et al. 2004). New approaches, using contemporary molecular techniques in conjunction with demographic data, can be extremely useful for improving the understanding of the dynamics, population structure and social biology of many invasive species (Taylor et al. 2000). All these parameters are important to quantify because the effective management of any species requires at least some basic understanding of their dynamics (Gabor et al. 1999; Dexter 2003; Zenger, Eldridge & Cooper 2003).

World-wide, but particularly in Australia, feral pigs Sus scrofa (Linnaeus 1758) can pose a significant threat to biodiversity, agriculture and public health (Choquenot, McIlroy & Korn 1996; Long 2003; Vernesi et al. 2003). In Australia, feral pigs are estimated, conservatively, to cause losses to agriculture costing $100 million annually. Feral pigs can also act as reservoirs or transmission ‘vectors’ for a number of endemic and exotic diseases capable of affecting livestock, wildlife and people. Endemic diseases transmitted by feral pigs in Australia include leptospirosis and brucellosis; they are also capable of transmitting exotic diseases such as foot-and-mouth disease and Japanese encephalitis (Choquenot, McIlroy & Korn 1996; Dexter 2003; Caley & Hone 2004). Feral pigs are now found across approximately 40% of Australia because of natural expansion, intentional release and escapes or abandonment. In the 1990s, it was estimated that there were 13–23 million feral pigs in Australia (Hone 1990) but their range is still increasing in many regions (Twigg 2003). Densities can range from 0·1 to > 20 pigs km−2 (Choquenot, McIlroy & Korn 1996). South-western Australia has been described as a ‘biological hotspot’, one of the 10 most biologically diverse regions world-wide (Myers et al. 2000), and feral pigs are perceived as a major threat to biodiversity and agricultural values.

Information on the genetic structure of free-ranging feral pig populations is largely limited to two main studies, one in the USA (Gabor et al. 1999) and the other in Europe (Vernesi et al. 2003). Very little is known about the genetic structure of feral pig populations in Australia. We used a relatively new Bayesian clustering (assignment) procedure (Pritchard, Stefens & Donnelly 2000) to interpret the genetic profiles of pigs in order to identify the inferred population structure from the observed genotypes. This approach has a number of advantages over the use of traditional single divergence measures that compare allele frequencies (Sunnucks 2000; Manel, Berthier & Luikart 2002). First, Bayesian clustering is based on individual multilocus genotypes and requires no assumed prior knowledge of the population structure. Secondly, under the assumption that an individual's genotype comes from unknown groups (i.e. the inferred populations) at Hardy–Weinberg equilibrium with all loci in linkage equilibrium, such clustering models reconstruct the number and genetic composition of these groups (Pritchard, Stefens & Donnelly 2000). They also account for differences in diversity between candidate populations and the sampling error associated with the observed allele frequencies. Thirdly, the Bayesian clustering procedure of Pritchard, Stefens & Donnelly (2000) avoids any bias introduced by null allele frequencies, and is therefore considered the most appropriate approach for analysing recently established invading populations (Davies, Villablanca & Roderick 1999).

Additionally, we used a range of descriptive genetic measures to assess the effectiveness of past control operations of feral pigs by identifying any genetic bottlenecks within the pig populations. The presence or absence of genetic bottlenecks can be used to infer where control operations have been effective or not. We combined the available genetic information with our basic demographic data and used the ‘model’ outcomes as a basis for providing recommendations for more effective control strategies for feral pigs in south-western Australia. We also discussed the importance of such an approach for assessing the role, potential or otherwise, that feral pigs may play in the transmission of endemic or exotic diseases. Finally, we commented on the utility of our suggested approach in helping to unravel the dynamics of other free-ranging animal populations.

Materials and methods

tissue sampling and demographic data

Tissue samples were collected from feral pigs at 18 sampling locations in the south-west corner of Australia between May 2001 and July 2003 (Fig. 1). These samples were collected during routine control operations by local pest animal management agencies and private landowners. Sampling involved the collection of a small tissue sample (skin, muscle or liver stored in a solution of 20% dimethyl sulphoxide, saturated with sodium chloride) for molecular analysis. We also collated basic demographic (age, sex, weight, date of collection) and location information for each animal.

Figure 1.

Map of the study site, showing the six south-west populations (b) containing the 18 south-west sample groups, including Northampton. Sample group numbers correspond to Table 2. The three drinking water catchment rivers analysed for internal genetic structure are also shown (c). Locations of pigs from domestic (commercial) piggeries are not included.

MOLECULAR ANALYSES

Complete genotypes were determined for 247 adult feral pigs from south-west Western Australia, 20 adult feral pigs from Northampton in mid-west Western Australia (Fig. 1) and nine adult pigs (Landrace and Large white breeds) from local commercial piggeries. Fourteen polymorphic microsatellite loci that had previously been shown to be highly informative for S. scrofa (Alexander, Rohrer & Beattie 1996) were used (Table 1). These microsatellites have been extensively characterized in pigs (Martinez et al. 2000; Lowden et al. 2002; Vernesi et al. 2003) and were designed as part of the Pig Genome Mapping Project. These loci were chosen from recommendations for pig biodiversity studies, as all loci are highly polymorphic, not linked and there was no indication of null alleles (Alexander, Rohrer & Beattie 1996). When markers occurred on the same chromosome they were chosen with a minimal distance of 30 centimorgans (cM).

Table 1.  Summary statistics for 276 adult individuals at each of the 14 microsatellite loci from feral pigs in the south-west of Australia (i.e. from all sampling groups). Data include the observed number of alleles (NA; sensuKimura & Crow 1964), observed (HO) and Nei's (1978) unbiased (HE) estimate of heterozygosity. HE values closer to unity indicate a high level of heterozygosity
Genetic markerNAHOHE
SW936 60·6470·689
S0026 50·3670·585
SW240 80·7560·833
SW951 40·3930·504
S0155 50·5600·683
SW632 90·5950·660
S0002110·5090·605
S0068120·6020·757
SW122 70·6620·783
SW911 70·5130·639
S0005170·7100·794
S0090 70·5240·649
SW857 80·5890·683
S0226 80·6260·655
Mean (± SD)8·14 ± 3·370·575 ± 0·1090·680 ± 0·089

The 14 microsatellite loci were amplified using four multiplex polymerase chain reactions (PCR) per individual. Reactions were carried out in a 10-µL volume under the following conditions: approximately 50 ng template DNA, 10 ρmoles of each primer (labelled with HEX, FAM and TET fluorescent dyes), 0·6 U of Taq polymerase, 0·2 mm of each dNTP, 1 × reaction buffer, 0·1 mm bovine serum albumin and 1·5 mm MgCl2. PCR conditions included an initial denaturation for 5 min followed by 35 cycles at 94 °C for 45 s, 55 °C for 30 s and 72 °C for 45 s. This was followed by an extension step at 72 °C for 10 min. Multiplex reactions were loaded together. DNA fragments were separated on a 5% polyacrylamide gel using an ABI 377 automatic sequencer (Applied Biosystems, Melbourne, Australia). Size was determined by co-running a size standard (TAMRA-350; Applied Biosystems, Melbourne, Australia). DNA fragments were scored manually, with the aid of Genescan (Applied Biosystems). Descriptive statistics (number of alleles, observed and expected heterozygosities) were also generated using popgene (version 1.31; available from http://www.ualberta.ca/~fyeh/).

population structure

Assignment tests were used to identify genetic structure and to assign individuals to their likely population of origin using structure (version 1·0; Pritchard, Stefens & Donnelly 2000). structure uses a full Bayesian assignment approach to determine: (i) the most likely number of inferred populations (K), based on the observed genotypes; (ii) the proportion of each predefined sample group contributed by each inferred population; and (iii) the proportion of each individual animal's genotype contributed by each inferred population. A predefined sample group is a discrete geographical area from which samples were obtained (Fig. 1), and an inferred population is a collection of sample groups that clustered together from the assignment results. The results generated were based on simulations from one to 10 (K = 1–10) inferred populations, using a burn-in period of 50 000 iterations, with 106 iterations of a Markov chain Monte Carlo simulation.

genetic diversity within populations

Evidence of recent population bottlenecks was investigated by testing for an excess in heterozygosity following Luikart & Cornuet (1998), using the program bottleneck (Piry, Luikart & Cornuet 1999). Due to the relatively small number of loci analysed (n = 14), a Wilcoxon sign-rank test was estimated. A mixed model of microsatellite mutation was assumed, with single-step mutations assumed to account for 90% of all mutation events and a variance among multiple steps of 12, as suggested by Piry, Luikart & Cornuet (1999) and Vernesi et al. (2003). The effective size for each population (Ne) was estimated from effective heterozygosity values, according to the methods of Ohta & Kimura (1973), for populations under mutation drift equilibrium. A generally accepted estimate of 1·0 × 10−3 was used for the mutation rate (µ) for the pure dinucleotide repeat microsatellites used in our study, as recommended by Weber & Wong (1993).

genetic diversity between populations

The hierarchical population structure was further defined by calculating two different estimators of genetic differentiation, FST (θ) and RST, using the program fstat (Goudet 2001). FST is considered a better estimator of migration rate (Nm) than RST for studies such as this, involving small sample sizes and low numbers of loci being scored (n < 20; Gaggiotti et al. 1999). However, both FST and RST values are commonly calculated as measures of genetic divergence when examining population structure (Vernesi et al. 2003). As such, we used only FST values for the calculation of Nm but calculated both estimators for the analysis of population structure. We also used FST and RST values to estimate the extent of gene flow and intrapopulation differences between upstream and downstream areas of three relatively long river systems, for which we had sufficient samples (n = 18–24 adult pigs). These three rivers supply major water catchment dams within the Perth Hills, Serpentine and Dandalup regions (Fig. 1c).

The effective number of migrants between any two inferred populations per generation (Nm) was estimated using FST values according to the methods of Cockerham & Weir (1993). An independent approach, using the ‘private allele’ method (Barton & Slatkin 1986), was also used to estimate migration rate, with genepop 3.3 (Raymond & Rousset 1995). All values are given as mean ± 1 SD.

Results

descriptive statistics

All the sample groups were polymorphic at all loci. All loci analysed were moderately variable, containing between four and 17 alleles per locus (8·14 ± 3·37), with heterozygosity (HE) ranging from 0·504 to 0·833 (0·680 ± 0·089; Table 1).

feral pig population structure

We detected eight discrete inferred populations (Fig. 1b and Table 2), which included individuals from the Northampton area and a small sample (n = 9) of local domestic pigs. Both the Northampton and domestic pig sample groups formed discrete inferred populations (Table 2), each consisting of between one and six geographically clustered sample groups (Fig. 1b). Each sample group, with the exception of Frankland, assigned to the inferred population contributed the highest percentage of their genotypes (Table 2). The Frankland sample group was assigned to the Muir inferred population rather than to the Collie inferred population because a large number of feral pigs in the Frankland group were suspected to have been illegally dumped into the Frankland area from the Collie area (Hampton 2003). Inferred populations will hereafter be referred to as populations.

Table 2.  Estimated genetic contribution (given as a proportion) of each of the eight inferred (K = 8) population clusters to each of the 20 sample groups, generated using structure (Pritchard, Stefens & Donnelly 2000). Values in bold indicate the most likely inferred population of origin, for each sample group, and n indicates the number of individuals tested. The number given for each sample group corresponds to geographical locations marked on Fig. 1
Inferred population/sample groupInferred population cluster
12345678
Perth Hills (n = 29)
1. Helena0·6180·0510·0680·0260·0090·0180·2040·007
2. Canning0·6050·0130·0520·0690·020·0290·0210·192
3. Churchman0·9110·0130·0310·0080·0060·0100·0120·009
4. Wungong0·9520·0040·0110·0060·0080·0100·0040·004
Serpentine (n = 46)
5. Serpentine0·1500·7360·0320·0260·0140·0150·0140·013
Dandalup (n = 56)
6. North Dandalup0·0120·0040·8270·0150·0070·0350·0940·007
7. South Dandalup0·1700·0240·5390·1300·0180·0500·0550·013
Collie (n = 16)
8. Samson Brook0·0140·0170·0160·8330·0710·0060·0210·022
9. Harvey0·0310·0230·0490·4100·1880·2000·0110·089
10. Wellington0·0060·0100·0070·9210·0110·0090·0040·032
11. McAlinden0·0170·0070·0150·6820·2600·0120·0050·003
Muir (n = 66)
12. Barlee0·0210·0050·0250·0680·8160·0540·0050·006
13. Blackwood0·0100·0030·0070·0260·9410·0040·0050·003
14. Tone River0·0020·0040·0050·0170·9570·0050·0040·006
15. Shannon0·0040·0050·0040·0350·9320·0090·0030·007
16. Lake Muir0·0130·0050·0500·0180·8960·0090·0060·003
17. Frankland0·0140·0060·0150·4740·3630·1170·0070·005
Denbarker (n = 34)
18. Denbarker0·0120·0060·0110·0100·0120·9400·0060·004
Domestic (n = 9)0·0050·0080·0070·0050·0040·0070·9580·006
Northampton (n = 20)0·0030·0050·0040·0050·0030·0050·0090·966

genetic diversity within populations

Overall, the level of genetic diversity (measured as expected heterozygosity) was similar within each population, with HE values ranging from 0·464 to 0·643 and a mean of 0·680 (± 0·089; Table 3).

Table 3.  Measures of genetic variability: mean observed number of alleles (NA), mean observed (HO) and mean expected heterozygosity (HE) for eight pig populations. Populations are the same as those in Fig. 1. HE values closer to unity indicate a high level of heterozygosity
PopulationnNA (± SD)HO (± SD)HE (± SD)
Perth Hills294·57 ± 1·220·586 ± 0·1530·608 ± 0·096
Serpentine464·36 ± 1·340·637 ± 0·1010·618 ± 0·094
Dandalup565·00 ± 1·620·638 ± 0·1370·643 ± 0·098
Collie164·00 ± 1·410·484 ± 0·2070·539 ± 0·177
Muir664·14 ± 1·290·500 ± 0·2070·558 ± 0·202
Denbarker343·21 ± 0·800·599 ± 0·1410·540 ± 0·121
Northampton203·36 ± 1·340·446 ± 0·1750·464 ± 0·184
Domestics 95·50 ± 2·030·746 ± 0·1970·681 ± 0·108

detection of recent and long-term bottlenecks

Three populations (Perth Hills, Collie and Muir) showed no evidence of either recent (P > 0·05; Table 4) or long-term bottlenecks (e.g. allelic diversity, heterozygosity; Table 1). However, the populations at Serpentine, Dandalup and Denbarker all retained genetic signatures of recent bottlenecks, suggesting that these populations have all experienced severe reductions in numbers (i.e. bottlenecks) in their recent history. Despite this, these populations retained some of the highest levels of genetic diversity (HE = 0·540–0·643; Table 3) of any of the feral pig populations investigated, and had effective population sizes (Ne) ranging from 466 to 856 pigs (Table 4).

Table 4.  Effective population sizes (Ne) and the probability of recent bottlenecks in six feral pig populations from south-west Western Australia. NS indicates that the population did not show a significant signature of a recent bottleneck (P > 0·05)
PopulationEffective population size (Ne)Significance of bottleneck
Perth Hills688NS
Serpentine732P = 0·001
Dandalup856P = 0·039
Collie463NS
Muir515NS
Denbarker466P = 0·015

differentiation among and within populations

Global (pooled) estimates of FST (0·118 ± 0·045) and RST (0·132 ± 0·066) between the feral pig populations studied were similar and indicated moderate to high levels of genetic differentiation among these populations. Pairwise FST and RST values (Table 5) indicated moderate to high levels of differentiation between all pairs of populations (i.e. most values were c. 0·1), except between the Perth Hills and Dandalup populations (FST = 0·043; RST = 0·051) where genetic differentiation was low.

Table 5.  Pairwise RST (above the diagonal) and FST (below the diagonal) estimates of population differentiation among six feral pig populations, based upon the observed genotypes that were estimated from 14 microsatellite loci. Values above 0·1 indicate a high degree of genetic differentiation
 Perth HillsSerpentineDandalupCollieMuirDenbarker
Perth Hills0·1050·0510·0670·0570·160
Serpentine0·0950·1950·1580·1900·260
Dandalup0·0430·0990·1230·1120·114
Collie0·0940·1220·0750·0200·194
Muir0·0960·1620·0810·0800·169
Denbarker0·1530·1710·1190·1790·196

As suggested by the population clustering results, FST and RST estimates of genetic differentiation within sample groups comprising single river systems revealed low levels of differentiation between upstream and downstream areas, over distances comparable to those between populations (Fig. 1c and Table 6). Whereas populations from neighbouring watercourses were highly differentiated, genetic differentiation within populations was relatively weak (Table 6).

Table 6.  Differentiation between populations, measured in terms of distance (km) and pairwise FST and RST, within single watercourses for three of the feral pig sample groups, based upon the observed genotypes that were estimated from 14 microsatellite loci. n indicates the number of individuals tested
Sample location (river system)Size of river catchment (km2)PopulationDistance between upstream and downstream samples (km)nFSTRST
Canning789Perth Hills11·9180·0770·032
Serpentine654Serpentine17·6240·0340·023
South Dandalup311Dandalup19·2200·1230·010

Generally, Nm estimates based on private alleles were lower than those estimated using FST (θ)-based estimates. Nm values were moderate among the south-west populations for both the private alleles method (0·672 ± 0·384 migrants generation−1) and the FST-based method (2·23 ± 1·14 migrants generation−1). When the two estimates were averaged, a relatively low estimate of 1·451 migrants year−1 was reached. Absolute dispersal distance estimates ranged between 17·5 and 172 km year−1 from the private alleles approach (mean 61·7 km year−1), and these were similar to the FST (θ)-based estimates, which ranged from 39·6 to 423·8 km year−1 (mean 213·4 km year−1). When the two estimates were averaged, an estimate of 137·5 km year−1 was reached.

Discussion

On the basis of the observed genotypes, the six feral pig populations identified in south-west Western Australia were highly genetically differentiated. These results are consistent with existing historical knowledge of the regional distribution and biology of the feral pig populations studied, and with the only other comparable published results on Italian wild boars (Vernesi et al. 2003). Past surveys have shown that feral pigs have always had a discontinuous distribution in the south-west of Western Australia, with established populations generally restricted to the major watercourses and catchment dams. Very little pig activity has been recorded in the dry sclerophyll forest between the major river systems and swamps (Long 2003). This suggests that, despite the feral pigs’ description as a habitat generalist (Choquenot, McIlroy & Korn 1996), their requirement for permanent water sources has led to relatively specific habitat requirements in many areas of south-western Australia.

Animals displaying habitat specialization may be restricted in distribution according to the availability of that habitat. If patches of suitable habitat are surrounded by less favourable habitat, then populations may become geographically isolated due to the infrequency of migration events between populations. Such isolated populations may then become genetically differentiated as a result of the effects of inbreeding and genetic drift (Whitlock, Ingvarsson & Hatfield 2000). Given the historical distribution pattern and the relatively small home ranges of feral pigs in south-western Australia (Choquenot, McIlroy & Korn 1996), a high level of genetic structuring was not unexpected. High differentiation levels were even seen to occur between populations that were only 25 km apart (e.g. Serpentine and Dandalup populations; Fig. 1b).

There was a substantial difference in the geographical size of the identified populations (Table 2), with larger geographical populations occurring in the south (Collie and Muir; Fig. 1b). This probably reflects the relatively recent invasion of these southern regions by feral pigs (Long 2003; Twigg 2003). In the northern populations (Perth Hills, Serpentine and Dandalup), which have been infested with feral pigs since early European settlement (1870s; Long 2003), there were high levels of genetic differentiation between populations found on neighbouring watercourses, even those only 25 km apart (Fig. 1b). In contrast, in the southern populations (Collie, Muir and Denbarker), which have only been known for 5–30 years, feral pigs occurred in relatively large populations, with undifferentiated groups up to 100 km apart (Fig. 1b). The Denbarker population has only been known for the last 5 years, and from the very high FST and RST values calculated between this population and all others studied, it appears unrelated to the populations to the north and west. This population has probably resulted from a recent release of either domestic pigs or feral pigs from outside southern Western Australia, rather than invasion from a neighbouring population (i.e. Muir; Fig. 1b). The lack of differentiation within the Muir population, in particular, is not surprising given that feral pigs have only been known in reasonable numbers in this area for the last 10 years. Similarly, several other animal species (e.g. jaguars Panthera onca; Eizirik et al. 2001) that have undergone very recent geographical range expansions also show very little genetic structuring, as the ‘speed of the expansion’ does not allow sufficient time for extensive population differentiation (Cabe 1998; Davies, Villablanca & Roderick 1999).

The lack of any detectable genetic differentiation between pigs from the opposite ends of the three major south-west river systems studied (Canning, Serpentine and South Dandalup rivers; Fig. 1c) indicates that gene flow is relatively high along these watercourses, compared with the relatively low level of gene flow observed over similar distances between these rivers. The presence of feral pigs around the fringes of the important drinking water reservoirs located on the lower reaches of these three rivers is perceived as a major public health threat and, as such, there are ongoing control efforts. However, these control programmes have been undermined as a result of relatively rapid reinvasion. Our data suggest that this natural reinvasion is most probably from pigs dispersing from the upper reaches of each river system rather than from neighbouring systems. Therefore, to improve the effectiveness of management strategies, future control programmes should also focus on the upper reaches of each river system.

Population size estimates and ‘signatures’ of recent bottlenecks differed substantially between the six feral pig populations investigated. When the effective population sizes for the individual areas sampled (i.e. subpopulations) were considered within the inferred Muir population, the estimate for the Lake Muir subpopulation was substantially larger than the estimates for the three other nearby subpopulations (Tone River, Shannon and Frankland; Fig. 1b). Furthermore, genetic parentage data for these subpopulations revealed that boars from Lake Muir had sired litters in all three of the other subpopulations (Hampton 2003) but there was no indication of gene flow in the reverse direction. We suggest that the Lake Muir subpopulation was acting as the source for the reinvasion of the three nearby subpopulations. These three subpopulations had been subjected to feral pig control operations in the immediate past, but the impact of immigration from the Lake Muir ‘source’ population appears to have negated control efforts. To be more effective, control strategies need to include all potential sources of reinvasion.

Genetic bottlenecks are usually the result of populations being reduced to a low number of individuals, which ultimately reduces genetic diversity (Luikart & Cornuet 1998; Piry, Luikart & Cornuet 1999). Thus, we used evidence of genetic bottlenecking to infer the effectiveness of past feral pig control programmes in reducing population sizes. Recent bottlenecks were detected in the Serpentine, Dandalup and Denbarker populations, all of which had been the target of rigorous control programmes over the past 3 years (mainly trapping). This suggests that the current control strategies have been effective in reducing the number of pigs in these three populations. In contrast, other populations, such as Collie, where control programmes have been ad hoc over the same period, showed no evidence of recent bottlenecking. Similarly, for the Muir and Perth Hills populations, where less comprehensive control efforts have occurred over the past 3 years, there was no evidence that control programmes have had a long-term effect on the genetic composition of these populations. These ad hoc control programmes have been ineffective in reducing overall pest numbers in the medium term.

The relatively low Nm values revealed generally low natural migration rates between the six feral pig populations identified in south-west Western Australia. Thus, considering both the high levels of differentiation and the low migration rates between populations, management strategies would be most effective if each population was considered as a single management unit. That is, management strategies need to apply control measures to all areas infested with feral pigs, particularly identified sources, within each unit. Failure to recognize this will result in the inevitable reinvasion of controlled areas by those feral pigs remaining elsewhere within each geographical population. Conversely, reinvasion from neighbouring, independent pig populations will occur relatively slowly as the movement of pigs between the discrete populations is low.

These findings have important implications for contingency plans for directly transmitted exotic disease incursions. For example, our migration rates suggest that potential disease transmission will be slow between populations but may occur relatively quickly within populations, such as along major river systems. Disease containment boundaries would therefore need to allow for this aspect of feral pig behaviour. The estimated dispersal distances (mean 84 km year−1) also suggest that the containment areas would need to be relatively large to enhance the containment of any disease. Such scenarios do, however, assume that there is no deliberate, illegal relocation of feral pigs via human intervention (Twigg 2003).

The improved ability to define management units is probably the most important benefit provided by our approach, incorporating molecular techniques in the development of wildlife management strategies. For example, currently the biggest impediment to successful feral pig control programmes in Australia, including efforts to achieve localized eradication, has arguably been the inability to delineate natural population boundaries (Saunders & Bryant 1988). Without reliable knowledge of the spatial structure of feral pig populations, arbitrary boundaries are often chosen for establishing eradication/containment zones for exotic disease contingencies (Saunders & Bryant 1988) and/or routine control programmes. Although the ability of such boundaries to reflect the ‘true’ behaviour and structure of feral pig populations is unknown, they are unlikely to be always appropriate. We believe that the incorporation of molecular-based information, such as that derived from our study, into the appropriate decision processes would ultimately lead to improved exotic disease contingency plans, and management strategies, for feral pigs throughout Australia.

Our study, which is one of the few that has looked at ‘open’ populations of a relatively large mammal, indicates that molecular techniques can make a valuable contribution to the understanding of the dynamics of many free-ranging animal populations. We believe such an approach can make a significant contribution to the planning and implementation of wildlife management programmes, particularly those involving invasive pest animals, as it has been to aid the conservation of threatened species. As illustrated above, this utility would include the development of improved contingency plans for managing wildlife diseases, particularly where such diseases are exotic (e.g. a potential foot-and-mouth disease incursion to Australia). We were able to: (i) elucidate population structure; (ii) estimate dispersal rates; (iii) gauge the effectiveness of past control programmes; (iv) predict the likely role of feral pigs in wildlife disease spread; and (v) define management units for a free-ranging species. This enabled a greater understanding of the potential threat posed by feral pigs as vectors for directly transmitted diseases, and also how to manage this threat better. Clearly, other factors, such as contact rates and interspecific transmission, are equally important in formulating exotic disease contingency plans (Ramsey et al. 2002; Caley & Hone 2004). Adequate monitoring programmes also need to be in place (Choquenot, McIlroy & Korn 1996; Edwards et al. 2004) for the management of vertebrate pest species to be successful in the long term.

Acknowledgements

We thank all those people who assisted with the collection of tissue samples, including both government agency employees and private landholders. We are grateful for support from the US Department of Agriculture Pig Genome Coordination Project. This research was generously supported by the W.A. Water Corporation, Murdoch University and W.A. Department of Conservation and Land Management.

Ancillary