Genes and song: genetic and social connections in fragmented habitat in a woodland bird with limited...

Ecology, 93(7), 2012, pp. 1717–1727� 2012 by the Ecological Society of America

Genes and song: genetic and social connections in fragmented habitatin a woodland bird with limited dispersal

ALEXANDRA PAVLOVA,1,5 J. NEVIL AMOS,1 MARIA I. GORETSKAIA,2 IRINA R. BEME,2 KATHERINE L. BUCHANAN,3

NAOKO TAKEUCHI,1 JAMES Q. RADFORD,4,6 AND PAUL SUNNUCKS1

1School of Biological Sciences and Australian Centre for Biodiversity, Monash University, Melbourne, Clayton Campus,Clayton VIC 3800 Australia

2Faculty of Biology, Moscow State University, 1-12 Leninskie Gory, Moscow, Russia 1199913School of Life and Environmental Sciences, Faculty of Science and Technology, Deakin University, Locked Bag 20000,

Geelong VIC 3220 Australia4Landscape Ecology Research Group, School of Life and Environmental Sciences, Deakin University, 221 Burwood Highway,

Burwood VIC 3125 Australia

Abstract. Understanding the processes leading to population declines in fragmentedlandscapes is essential for successful conservation management. However, isolating theinfluence of disparate processes, and dispersal in particular, is challenging. The Grey Shrike-thrush, Colluricincla harmonica, is a sedentary woodland-dependent songbird, with learnedvocalizations whose incidence in suitable habitat patches falls disproportionally with decline intree cover in the landscape. Although it has been suggested that gaps in tree cover might act asbarriers to its dispersal, the species remains in many remnants of native vegetation inagricultural landscapes, suggesting that it may have responded to habitat removal andfragmentation by maintaining or even increasing dispersal distances. We quantifiedpopulation connectivity of the Grey Shrike-thrush in a system fragmented over more than120 years using genetic (microsatellites) and acoustic (song types) data. First, we tested forpopulation genetic and acoustic structure at regional and local scales in search of barriers todispersal or gene flow and signals of local spatial structuring indicative of restricted dispersalor localized acoustic similarity. Then we tested for effects of habitat loss and fragmentation ongenetic and acoustic connectivity by fitting alternative models of mobility (isolation-by-distance [the null model] and reduced and increased movement models) across treeless vs. treedareas. Birds within ;5 km of each other had more similar genotypes and song types than thosefarther away, suggesting that dispersal and song matching are limited in the region. Despiterestricted dispersal detected for females (but not males), populations appeared to be connectedby gene flow and displayed some cultural (acoustic) connectivity across the region.Fragmentation did not appear to impact greatly the dispersal of the Grey Shrike-thrush:none of the mobility models fit the genetic distances of males, whereas for females, anisolation-by-distance model could not be rejected in favor of the models of reduced orincreased movement through treeless gaps. However, dissimilarities of the song types weremore consistent with the model of reduced cultural connectivity through treeless areas,suggesting that fragmentation impedes song type sharing in the Grey Shrike-thrush. Our paperdemonstrates that habitat fragmentation hinders important population processes in anAustralian woodland bird even though its dispersal is not detectably impacted.

Key words: acoustics; central Victoria, Australia; Colluricincla harmonica; Grey Shrike-thrush;habitat fragmentation; isolation-by-distance; isolation-by-resistance; landscape bioacoustics; landscapegenetics; landscape resistance; population genetics; spatial autocorrelation.

INTRODUCTION

Understanding processes underlying patterns of pop-

ulation decline is critical for effective management of

populations in fragmented landscapes (Ewers and

Didham 2005, Lindenmayer et al. 2008). Species richness

and incidence of many individual species of woodland

birds decline with lowered habitat cover, and some

species disappear from apparently suitable habitat in

landscapes with tree cover below a species-specific

threshold (Radford et al. 2005, Radford and Bennett

2007). Dispersal is one vital and poorly understood

process that could influence population viability in this

system. Reduced dispersal as a result of habitat

fragmentation can arrest supplementation of declining

populations and colonization of vacant habitat, alter

social behavior and age class distributions, and result in

loss of genetic diversity, inbreeding, and increased risk

Manuscript received 19 October 2011; revised 19 January2012; accepted 23 January 2012. Corresponding Editor: W. D.Koenig.

5 E-mail: [email protected] Present address: Bush Heritage Australia, P.O. Box 329

Flinders Lane, Melbourne VIC 8009 Australia.

1717

of local extinction (Frankham 2005, Sunnucks 2011).

However, bird assemblages that remain in forest

fragments mainly comprise species that either increase

dispersal distances after fragmentation or are tolerant of

the non-forest matrix (Van Houtan et al. 2007, Lees and

Peres 2008), and evidence that dispersal distance may be

greater in fragmented habitat is fairly common (Schmu-

ki et al. 2006, Coulon et al. 2010). Therefore, even

species declining in fragmented landscapes could have

responded to fragmentation by maintaining or increas-

ing dispersal distances. Although many landscape

genetic studies explicitly test negative effects of frag-

mentation on dispersal (Shanahan et al. 2010), few have

tested the alternative model of increased dispersal.

Despite its utility and increasing use, a landscape

genetic approach can present challenges in distinguish-

ing alternative models when dispersal of the organisms is

restricted by distance, because alternative models are

often highly correlated (Balkenhol et al. 2009, Amos et

al. 2012). Therefore an independent data set on

population connectivity would be beneficial. Acoustic

(vocalization) data in birds have great potential to show

the effects of fragmentation on coherence of social

interactions over relatively short time frames (years to

decades), owing to rapid cultural transmission and

change (Laiolo 2010). Smaller and more isolated

populations may be associated with lower acoustic

diversity and higher differentiation among them because

there are fewer opportunities for learning from conspe-

cifics and retaining acoustic diversity (Laiolo and Tella

2005, 2006, 2007). In addition, neighboring birds can

form local communication networks (McGregor and

Dabelsteen 1996) and match their songs to those of their

neighbors (Catchpole and Slater 2008); thus a change in

the distance over which such matching occurs in

fragmented habitats could indicate disruptions in

acoustic connectivity. Few studies to date have used

acoustic data to examine the effects of habitat fragmen-

tation (Laiolo and Tella 2005, Robin et al. 2011) or to

test spatially explicit predictions about acoustic connec-

tivity of populations (Laiolo and Tella 2006), and those

that have tend to explore acoustic variation at large

geographic scales.

In this paper, we show that acoustic data can be used

successfully to detect effects of habitat fragmentation

on population connectivity at a relatively small spatial

scale, even when genetic data do not readily distinguish

among alternative models of dispersal. To do this, we

use genotypic and acoustic data for the Grey Shrike-

thrush Colluricincla harmonica; see Plate 1), a sedentary

woodland-dependent bird, whose incidence and abun-

dance decline disproportionately with decreasing tree

cover in the landscape (Radford and Bennett 2007) and

fall dramatically in landscapes with ,5% tree cover

(Amos et al. 2012). In a field study of mobility,

movements of individual Grey Shrike-thrushes was

significantly reduced by gaps (;170 m) in forest cover,

suggesting that loss of structural connectivity will

reduce dispersal for this species (Robertson and

Radford 2009). Nevertheless, the Grey Shrike-thrush

remains in remnants of native vegetation in agricultural

landscapes, suggesting that it may have responded to

fragmentation by maintaining, or even increasing, its

dispersal distances. On the other hand, fragmentation

could have reduced its dispersal, but not enough time

has passed since habitat clearance for stochastic effects

to result in local extinctions (Ford 2011). The Grey

Shrike-thrush is socially monogamous and both sexes

contribute almost equally to incubation, brooding, and

feeding nestlings (Higgins and Peter 2002, Stevens and

Watson 2005). Males and females both sing and there

is some evidence of localized song type matching,

matching songs to those being played during playback

experiments and mimicking other species (Higgins and

Peter 2002). Thus vocalizations are probably learned

over a bird’s lifetime (as shown for other species within

the family Pachycephalidae; Brown and Brown 1994).

The objectives of this study were to (1) test for the

presence of genetic and acoustic structure at regional

and local scales in search of restriction of dispersal or

gene flow and signals of local spatial structuring or

localized acoustic similarity; and (2) test for the effect

of habitat loss and fragmentation on genetic and

acoustic connectivity by fitting a set of alternative

models of mobility that cover all the logical possibil-

ities (isolation-by-distance [null], reduced, and in-

creased) across treeless vs. treed areas.

METHODS

Study region and landscape choice

The study region was ;10 000 km2 of central Victoria

in southeastern Australia, where only 19.2% tree cover

remains after extensive habitat clearance for timber,

mining, and agriculture, which started ;150 year ago

with the arrival of European settlers (Fig. 1). Remnant

native vegetation is principally box-ironbark forest

(dominated by grey box Eucalyptus microcarpa, red

ironbark E. tricarpa, and yellow gum E. leucoxylon) with

the intervening land composed mainly of cleared

farmland, with scattered trees in parts. Sampling

occurred within 12 103 10 km landscapes: nine of these

landscapes were studied by Radford et al. (2005) and

ranged in tree cover from 11% to 45% (herein

‘‘fragmented landscapes’’). Three ‘‘reference’’ landscapes

were selected with the highest available tree cover (72–

78%) to approximate continuous tree cover (Fig. 1;

Appendix A). Landscapes very low in cover (,10%)

were not included, due to low probability of encounter-

ing Grey Shrike-thrush (Radford and Bennett 2007).

Because there is usually a time lag between habitat loss

and local extinction (‘‘extinction debt’’; Ford et al. 2009,

Szabo et al. 2011), changes in spatial genetic or acoustic

structure in response to fragmentation could signal the

processes leading to declines in apparently stable,

fragmented populations.

ALEXANDRA PAVLOVA ET AL.1718 Ecology, Vol. 93, No. 7

Field methods, DNA sampling, extraction, and genotyping

Mist-netting was conducted at 63 sites across the 12

landscapes from November 2007 to February 2010;

nearly all sites were visited twice, ;6 months apart. We

captured 97 Grey Shrike-thrush individuals (56 males

and 41 females) from 39 sites (1–6 individuals per site)

spread across all landscapes (Fig. 1; Appendix A: Table

A1). Birds were banded, sexed, and aged as juvenile,

immature, or adult (following Rogers et al. 1986).

A small blood sample (,100 lL) was taken from a

brachial vein of each bird and stored in absolute ethanol

at�208C for genetic analysis. DNA was extracted using

a ‘‘salting out’’ protocol; samples were screened for 14

nuclear loci, including 11 microsatellites and three

length-variable exon-primed-intron-crossing regions

(details in Appendix B). GENEPOP 4.0.10 (Raymond

and Rousset 1995) was used to check loci for significant

deviations from Hardy–Weinberg (HW) and linkage

(LD) equilibria for each landscape and site; all loci were

retained for analyses because they had no detectable null

alleles and segregated independently (Appendix B). The

genetic assay applied had enough power to detect

reduction in dispersal that would theoretically be

sufficient to result in harmful inbreeding (Appendix B).

Field-based sex identification was supplemented by

applying the standard molecular protocol (Appendix B).

Detecting regional and local genetic structure

Presence of historic population structure (from allele

frequencies) was tested using analysis of molecular

variance (AMOVA; Excoffier et al. 1992) in Arlequin

3.5.1.2 (Excoffier and Lischer 2010) with landscapes,

sites, and individuals as three hierarchical levels.

Pairwise landscape FST values, which reflect differenti-

ation accumulated over several to many generations,

were computed in Arlequin. For the five largest

landscape samples (N . 11, a total of 71 individuals),

the number of first-generation immigrants into each

landscape was estimated using GENECLASS2 (Piry et

al. 2004).

Contemporary geographic structure (from genotypes)

was explored using TESS 2.3.1 (Chen et al. 2007;

Appendix C). Regional (the entire study area) and local-

scale (within-landscape) spatial genotypic autocorrela-

tion was explored in GENALEX 6.41 (Peakall and

Smouse 2006; Appendix C). For regional-scale analysis,

distance classes were 0–1 km (reflecting within-site

distances), .1–15 km (among-sites within landscape),

.15–50 km (neighboring landscapes), and .50–165 km

(across the study area). For local analysis, distance

classes were 0–1, .1–5, .5–10, and .10–15 km. Local

analyses were run on all samples and then on only those

from fragmented landscapes (,50% tree cover) to gain

insights on potential impacts of fragmentation on

genotypic structure, but low sample size (N ¼ 20)

precluded separate analysis of reference landscapes.

Acoustic recordings and data collection

Recordings were made from 16 September 2009 to 9

November 2009, during the breeding season of the Grey

Shrike-thrush. Most individuals were unmarked (and

sex unknown); hence, each bird was recorded during a

single visit to avoid resampling the same individual.

Sequences of songs (of 5–15 minutes in length) were

recorded at singing posts. We recorded 103 individual

birds in 29 sites (1–11 birds recorded per site) located in

six landscapes that represented a range of tree cover

(Fig. 1; Appendix A: Table A2).

Song type variation, which is likely to be culturally

transmitted (through learning or song matching) and

FIG. 1. Study area, sampling sites, and sample sizes for genetic (black circles) and acoustic (small while squares) samples;genetic sample size precedes an acoustic one for sites where both data were collected (black circles within white squares). Largesquares are the 10 3 10 km landscapes, and gray shading shows native vegetation cover. The inset map shows the location of thestudy area in Australia.

July 2012 1719GENES AND SONGS IN FRAGMENTED LANDSCAPES

often shows geographic structure in birds (reviewed in

Catchpole and Slater 2008), was assessed in this study.

Songs were analyzed with AVISOFT SASLAB PRO (R.

Specht, Berlin, Germany; Appendix C; software avail-

able online).7 Presence or absence of unique song types in

each bird’s repertoire was recorded. Song types, defined

as a visually distinct sequence of syllables, were

identified visually from sound spectrograms by a

consensus of two researchers, M. I. Goretskaia and

I. R. Beme (see Fig. 2 for examples). The number of

song types in an individual’s repertoire is likely to vary

with age and sex (which we did not control for), but

published data on the Grey Shrike-thrush’s acoustic

communications is lacking. Because recordings were not

designed to capture the whole repertoires of all birds, the

number of song types detected in each site or landscape

was correlated with the number of analyzed birds (R2¼0.63, P , 0.001 for sites; R2 ¼ 0.73, P ¼ 0.02 for

landscapes) and songs (R2¼0.74, P , 0.001 for sites; R2

¼0.78, P¼0.01 for landscapes). To standardize the song

type data, only a subsample of 10 consecutive songs per

bird to a maximum of 10 randomly chosen birds per

FIG. 2. Sound spectrograms showing examples of four distinct song types observed for the Grey Shrike-thrush (Colluricinclaharmonica) during this study.

7 http://www.avisoft.com/soundanalysis.htm


landscape (56 birds, 553 songs overall) was used for song

type analyses (Appendix A: Table A2); from this subset

the number of song types per landscape and the number

of shared song types (those represented in a song of

more than one bird) per landscape were calculated.

Detecting regional and local acoustic structure

Analysis of acoustic spatial autocorrelation can reveal

patterns of restricted acoustic connectivity (Laiolo and

Tella 2006). We used spatial autocorrelation analysis to

explore whether acoustic similarity is higher for closer

neighbors, as would be expected if communication were

limited by distance or if birds matched their songs to local

neighbors (Catchpole and Slater 2008). R 2.13.0 (R

Development Core Team 2011) was used to compute

geographic distances and acoustic dissimilarities ex-

pressed as Bray-Curtis dissimilarities (appropriate for

presence–absence song type data) for each pair of birds

(Appendix C). Acoustic spatial autocorrelation was

explored in GENALEX 6.4 on regional (study area)

and local spatial scales. The same distance classes as for

genetic analysis were used (except that the upper bound

was 124 km, the maximum distance between acoustic

recordings). Local analyses were run on all data and on

fragmented landscapes only, but low sample size in the

reference landscape (N¼ 10) precluded separate analysis.

Testing landscape models of gene flow and song flow

Modeling movement through landscapes with differ-

ent context-dependent ‘‘resistances’’ using electrical

circuit theory (isolation-by-resistance; McRae 2006) is

a powerful approach to describe gene flow and mobility

through different land uses (McRae and Beier 2007); it

builds maps of landscape resistance to movement

(resistance surfaces) by integrating across multiple

movement paths. Using Circuitscape 3.5.1 (McRae et

al. 2008, McRae and Shah 2008), Amos et al. (2012)

classified each 25 3 25 m cell in our study region as

either treeless or treed, generated resistance surfaces for

(1) the null model of isolation-by-distance (IBD, which

assumes uniform resistances across the grid: resistance

of treed habitat ¼ 1, treeless ¼ 1) and (2) the model of

reduced movement through treeless areas (RM, which

assumes doubled resistance of treeless cells: treed 1,

treeless 2), and calculated pairwise resistances for each

pair of sampling sites. Here we use genetic and acoustic

distances to test three models (methods as in Amos et al.

2012): (1) IBD; (2) reduced movement through treeless

areas (RM), which predicts that less dispersal or ‘‘song

flow’’ will occur through treeless than treed areas; and

(3) increased movement through treeless areas (IM),

which assumes that treeless cells have lower resistance

(treed 2, treeless 1) and thus higher gene flow or song

flow across treeless areas.

Pairwise resistances from all three isolation-by-resis-

tance models were tested for correlation with pairwise

individual genetic (Smouse and Peakall 1999) and song

type (Bray-Curtis) dissimilarities using Mantel tests. Then

partial Mantel tests were used to separate the effects of

RM and IM models from those due to geographicdistances (IBD): RM or IM models were considered to fit

better than IBD if partial Mantel test conditioned ongeographic distance resulted in a significant r. Mantel and

partial Mantel tests were performed in R (Appendix C);those involving genetic distances were performed on alldata, and with sexes separated.

RESULTS

Gene flow across the study areabut detectable local structure

Unrestricted historic and contemporary gene flow

across the study area were inferred from analyses ofallele frequencies and genotypes, respectively. AMOVA

showed that the majority of variation in allele frequen-cies was explained by intra-individual differences

(93.5%, P , 0.001), some by differences amongindividuals within sites (4.2%, P ¼ 0.01), and little bysites or landscapes (1.43% and 0.92%, respectively, P .

0.05). Six out of 60 pairwise FST values (allele frequencydifferentiation) were significant, four of them involving

the Havelock sample (FST ¼ 0.02, 0.02, 0.05, and 0.03with Glenalbyn, Redcastle, Rushworth, and Wehla,

respectively; P , 0.05), Glenalbyn-Axe Creek (FST ¼0.03, P¼ 0.02), and Stuart Mill-Murchison (FST¼ 0.04,

P¼ 0.03). Nearby landscapes appeared to be connectedby gene flow sufficient to prevent harmful effects of local

inbreeding (Lowe and Allendorf 2010), because noneighboring landscape pair showed significant differen-

tiation, and differentiation caused by strongly limitedgene flow since deforestation 150 years ago would be

expected to be detected by our genetic assay (AppendixB). TESS did not detect any geographic clustering of

genotypes, indicating no strong barriers to gene flowwithin the study area.

Significant spatial genetic autocorrelation was detect-ed at the regional (study-wide) scale: birds at a site

(distances 0–1 km apart) and a landscape (.1–15 km)were genotypically more similar to each other thanrandom, whereas birds sampled at larger distances (.50

km) were less similar (Fig. 3; Appendix D: Table D1).This decreasing genotypic similarity with distance was

probably generated by restricted dispersal rather thantemporary clustering of pre-dispersive immatures in

natal territories, because the patterns held when adultsalone were analyzed (Appendix D: Table D1). In local-

scale analysis of all landscapes, birds at distances of upto 5 km, but not farther, were genotypically more similar

than random (Fig. 3; Appendix D: Table D2); localgenotypic similarity extended to 10 km when only

fragmented landscapes were analyzed (Appendix D:Table D2).

Evidence for male-biased dispersal

Although spatial autocorrelation patterns for males

and females were not significantly different overall (P¼0.79) or for any distance class (P . 0.5), local genetic


structure was more pronounced in females, consistent

with male-biased dispersal: spatial autocorrelation was

not significant at any distance class for males, but was

significantly positive at the site scale (0–1 km) and

significantly negative at the regional scale (.51 km) for

females, and genetic similarity at a site (0–1 km) was

lower for males than females (albeit not significantly so;

males, rM¼ 0.43; females, rF¼ 0.81; Appendix D: Table

D1). Assignment tests presented further evidence of

higher mobility in males: the two putative first-

generation immigrants (of 71 individuals) detected by

GENECLASS2 were males, one adult (Glenalbyn) and

one immature (Wehla).

Regional and local acoustic structure and song type sharing

Overall, 69 song types were detected (mean: 1.9 types/

bird; range: 1–4 types/bird); of them, 11 were shared by

birds from two (N¼ 9) or three (N¼ 2) landscapes that

are up to ;120 km apart (e.g., song type 12 was shared

by birds from Wehla, Havelock, and Redcastle; Appen-

dix A: Table A2). At the regional scale, significant

spatial autocorrelation was detected for song types (Fig.

3; Appendix D: Table D1): song types were significantly

alike at a site and within a landscape, and less similar at

large distances (.50 km), conforming to an isolation-

by-distance pattern. Local-scale analyses of all land-

scapes showed that significant autocorrelation of song

types within landscapes was limited to 5-km distances

(as for genotypic similarity; Fig. 3; Appendix D: Table

D2); this was also true for fragmented landscapes only,

but song types at .5–10 km distances were significantly

dissimilar (Appendix D: Table D2).

Tests of landscape models

The model of isolation-by-distance (IBD) provided

the best fit for pairwise genotypic distances for all

individuals (Mantel r ¼ 0.046, P ¼ 0.034) and females

(Mantel r¼ 0.105, P¼ 0.007), but was not significant for

males (Mantel r¼ 0.055, P¼ 0.075; Table 1), supporting

greater male dispersal distances. Female genotypic

distances also correlated with resistances derived from

two contrasting landscape models of reduced (RM

Mantel r ¼ 0.1, P ¼ 0.039) and increased (IM Mantel r

¼ 0.098, P ¼ 0.008; Table 1) dispersal through treeless

cells, but none of the partial Mantel tests resulted in a

significant correlation (Table 1). Thus, the model of

isolation-by-distance cannot be rejected in favor of

models of reduced or increased dispersal through cleared

FIG. 3. Study-scale spatial autocorrelation analysis of genotypes and song types for regional (top row) and local (bottom row)spatial scales (x-axis endpoints of distance classes). The spatial autocorrelation coefficient r is on the y-axis; r . 0 indicates positivespatial autocorrelation (individuals at a given distance class are more similar than at random), r , 0 indicates negative spatialautocorrelation (less similar than at random). The shaded areas outline ranges of r under random association (95% confidenceintervals about zero) in 999 permutations; error bars show 95% confidence intervals estimated with 999 bootstraps; the asteriskindicates significance of r (P , 0.05) according to permutation tests.


areas for the genotypic data of females. Male dispersal

does not appear to be impacted by distances or

landscape resistances (fragmentation) on the scale of

our study, as none of the mobility models were

supported (Table 1).

Pairwise song type dissimilarity was correlated with

all landscape models in Mantel tests (IBD Mantel r ¼0.15, RM Mantel r¼ 0.16, and IM Mantel r¼ 0.13; P ,

0.001; Table 1), but in partial Mantel tests, the model of

reduced mobility through cleared areas provided signif-

icant fit (RM partial Mantel r ¼ 0.051, P ¼ 0.036),

suggesting that habitat loss and fragmentation impede

song type sharing in the Grey Shrike-thrush.

DISCUSSION

Locally restricted dispersal does not prevent regional

genetic connectivity

Despite apparently limited dispersal of the Grey

Shrike-thrush at a local scale (Fig. 3) resulting in overall

isolation-by-distance effect (Table 1), we found evidence

of gene flow across the study area, reflected in a lack of

population genetic structure (AMOVA, TESS) and no

significant pairwise FST between neighboring landscapes.

This suggests that there is mobility, at least of a low

level, at the regional scale. Lack of differentiation is not

likely to result from a time lag effect (insufficient time

for differentiation), because arrest or significant hin-

drance of gene flow (to ,1 migrant per generation,

which can lead to harmful inbreeding; Lowe and

Allendorf 2010) across areas cleared during the gold

rush was expected to be manifested in significant

pairwise FST and overall genotypic structure (Appendix

B; Harrison et al., in press). Presence of putative

first-generation immigrants in low-cover landscapes

(GENECLASS2) suggests that between-landscape

movements are not uncommon, including into low-

cover landscapes. Restricted dispersal accompanied by

gene flow is possible, because gene flow accumulates

over multiple generations and the effects of successful

dispersal events are compounded, even if individual

dispersal distances are short. Further, it is possible that

gene flow occurs primarily via males; male-biased

dispersal was supported by higher female spatial

autocorrelation structure, the significant fit of female

(but not male) genotypic distances to the model of

isolation-by-distance (Table 1), the only immature

recaptured at the same site as an adult being a female,

and the two inferred first-generation immigrants being

males. Genetic connectivity for the Grey Shrike-thrush

populations contrasts with the expectation of low

functional connectivity for this species in a fragmented

landscape (Robertson and Radford 2009, Amos et al.

2012), but accords with results for other woodland-

dependent birds in our study area: the Eastern Yellow

Robin Eopsaltria australis, Weebill Smicrornis brevi-

rostris, and Spotted Pardalote Pardalotus punctatus

(Harrison et al., in press). We conclude that at the

spatial scale studied here (;200 3 100 km), populations

of the Grey Shrike-thrush remain connected by gene

flow despite habitat loss.

Social connectivity across the study region and presence

of local communication networks

Acoustic analyses suggested some cultural connectiv-

ity at the regional scale, seen in presence of the same song

types in distant landscapes, notwithstanding local

TABLE 1. Summary of tests of three models of mobility through fragmented landscapes using genetic and acoustic data.

Birddistance

Landscapemodel

Landscaperesistance values(treed : treeless)

Mantel tests Partial Mantel tests

Bird distance 3landscape model

Bird distance 3 landscapemodel (RM or IM),IBD partialled out

Bird distance 3 IBD,landscape model

(RM or IM) partialled out

r P r P r P

Genotypic

All IBD 1:1 0.046 0.034All RM 1:2 0.037 0.132 �0.019 0.615 0.034 0.289All IM 2:1 0.049 0.044 0.017 0.407 0.004 0.463Females IBD 1:1 0.105 0.007Females RM 1:2 0.100 0.039 0.000 0.484 0.033 0.408Females IM 2:1 0.098 0.008 0.004 0.527 0.038 0.338Males IBD 1:1 0.055 0.075Males RM 1:2 0.031 0.256 �0.058 0.739 0.073 0.171Males IM 2:1 0.072 0.070 0.051 0.269 �0.022 0.596

Song types IBD 1:1 0.153 0.000RM 1:2 0.161 0.000 0.051 0.036 0.006 0.406IM 2:1 0.126 0.000 �0.052 0.967 0.103 0.000

Notes: Mantel tests show correlations between matrices of dissimilarities for all pairs of individuals (bird distance based ongenotypic distance or song type dissimilarity) and pairwise resistances from isolation-by-resistance (Circuitscape) models: IBD,isolation-by-distance; RM, reduced movement through treeless areas; IM, increased movement through treeless areas. PartialMantel tests show partial correlations separating the effects of RM and IM from those of IBD. Significant Mantel r values are inbold; P values are the probability for one-tailed test of obtaining a value greater than the observed r under null model of nocorrelation (9999 permutations).


similarities (Fig. 3). Patterns of local similarity of song

features across geographic scales could arise, for

example, from song learning before or after dispersal

or matching of song structure and song types to

neighbors (Catchpole and Slater 2008). Although imma-

ture Grey Shrike-thrushes can sing while still in natal

territories (Higgins and Peter 2002), the timing of their

song development is not known. Studies of other

songbird species routinely suggest that birds match their

songs to neighboring individuals after dispersal (Payne

and Payne 1993, Bell et al. 1998), and field observations

of the Grey Shrike-thrush show that the birds often sing

the same song type as the one they have just heard from

either a neighbor or a partner (M. Goretskaia, unpub-

lished data). Gradual decline of similarity of Grey Shrike-

thrush’s song types with distance at a local scale is

consistent with social synchronization of songs owing to

neighbors matching each other and forming a commu-

nication network (McGregor and Dabelsteen 1996).

Concordance of spatial autocorrelation patterns of song

types and genotypes in the Grey Shrike-thrush appears to

be due to localized song matching accompanied by

unrelated patterns of restricted dispersal. Alternatively,

concordance could be due to song types being transmit-

ted by dispersing individuals with pre-dispersive learn-

ing, although this explanation is less likely, as restricted

dispersal was not detected for males (Table 1) and song

matching to local neighbors was observed in the field.

Effect of habitat loss and fragmentation on genetic

and social connectivity

Contrary to the expectation from experiments sug-

gesting that gaps in habitat cover are barriers to

movement of the Grey Shrike-thrush (Robertson and

Radford 2009), the model of reduced mobility through

treeless areas was not the best-supported model for the

genetic data set; in addition, no evidence of reduced

genetic or acoustic diversity in low-cover landscapes was

found (Appendices A, B, and E). Although data on

female genotypic distances did fit the model of reduced

mobility through cleared areas by Mantel test, they also

fit the models of isolation-by-distance and increased

mobility, and support for these models could not be

distinguished by partial Mantel tests due to high

correlations between the alternative resistance matrices

(Amos et al. 2012). Our results suggest that isolation-by-

distance shapes the genotypic structure of the Grey

Shrike-thrush females, and habitat loss and fragmenta-

tion do not have a detectable impact on female dispersal.

On the other hand, lack of evidence of any spatial

genotypic patterns in males suggests that on the spatial

scale studied here (;100 3 200 km), the dispersal of the

Grey Shrike-thrush males is neither restricted by

distance or habitat gaps, nor facilitated by habitat loss

and fragmentation. It is possible that current levels of

fragmentation, where widespread scattered trees provide

connectivity among habitat patches (Robertson and

Radford 2009), is at a spatial scale smaller than that of

natural dispersal of the Grey Shrike-thrush. It is also

possible that female dispersal distances are unaffected by

fragmentation because they are naturally extremely

small. According to this scenario, genetic connectivity

provided by males will be enough to avoid negative

impacts of inbreeding, but will fail to rescue populations

in which females went extinct locally (Cooper et al.

2002). If this were true, local genotypic similarities of

PLATE 1. Female Grey Shrike-thrush in nest in bark crevice on red ironbark tree. Photo credit: J. N. Amos.


females would not be expected to extend beyond the site

level, especially in fragmented landscapes. However,

they do for up to 5 km for genotypic data from all

landscapes (Fig. 3) and for up to 10 km for data

restricted to fragmented landscapes only (Appendix D:

Table D2). Thus, it is more likely that the spatial scale of

current fragmentation in the region is smaller than the

spatial scale of dispersal for both sexes. Finally, low

sample sizes, despite considerable field effort, might be

partially responsible for the lack of detectable spatial

genotypic structure for males (weak IBD was hinted at

by the gradual decline of the spatial autocorrelation

coefficient over distance and significant correlogram

heterogeneity; Appendix D: Table D1) and for the

inability to reject the null model for the females. Larger

sample sizes would increase the power of partial Mantel

tests to distinguish between alternative mobility models,

given the high correlation of all three resistance surfaces

(Amos et al. 2012). Alternatively, a novel application of

other statistical approaches (Legendre and Fortin 2010,

Storfer et al. 2010) may result in increased power

without the need for greater sample sizes. Low sample

sizes for reference landscapes also precluded direct

comparison of spatial autocorrelation patterns between

fragmented and continuous landscapes, but lower

genotypic similarity at a site and larger distances at

which birds remain significantly similar in fragmented

landscapes compared to all landscapes hint at increased

dispersal distances in fragments (Appendix D: Table

D2).

In contrast to the genetic results, the model of reduced

movement through treeless cells best fit the song type

data, because it explained variation beyond that

explained by the null model of isolation-by-distance

(RM partial Mantel test; Table 1). This result (also

supported by faster decay of song type similarity over

distance for fragmented landscapes compared to all

landscapes; Appendix D: Table D2) means that birds

share fewer song types across cleared areas compared to

the forested areas. It is possible that song matching is

lower across gaps merely because birds have no or fewer

neighbors to match in cleared areas. The density of

territories in low-cover landscapes is lower, even within

forested areas (Radford et al. 2005), so that each bird in

these landscapes has fewer neighbors to interact with

and to learn from (Catchpole and Slater 2008). In

another songbird, Dupont’s Lark (Chersophilus dupon-

ti ), gaps in habitat have been shown to hinder cultural

transmission of song types over distances, resulting in an

increased differentiation between birds on the opposite

side of gaps (due to lack of interactions), and increasing

matching between neighbors (owing to stronger compe-

tition for remaining resources; Laiolo and Tella 2005). It

is possible that stronger competition in continuous

habitat leads to higher song matching of the Grey

Shrike-thrush across habitat. Reduced dispersal through

gaps in habitat of birds with pre-dispersive learning

could also explain reduced song sharing across treeless

areas, but this scenario was not supported by genetic

data or field observations of song matching. Moreover,

males, which our data, surprisingly, suggest are the

dispersing sex in the Grey Shrike-thrush, did not display

even significant isolation-by-distance on the spatial scale

of our study. Further work on the timing of song learning

in this species is required to allow a clear interpretation of

this result. If this species is an ‘‘open-ended learner,’’ as

are other species within the family Pachycephalidae,

which learn song types throughout their life (Brown and

Brown 1994), then the reduced sharing in cleared areas

may well represent reduced song transmission, with

consequences for local social interactions.

Our song type data, collected from a short sequence of

birds’ songs, captured only a fraction of the total

variation in song type for the sampled birds. It is likely

that longer recordings of the same bird would detect

more song types and more sharing within and between

landscapes. Nevertheless, the pattern of decreasing song

type similarity with distance would probably remain

with longer recording, as no known systematic bias was

introduced by subsampling. Unfortunately, in this study

we have no information on the age or the sex of the

recorded birds (from the few banded birds recorded, we

know that we have sampled both sexes), but it appears

unlikely that systematic differences in either trait could

yield a result similar to that observed in our study.

Despite these limitations, our acoustic song type data

provide evidence of reduced acoustic connectivity in

response to habitat loss and fragmentation. Further data

on song sharing, song learning, and cultural connections

(number of singing neighbors) might clarify the exact

mechanism for this effect and its consequences for

population viability, given the important role of singing

in resource and mate acquisition. For example, it has

been suggested that acoustic differentiation can lead to

reproductive isolation of local pools of individuals, for

instance, as a result of female preference for local

acoustic dialects over those from farther away (Grant

and Grant 1996), or as a by-product of ecological

selection (Schluter 2001). On the other hand, low

variability in song type, expected for populations with

lower densities, may render the male unattractive to

females, with detrimental consequences for individual

fitness and population demography (Blumstein 1998).

Our data show that current levels of habitat loss and

fragmentation within box-ironbark forest of central

Victoria, Australia have not resulted in detectably

reduced dispersal of the Grey Shrike-thrush, but alter

behavioral processes by impeding cultural transmission

of song types through the agricultural matrix. Inevita-

bly, given the necessity of our study design, our

sampling did not extend to landscapes with extremely

low tree cover (,10%), where distances between suitable

fragments may be larger and present stronger filters to

dispersal. However, by considering all pairwise distances

between individuals (some of which are between-

landscape distances across treeless areas greater than


gaps within low-cover landscapes), the landscape genetic

approach overcomes the problem of landscape-based

inferences. Thus, it appears that the demographic

decline of the Grey Shrike-thrush from landscapes with

the lowest levels of tree cover is due to factors other than

strongly reduced dispersal or barriers to movement. The

Grey Shrike-thrush is a successful nest predator (Major

et al. 1999), and much of its diet consists of other

animals, including birds and their eggs (Higgins and

Peter 2002). It is possible that the widespread disap-

pearance of fauna with habitat loss (Lindenmayer et al.

2005, Radford et al. 2005) drives disappearance of the

Grey Shrike-thrush by resulting in low resource avail-

ability in landscapes with low tree cover. Given that

management often has to compromise between recon-

necting patches and increasing or improving habitat, for

the Grey Shrike-thrush the latter goal appears to be

more important.

ACKNOWLEDGMENTS

We thank the Australian Research Council Linkage Grant(LP0776322), the Victorian Department of Sustainability andEnvironment (DSE), the Museum of Victoria, the VictorianDepartment of Primary Industries, Parks Victoria, the NorthCentral Catchment Management Authority, and the GoulburnBroken Catchment Management Authority for funding. Dea-kin University, Birds Australia, and the Holsworth WildlifeResearch Endowment provided additional support. Sampleswere collected under DSE permit number 10004294 (WildlifeAct 1975 and the National Parks Act 1975), DSE permitnumber NWF10455 (section 52 of the Forest Act 1958), theAustralian Bird and Bat Banding Scheme, and MonashUniversity ethics processes (BSCI/2007/07). We thank theLP0776322 project team and volunteers for assistance in thefield, Katherine Harrisson for assistance in the lab, and JimThomson for statistical advice. Computationally intensiveanalyses (TESS, Circuitscape) were performed on MonashSun Grid. We also thank two anonymous reviewers for theirinsightful comments on an earlier draft of the manuscript.

LITERATURE CITED

Amos, J. N., A. F. Bennett, R. Mac Nally, G. Newell, A.Pavlova, J. Q. Radford, J. R. Thomson, M. White, and P.Sunnucks. 2012. Predicting landscape-genetic consequencesof habitat loss, fragmentation and mobility for multiplespecies of woodland birds. PLoS ONE 7(2):e30888.

Balkenhol, N., F. Gugerli, S. A. Cushman, L. P. Waits, A.Coulon, J. W. Arntzen, R. Holderegger, and H. H. Wagner.2009. Identifying future research needs in landscape genetics:where to from here? Landscape Ecology 24:455–463.

Bell, D. A., P. W. Trail, and L. F. Baptista. 1998. Song learningand vocal tradition in Nuttall’s white-crowned sparrows.Animal Behaviour 55:939–956.

Blumstein, D. T. 1998. Female preferences and effectivepopulation size. Animal Conservation 1:173–177.

Brown, R. J., and M. N. Brown. 1994. Matched song andduetting by a breeding pair of golden whistlers Pachycephalapectoralis. Emu 94:58–59.

Catchpole, C. K., and P. J. B. Slater. 2008. Bird song.Biological themes and variations. Second edition. CambridgeUniversity Press, Cambridge, UK.

Chen, C., E. Durand, F. Forbes, and O. Francois. 2007.Bayesian clustering algorithms ascertaining spatial popula-tion structure: a new computer program and a comparisonstudy. Molecular Ecology Notes 7:747–756.

Cooper, C. B., J. R. Walters, and H. Ford. 2002. Effects ofremnant size and connectivity on the response of BrownTreecreepers to habitat fragmentation. Emu 102:249–256.

Coulon, A., J. W. Fitzpatrick, R. Bowman, and I. J. Lovette.2010. Effects of habitat fragmentation on effective dispersalof Florida scrub-jays. Conservation Biology 24:1080–1088.

Ewers, R. M., and R. K. Didham. 2005. Confounding factors inthe detection of species responses to habitat fragmentation.Biological Reviews 81:117.

Excoffier, L., and H. E. L. Lischer. 2010. Arlequin suite ver 3.5:a new series of programs to perform population geneticsanalyses under Linux and Windows. Molecular EcologyResources 10:564–567.

Excoffier, L., P. E. Smouse, and J. M. Quattro. 1992. Analysisof molecular variance inferred from metric distances amongDNA haplotypes: application to human mitochondrial DNArestriction data. Genetics 131:479–491.

Ford, H. A. 2011. The causes of decline of birds of eucalyptwoodlands: advances in our knowledge over the last 10 years.Emu 111:1–9.

Ford, H. A., J. R. Walters, C. B. Cooper, S. J. S. Debus, andV. A. J. Doerr. 2009. Extinction debt or habitat change? —Ongoing losses of woodland birds in north-eastern NewSouth Wales, Australia. Biological Conservation 142:3182–3190.

Frankham, R. 2005. Genetics and extinction. BiologicalConservation 126:131–140.

Grant, B. R., and P. R. Grant. 1996. Cultural inheritance ofsong and its role in the evolution of Darwin’s finches.Evolution 2471–2487.

Harrison, K., A. Pavlova, J. N. Amos, N. Takeuchi, A. Lill, J. Q.Radford, and P. Sunnucks. In press. Fine-scale effects ofhabitat loss and fragmentation despite large-scale gene flow forsome regionally declining woodland bird species. LandscapeEcology.

Higgins, P. J., and J. M. Peter, editors. 2002. Handbook ofAustralian, New Zealand and Antarctic Birds. Volume 6:Pardalotes to Shrike-thrushes. Oxford University Press,Melbourne, Australia.

Laiolo, P. 2010. The emerging significance of bioacoustics inanimal species conservation. Biological Conservation143:1635–1645.

Laiolo, P., and J. L. Tella. 2005. Habitat fragmentation affectsculture transmission: patterns of song matching in Dupont’slark. Journal of Applied Ecology 42:1183–1193.

Laiolo, P., and J. L. Tella. 2006. Landscape bioacoustics allowdetection of the effects of habitat patchiness on populationstructure. Ecology 87:1203–1214.

Laiolo, P., and J. L. Tella. 2007. Erosion of animal cultures infragmented landscapes. Frontiers in Ecology and theEnvironment 5:68–72.

Lees, A. C., and C. A. Peres. 2008. Avian life-historydeterminants of local extinction risk in a hyper-fragmentedneotropical forest landscape. Animal Conservation 11:128–137.

Legendre, P., and M.-J. Fortin. 2010. Comparison of theMantel test and alternative approaches for detecting complexmultivariate relationships in the spatial analysis of geneticdata. Molecular Ecology Resources 10:831–844.

Lindenmayer, D., J. Fischer, and R. Cunningham. 2005. Nativevegetation cover thresholds associated with species responses.Biological Conservation 124:311–316.

Lindenmayer, D., et al. 2008. A checklist for ecologicalmanagement of landscapes for conservation. Ecology Letters11:78–91.

Lowe, W. H., and F. W. Allendorf. 2010. What can genetics tellus about population connectivity? Molecular Ecology19:3038–3051.

Major, R. E., F. J. Christie, G. Gowing, and T. J. Ivison. 1999.Elevated rates of predation on artificial nests in linear stripsof habitat. Journal of Field Ornithology 70:351–364.


McGregor, P. K., and T. Dabelsteen. 1996. Communicationnetworks. Pages 409–425 in D. E. Kroodsma and E. H.Miller, editors. Ecology and evolution of acoustic commu-nication in birds. Cornell University Press, Ithaca, NewYork, USA.

McRae, B. H. 2006. Isolation by resistance. Evolution 60:1551–1561.

McRae, B. H., and P. Beier. 2007. Circuit theory predicts geneflow in plant and animal populations. Proceedings of theNational Academy of Sciences USA 104:19885–19890.

McRae, B. H., B. G. Dickson, T. H. Keitt, and V. B. Shah.2008. Using circuit theory to model connectivity in ecology,evolution, and conservation. Ecology 89:2712–2724.

McRae, B. H., and V. B. Shah. 2008. Circuitscape user’s guide.University of California, Santa Barbara, California, USA.http://www.circuitscape.org

Payne, R. B., and L. L. Payne. 1993. Song copying and culturaltransmission in indigo buntings. Animal Behaviour 46:1045–1065.

Peakall, R., and P. E. Smouse. 2006. GenAlEx 6: geneticanalysis in Excel. Population genetic software for teachingand research. Molecular Ecology Notes 6:288–295.

Piry, S., A. Alapetite, J. M. Cornuet, D. Paetkau, L. Baudouin,and A. Estoup. 2004. GENECLASS2: A software for geneticassignment and first-generation migrant detection. Journal ofHeredity 95:536–539.

R Development Core Team. 2011. R: a language andenvironment for statistical computing. R Foundation forStatistical Computing, Vienna, Austria.

Radford, J. Q., and A. F. Bennett. 2007. The relativeimportance of landscape properties for woodland birds inagricultural environments. Journal of Applied Ecology44:737–747.

Radford, J. Q., A. F. Bennett, and G. J. Cheers. 2005.Landscape-level thresholds of habitat cover for woodland-dependent birds. Biological Conservation 124:317–337.

Raymond, M., and F. Rousset. 1995. GENEPOP (version 1.2):a population genetics software for exact tests and ecumen-icism. Journal of Heredity 86:248–249.

Robertson, O. J., and J. Q. Radford. 2009. Gap-crossingdecisions of forest birds in a fragmented landscape. AustralEcology 34:435–446.

Robin, V. V., M. Katti, C. Purushotham, A. Sancheti, and A.Sinha. 2011. Singing in the sky: song variation in an endemicbird on the sky islands of southern India. Animal Behaviour82:513–520.

Rogers, K., A. Rogers, and D. Rogers. 1986. Bander’s aid: aguide to ageing and sexing bush birds. A. Rogers, St.Andrews, Victoria, Australia.

Schluter, D. 2001. Ecology and the origin of species. Trends inEcology and Evolution 16:372–380.

Schmuki, C., C. Vorburger, D. Runciman, S. Maceachern, andP. Sunnucks. 2006. When log-dwellers meet loggers: impactsof forest fragmentation on two endemic log-dwelling beetlesin southeastern Australia. Molecular Ecology 15:1481–1492.

Shanahan, D. F., H. P. Possingham, and C. Riginos. 2010.Models based on individual level movement predict spatialpatterns of genetic relatedness for two Australian forestbirds. Landscape Ecology 26:137–148.

Smouse, P. E., and R. Peakall. 1999. Spatial autocorrelationanalysis of individual multiallele and multilocus geneticstructure. Heredity 82:561–573.

Stevens, H. C., and D. M. Watson. 2005. Breeding biology ofthe Grey Shrike-thrush (Colluricincla harmonica). Emu105:223–231.

Storfer, A., M. A. Murphy, S. F. Spear, R. Holderegger, andL. P. Waits. 2010. Landscape genetics: where are we now?Molecular Ecology 19:3496–3514.

Sunnucks, P. 2011. Towards modelling persistence of woodlandbirds: the role of genetics. Emu 111:19–39.

Szabo, J. K., P. A. Vesk, P. W. J. Baxter, and H. P.Possingham. 2011. Paying the extinction debt: woodlandbirds in the Mount Lofty Ranges, South Australia. Emu111:59–70.

Van Houtan, K. S., S. L. Pimm, J. M. Halley, R. O.Bierregaard, Jr., and T. E. Lovejoy. 2007. Dispersal ofAmazonian birds in continuous and fragmented forest.Ecology Letters 10:219–229.

SUPPLEMENTAL MATERIAL

Appendix A

Sampling details and population statistics for genetic and acoustic data (Ecological Archives E093-149-A1).

Appendix B

Molecular methods and tests for the effect of habitat cover on genetic and acoustic diversity (Ecological Archives E093-149-A2).

Appendix C

Details of genetic and acoustic analyses (Ecological Archives E093-149-A3).

Appendix D

Results of spatial autocorrelation analyses for genetic and acoustic data on regional (study scale) and local scale (EcologicalArchives E093-149-A4).

Appendix E

Relationship between per-landscape estimates of genetic diversity (allelic richness, expected heterozygosity, effective populationsize), within-landscape autocorrelation coefficient, and landscape tree cover (Ecological Archives E093-149-A5).


Genes and song: genetic and social connections in fragmented habitat in a woodland bird with limited...

Documents

Transcript of Genes and song: genetic and social connections in fragmented habitat in a woodland bird with limited...