The tales of two geckos: does dispersal prevent extinction in recently fragmented populations?
M. HOEHN,
*†
S . D . SARRE
†
and K. HENLE
*
*
Helmholtz Centre for Environmental Research — UFZ, Department of Conservation Biology, Permoserstrasse 15, 04318 Leipzig, Germany
, †
Applied Ecology Research Group, University of Canberra, ACT 2601, Australia
Abstract
Although habitat loss and fragmentation threaten species throughout the world and are amajor threat to biodiversity, it is apparent that some species are at greater risk of extinctionin fragmented landscapes than others. Identification of these species and the characteristicsthat make them sensitive to habitat fragmentation has important implications for conser-vation management. Here, we present a comparative study of the population genetic structureof two arboreal gecko species (
Oedura reticulata
and
Gehyra variegata
) in fragmented andcontinuous woodlands. The species differ in their level of persistence in remnant vegetationpatches (the former exhibiting a higher extinction rate than the latter). Previous demographicand modelling studies of these two species have suggested that their difference in persistencelevels may be due, in part, to differences in dispersal abilities with
G. variegata
expectedto have higher dispersal rates than
O. reticulata
. We tested this hypothesis and genotypeda total of 345
O. reticulata
from 12 sites and 353
G. variegata
from 13 sites at nine micro-satellite loci. We showed that
O. reticulata
exhibits elevated levels of structure (
F
ST
= 0.102vs. 0.044), lower levels of genetic diversity (
H
E
= 0.79 vs. 0.88), and fewer misassignments(20% vs. 30%) than similarly fragmented populations of
G. variegata
, while all these parameterswere fairly similar for the two species in the continuous forest populations (
F
ST
= 0.003 vs.0.004,
H
E
= 0.89 vs. 0.89, misassignments: 58% vs. 53%, respectively). For both species, geneticstructure was higher and genetic diversity was lower among fragmented populations thanamong those in the nature reserves. In addition, assignment tests and spatial autocorrelationrevealed that small distances of about 500 m through fragmented landscapes are a barrierto
O. reticulata
but not for
G. variegata
. These data support our hypothesis that
G. variegata
disperse more readily and more frequently than
O. reticulata
and that dispersal and habitatspecialization are critical factors in the persistence of species in habitat remnants.
Keywords
: continuous, dispersal, effective population size, geckos, generalist, habitat fragmentation,microsatellites, population genetics, specialization
Received 9 December 2006; revision accepted 20 March 2007
Introduction
Habitat loss and fragmentation threaten species through-out the world and are a major threat to biodiversity(Groombridge 1992; WCMC 1992). When formerlycontiguous vegetation becomes fragmented through landclearing, small isolated populations will result for manyspecies. Emerging empirical evidence suggests that somespecies are at greater risk of extinction in fragmented
landscapes than others. Identification of these species andthe characteristics that make them sensitive to habitatfragmentation has important implications for the manage-ment of species as well as contributing to our understandingof ecological and evolutionary theory (Sarre
et al
. 1995;MacNally
et al
. 2000; Davies
et al
. 2000, 2001, 2004; Henle
et al
. 2004a, b; Schmuki
et al
. 2006).Dispersal between an individual’s birthplace and that of
its offspring is one of the most important life-history traitsin species persistence (Koenig
et al
. 1996; Clobert
et al
. 2001;Sumner
et al
. 2001). Dispersal will usually result in geneflow, the movement and integration of alleles from one
Correspondence: Marion Hoehn, Fax: +49 341 235 3191; E-mail:[email protected]
3300
M . H O E H N , S . D . S A R R E and K . H E N L E
population into another. Gene flow protects small popula-tions from processes such as reduction in the number ofalleles, reduction in heterozygosity, and reduction ingenetic diversity through genetic drift and inbreeding(Frankham
et al
. 2002; Keller & Waller 2002). Althoughdispersal used to be difficult to study, genetic markersprovide a powerful tool for obtaining indirect estimates ofdispersal and gene flow in natural populations. Recentadvances in statistical techniques and the analysis of micro-satellite data promise to simplify studies of dispersal andcolonization. For instance, assignment testing can improvethe resolution of dispersal patterns by assigning individu-als to their most likely population of origin (Paetkau
et al
.1995; Rannala & Mountain 1997; Cornuet
et al
. 1999; Pritchard
et al
. 2000). Like
F
-statistics, the basis of assignment protocolsis an estimate of the allele frequency within populationsand therefore requires an a priori identification of popula-tion boundaries. In contrast, analytical methods that do notrequire information about population boundaries, such asisolation-by-distance analysis and spatial autocorrelation,can use genetic and geographical information to describedispersal and genetic spatial structure in a two-dimensionallandscape (Rousset 1997; Peakall & Smouse 2001; Double
et al
. 2005).Here, we present a comparative study of the genetic
structure of two gecko species in fragmented and continuouswoodlands. The species occur sympatrically, but haveresponded differently to the fragmentation of their wood-land habitat.
Gehyra variegata
(tree dtella) is abundant andwidespread throughout the southern half of Australia. It isa habitat generalist, and can be found on trees, logs, fallentimber, shrubs, rocks, and in highly disturbed habitat.
Oedura reticulata
(reticulated velvet gecko) is endemic tothe southwest of Western Australia and is a habitat specialist.It is limited in its habitat range, being exclusively arborealand restricted to smooth-barked
Eucalyptus
woodlands.Both species have been subjected to severe populationfragmentation in the Western Australian wheatbelt, wherenumerous very small populations persist in patches ofwoodland surrounded by land used for wheat crops. Insummer, the intervening matrix is often reduced to sparsestubble or even predominantly bare earth (How & Kitchener1983; Kitchener
et al
. 1988; Sarre 1995a, b, 1998; Sarre
et al
.1995).
In a previous empirical study, Sarre
et al
. (1995) demon-strated that
G. variegata
showed a markedly higher persistencethan
O. reticulata
in habitat remnants (97% remnant occu-pancy vs. 72%). In addition, studies 10 years apart (1991and 2001) documented extinction of
O. reticulata
in three of33 habitat remnants, whereas
G. variegata
has persisted atcomparable population sizes in all 33 remnants over thesame time period (M. Hoehn, unpublished data). Individual-based population viability models developed for bothspecies and including demographic and environmental
stochasticity also show contrasting expectations for thetwo species (Sarre
et al
. 1996; Wiegand
et al
. 2001, 2002).Whereas rates of persistence observed were lower for
O.reticulata
than those predicted by the population viabilitymodel, the opposite was the case for
G. variegata
. One ofthe reasons for the discrepancies between predicted andobserved distributions may be that the models did notinclude an expectation of dispersal between remnants foreither species. While the dispersal capability of
O. reticulata
is believed to be low, pitfall trapping and anecdotal evidencesuggest that movement by
G. variegata
is also infrequent(Sarre
et al
. 1995; R. How, personal communication).Nevertheless, occasional longer-distance dispersal isknown for
G. variegata
(Moritz 1987; K. Henle & B. Gruber,unpublished data.). Sarre
et al
. (1996) suggested that thediscrepancy between the observed and modelled persistencerates could be best explained by greater rates of movementfor
G. variegata
than for
O. reticulata
.To investigate further this dichotomy of responses to
fragmentation, we used microsatellite DNA markers tocompare the genetic structure and rates of dispersal in thetwo gecko species. Our expectation is that the high level ofpersistence exhibited by
G. variegata
is a result of a higherdispersal rate compared with
O. reticulata
. Thus, we pre-dicted higher levels of genetic differentiation and lower ratesof dispersal among
O. reticulata
than among
G. variegata
populations. We surveyed the genetic structure in fragmentedand continuous populations of these two species to test ourhypotheses. In addition, we expected a loss of geneticdiversity and a higher level of genetic structure for bothspecies in habitat fragments.
Materials and methods
Study area and sampling
The study area was located between Kellerberrin andTrayning in the Western Australian wheatbelt (Fig. 1).Large areas of native vegetation have been removed fromthe region and replaced by agricultural crops, pastures,and livestock. Since 1900, approximately 93% of the originalvegetation has been cleared and the remnant vegetation isdistributed over thousands of patches of varying size(Saunders & Hobbs 1991; Hobbs 1993; Saunders
et al
. 1993).Tissue samples were collected from the two gecko spe-
cies during the summer months, from November 2000until March 2001, in December 2003, and in November2005. A total of 345
Oedura reticulata
individuals from sixhabitat fragments and six locations in two nature reservesand a total of 353
Gehyra variegata
individuals from sevenhabitat fragments and six locations in two nature reserveswere sampled for genetic analysis. The majority of contin-uous habitat has been cleared in the area, but three sites perspecies were located within 1 km in the North Bandee
D I S P E R S A L A N D E X T I N C T I O N I N F R A G M E N T E D G E C K O P O P U L A T I O N S
Nature Reserve, which comprises 174 ha of continuouswoodland habitat. Another three populations of eachspecies where located within 1.2 km in the KorrelockingNature Reserve (259 ha).
Lizards were located at night using head-torches andcaptured by hand. The tip of the tail of each individual wasremoved and stored in liquid nitrogen. In all fragments,25–30 samples were collected with the exception of frag-ment population 7 (
N
= 13) for
G. variegata
, where no fur-ther individuals could be located. The sample populationswere labelled independently for each species, with onehabitat fragment (or ‘patch’ or ‘remnant’) equivalent to onesample population. The
O. reticulata
populations werelabelled ORF1-6, and the
G. variegata
populations GVF1-7(Table 1). In two cases, a habitat remnant was used as asample population for both species, that is
G. variegata
popu-lation GVF3 inhabited the same fragment as
O. reticulata
population ORF5, and similarly,
G. variegata
populationGVF6 occupied the same habitat patch as
O. reticulata
population ORF4. All other sample populations were fromseparate fragments, with no overlap between the twospecies. Continuous woodland sites were used as samplepopulations for both species and were numbered ORC1-3and GVC1-3 in the North Bandee Nature Reserve andORC4-6 and GVC4-6 in the Korrelocking Nature Reserve.
The presence and absence of
O. reticulata
and
G. variegata
populations in the study area was determined in previoussurveys (Sarre
et al
. 1995; M. Hoehn, unpublished data).This distribution data was used to design the following
sampling strategies (semi-experimental approach) forstudying assignment and dispersal between pairs of habi-tat patches on a fine geographical scale. We selected threepairs of
O. reticulata
populations, separated by 150, 550,and 580 m, and three pairs of
G. variegata
populations, sep-arated by 150, 300, and 1000 m (Table 1). In most cases thedistance to other extant populations was large and we con-sider that dispersal from individuals of other populationsinto the sample populations was likely to be very low.However, two neighbouring
O. reticulata
populations (ORF4and ORF1) were close to the same roadside vegetation(150 m and 230 m distant), which could harbour a potentialsource population of
O. reticulata
or function as a corridor.Two neighbouring
G. variegata
populations were also closeto roadside vegetation (0 m and 350 m distant), but in thiscase the roadside vegetation did not create a connectionbetween the neighbouring habitat fragments. All othersample populations, for both species, were separated fromroadside vegetation by distances exceeding 400 m. In theNorth Bandee Nature Reserve, we selected two pairs ofpopulations separated by 400 m and 650 m and in theKorrelocking Nature Reserve, two pairs of populationsseparated by 700 m and 500 m for each species.
The census population size was determined with theprogram
capture
(Otis
et al
. 1978) for another study (Hoehn2006). The values ranged from 33 to 197 individuals for
O. reticulata
and from 18 to 266 for
G. variegata
and we selectedpopulations so that the mean population size (124 and 95,respectively) would be similar between species.
Fig. 1 Map of the study area and locationof sites in the Western Australian wheat-belt. Oedura reticulata populations inhabitfragments labelled ORF1-6 and continuoussites labelled ORC1-3 (North Bandee NatureReserve), Gehyra variegata fragments arelabelled GVF1-7 and continuous sitesGVC1-3 (North Bandee Nature Reserve).Korrelocking Nature Reserve with siteslabelled ORC4-6 and GVC4-6 are not shown.
3302
M . H O E H N , S . D . S A R R E and K . H E N L E
DNA was extracted from the tip of the tail of each indi-vidual using the Chelex extraction method (Walsh
et al
.1991). We genotyped individuals of
O. reticulata
usingnine tetranucleotide microsatellite loci developed from anenriched library for this species (OR205, OR220, OR266,OR6F4, OR10H7, OR11G3, OR12D7, OR12D9, OR14A7)(Hoehn & Sarre 2005). For
G. variegata
, we genotypedindividuals using nine tetranucleotide microsatellite markerscloned from an enriched library for this species (GV1C5,GV1C10, GV1F1, GV3B5, GV3C6, GV3E10, GV4B6, GV4G6,GV4C9) (Hoehn & Sarre 2006). Polymerase chain reaction(PCR) amplification and genotyping using the BeckmanCoulter CEQ 8000 were performed according to conditionsdescribed in Hoehn & Sarre (2005) and Hoehn & Sarre(2006).
Statistical analysis
Descriptive statistics — the allelic richness (
A
n
) (number ofalleles corrected by sample size, based on a minimumsample size of 13 individuals for both species) and theobserved (
H
O
) and expected (
H
E
) heterozygosities per
locus — were calculated using
fstat
2.9.3 (Goudet 1995,2001). Global tests for deviation from Hardy–Weinbergwere performed. In addition, we tested heterozygote deficitsper locus and habitat fragment employing a sequentialBonferroni correction (
fstat
2.9.3
)
.The population genetic structure was investigated by
estimating
F
ST
using the method of Weir & Cockerham(1984) in the program
fstat
2.9.3. The significance of mean
F
-statistics was assessed by constructing 95% confidenceintervals (CI) by jackknifing across loci using
genetix
4.01(Belkhir
et al
. 2001). For the comparison of mean
F
ST
between species and between landscapes,
t
-tests for statis-tical significance were performed. Interpretation of geneticdifferentiation values between species is often problematicbecause of their dependence on the level of genetic variation.To address this, we applied a standardized measure ofgenetic differentiation (Hedrick 1999, 2005). The standardized
G
ST
,
is defined as:
(1)
where
G
ST
is an estimate of
F
ST
for a locus with multiplealleles and
Table 1 Habitat fragments and continuous forest sites sampled (C1-3 North Bandee Nature Reserve, C4-6 Korrelocking Nature Reserve)and reference number (ref) referring to the study of Sarre et al. (1995). The number of individuals genotyped (samples), fragment size (sizein ha), population size (pop size) calculated by the program capture (Hoehn 2006), distances to neighbouring fragments (distance NF) andthe road side vegetation (distance RSV) and, approximate time (years) since isolation from Sarre (1995a)
Fragment Ref Species Samples Size Pop size Distance NF Distance RSV Isolation
ORF1 170 Oedura 30 0.5 197 580 m 230 m 90ORF4 — Oedura 27 0.4 35 580 m 150 m 90ORF2 s85 Oedura 30 0.8 168 550 m 1000 m 60ORF3 s84 Oedura 27 0.4 33 550 m 500 m 60ORF5 442 Oedura 30 5.4 167 150 m 450 m 90ORF6 — Oedura 30 2 141 150 m 450 m 90ORC1 — Oedura 30 — — 400 m — —ORC2 — Oedura 17 — — 400 m/650 m — —ORC3 — Oedura 30 — — 650 m — —ORC4 — Oedura 31 — — 700 m — —ORC5 — Oedura 32 — — 700 m/500 m — —ORC6 — Oedura 31 — — 500 m — —GVF1 171 Gehyra 30 0.3 — 300 m 350 m 90GVF2 — Gehyra 30 1.4 76 300 m 0 m 90GVF3 442 Gehyra 30 5.4 266 150 m 450 m 90GVF4 — Gehyra 30 2 127 150 m 450 m 90GVF5 169 Gehyra 30 0.3 50 1000 m 400 m 80GVF6 s143 Gehyra 25 0.5 30 1000 m 1000 m 80GVF7 168 Gehyra 13 0.6 18 400 m 250 m 80GVC1 — Gehyra 30 — — 400 m/650 m — —GVC2 — Gehyra 24 — — 650 m — —GVC3 — Gehyra 29 — — 700 m — —GVC4 — Gehyra 33 — — 700 m/500 m — —GVC5 — Gehyra 23 — — 500 m — —GVC6 — Gehyra 27 — — 400 m — —
′GST
′ =G
GGST
ST
ST
(max)
D I S P E R S A L A N D E X T I N C T I O N I N F R A G M E N T E D G E C K O P O P U L A T I O N S 3303
where k is the number of equal-sized subpopulationsand HS is the average subpopulation Hardy–Weinbergheterozygosity. Pairwise values were calculated foreach species. The significance of between species andbetween landscapes was assessed by calculating the mean
and performing a t-test.Isolation by distance was tested through linear regres-
sion of FST/(1 – FST) against the geographical distancebetween pairs of habitat fragments (Rousset 1997). TheMantel test option in fstat 2.9.3 was used to assess thesignificance of correlation between matrices of genetic dif-ferentiation and geographical distance between sampledpopulations. Geographical straight-line distance betweenthe edges of the fragments within the study area was deter-mined using the map generated by M. G. Brooker, CSIRO,Division of Wildlife and Ecology, Perth and Department ofAgriculture, Western Australia.
We applied spatial autocorrelation (SA) analysis,implemented in genalex 6.0 (Peakall & Smouse 2001),as an additional method to estimate if dispersal and geneflow are limited. Unlike traditional SA (Sokal & Oden 1978),this technique employs a multivariate approach thatstrengthens the spatial signal and reduces the noise. Usingpairwise geographical and pairwise squared genetic distancematrices, it generates an autocorrelation coefficient, r,which is closely related to Moran’s I (Sokal & Oden 1978).The autocorrelation coefficient provides a measure of thegenetic similarity between pairs of individuals whose geo-graphical separation falls within the specified distance class.Significant positive autocorrelation implies that individualswithin a particular distance class are more genetically similarthan expected by random. Tests for statistical significancewere performed by random permutation with a MonteCarlo simulation performed with 1000 permutations and1000 bootstraps for estimation of 95% confidence intervals.We performed this analysis separately for habitat fragmentsand nature reserves using the even-distance class option(using 25 distance classes, 100 m each).
To provide an index of dispersal rates between popula-tions, assignments were conducted using the programsstructure 2.1 (Pritchard et al. 2000) and geneclass 1.0.02(Cornuet et al. 1999). Using the Bayesian clustering methodimplemented in structure 2.1, six populations (threepopulation pairs) in the fragments and six populations (fourpopulation pairs) in the continuous forest were analysedfor each species. Population GVF7 was omitted from thisanalysis because of the small sample size (N = 13) andbecause dispersal was assessed between pairs of popula-tions. For every population pair, five independent runs ofK = 2 were performed at 200 000 Markov chain MonteCarlo repetitions and 100 000 burn-in periods. No prior
information about the potential source populations wasincorporated, admixture was assumed and the model withcorrelated allele frequency was selected. For some popula-tion pairs, structure 2.1 failed to confidently assign anyindividual to a single genetic population. Thus, in a secondstep of the analysis, runs were performed at K = 1–5 toassess if some of the paired sites actually constitute a singlepanmictic population.
The Bayesian method of likelihood-based assignmenttests were implemented using geneclass 1.0.02 (Rannala& Mountain 1997; Cornuet et al. 1999). In a first step, indi-viduals were directly assigned to the population for whichtheir probability of belonging is highest. In a second step,a probability that the individual belongs to each populationwas simulated and individuals were assigned to the popu-lation for which their probability of belonging is highest,and where the arbitrary threshold value is above P = 0.05.
Results
A total of 345 Oedura reticulata from 12 sites (six fragmentsand six continuous forest sites) and 353 Gehyra variegatafrom 13 sites (seven fragments and six continuous forestsites) were genotyped at nine microsatellite loci. Four ofthe 108 tests for Hardy–Weinberg equilibrium in O. reticulataand six of the 117 tests in G. variegata showed significantdeviation from expected genotype frequencies afterBonferroni correction; all were due to a deficiency ofheterozygotes. The heterozygote deficiency may be due tonull alleles; although there does not appear to be anyconsistent pattern of Hardy–Weinberg deviation amongloci, or populations. No linkage disequilibrium was detectedbetween loci within populations at any locality and there-fore independence among loci has been assumed in thesubsequent analyses.
Genetic diversity
In O. reticulata, allelic richness (An) per site and locus rangedfrom 4.00 to 19.26 in fragments and from 6.00 to 17.53 incontinuous forest sites. We detected a highly significantdifference in the mean allelic richness between fragmentedand continuous forest sites (mean 7.97 and 11.36, respectively,Mann–Whitney U-tests: P < 0.001) (Fig. 2a). In G. variegata,allelic richness (An) per site and locus ranged from 4.74 to16.00 in fragments and from 7.44 to 16.42 in continuousforest sites. The difference in the allelic richness betweenfragments and continuous forest sites was significant (mean9.95 and 10.77, respectively, P < 0.05). Our analysis showedthat fragmented populations of O. reticulata have significantlylower allelic richness than fragmented populations of G.variegata (P < 0.001), while there is no significant differencein allelic richness between populations of the two speciesin the nature reserves (P = 0.48) (Fig. 2b).
G
k Hk HST
S
S(max)
( )( )
=− −
− +1 1
1
′GST
′GST
′GST
3304 M . H O E H N , S . D . S A R R E and K . H E N L E
Fig. 2 Comparison of allelic richness (An),expected (HE) heterozygosity, FST and estimates (a) between fragments andcontinuous forest site for Oedura reticulata(above) and Gehyra variegata (below) and(b) between Oedura reticulata and Gehyravariegata in fragments (above) andcontinuous forest site (below).
′GST
D I S P E R S A L A N D E X T I N C T I O N I N F R A G M E N T E D G E C K O P O P U L A T I O N S 3305
Average expected heterozygosity (HE) observed in O.reticulata ranged from 0.69 to 0.87 in fragments (mean HE:0.79) and from 0.85 to 0.92 in continuous forest sites (meanHE: 0.89). In G. variegata, expected heterozygosity (HE)averaged across loci ranged from 0.85 to 0.90 in fragments(mean HE: 0.88) and from 0.88 to 0.91 in continuous forestsites (mean HE: 0.89). Populations of both species showeda significant difference in heterozygosity between fragmentsand continuous forest sites (Fig. 2a). However, the level ofsignificance for O. reticulata (P < 0.001) was considerablyhigher than that for G. variegata (P < 0.05). Furthermore,tests revealed that fragment populations of O. reticulatahave significantly lower heterozygosity than fragmentpopulations of G. variegata (P < 0.001), while there is no sig-nificant difference between populations of either species inthe continuous forests (P = 0.42) (Fig. 2b).
Population differentiation
Both gecko species showed significant genotypic differ-entiation among fragment populations (P < 0.01). However,FST measures revealed higher levels of subdivision amongfragmented populations of O. reticulata (FST = 0.102; 95% CI0.086–0.119) than among fragmented populations of G.variegata (FST = 0.044; 95% CI 0.037–0.050, t-test: P < 0.001).Conversely, no significant difference in subdivision wasobserved between species in the continuous woodland(O. reticulata: FST = 0.003; 95% CI 0.001–0.005; G. variegata:FST = 0.004; 95% CI 0.001–0.007, t-test: P = 0.26) and therewas no significant genotypic differentiation among con-tinuous populations (Fig. 2b).
In O. reticulata, pairwise FST estimates ranged from 0.041to 0.163 among fragment populations and from 0.001 to0.007 among populations in the continuous nature reserves.The level of differentiation was significantly higher in frag-mented (FST = 0.109; 95% CI 0.079–0.140) compared withcontinuous populations (FST = 0.003; 95% CI 0.001–0.005, t-test: P < 0.001) (Fig. 2a). The same trend was apparent in G.variegata with pairwise FST estimates ranging from 0.007 to0.081 among fragment populations and from 0.000 to 0.009among populations in the continuous woodland. Differen-tiation among fragmented populations (FST = 0.025; 95%CI 0.010–0.041) appeared to be significantly higher thanamong continuous populations (FST = 0.004; 95% CI 0.001–0.007, t-test: P < 0.01) (Fig. 2a). For the previous comparison,we included pairwise FST estimates only from fragmentpopulations that were in close proximity to each other. Themean geographical distance between these fragments was670 m for O. reticulata and 650 m for G. variegata comparedwith a mean geographical distance of 750 m among thecontinuous forest sites for both species.
Using the standardized values of GST, the level of differ-entiation was also higher in O. reticulata ( = 0.45, 95% CIs:0.40–0.49) than in G. variegata ( = 0.34, 95% CIs: 0.29–
0.39). Tests showed that there was a significant differencebetween species (t-test, P < 0.01). All other test followedthe trend shown by the FST values (Fig. 2b).
Most pairs of sites exhibited a fairly similar range of FSTvalues, although for O. reticulata population, ORF3 hadlarger values on average than the other sites. In addition,ORF3 seemed to have lower allelic richness and heterozy-gosity. Therefore, we performed all calculations withoutthe outlier population and found that differences betweenthe species in An and HS and FST were still significant(P < 0.001) with the exception that the standardizedvalue of was only approaching significance (t-test,P = 0.06) compared with P < 0.05 when all populations areincluded.
Isolation by distance
A positive association between genetic differentiation(FST/1 – FST) and the ln of geographical distance separatingsamples (P < 0.01) was observed for G. variegata. But nosuch pattern was apparent for O. reticulata with or withoutthe outlier population of ORF3 when total geographicaldistance was taken into account, suggesting that distancebetween remnants had no influence on the genetic structurein this species (Fig. 3).
′GST
′GST
Fig. 3 Relationship between the logarithms of geographicaldistances and genetic differentiation estimated as FST/(1 – FST) forOedura reticulata (above) and Gehyra variegata (below).
′GST
3306 M . H O E H N , S . D . S A R R E and K . H E N L E
The outcome of the spatial autocorrelation analysis of thetwo gecko species is presented in Fig. 4. The correlogramproduced for the fragment populations in G. variegataindicates a significant positive correlation in the firstthree distance classes, 100 m (r = 0.070, P = 0.01), 200 m(r = 0.027, P = 0.01) and 400 m (r = 0.033, P = 0.01), with anintercept of 951 m (Fig. 4a). An autocorrelation focusing onthe nature reserve populations revealed no significant spatialstructure in any of the distance classes. For convenience,Fig. 4(b) only shows the results from the populations inthe Korrelocking Nature Reserve, but results from theNorth Bandee Nature Reserve are comparable. The spatialautocorrelation of fragment populations for O. reticulatashowed contrasting results to the isolation-by-distanceanalysis. The analysis revealed a significant positive geneticstructure up to a distance of 518 m, with positive r in thedistance classes, 100 m (r = 0.15, P = 0.01) and 200 m (r = 0.03,P = 0.01). Spatial autocorrelation seems to have a higher
potential than isolation-by-distance analysis for identifyingstructure at a genetic fine-scale level. Nevertheless, theanalysis of the continuous forest populations showed nosign of significant positive correlation.
Assignment
Both the Bayesian clustering method (structure 2.1) andthe direct Bayesian method (geneclass 1.0.02) producedconsistent results that could be used to estimate dispersalprobabilities. Figure 5 shows the results from the Bayesianclustering method (structure 2.1). We observed a higherpercentage of misassignments in fragmented populationsof G. variegata (mean of 30% for both structure 2.1 andgeneclass 1.0.02 methods, 0.070 and 0.045 standard error(SE), respectively, subsequent data always presented inthis order) than of O. reticulata (mean of 15% and 20%,0.047 and 0.096 SE, respectively). This occurred eventhough the populations of O. reticulata were in slightlycloser proximity to each other. For O. reticulata, the
Fig. 4 (a) Spatial autocorrelation corre-lograms for fragmented Oedura reticulatapopulations (above) and fragmented Gehyravariegata populations (below). (b) Spatialautocorrelation correlograms for continuousOedura reticulata populations (above) andcontinuous Gehyra variegata populations(below). Distances are in metres and onlypopulations from the Korrelocking NatureReserve are shown. The permuted 95%confidence interval (dashed lines) and thebootstrapped 95% confidence error bars arealso shown. The autocorrelation coefficient,r, provides a measure of the geneticsimilarity between pairs of individuals andsignificant positive autocorrelation impliesthat individuals within a particular distanceclass are more genetically similar thanexpected by random.
D I S P E R S A L A N D E X T I N C T I O N I N F R A G M E N T E D G E C K O P O P U L A T I O N S 3307
assignment tests revealed that the two closest populations(ORF5 and ORF6: 150 m) had the highest percentage ofmisassigned individuals (27% and 50%, respectively).Populations ORF1 and ORF4 and populations ORF2 andORF3 are isolated by 580 m and 550 m, respectively, andthe assignment tests revealed that between 6% and 19%were misassigned for the first population pair, and between0 and 4% for the second. For G. variegata, the highestpercentage (between 40% and 50%) of misassigned indi-viduals was between the two populations GVF3 and GVF4,which were closest in distance (150 m). Populations GVF1and GVF2 are isolated by 300 m and had a misassignment
rate between 17% and 27%. Populations GVF5 and GVF6were 1000 m apart and showed relatively high misassign-ment rates (between 8% and 40%). In the continuous forestpopulations, we observed a higher percentage of misassign-ments for both species (e.g. Korrelocking Nature Reserve:O. reticulata: mean of 50% and 58%, 0.00 and 0.038 SE,respectively; G. variegata: mean of 50% and 53%, 0.00 and0.063 SE, respectively) than in the fragmented populations.Data for the North Bandee Nature Reserve is corresponding.
Using the Bayesian approach with a threshold level ofP ≤ 0.05 for the habitat fragments, only 77 of 174 individualsof O. reticulata and even fewer individuals of G. variegata
Fig. 5 Average probability for the individuals of (a) Oedura reticulata and (b) Gehyra variegata populations having originated in the sourcepopulation or a neighbouring population derived from data of the Bayesian clustering method (structure 2.1). Neighbouring populationpairs are ORF1/ORF4; ORF2/ORF3; ORF5/ORF6 for Oedura reticulate and GVF1/GVF2; GVF3/GVF4; GVF5/GVF6 for Gehyra variegata.
3308 M . H O E H N , S . D . S A R R E and K . H E N L E
(25 of 171) could be assigned. Thus, this level of constraintprovided little information: we identified only nine of 77assigned O. reticulata as dispersers (mean = 3.0 per popu-lation pair) and six of 25 assigned G. variegata (mean = 2.0per population pair). The majority of the O. reticulatadispersers (five) were identified between the two closestpopulations (ORF5 and ORF6) whereas the six G. variegatadispersers were more evenly distributed among the threepopulation pairs. In the Korrelocking Nature Reserve, onlyseven of 12 O. reticulata were assigned as dispersers (mean= 3.5 per population pair) and four of four G. variegata wereassigned (mean = 2.0 per population pair).
For population pairs ORF5/ORF6 and GVF3/GVF4,and for all the population pairs in the two nature reserves,structure 2.1 failed to confidently assign any individualto a single population. Thus, we estimated the number ofclusters (K) for these populations. Log-likelihood valueswere found to be similarly low for K = 1 and K = 2 or lowerfor K = 1. For comparison, we also calculated the numberof K for other population pairs, where log-likelihoodvalues were low for K ≥ 2, but substantially higher forK = 1. This result indicates that the fragments, which arein close geographical distances, and the two nature reservesconstitute a single panmictic population for both species.
Discussion
Dispersal is a key trait of species that avoid extinctionfollowing habitat fragmentation. When populations becomefragmented, dispersal between patches can provide a‘top-up’ or ‘rescue effect’ for small, resident populations,which reduces the probability of local extinction (Brown& Kodric-Brown 1977; Hanski 1999). Understanding theimpacts of human-induced fragmentation upon dispersalas opposed to the impacts of other environmental changesis problematic because of the nearly universal lack ofprefragmentation data in naturally occurring populations(Srikwan & Woodruff 2000; Sumner et al. 2004). In thiscontext, comparative molecular approaches that contrastfragmented and nonfragmented populations, combinedwith detailed field and modelling studies, provide one ofthe few ways of exploring such an important topic (Young& Clarke 2000; Stow et al. 2001; Caizergues et al. 2003;Segelbacher et al. 2003; Williams et al. 2003; Stow & Sunnucks2004a, b).
Our fine-scale microsatellite DNA analysis of two speciesof gecko (Oedura reticulata and Gehyra variegata) demon-strates the usefulness of such an approach. Clear hypothesesemerged from long-term field and modelling studiesand we have demonstrated that the genetic structure offragmented O. reticulata populations is significantly higherthan in G. variegata. In addition, assignment tests revealedthat G. variegata has higher rates of interpatch dispersalcompared with O. reticulata suggesting that small distances
of about 500 m are a barrier to O. reticulata but not for G.variegata. We also observed an isolation-by-distance effectin FST values in G. variegata but not in O. reticulata, whichtogether with spatial autocorrelation analysis suggests thatG. variegata can disperse on a scale approaching a kilometrerather then the few hundred metres achieved by O. reticulata.Overall, O. reticulata exhibits lower levels of heterozygosity,lower allelic richness, elevated levels of structure, fewermisassignments than similarly fragmented populations ofG. variegata, while all these parameters were fairly similarin the continuous forest populations. These findings are ofecological significance to the distribution of populationsthroughout the landscape of the Kellerberrin wheatbelt area.From these results recommendations for possible manage-ment action to prevent extinctions can be derived. Here, wediscuss these patterns and their methodological limitations.
Structure in fragmented and nonfragmented populations
Our analyses, incorporating comparisons among fragmentedand nonfragmented populations and between specieswith contrasting levels of persistence and speciality, revealseveral clear patterns concerning the impact of frag-mentation upon these two species. First, in continuous andrelatively pristine habitat (nature reserves), the geneticstructure and the percentage of misassignments of thetwo species are indistinguishable from each other on ageographical scale of 1–1.2 km. On this scale, geneticstructure was very low for both species (FST = 0.003 OR;0.004 GV). Spatial autocorrelation might be a better approachfor the detection of fine-scale genetic structure when othermethods fail to identify genetic patterns (Double et al. 2005).However, there was no significant positive correlationbetween any of the nature reserve populations. In addition,the number of populations in the nature reserve wasestimated with the program structure 2.1 and the resultindicated that the populations in each of the two naturereserves are approaching panmixis. Second, genetic structurein both species was clearly influenced by fragmentationthrough land clearance for agriculture. For both species,levels of FST were higher and the percentage of mis-assignments was lower among fragmented populationsthan among those in the nature reserves. While wastwice as high in O. reticulata than in G. variegata, theimpacts of fragmentation were still evident in G. variegatawith significantly higher levels of FST among fragmentedpopulations than among nature reserve populations.
Dispersal as a function of genetic and geographical distance
Isolation by distance. The relationship between genetic dif-ferentiation and geographical distance contributes to ourunderstanding whether genetic differentiation is a result of
′GST
D I S P E R S A L A N D E X T I N C T I O N I N F R A G M E N T E D G E C K O P O P U L A T I O N S 3309
limited dispersal or more complex demographic factors(Leblois et al. 2000; Brouat et al. 2003). In general, it has beenargued that if dispersal is limited by distance, at equilibriumbetween mutation, migration, and drift genetic andgeographical distance should be positively correlated. Anegative correlation or no isolation by distance is usuallyinterpreted as evidence that dispersal is not limited bydistance, such that the sampled region functions effectivelyas one large population (Rousset 1997, 2001; Leblois et al.2000; Brouat et al. 2003). However, there are at least twoother reasons why a lack of a significant pattern of isolationby distance may occur. First, gene flow may be very lowover the distance sampled such that populations areessentially isolated, and allele frequencies are determinedby drift. This was recently demonstrated in the alpinebutterfly, Parnassius smintheus, because of reduced con-nectivity of habitat (Keyghobadi et al. 2005). Second, theMantel test might not have enough power to detect geneticstructure at a fine scale.
On the basis of isolation-by-distance analysis alone, wewould interpret that the correlation between geographicaland genetic distance for G. variegata but not for O. reticulatasuggests that G. variegata moves between remnants andO. reticulata does not. The alternative explanation, thatO. reticulata moves on a scale of 10 s of km, is contradictedby the lack of high numbers of misassigned individualsamong closely located populations of O. reticulata. Incontrast, genetic differentiation is positively correlatedwith total geographical distance in G. variegata, whichconforms to the theoretical expectations that the species isdispersing, but that dispersal is limited by distance.
Spatial autocorrelation. The potential contribution of spatialautocorrelation analysis in combination with hypervariableDNA markers has been overlooked in animal studies(Peakall et al. 2003; Double et al. 2005). While the isolation-by-distance analysis failed to identify a landscape-relatedgenetic pattern in O. reticulata, the autocorrelation showedthat genetic structure is present at two distance classes,100 m and 200 m. The intercept occurs at a relatively shortdistance of 518 m, which indicates a low level of dispersaland a small neighbourhood size (Vignieri 2005). Still, thisresult is in rough agreement with the conclusions fromthe isolation-by-distance and assignment analysis. Theperformance of the isolation-by-distance analysis mighthave been compromised by the clumping of the samples.The observed spatial pattern in G. variegata implies that thefragmented landscape is also a barrier to this species, butthat dispersal occurs at a larger scale than in O. reticulata.The spatial autocorrelation remains positive up to 400 mand then begins to decline with an intercept of 951 m. Inboth species, no significant genetic structure has beenidentified in the continuous woodlands suggesting that thepopulations are panmictic.
Assignment. The benefit of the first analysis in our study,the most-likely option, is that gecko individuals do notnecessarily need to be assigned with a high accuracy toobtain an overall estimate of the dispersal probability. Ina second approach for identifying an exact number ofdispersers, the assignment test was then performed at athreshold of P ≤ 0.05. We were able to estimate the percentageof misassignments, which was on average lower for O.reticulata than for G. variegata, suggesting that the numberof dispersers in G. variegata is nearly twice that in O. reticulata.For both species, the percentage of misassignments was onaverage higher between the continuous forest populationsthan between the fragments.
With the semi-experimental approach, we were also ableto determine a rough distance of separation. For O. reticulataa distance of 150 m through cleared habitat did not appearto represent a barrier. Movement between the two pairs ofdistant populations (ORF1/ORF4 and ORF2/ORF3) wassubstantially lower than for those separated by only 150 mindicating that 500–600 m is sufficient to prevent evenmodest dispersal. Three of four individuals detected asimmigrants between these distant pairs were detected ashaving moved between population pair ORF1/ORF4. Thesetwo patches are adjacent to woodland roadside vegetationthat may act as a corridor for O. reticulata movement. Thispossibility deserves future research as it suggests a potentialrole for corridors in the conservation of species.
For G. variegata, the isolation of remnants by 150–300 m ofcleared matrix did not stop migrants from moving throughthe modified landscape. At least one individual covereddistances of up to 1000 m between habitat patches. Datafrom the isolation-by-distance and spatial autocorrelationanalyses suggest that movement over 1 km still occurs. Dis-persal requires individuals to pass through rural landscapesin which the native vegetation has been removed and replacedby crops or stubble after harvest. The environment containsa high density of introduced predators, and a scarcity ofsuitable hiding places (even rocks and fences) that may beused as temporary shelter for the geckos. Hence, successfuldispersal under these circumstances is likely to be a rare event.
We discovered that the number of identified dispersersby assignment test where a threshold was established wassimilar for the two gecko species or even slightly higher inO. reticulata similar in fragmented and continuous land-scapes. Theoretical and empirical research imply that thereis a positive relationship between the level of differenti-ation (FST) and the ability to correctly assign individualsto their natal population (Eldridge et al. 2001; Berry et al.2004). We suggest that the assignment test was more likelyto correctly identify the dispersers for O. reticulata, whichhad an FST of 0.10, than for G. variegata with an FST of 0.04.Therefore, we assume that in the species G. variegata, the actualnumber of dispersers is likely to be higher than estimated.The same might be true for the estimated number of
3310 M . H O E H N , S . D . S A R R E and K . H E N L E
dispersers between population pairs ORF5/6 and GVF3/4,which are in close proximity and between the populationpairs in the continuous forest. The differentiation betweenthese populations is lower and consequently the number ofdispersers could be underestimated. In addition, furtherstructure 2.1 analysis showed that these populationsconsist of one (K = 1) panmictic population.
Conclusion and conservation implications
In the study area, five other terrestrial gecko species(Crenadactylus ocellatus, Diplodactylus granariensis, D. maini,D. pulcher, and D. spinigerus) have already become extinctin most remnant woodlands, presumably as a result of habitatclearing and agriculture. We suggest that the differencesin the persistence of the two remaining gecko species aremainly attributable to their different dispersal ability througha matrix of habitat more suitable for G. variegata than for O.reticulata. Similarly, in western Japan, it was demonstratedthat the house-dwelling gecko Gekko japonicus had higher ratesof gene flow between habitat fragments than its sympatriccounterpart, Gekko tawaensis, a species that is not presentin human constructions. It was argued that this was pro-moted by human-mediated transport (Toda et al. 2003). Thefew available comparative genetic studies of two or moresympatric species support our conclusions (lizards: Branchet al. 2003; small mammals: Matocq et al. 2000; Ehrich et al.2001a, b; insects: Monaghan et al. 2002; Brouat et al. 2003).
From a conservation perspective, the specialist speciesO. reticulata might be a good genetic indicator species formonitoring the impact of anthropogenic perturbations inthe Western Australian wheatbelt. Although single speciescannot fully represent the fauna of fragmented habitats, itmight be a pragmatic approach to genetically monitora species which is especially sensitive to habitat frag-mentation. Typically, conservation managers are required tominimize management costs and time, and genetic monitoringis still relatively expensive and time-consuming.
Acknowledgements
We thank Lachlan Farrington, Alex Quinn, Oliver Berry, and NiccyAitken for laboratory assistance and advice on laboratory tech-niques. Thanks to Bernd Gruber for comments on earlier drafts ofthis manuscript. Thanks to Alex Quinn and Niccy Aitken forproofreading. This work was funded by the DAAD, UniversitätErlangen-Nürnberg and a Collaborative Industry Grant from theUniversity of Canberra, Australia and the UFZ Centre for En-vironmental Research, Leipzig-Halle, Germany.
References
Belkhir K, Borsa P, Chikhi L, Raufaste N, Bonhomme F (2001)GENETIX 4.02, Logiciel Sous Windows TM Pour la Genetique DesPopulation. Laboratoire Genome Population, Interactions: CNRSUMR 5000, University of Montpellier, Montpellier, France.
Berry O, Tocher MD, Sarre SD (2004) Can assignment tests meas-ure dispersal? Molecular Ecology, 13, 551–561.
Branch LC, Clark A-M, Moler PE, Bowen BW (2003) Habitat spe-cificity, fragmented landscapes, and the phylogeography ofthree lizards in Florida scrub. Conservation Genetics, 4, 199–212.
Brouat C, Sennedot F, Audiot P, Leblois R, Rasplus JY (2003) Fine-scale genetic structure of two carabid species with contrastedlevels of habitat specialization. Molecular Ecology, 12, 1731–1745.
Brown JH, Kodric-Brown A (1977) Turnover rates in insularbiogeography: effect of immigration on extinction. Ecology, 58,445–449.
Caizergues A, Ratti O, Helle P (2003) Population genetic structureof male black grouse (Tetrao tetrix L.) in fragmented vs. continuouslandscapes. Molecular Ecology, 12, 2297–2305.
Clobert J, Danchin E, Dhondt AA (2001) Dispersal. Oxford Univer-sity Press, Oxford.
Cornuet JM, Piry S, Luikart G, Estoup A, Solignac M (1999) Newmethods employing multilocus genotypes to select or excludepopulations as origins of individuals. Genetics, 153, 1989–2000.
Davies KF, Margules CR, Lawrence KF (2000) Which traitsof species predict population declines in experimental forestfragments? Ecology, 81, 1450–1461.
Davies KF, Melbourne BA, Margules CR (2001) Effects of within-and between-patch processes on community dynamics in afragmentation experiment. Ecology, 82, 1830–1846.
Davies KF, Margules CR, Lawrence JF (2004) A synergistic effectputs rare, specialized species at greater risk of extinction. Ecology,85, 265–271.
Double MC, Peakall R, Beck NR (2005) Dispersal, philopatry, andinfidelity: dissecting local genetic structure in superb fairy-wrens (Malurus cyaneus). Evolution, 59, 625–635.
Ehrich D, Krebs CJ, Kenney AJ, Stenseth NC (2001a) Comparingthe genetic population structure of two species of arcticlemmings: more local differentiation in Lemmus trimucronatusthan in Dicrostonyx groenlandicus. Oikos, 94, 143–150.
Eldridge MDB, Kinnear JE, Onus ML (2001) Source population ofdispersing rock-wallabies (Petrogale lateralis) identified by assign-ment tests on multilocus genotypic data. Molecular Ecology, 10,2867–2876.
Frankham R, Ballou JD, Briscoe DA (2002) Introduction to Conser-vation Genetics. Cambridge University Press, Cambridge, UK.
Goudet J (1995) fstat version 1.2: a computer program to calculateF-statistics. Journal of Heredity, 86, 485–486.
Goudet J (2001) FSTAT: A Program to Estimate and Test Gene Diversitiesand Fixation Indices (Version 2.9.3). Available at URL: www.unil.ch/izea/softwares/fstat.html
Groombridge B (1992) Global Biodiversity. Chapman & Hall, London.Hanski I (1999) Metapopulation Ecology. Oxford University Press,
Oxford, New York.Hedrick PW (1999) Perspective: highly variable loci and their inter-
pretation in evolution and conservation. Evolution, 53, 313–318.Hedrick PW (2005) A standardized genetic differentiation measure.
dictors of species sensitivity to fragmentation. Biodiversity andConservation, 13, 207–251.
Henle K, Lindenmayer DB, Margules CR, Saunders DA, Wissel C(2004b) Species survival in fragmented landscapes: where arewe now? Biodiversity and Conservation, 13, 1–8.
D I S P E R S A L A N D E X T I N C T I O N I N F R A G M E N T E D G E C K O P O P U L A T I O N S 3311
Hobbs RJ (1993) Effects of landscape fragmentation on ecosystemprocesses in the Western Australian wheatbelt. Biological Con-servation, 64, 193–201.
Hoehn M (2006) Conservation genetics of two gecko species in afragmented landscape. PhD Thesis. University of Leipzig,Leipzig, Germany.
Hoehn M, Sarre S (2005) Tetranucleotide microsatellites in thegecko Oedura reticulata isolated from an enriched library.Molecular Ecology Notes, 5, 730–732.
Hoehn M, Sarre S (2006) Microsatellite DNA markers for Australiangeckos. Conservation Genetics, 7, 795–798.
How RA, Kitchener DJ (1983) The biology of the Gecko Oedurareticulata Bustard, in a small habitat isolate in the WesternAustralian wheatbelt. Australian Wildlife Research, 10, 543–556.
Keller LF, Waller DM (2002) Inbreeding effects in wild populations.Trends in Ecology & Evolution, 17, 230–241.
Keyghobadi N, Roland J, Strobeck C (2005) Genetic differentiationand gene flow among populations of the alpine butterfly,Parnassius smintheus, vary with landscape connectivity. MolecularEcology, 14, 1897–1909.
Kitchener DJ, How RA, Dell J (1988) Biology of Oedura reticulataand Gehyra variegata (Gekkonidae) in an isolated woodland ofWestern Australia. Journal of Herpetology, 22, 401–412.
Koenig WD, VanVuren D, Hooge PN (1996) Detectability, philo-patry, and the distribution of dispersal distances in vertebrates.Trends in Ecology & Evolution, 11, 514–517.
Leblois R, Rousset F, Tikel D, Moritz C, Estoup A (2000) Absenceof evidence for isolation by distance in an expanding cane toad(Bufo marinus) population: an individual-based analysis ofmicrosatellite genotypes. Molecular Ecology, 9, 1905–1909.
MacNally R, Bennett AF, Horrocks G (2000) Forecasting theimpacts of habitat fragmentation. Evaluation of species-specificpredictions of the impact of habitat fragmentation on birds inthe box-ironbark forests of central Victoria, Australia. BiologicalConservation, 95, 7–29.
Matocq MD, Patton JL, da Silva MNF (2000) Population geneticstructure of two ecologically distinct Amazonian spiny rats:separating history and current ecology. Evolution, 54, 1423–1432.
Monaghan MT, Spaak P, Robinson CT, Ward JV (2002) Populationgenetic structure of 3 alpine stream insects: influences of geneflow, demographics, and habitat fragmentation. Journal of theNorth American Benthological Society, 21, 114–131.
Moritz C (1987) The population biology of Gehyra (Gekkonidae). 2.Dispersal and colonisation. Australian Journal of Zoology, 35,587–596.
Otis DL, Burnham KP, White GC, Anderson DR (1978) Statisticalinference for capture-recapture data in closed animal populations.Wildlife Monographs, 62.
Paetkau D, Calvert W, Stirling I, Strobeck C (1995) Microsateliteanalysis of population structure in Canadian polar bears.Molecular Ecology, 4, 354.
Peakall R, Smouse PE (2001) GENALEX version 5: Genetic Analysis inExcel. Population Genetic Software for Teaching and Research.Australian National University, Canberra, Australia.
Peakall R, Ruibal M, Lindenmayer DB (2003) Spatial autocorrela-tion analysis offers new insights into gene flow in the Australianbush rat, Rattus fuscipes. Evolution, 57, 1182–1195.
Pritchard JK, Stephens M, Donnelly P (2000) Inference of populationstructure using multilocus genotype data. Genetics, 155, 945–959.
Rannala B, Mountain JL (1997) Detecting immigration by usingmultilocus genotypes. Proceedings of the National Academy ofSciences, USA, 94, 9197–9201.
Rousset F (1997) Genetic differentiation and estimation of geneflow from F-statistics under isolation by distance. Genetics, 145,1219–1228.
Rousset F (2001) Genetic approaches to the estimation of dispersalrates. In: Dispersal (eds Clobert J, Danchin E, Dhondt AA),pp. 18–28. Oxford University Press, Oxford, New York.
Sarre S (1995a) Size and structure of populations of Oedura reticu-lata (Gekkonidae) in woodland remnants: implications for thefuture regional distribution of a currently common species.Journal of Ecology, 20, 288–298.
Sarre S (1995b) Mitochondrial DNA variation among populationof Oedura reticulata (Gekkonidae) in remnant vegetation: impli-cations for metapopulation structure and population decline.Molecular Ecology, 4, 395–405.
Sarre S (1998) Demographics and population persistence of Gehyravariegata (Gekkonidae) following habitat fragmentation. Journalof Herpetology, 32, 153–162.
Sarre S, Smith GT, Meyers JA (1995) Persistence of two species ofgecko (Oedura reticulata and Gehyra variegata) in remnant habitat.Biological Conservation, 71, 25–33.
Sarre S, Wiegand K, Henle K (1996) The conservation biology of aspecialist and generalist gecko in the fragmented landscape of theWestern Australian wheatbelt. In: Species Survival in FragmentedLandscapes: from the Local to the Landscape Level (eds Settele J,Margules C, Poschlod P, Henle K), pp. 39–51. Kluwer AcademicPublishers, Dordrecht, The Netherlands.
Saunders DA, Hobbs RJ (1991) Nature Conservation 2: The Roleof Corridors. Surrey Beatty and Sons Pty Ltd, Chipping Norton,Australia.
Saunders DA, Hobbs RJ, Arnold GW (1993) The Kellerberrinproject on fragmented landscapes: a review of current information.Biological Conservation, 64, 185–192.
Schmuki C, Vorburger C, Runciman D (2006) When log-dwellersmeet loggers: impacts of forest fragmentation on two endemiclog-dwelling beetles in southeastern Australia. Molecular Ecology,15, 1481–1492.
Segelbacher G, Hoglund J, Storch I (2003) From connectivity toisolation: genetic consequences of population fragmentation incapercaillie across Europe. Molecular Ecology, 12, 1773–1780.
Sokal RR, Oden NL (1978) Spatial autocorrelation in biology. 1.Methodology. Biological Journal of the Linnean Society, 10, 199–228.
Srikwan S, Woodruff DS (2000) Genetic erosion in isolated small-mammal populations following rainforest fragmentation. In:Genetics, Demography and Viability of Fragmented Populations (edsYoung A, Clarke GM), pp. 149–172. Cambridge UniversityPress, Cambridge, UK.
Stow AJ, Sunnucks P (2004a) High mate and site fidelity inCunningham’s skinks (Egernia cunninghami) in natural andfragmented habitat. Molecular Ecology, 13, 419–430.
Stow AJ, Sunnucks P (2004b) Inbreeding avoidance in Cunning-ham’s skinks (Egernia cunninghami) in natural and fragmentedhabitat. Molecular Ecology, 13, 443–447.
Stow AJ, Sunnucks P, Briscoe DA, Gardner MG (2001) The impactof habitat fragmentation on dispersal of Cunningham’s skink(Egernia cunninghami): evidence from allelic and genotypic analysesof microsatellites. Molecular Ecology, 10, 867–878.
Sumner J, Rousset F, Estoup A, Moritz C (2001) ‘Neighbourhood’size, dispersal and density estimates in the prickly forest skink(Gnypetoscincus queenslandiae) using individual genetic anddemographic methods. Molecular Ecology, 10, 1917–1927.
Sumner J, Jessop T, Paetkau D, Moritz C (2004) Limited effect ofanthropogenic habitat fragmentation on molecular diversity in
3312 M . H O E H N , S . D . S A R R E and K . H E N L E
a rain forest skink, Gnypetoscincus queenslandiae. Molecular Ecol-ogy, 13, 259–269.
Toda M, Hikida T, Okada S, Ota H (2003) Contrasting patterns ofgenetic variation in the two sympatric geckos Gekko tawaensisand G. japonicus (Reptilia: Squamata) from western Japan, asrevealed by allozyme analyses. Heredity, 90, 90–97.
Vignieri SN (2005) Streams over mountains: influence of riparianconnectivity on gene flow in the Pacific jumping mouse (Zapustrinotatus) Molecular Ecology, 14, 1925–1937.
Walsh PS, Metzger DA, Higuchi R (1991) Chelex-100 as a mediumfor simple extraction of DNA for PCR-based typing from forensicmaterial. Biotechnology, 10, 506–513.
WCMC (1992) Global Biodiversity: Status of the Earth’s LivingResources. Chapman & Hall, London.
Weir BS, Cockerham CC (1984) Estimating F-statistics for theanalysis of population structure. Evolution, 38, 1358–1370.
Wiegand K, Sarre SD, Henle K et al. (2001) Demographic stochasticity
does not predict persistence of gecko populations. EcologicalApplications, 11, 1738–1749.
Wiegand K, Henle K, Sarre SD (2002) Extinction and spatial struc-ture in simulation models. Conservation Biology, 16, 117–128.
Williams BL, Brawn JD, Paige KN (2003) Landscape scale geneticeffects of habitat fragmentation on a high gene flow species:Speyeria idalia (Nymphalidae). Molecular Ecology, 12, 11–20.
Young A, Clarke GM (2000) Genetics, Demography and Viability ofFragmented Populations. Cambridge University Press, Cambridge,UK.
The study presented here is part of Marion Hoehn’s PhD research.She is interested in the application of molecular approaches toquestions in conservation and evolution, especially with the focuson reptiles.
