SEARCH

SEARCH BY CITATION

Keywords:

  • Antechinus flavipes;
  • connectivity;
  • fragmentation;
  • gene flow;
  • GIS;
  • microsatellite;
  • river;
  • vegetation corridor

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information
  • 1
    In much of the world, fauna has been adversely affected by human actions, including conversion of forests to farmland, logging and regulation of river flows. Landscape genetics data can provide information about dispersal and gene flow across the landscape, identifying barriers and facilitators of gene flow. Landscapes of central Victoria, Australia, have been altered extensively in the last 160 years. Much vegetation has been cleared or degraded, and only forest patches of mainly re-growth remain, yet some forest-dependent species like the yellow-footed antechinus Antechinus flavipes persist. The antechinus has good dispersal capabilities and is the only native, small, carnivorous mammal on most floodplains. We use antechinus as a model to understand species persistence in fragmented landscapes.
  • 2
    We analysed variation at 11 microsatellite loci and the control region of mitochondrial DNA to infer past and contemporary gene flow among A. flavipes populations. To explore genetic connectivity, we used least-cost path methods, which assign different ‘friction’ costs to vegetation, cleared land, roads and rivers.
  • 3
    Populations from 11 forests formed six distinct genetic groups, and with few exceptions, animals from nearby forests clustered together despite the intervening Murray River or farmland with only narrow vegetation corridors between them.
  • 4
    Genetic connectivity was aided by corridors of vegetation and inhibited by cleared land.
  • 5
    Synthesis and applications. Our approach, capitalizing on inferences on both historic and contemporary gene flow, provides management agencies with key information on metapopulation dynamics in landscapes. Rather than merely maintaining existing vegetation upon which this (and many other) species depend, the genetic information also informs where future plantings should be prioritized to facilitate demographic and genetic exchange among sub-populations of species. Moreover, the decline in condition (‘health’) of riparian trees in this region must be reversed by provision of flooding flows; otherwise, metapopulation dynamics will become even more disarticulated than at present.

Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Destruction of habitats is one of the main causes of decline of many species (Wilcove et al. 1998; Mac Nally 2007). Many habitats are degraded or fragmented (e.g. Bennett et al. 1998). Our understanding of factors affecting population persistence within fragmented landscapes may be improved through landscape genetics (Manel et al. 2003; Waples & Gaggiotti 2006; Storfer et al. 2007). In landscape genetics, the effects on gene flow of natural and artificial features and of habitat characteristics are explored (e.g. Eriksson et al. 2004; Schmuki et al. 2006). This knowledge, which is impossible to obtain by conventional methods for tracking organisms through landscapes (e.g. mark–recapture), can be used to assist managers in whole-landscape planning to address the problems of broad landscape change, and to re-establish the natural connections between metapopulation elements.

Vegetation corridors, such as roadside reserves and riparian zones, may provide genetic connectivity between fragmented populations. When a surrounding matrix is a ‘hostile’ barrier, dispersal may only be possible through ‘friendly’ remnant vegetation. Existing studies provide contrary results for gene flow through corridors despite the presence of animals within corridors (c.f. Mech & Hallett 2001, Horskins, Mather & Wilson 2006). Therefore, the effectiveness of corridors is likely to be species-specific and should be assessed through genetic analyses. This information may provide both insight and guidance for natural resource managers who are responsible for conserving native species in massively altered landscapes.

In south-eastern Australia, much native vegetation has been cleared over the past 160 years for agriculture and timber extraction (Walker, Bullen & Williams 1993). Almost all remaining forests north of the Great Dividing Range in Victoria have been extensively disturbed and occur now as patches embedded in farmland. The remaining vegetation consists mostly of re-growth of box and ironbark forests (dominated by grey box Eucalyptus microcarpa and red ironbark E. sideroxylon or E. tricarpa) on the slopes and plains, and river red gum Eucalyptus camaldulensis forests on floodplains (ECC 1997).

Our work is a first step towards an understanding of the effectiveness of vegetation strips in promoting connectivity among populations of native species in the region and thus informing landscape management. We carried out a landscape genetics analysis of the yellow-footed antechinus Antechinus flavipes (Waterhouse 1828). Landscape connectivity is particularly important for this species because all males of A. flavipes die following an annual 2-week mating season (Woolley 1966). Therefore, total breeding failure in one breeding season will result in patch-scale population extinction unless there is immigration. Connectivity may also allow A. flavipes to track environmental changes, such as increased prey abundance on floodplains after floods (Ballinger, Mac Nally & Lake 2005), and will allow populations to undergo demographic recovery following drought (Mac Nally & Horrocks 2008) or forest fires.

Antechinus flavipes is the only native, small, carnivorous mammal (Marsupialia) on most floodplains in south-eastern Australia. In Victoria, A. flavipes is found in different habitat types (e.g. box–ironbark and river red gum forests) in a broad geographic band north of the Great Dividing Range and along the Murray River. It is considered to have high dispersal capabilities (Coates 1995; Marchesan & Carthew 2004), and although the Murray River is unlikely to be a barrier, towns and cleared land appear to be (Lada, Mac Nally & Taylor 2008a). It is possible that elongated remnants of vegetation, such as roadsides and riparian zones, provide genetic connectivity between populations.

Here, we analyse the control region of mitochondrial DNA (mtDNA) and 11 microsatellite markers to determine past and current gene flow in A. flavipes inhabiting river red gum and box–ironbark forests embedded in an agricultural production landscape. We analyze the nature and distribution of mtDNA sequence diversity among all forests as a measure of long-term population connectivity. We also use least-cost path methods in geographic information systems (GIS) to identify geographic features affecting gene flow in A. flavipes. We ask:

  • 1
    Are there genetically distinct populations that may require special conservation measures?
  • 2
    Do local populations separated by cleared land exchange migrants?
  • 3
    Which landscape features affect genetic connectivity of these populations?
  • 4
    Which management options are likely to be effective?

Materials and methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

study sites and animal capture

Animals were trapped in the austral autumn and winter in 2004 and 2005 at randomly selected 0·25 ha sites (see Lada, Mac Nally & Taylor 2007a) in eight river red gum forests and three box–ironbark forests (Fig. 1). There were 13–150 sites per forest (569 in total) and four sites in a box–ironbark/Blakely's red gum (Eucalyptus blakelyi) ecotone (Warbyecotone) (Fig. 1).

image

Figure 1. Study areas and Bayesian population assignment analysis of individuals of Antechinus flavipes in 2005 for K = 7 (Structure; Pritchard & Wen 2003). Seven clusters were inferred and are denoted by different colours and patterns. Q is a vector of q values; each q is the proportion of a genotype that has an ancestry in a given cluster. The partition of colours/patterns in each column represents each q value for a given genotype. Geographic origins of DNA samples are indicated by arrows. F denotes forest; Gunbower (centred on 35°46′S, 144°16′E), Campbells Island (35°34′S, 144°04′E), Guttrum (35°35′S, 144°04′E), Koondrook (35°44′S, 144°17′E), Barmah (35°52′S, 145°01′E), Millewa (35°50′S, 145°00′E), Rushworth (36°39′S, 145°01′E), Reedy Lake (36°42′S, 145°05′E), Warby (36°13′S, 146°11′E), Ovens (36°13′S, 146°15′E, 13 sites) and Chiltern (36°08′S, 146°36′E).

Download figure to PowerPoint

Five, six or 10 small-mammal Elliott traps (33 × 10 × 9 cm) were placed at each site for 1–5 nights and checked at dawn and dusk. Traps contained bedding and bait (peanut butter, quick oats and honey), were in plastic bags and camouflaged with bark. Animals were handled in humane manner and released at the point of capture; research was authorized by Monash University Biological Sciences Animal Ethics Committee. Ear-biopsy (2 mm) tissue from trapped Antechinus was placed in 96% ethanol for subsequent genetic analysis. A global positioning system (GPS 72R, Garmin, Olathe, KS, USA) was used to record geographic locations of animals in traps to within 10 m. Individuals within 1·5 km (the maximum male natal dispersal distance in continuous forest; see Lada et al. 2007a) of each other through a continuous forest were considered as one sampling unit for a least-cost path analysis. Each unit was assigned coordinates corresponding to the geographic centre of all the within-unit capture locations.

genetic analyses: testing markers, estimating effective population size and genetic diversity

We used a suite of analytical tools (see Supporting Information Appendix S1). DNA was extracted from 744 ear biopsies, and 11 microsatellite loci (Aa1A, Aa2B, Aa4A, Aa4D, Aa4 K, Aa7D, Aa7F, Aa7H, Aa7 K, Aa7O and Aa7Q) were amplified in polymerase chain reaction (PCR; Banks et al. 2005a), visualized and quality-checked as described in Lada et al. (2007a). We also amplified 500 bp of the control region of mtDNA in all samples, analysed the reactions by single-strand conformation polymorphism (SSCP), and sequenced several representatives of each SSCP morph from each forest.

All animals trapped in 1 year in the same forest were treated as one sampling unit (22 units) for tests of linkage equilibrium (Arlequin 3·0 Excoffier, Laval & Schneider 2005) and for tests of departure from Hardy–Weinberg equilibrium (GenePop 3·4 Raymond & Rousset 1995).

We estimated effective population sizes (Ne) for six regions defined by the Structure analysis (see below), using the linkage disequilibrium method (Hill 1981) in NeEstimator 1·3 (Peel, Ovenden & Peel 2004).

Gene diversity (HS), observed heterozygosity (HO) and allelic richness (AR) were estimated using the fstat software (Goudet 1995). AR represents the number of alleles per sample standardized for n, in this case the smallest sample size of 12. We used 15 000 permutations to test whether AR and HS were significantly: (i) different among the six regions; and (ii) higher in Koondrook-Gunbower compared to nearby Campbells-Guttrum.

patterns of historic gene flow and population dynamics

To infer historic female gene flow among the five regions, we constructed a network of mtDNA sequence haplotypes (see Supporting Information Appendix S2) using the median-joining algorithm (Bandelt, Forster & Roehl 1999) in Network 4·2 (Fluxus Technology Ltd). Divergence times (with the maximum-scaled migration rate Mmax of 0 or 10) between Chiltern and each of its nearest populations (Warby–Ovens and Millewa–Barmah) were estimated in MDIV (Nielsen & Wakeley 2001) from mtDNA sequences using the Hasegawa–Kishino–Yano model of evolution.

gene flow and landscape features: farmland, vegetation and rivers

Bayesian assignment analyses (Structure 2·1 Pritchard, Stephens & Donnelly 2000) were performed on microsatellite genotypes of all animals trapped in 2005 without using a priori information. This analysis clusters genotypes so that Hardy–Weinberg equilibrium is maximized and linkage disequilibrium (LD) minimized. Genotypes alone are used; geographic locations are ignored. We ran three chains of the model with admixture of ancestry and correlation of allele frequencies, for number of populations (K) from 1–10, for 150 000 burns-in, followed by 1 million samples. Chain convergence was checked by comparing values of parameters among independent runs of the same K (Pritchard et al. 2000). Selection of the most likely number of clusters (K) was based on the methods of Pritchard & Wen (2003).

Estimates of microsatellite pairwise FST values were used to assess differences in allele frequencies between sampling units (entire forests in single years) and tested statistically using Arlequin 3·1 (Excoffier et al. 2005) for all sampling units with n ≥ 10 (16 units). To allow simultaneous visualization of differentiation among all sampling units, we constructed trees of pairwise FST values using mega 3·1 (Kumar, Tamura & Nei 2004) with the neighbour-joining algorithm.

Least Cost Paths (LCPs) were constructed between all pairs of 2005 sampling units (N = 15) that consisted of ≥ 10 trapped animals (Fig. 5). There was one sampling unit per forest, except for Millewa (2), Gunbower (5) and Rushworth (0, insufficient captures). A cost layer was derived, using ArcGIS 9 (ESRI), from a raster layer showing presence or absence of vegetation at 10-m resolution (TREE25, Land Victoria 2004), a vector layer of roads (including towns as networks of roads; VicMAP Transport, Land Victoria 2006), and a vector layer of watercourses (Topo250 Hydrography, Geosciences Australia 2006). Elevation varies little across the study area. We converted all layers to raster format with 50 m cell size for cost path analysis. In the cost layer, each 50 × 50 m pixel (< 0·25 of antechinus home range, Coates 1995) was assigned a ‘friction’ value based on ease of movement as follows. Yellow-footed antechinuses are forest-dwelling animals whose abundance is positively associated with habitat complexity (Lada et al. 2007b; Mac Nally & Horrocks 2008), and thus, the lowest cost was assigned to vegetation; LCP in a continuous forest is a straight line. We initially assigned relative costs (n.b. denoted ‘original’) as vegetation < [cleared land = unsealed road (or bridge) = small creek] < sealed road (or bridge) < small river < large river (see Supporting Information Appendix S3).

image

Figure 5. Least-cost paths between sampling units of Antechinus flavipes and localized FST values for each unit underneath its name. G denotes Gunbower.

Download figure to PowerPoint

Three measures of geographic distance among sampling units were calculated: Euclidean (straight line) distances were obtained from GenAlEx 6·0 (Peakall & Smouse 2006), while costs and lengths of the LCP were derived using ArcGIS 9 (ESRI). These three types of pairwise distances then were loge-transformed and each correlated with pairwise FST /(1 – FST) using Mantel tests (Mantel 1967) in GenAlEx 6·0 (α = 0·05 and 999 permutations). These tests were conducted between: (i) all sampling units; and (ii) at a small spatial scale, between five sampling units in Gunbower forest, which provides information on isolation by distance in continuous forest (with some creeks and unsealed roads). Partial Mantel tests were conducted for the larger scale using Vegan 1·6 (Oksanen et al. 2006) in r (r Development Core Team 2005) to determine which geographic distance measure related best to genetic distance.

Mantel and partial Mantel tests were repeated using cost of LCP derived from vegetation-only cost layers (i.e. costs of rivers and roads = cost of no vegetation). Five cost layers were used, in which cost of moving through unvegetated (farmland, roads and rivers) relative to vegetated areas was, respectively: double (farm2);  × 5 (farm5, similar to the original but excluding roads/rivers); × 10 (farm10); × 100 (farm100); and half (i.e. farmland more favourable that forests, which is considered very unlikely). For each single-factor cost layer, we tested whether cost of LCP or Euclidean distance was more highly correlated with transformed pairwise FST values. Partial Mantel tests were conducted also for LCP costs derived from layers: farm5 and original layer (testing contribution of rivers and roads to the overall cost); farm2 and farm100; and farm10 and farm100.

GESTE 1·0 (Foll & Gaggiotti 2006) was used to test which geographic factors might affect genetic structure of populations. Genetic and geographic variables are combined in GESTE in a hierarchical Bayesian analysis to estimate localized FST for each sampling unit (i.e. not a pairwise value, but an overall indicator of genetic distinctness of each sampling unit) and generalized linear models are used to identify environmental factors influencing genetic structure. The 15 sampling units from the entire study area were treated as a metapopulation. We used allele frequencies in the 15 sampling units and tested the standard nine GESTE models (Table 1) for two factors: average Euclidean distance and mean LCP cost (farm100). Mean LCP cost for each sampling unit was the average of pairwise LCP costs from the unit to each of the other 14 units. The procedure was repeated with average Euclidean distance and mean LCP length. Burns-in consisted of 10 pilot chains each with 5000 iterations. Posterior probabilities were determined from a long sample chain, after 100 000 iterations. We repeated this procedure twice and compared the results among the runs for consistency. The BOA package (Smith 2005) in r was used to check for model convergence (Heidelberger & Welch 1983). The superior model had the highest posterior probability and if probabilities for > 1 model were similar, then GESTE analyses were conducted with each factor separately for the first three models (Table 1) to determine which of the three distances best explained genetic differentiation among sampling units.

Table 1.  Nine GESTE models (Foll & Gaggiotti 2006) to explain genetic differentiation in terms of landscape variables, and their posterior probabilities. LCP cost is the mean cost of least-cost path from each sampling unit to 14 other sampling units of Antechinus flavipes. LCP length is the mean length of least-cost path. LCP was derived from farm100 cost layer in which absence of vegetation was 100 times more costly than presence of vegetation
Model descriptionPosterior probability
Two factors
1 The constant regression term only = null model, neither LCP cost nor Euclidean distance affects genetic differentiation0·25
2 LCP cost, only LCP cost influences genetic differentiation0
3 The constant and LCP cost0·27
4 Euclidean distance0
5 The constant and Euclidean distance0·15
6 LCP cost and Euclidean distance0
7 The constant, LCP cost and Euclidean distance0·22
8 LCP cost, Euclidean distance and interaction between LCP and Euclidean distance0
9 All0·09
One factor – LCP cost
1 the constant regression term only = null model0·15
2 LCP cost,0
3 The constant and LCP cost0·85
One factor – LCP length
1 The constant regression term only = null model0·24
2 LCP length0
3 The constant and LCP length0·76
One factor – Euclidean distance
1 The constant regression term only = null model0·34
2 Euclidean distance0
3 The constant and Euclidean distance0·66

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

testing markers: linkage equilibrium and hardy–weinberg equilibrium

In 242 tests, there were 17 deviations from Hardy–Weinberg equilibrium, 15 of which were due to homozygote excess at seven loci. After sequential Bonferroni corrections (Rice 1989), four deviations remained: loci 4K, 4D, 7F, in Ovens, Koondrook and Gunbower forests, respectively, in 2005 (homozygote excess), and locus 4K in Chiltern forest (heterozygote excess) in 2004. All individuals produced ≥ 1 band at all loci (i.e. there were no candidate null homozygotes). In parentage analyses, there were no mismatches consistent with null alleles at the loci showing homozygous excess between pairs of individuals that otherwise matched (see Lada et al. 2007a). Null alleles if present must therefore be at very low frequency.

We found four significant linkage disequilibria (see Lada et al. 2007a): 7F-7K in both years and 4K-7H in Chiltern (2004); and 1A-7F in Ovens (2004). It seemed reasonable to retain all loci given the transitory and uncommon nature of linkage disequilibria and the amount of information lost if loci 7F and 4K were removed.

patterns of historic gene flow and population dynamics

Thirty-two mtDNA sequence haplotypes were identified (biggest difference = 2%), 24 of which were unique to a region (Fig. 2; Supporting Information Appendix S4). Seven of these were from Chiltern, which shared only three of its 10 haplotypes with other populations: haplotype 23 in Millewa–Barmah, haplotype 19 in Millewa–Barmah and Rushworth–Reedy and, the very common and internal on the network and thus, probably ancestral, haplotype 2, found in all regions (Fig. 2). Other than the latter, all haplotypes in the Warby–Ovens region were unique (Fig. 2). Campbells–Guttrum–Koondrook–Gunbower had seven unique haplotypes but shared haplotypes 10 and 7 with Millewa–Barmah, and haplotypes 14 and 21 with the Reedy–Rushworth region (Fig. 2). Haplotype 3, which differed from the ancestral haplotype 2 by only 1 bp, was the second most common haplotype in the study area, but was not found in Warby–Ovens and Chiltern (Fig. 2). Overall, substantial retention of the ancestral haplotype was coupled with the presence of unique haplotypes in all regions, suggesting occurrence of relatively recent mutations and restrictions to female-mediated gene flow among regions. Possession of mostly unique haplotypes in the Chiltern and Warby–Ovens regions suggests that these populations may have been isolated from other regions for the longest time (Fig. 2; Supporting Information Appendix S4). With zero migration, the divergence time between Chiltern and Warby–Ovens was estimated at T = 1 = 0·5Ne, which was relatively longer than the estimated T = 0·2 = 0·1Ne between Chiltern and Millewa–Barmah.

image

Figure 2. Network of mitochondrial DNA sequence haplotypes in Antechinus flavipes from five regions in south-eastern Australia: (Ca-T-K-G) Campbells–Guttrum–Koondrook–Gunbower, (M-B) Millewa–Barmah, (Rus-RL) Rushworth–Reedy Lake, (W-Ov) Warby–Ovens and Chiltern. Each circle represents one haplotype, and the size of the circle is proportional to the overall number of individuals with that haplotype. The size and colour/pattern of each pie slice represents the number of animals with that haplotype in each region. Grey rectangles indicate 1-bp mutations and stars represent undetected haplotypes. All haplotypes found in a forest are listed in order from the most to least common within a region. Unique haplotypes are in italic bold and sample sizes are in brackets.

Download figure to PowerPoint

effective population sizes and genetic diversity

The Warby–Ovens region had the lowest values for both Ne and allelic richness, while the Campbells–Guttrum and Rushworth–Reedy regions had the lowest gene diversity and relatively low Ne (Table 2). Allelic richness was significantly heterogeneous (P < 0·02), among the six regions, but gene diversity was not (P > 0·1). Allelic richness and gene diversity were significantly lower in Campbells–Guttrum than in the nearby Gunbower–Koondrook (P < 0·01) (Table 2).

Table 2.  Mean genetic diversity parameters and effective population size (Ne) in Antechinus flavipes from six regions (five pairs of nearby forests and the Chiltern forest). Forest areas were calculated from a vegetation layer in ArcGIS. LD is the linkage disequilibrium method for estimating Ne; gene diversity, HS; observed heterozygosity, HO
RegionForest area (ha)LD Ne95% CIAllelic richness (n = 12)HSHO
Warby–Ovens 9020  55 42, 756·700·7830·760
Campbells–Guttrum 5370 124 86, 2146·760·7710·758
Rushworth–Reedy43720 146 84, 4736·770·7730·773
Chiltern 2610 175123, 2107·480·8250·824
Koondrook–Gunbower51930 534423, 7157·890·8330·819
Millewa–Barmah687201454323, ∞7·570·7950·798

gene flow and landscape features: farmland, vegetation and rivers

Structure analyses of the 2005 genotypic data suggested K = 8 (Fig. 3) as the most likely number of clusters. However, K = 7 produced similar regional structuring, and thus, for simplicity, is presented here (Fig. 1). Five regions each had strong membership with a different, single cluster (population q values, denoting proportional cluster membership, ranging from 0·61–0·87): (i) Guttrum–Campbells; (ii) Ovens–Warby; (iii) Reedy–Rushworth; (iv) Barmah–Millewa; and (v) Chiltern (Fig. 1). Low similarity in genetic cluster membership among these five regions suggested lack of genetic connectivity among them (Fig. 1), but connectivity typically was evident between forests within each region. Gunbower–Koondrook was genetically highly differentiated from all other regions, including the nearby Guttrum–Campbells. Structure results were unaltered for K between seven and 10, in the sense that all new clusters were absorbed into Gunbower–Koondrook with uniform q values, indicating that these clusters were artificial.

image

Figure 3. The number of likely clusters (K) versus estimated ln of probability of data [ln Pr(X|K)] in the Structure analysis. The largest SD (15·7) was for K = 10, yet too small to be represented on the graph.

Download figure to PowerPoint

Microsatellite pairwise FST values for 2005 were significantly > 0 between all forests. Due to small sample sizes, five forests were not tested in 2004 (Fig. 4; Supporting Information Table S1). The tree of pairwise FST values (Fig. 4) shows that the Chiltern population may be genetically more similar to Barmah (146 km away) than to Ovens (29 km). Pairwise FST values were low (0·015–0·029) between local populations with the exception of Guttrum–Gunbower (0·045) and Campbells–Koondrook (0·060).

image

Figure 4. Neighbour-joining tree of pairwise FST values between sampling units of Antechinus flavipes. Sample sizes are in brackets; a thick, black line next to a branch indicates a non-significant pairwise FST value.

Download figure to PowerPoint

Transformed pairwise FST values among 15 sampling units (Fig. 5; Supporting Information Table S2) were significantly correlated with each of transformed Euclidean distances, original LCP costs, and original LCP lengths (Mantel tests, Table 3). LCP cost was significantly correlated with genetic distance corrected for Euclidean distance but Euclidean distance was not correlated with genetic distance conditioned on LCP cost (partial Mantel tests, Table 4). Thus, LCP cost explained additional variation not explained by Euclidean distance. The significant correlation between genetic distance and LCP cost remained after controlling for LCP length, but not vice versa (Table 4). LCP distance was marginally correlated with genetic distance corrected for Euclidean distance, and there was no correlation between Euclidean distance and genetic distance corrected for LCP distance (Table 4). Thus, variation in genetic distance is correlated with LCP cost > LCP distance > Euclidean distance.

Table 3.  Correlation coefficients (from Mantel tests) in 2005 among transformed variables: pairwise FST values (FST), Euclidean distances, original LCP cost (LCP cost) and original LCP length (LCP length). All correlations were significant, (P < 0·001)
 FSTEuclideanLCP cost
Euclidean0·674  
LCP cost0·7050·990 
LCP length0·6840·9940·995
Table 4.  Partial Mantel tests in 2005 among transformed variables: pairwise FST values (FST), Euclidean distances, original LCP cost (LCP cost) and original LCP length (LCP length)
Correlation betweenConditioned onr correlation coefficientP value
  • *

    significant correlation.

LCP cost and FSTEuclidean distance 0·357*0·002
Euclidean distance and FSTLCP cost–0·2330·952
LCP cost and FSTLCP length 0·321*0·010
LCP length and FSTLCP cost–0·2290·958
LCP length and FSTEuclidean distance 0·1770·106
Euclidean distance and FSTLCP length–0·0810·681

Similar results were obtained for all cost layers in which frictional cost of no-vegetation > cost of vegetation. Correlation between LCP costs (e.g. Table S3) and genetic distances increased with cost for no-vegetation, and was greatest for farm100 (r = 0·747, P < 0·001), which was significantly greater than for either farm10 or farm2. LCP costs from both the original and farm5 layers explained genetic distances equally well, and thus, at this spatial scale, cost of moving through farmland seems important, unlike that of roads and rivers. For the counterintuitive layer (cost of vegetation > no-vegetation), after we controlled for Euclidean distances, there was no correlation between LCP costs and genetic distances.

In the GESTE analysis at the largest spatial scale, no model seemed superior to the others when Euclidean distance and farm100 cost of LCP were tested simultaneously (Table 1, Fig. 5). Models 1 (constant), 3 (constant + cost of LCP) and 7 (constant + cost of LCP + Euclidean distance) had the highest posterior probabilities (Table 1). Similar results were obtained when length of LCP and Euclidean distance were tested (models 1, 3, 5 and 9 favoured, results not shown). When each factor was tested separately, model 3 (constant plus factor) was best in each case (Table 1).

In the only large continuous forest in which such tests were possible (Gunbower), transformed genetic distances were correlated significantly with each of transformed Euclidean distances (r = 0·938, P < 0·010), LCP lengths (r = 0·935, P < 0·012) and LCP costs (r = 0·934, P < 0·003), indicating isolation by distance.

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Massive clearance of river red gum and box–ironbark forests and woodlands in south-eastern Australia has left relatively small remnants of forest, most of which are re-growth and are still exploited (ECC 1997). Biodiversity management in these forests requires information about resident populations, which we have partly provided through genetic analysis of one key species. We have shown that genetic connectivity is aided by corridors of vegetation and inhibited by cleared land. These outcomes may contribute to landscape planning associated with proposals for ‘biolinks’– broad swathes of vegetation linking existing remnants – across northern Victoria (Brereton, Bennett & Mansergh 1995).

longer-term isolation of forests

MtDNA sequence data are more informative over a longer time-scale than are microsatellite genotypes. Differentiation at both nuclear and mtDNA markers compared to nuclear only may indicate longer-term barriers to gene flow. Our results suggest the longest isolation from populations in other forests for Chiltern and Warby–Ovens, where most mtDNA haplotypes were unique, and similar to one another in sequence and possibly arising from local mutation events (Fig. 2). It is plausible that the current distribution of mtDNA haplotypes is due to substantial landscape modification by Europeans over the past 160 years. Sampling on the eastern bank of the Ovens River and along roadsides is required to investigate the potential for barrier effects of farmland, which has replaced antechinus-suitable habitat (‘open forest plain’, Mitchell 1838) between the Ovens River and Chiltern.

do local populations separated by cleared land exchange migrants? what landscape features affect genetic connectivity of these populations?

Genetic differences among sampling units were correlated with cost of least-cost path (LCP), length of LCP and Euclidean distances (Tables 1, 3 and 4). Controlling for the other distance measures, cost of LCP explained additional variation in genetic differences. This is consistent with gene flow for A. flavipes being assisted by vegetation corridors and impaired by farmland because the cost of LCP had been generated with cost for vegetation < cost for farmland. The cost of LCP incorporated distance and‘friction’, whereas the other two measures were distances only. Banks et al. (2005b) reported that native eucalypt corridors within a pine plantation provided higher genetic connectivity among fragments than did pine itself, for A. agilis, a close relative of A. flavipes.

Rivers generally seem to be less of a barrier to gene flow than farmland. Although large rivers were assigned the highest friction cost, they contributed little to the overall cost because: (i) paths crossed rivers relatively rarely; (ii) rivers are relatively narrow (< 200 m), while farmland barriers stretch for tens of kilometres; (iii) LCPs rarely followed bridges (Fig. 5); and (iv) correlation between costs of LCP and genetic distances was the same irrespective of river costs. Most LCPs were equivalent to river distances because, due to extensive clearance of vegetation, only linear riparian corridors remain between most of our study populations. Such habitats may be disrupted more easily than less linear ones, for example, lack of gene flow between Gunbower and Guttrum (different genetic clusters in Structure). However, there is genetic connectivity between Ovens and Warby (one genetic cluster). Gunbower and Guttrum are 11 km apart along the Murray River and are separated by a river anabranch, a sealed road and a town, while floodplain Ovens and upland Warby are 6 km apart and are separated by a river anabranch, a sealed road and cleared land. This contrast in gene flow between the pairs of forests is remarkable and calls for more studies into the effects of dissecting features on floodplains on gene flow. This may provide important information for management of landscapes containing predominantly linear remnants of habitats.

gene flow and high genetic diversity

We have shown that populations of one species of small mammal can survive in a massively altered habitat and retain relatively high genetic diversity due to large Ne. Genetic connectivity is important for maximizing Ne(Frankham, Ballou & Briscoe 2002), particularly if effective population sizes are small and/or genetic diversity is lower in one or both of the sites. For example, although allelic richness typically was relatively high in this study, it was significantly lower in the Campbells–Guttrum region (low Ne) than in the nearby Koondrook–Gunbower region (high Ne), to which it was no longer genetically connected (Lada et al. 2008a). Other studies have shown reductions in genetic diversity concomitant with fragmentation. For example, populations of Rana dalmatina, separated by roads for 20 years, had significantly lower allelic richness and were more differentiated compared to non-fragmented populations (Lesbarreres et al. 2006).

implications for management

The yellow-footed antechinus appears to be persisting, if not necessarily thriving, in the much modified landscapes that we considered. However, results for this species may provide an overly optimistic picture for the vertebrates of the region generally. Work on other species, at a similar spatial scale and in the same forests, would help to show how general is the degree of population connectivity afforded to A. flavipes by the intervening landscape structures. Candidate species that co-occur with A. flavipes in the study area, and whose threatened status may be in part due to habitat modification, include the brown treecreeper Climacteris picumnus (vulnerable), the squirrel glider Petaurus norfolcensis (endangered) and the brush-tailed phascogale Phascogale tapoatafa (vulnerable) (ECC 1997). Given the extinction of several species of native mammals in the last 160 years in the study region, and conservation concern regarding others (Bennett et al. 1998), such information will assist conservation managers in planning to prevent further local extinctions and maintain evolutionary potential.

Notwithstanding different vegetation types, the Murray River or intervening farmland, animals from the following forests should be managed jointly to maintain and/or improve genetic connectivity between them: Rushworth with Reedy Lake, Ovens with Warby Range, Barmah with Millewa, Guttrum with Campbells, and Gunbower with Koondrook. The Rushworth population may be of concern (low trapping rates, Lada, Mac Nally & Taylor 2008b,c) and may rely on immigrants from the nearby, ‘healthier’ population at Reedy Lake. Similarly, Warby would benefit from increased connectivity with Ovens.

Information derived from this study also provides managers with the necessary justification to allocate scarce resources to maintaining the condition of floodplain forests of the southern Murray–Darling Basin. The region has experienced significant drying since the mid-1990s (and even earlier, Cai & Cowan 2008), and along with ongoing high water extractions for irrigation and urban use, pronounced declines in tree health have occurred (Cunningham et al. 2007). Federal and state governments have responded with The Living Murray initiative (http://www.environment.gov.au/water/mdb/lmi.html; accessed 13 June 2008), which involves a plan to provide environmental watering to the floodplain forests to improve tree health and encourage recruitment. The initiative focuses on five major remnant forests (‘icon sites’), but our work shows that the connecting riparian corridors are equally crucial from the landscape perspective. These corridors are not specifically considered in the initiative, but if tree death is widespread in the corridors, connectivity across the landscape probably would be abolished leading to a calamitous decline in landscape function from a biodiversity viewpoint.

By using different landscape genetics methods, as well as nuclear and mtDNA markers, we have provided insights into population connectivity in fragmented systems. We recommend taking similar approaches when investigating these issues in other species and landscapes, particularly since distinguishing levels of historic and current gene flow is essential for evaluating anthropogenic effects.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

The work was supported by a Monash Small Grant, the Hermon Slade Foundation and the ARC (F19804210, A19927168). Thanks to Anna Lada, Peter Lada and Quentin Lang for assistance with fieldwork, Leigh Privett for maps and Alexandra Pavlova, editors and reviewers for comments on how to improve the manuscript. This work was carried out under ACEC of NSW ACA, NSW DPI ARA, permits: BSCI/2003/02 (MUBS AEC), 10002325 (DSE), XX23517 (FC of NSW) and S10252 (NSW NPWS). This is publication no. 109 from the Australian Centre for Biodiversity.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information
  • Ballinger, A., Mac Nally, R. & Lake, P.S. (2005) Immediate and longer-term effects of managed flooding on floodplain invertebrate assemblages in south-eastern Australia: generation and maintenance of a mosaic landscape. Freshwater Biology, 50, 11901205.
  • Bandelt, H.-J., Forster, P. & Roehl, A. (1999) Median-joining networks for inferring intraspecific phylogenies. Biology & Evolution, 16, 3748.
  • Banks, S.C., Finlayson, G.R., Lawson, S.J., Lindenmayer, D.B., Paetkau, D., Ward, S.J. & Taylor, A.C. (2005a) The effects of habitat fragmentation due to forestry plantation establishment on the demography and genetic variation of a marsupial carnivore, Antechinus agilis. Biological Conservation, 122, 581597.
  • Banks, S.C., Lindenmayer, D.B., Ward, S.J. & Taylor, A.C. (2005b) The effects of habitat fragmentation via forestry plantation establishment on spatial genotypic structure in the small marsupial carnivore, Antechinus agilis. Molecular Ecology, 14, 16671680.
  • Bennett, A., Brown, G., Lumsden, L., Hespe, D., Krasna, S. & Silins, J. (1998) Fragments for the Future. Wildlife in the Victorian Riverina (the Northern Plains). The Department of Natural Resources and Environment, East Melbourne, Australia.
  • Brereton, R., Bennett, S. & Mansergh, I. (1995) Enhanced greenhouse climate change and its potential effect on selected fauna of south-eastern Australia – a trend analysis. Biological Conservation, 72, 339354.
  • Cai, W. & Cowan, T. (2008) Dynamics of late autumn rainfall reduction over southeastern Australia. Geophysical Research Letters, 35, L09708.
  • Coates, T. (1995) Reproductive ecology of the yellow-footed antechinus, Antechinus flavipes (Waterhouse), in north-east Victoria. Ph.D., Monash University, Melbourne, Australia.
  • Cunningham, S., Mac Nally, R., White, M., Read, J., Baker, P., Thomson, J. & Griffioen, P. 2007. Mapping the current condition of river red gum (Eucalyptus camaldulensis) stands along the Victorian Murray River floodplain. Occasional Papers of the Australian Centre for Biodiversity, Monash University, Melbourne, Australia.
  • ECC (1997) Box–ironbark. Forest and Woodlands Investigation. Resources and Issues Report. Environment Conservation Council (Victoria), Melbourne, Australia.
  • Eriksson, J., Hohmann, G., Boesch, C. & Vigilant, L. (2004) Rivers influence the population genetic structure of bonobos (Pan paniscus). Molecular Ecology, 13, 34253435.
  • Excoffier, L., Laval, G. & Schneider, S. (2005) Arlequin (version 3·0): an integrated software package for population genetics data analysis. Evolutionary Bioinformatics Online, 1, 4750.
  • Foll, M. & Gaggiotti, O. (2006) Identifying the environmental factors that determine the genetic structure of populations. Genetics, 174, 875891.
  • Frankham, R., Ballou, J.D. & Briscoe, D.A. (2002) Introduction to Conservation Genetics. Cambridge University Press, Cambridge, UK.
  • Geosciences Australia (2006) Geodata Topo 250 K Series 3 Users Guide, Vol. 2006. Geosciences Australia, Canberra, Australia.
  • Goudet, J. (1995) fstat (version 1·2): a computer program to calculate F-statistics. Journal of Heredity, 86, 485486.
  • Heidelberger, P. & Welch, P. (1983) Simulation run length control in the presence of an initial transient. Operations Research, 31, 11091144.
  • Hill, W.G. (1981) Estimation of effective population size from data on linkage disequilibrium. Genetical Research, 38, 209216.
  • Horskins, K., Mather, P.B. & Wilson, J.C. (2006) Corridors and connectivity: when use and function do not equate. Landscape Ecology, 21, 641655.
  • Kumar, S., Tamura, K. & Nei, M. (2004) mega 3: integrated software for Molecular Evolutionary Genetics Analysis and sequence alignment. Briefings in Bioinformatics, 5, 150163.
  • Lada, H., Mac Nally, R. & Taylor, A.C. (2007a) Genetic reconstruction of population dynamics of a carnivorous marsupial (Antechinus flavipes) in response to floods. Molecular Ecology, 16, 29342947.
  • Lada, H., Thomson, J.R., Mac Nally, R., Horrocks, G. & Taylor, A.C. (2007b) Evaluating simultaneous impacts of three anthropogenic effects on a floodplain-dwelling marsupial Antechinus flavipes. Biological Conservation, 134, 527536.
  • Lada, H., Mac Nally, R. & Taylor, A.C. (2008a) Distinguishing past from present gene flow along and across a river: the case of the carnivorous marsupial (Antechinus flavipes) on southern Australian floodplains. Conservation Genetics, 9, 569580.
  • Lada, H., Mac Nally, R. & Taylor, A.C. (2008b) Responses of a carnivorous marsupial (Antechinus flavipes) to local habitat factors in two forest types. Journal of Mammalogy, 89, 398407.
  • Lada, H., Mac Nally, R. & Taylor, A.C. (2008c) Phenotype and gene flow in a marsupial (Antechinus flavipes) in contrasting habitats. Biological Journal of the Linnean Society, 94, 303314.
  • Land Victoria (2004) Corporate Geospatial Data Library: TREEDEN25. Land Victoria, Helbourne, Australia.
  • Land Victoria (2006) VicMap Transport Product Description version 3·0. Land Victoria, Helbourne, Australia.
  • Lesbarreres, D., Primmer, C.R., Lode, T. & Merila, J. (2006) The effects of 20 years of highway presence on the genetic structure of Rana dalmatina populations. Ecoscience, 13, 531538.
  • Mac Nally, R. (2007) Use of the abundance spectrum and relative-abundance distributions to analyze assemblage change in massively altered landscapes. The American Naturalist, 170, 319330.
  • Mac Nally, R. & Horrocks, G. (2008) Longer-term responses of a floodplain-dwelling marsupial to experimental manipulation of fallen timber loads. Basic & Applied Ecology, 9, 458465.
  • Manel, S., Schwartz, M.K., Luikart, G. & Taberlet, P. (2003) Landscape genetics: combining landscape ecology and population genetics. Trends in Ecology & Evolution, 18, 189197.
  • Mantel, N. (1967) The detection of disease clustering and generalized regression approach. Cancer Research, 27, 209220.
  • Marchesan, D. & Carthew, S.M. (2004) Autecology of the yellow-footed antechinus (Antechinus flavipes) in a fragmented landscape in southern Australia. Wildlife Research, 31, 273282.
  • Mech, S.G. & Hallett, J.G. (2001) Evaluating the effectiveness of corridors: a genetic approach. Conservation Biology, 15, 467474.
  • Mitchell, T.L. (1838) Three Expeditions into the Interior of Eastern Australia: With Descriptions of the Recently Explored Region of Australia Felix and of the Present Colony of New South Wales. T & W Boone, London.
  • Nielsen, R. & Wakeley, J. (2001) Distinguishing migration from isolation: a Markov chain Monte Carlo approach. Genetics, 158, 885896.
  • Oksanen, J., Kindt, R., Legendre, P. & O’Hara, R.B. (2006) Vegan: Community Ecology Package version 1·6, University of Oulu, Finland.
  • Peakall, R. & Smouse, P.E. (2006) GenAlEx 6: genetic analysis in Excel. Population genetic software for teaching and research. Molecular Ecology Notes, 6, 288295.
  • Peel, D., Ovenden, J.R. & Peel, S.L. (2004) NeEstimator: Software for Estimating Effective Population Size. Queensland Government, Department of Primary Industries and Fisheries, Brisbane, Australia.
  • Pritchard, J.K. & Wen, W. (2003) Documentation for Structure software: version 2, University of Chicago, Chicago, IL, USA.
  • Pritchard, J.K., Stephens, M. & Donnelly, P. (2000) Inference of population structure using multilocus genotype data. Genetics, 155, 945959.
  • R Development Core Team (2005) r: A Language and Environment for Statistical Computing. r foundation for Statistical Computing, Vienna, Austria.
  • Raymond, M. & Rousset, F. (1995) GenePop (version 1·2): population genetics software for exact tests and ecumenicism. Journal of Heredity, 86, 248249.
  • Rice, W.R. (1989) Analyzing tables of statistical tests. Evolution, 43, 223225.
  • Schmuki, C., Vorburger, C., Runciman, D., Maceachern, S. & Sunnucks, P. (2006) When log-dwellers meet loggers: impacts of forest fragmentation on two endemic log-dwelling beetles in southeastern Australia. Molecular Ecology, 15, 14811492.
  • Smith, B.J. (2005) Bayesian Output Analysis Program (BOA) for MCMC. University of Iowa, Iowa, IA, USA.
  • Storfer, A., Murphy, M., Evans, J., Goldberg, C., Robinson, S., Spear, S., Dezzani, R., Delmelle, E., Vierling, L. & Waits, L. (2007) Putting the ‘landscape’ in landscape genetics. Heredity, 98, 128142.
  • Walker, J., Bullen, F. & Williams, B.G. (1993) Ecohydrological changes in the Murray–Darling Basin. I. The number of trees cleared over two centuries. Journal of Applied Ecology, 30, 265273.
  • Waples, R.S. & Gaggiotti, O. (2006) What is a population? An empirical evaluation of some genetic methods for identifying the number of gene pools and their degree of connectivity. Molecular Ecology, 15, 14191439.
  • Wilcove, D.S., Rothstein, D., Dubow, J., Phillips, A. & Losos, E. (1998) Quantifying threats to imperiled species in the United States. Bioscience, 48, 607615.
  • Woolley, P. (1966) Reproduction in Antechinus spp. and other dasyurid marsupials. Symposia of the Zoological Society of London, 15, 281294.

Supporting Information

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Appendix S1. Analytical tools used.

Appendix S2. Exclusions of samples of mtDNA haplotypes.

Appendix S3. Ranking of costs for least-cost paths.

Appendix S4. Distribution and frequencies of mtDNA sequence haplotypes.

Table S1. Microsatellite pairwise FST values between 16 sampling units.

Table S2. Matrix of transformed pairwise FST values between 15 sampling units in 2005.

Table S3. Matrix of loge-transformed costs of least-cost paths (farm100) between 15 sampling units in 2005.

Please note: Wiley-Blackwell are not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.

FilenameFormatSizeDescription
JPE_1563_sm_AppendixS1.doc20KSupporting info item
JPE_1563_sm_AppendixS2.doc22KSupporting info item
JPE_1563_sm_AppendixS3.doc21KSupporting info item
JPE_1563_sm_AppendixS4.doc33KSupporting info item
JPE_1563_sm_TableS1.doc30KSupporting info item
JPE_1563_sm_TableS2.xls11KSupporting info item
JPE_1563_sm_TableS3.xls11KSupporting info item

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.