Reduced gene flow in a vulnerable species reflects two centuries of habitat loss and fragmentation

Abstract

Understanding the effects of landscape modification on gene flow of fauna is central to informing conservation strategies that promote functional landscape connectivity and population persistence. We explored the effects of large‐scale habitat loss and fragmentation on spatial and temporal patterns of gene flow in a threatened Australian woodland bird: the Grey‐crowned Babbler Pomatostomus temporalis. Using microsatellite data, we (1) investigated historical (i.e., pre‐fragmentation) and contemporary (i.e., post‐fragmentation) levels of gene flow among subpopulations and/or regions, (2) identified first‐generation migrants and likely dispersal events, (3) tested for signatures of genetic bottlenecks, (4) estimated contemporary and historical effective population sizes, and (5) explored the relative influences of drift and migration in shaping contemporary population structure. Results indicated that the functional connectivity of landscapes used by the Grey‐crowned Babbler is severely compromised in the study area. The proportion of individuals that were recent immigrants among all subpopulations were low. Habitat fragmentation has led to a clear division between subpopulations in the east and west, and the patterns of gene flow exchange between these two regions have changed over time. The effective population size estimates for these two regions are now well below that required for long‐term population viability (N_e < 100). Demographic history models indicate that genetic drift was a greater influence on subpopulations than gene flow, and most subpopulations show signatures of bottlenecks. Translocations to promote gene flow and boost genetic diversity in the short term and targeted habitat restoration to improve landscape functional connectivity in the long term represent promising conservation management strategies that will likely have benefits for many other woodland bird species.

Introduction

Habitat loss and fragmentation have a substantial influence on the structure and viability of animal populations (Hanski et al. , Villard et al. , Ortego et al. ). Landscape‐scale anthropogenic habitat modification can fragment populations into small, isolated subunits that are at an increased risk of local patch extinction (Hanski , Saccheri et al. , Fuhlendorf et al. , Banks et al. ). Small populations lose genetic diversity through random genetic drift, leaving them vulnerable to the negative effects of inbreeding and reducing their capacity to adapt to environmental change (Saccheri et al. , O'Grady et al. , Pavlacky et al. ). In populations with small effective population sizes (N_e <100; N_e, heuristically, is the number of individuals that contribute to the next generation), a population is expected to experience inbreeding depression and loss of critical functional genes (e.g., immunity) through genetic drift over a period of five, or fewer, generations (Frankham , Frankham et al. ). Reduced fitness as a result of inbreeding can have negative implications for a species’ reproductive rate, population size, and likelihood of long‐term population persistence (Keller ). Although population sizes when N_e > 100 should limit loss of fitness over five generations to ≤10%, it is widely accepted that much larger population sizes (e.g., N_e > 1000) are required to maintain a population's ability to adapt to environmental change (Jamieson and Allendorf , Frankham et al. ).

Dispersal of individuals promotes gene flow among habitat patches and is crucial for recolonizing suitable vacant habitat, maintaining genetic diversity, and mitigating extinction risk (Bowler and Benton ). The degree to which landscapes facilitate the movement of populations, individuals, and ultimately genes (Taylor et al. ) is influenced by landscape connectivity. Landscape connectivity has two components: Structural connectivity refers to the physical elements and configuration of the landscape, while functional connectivity refers to an animal's ability to move through the landscape (Tischendorf and Fahrig ). While it follows that structural connectivity influences functional connectivity, functional connectivity is a more direct measure of the capacity of a population to persist in modified landscapes (Uezu et al. , FitzGibbon et al. ). Thus, an understanding of functional connectivity in fragmented landscapes can be central to the successful implementation of conservation management actions for threatened taxa (Fahrig , Sunnucks ).

Increasing rates of gene flow among vulnerable and declining populations (e.g., via genetic rescue or genetic restoration) can counteract genetic drift, reduce inbreeding depression, and boost genetic diversity (Frankham , Hoffmann et al. , Whiteley et al. ). There is a growing body of evidence that demonstrates the positive outcomes of gene flow, including genetic rescue, for small, inbred populations (Frankham , Whiteley et al. ). Maintaining large metapopulations and promoting functional connectivity between small and isolated population subunits (i.e., maintaining metapopulation processes) is therefore predicted to promote species persistence under increasing human‐induced pressures from landscape modification, extreme events, and climate change uncertainties (Nimmo et al. ).

The Grey‐crowned Babbler Pomatostomus temporalis is a cooperatively breeding woodland bird found on mainland Australia (i.e., excluding the island State of Tasmania) and in southern New Guinea, and which has been adversely affected by human‐induced reductions in landscape connectivity (Adam and Robinson , Blackmore et al. , Stevens et al. ). Cooperatively breeding birds typically have much smaller effective breeding populations than pair breeding species when compared to total population size (Frankham ). As such, cooperative breeders present ideal models with which to investigate the effects of reduced gene flow on populations before they become critically endangered.

We investigated the effects of landscape‐scale habitat loss and fragmentation on spatial and temporal patterns of gene flow in a threatened woodland bird. We analyzed microsatellite data of the Grey‐crowned Babbler to (1) investigate the levels of historical (i.e., pre‐fragmentation) and contemporary (i.e., post‐fragmentation) gene flow among subpopulations and/or regions, (2) identify first‐generation migrants and likely dispersal events, (3) screen for signatures of genetic bottlenecks, (4) estimate contemporary and historical effective population sizes, and (5) explore the relative influences of drift and migration in shaping contemporary population structure. In doing so, our primary goal is to provide recommendations for management actions that would promote functional connectivity and population persistence in this and other woodland‐dependent bird species.

Materials and Methods

Study species

The Grey‐crowned Babbler was historically common across much of eastern Australia (Department of Environment and Heritage ), but has undergone a major range contraction and population declines of over 90% across the southern extent of its distribution as a consequence of habitat loss and fragmentation (Robinson , , Environment Conservation Council , Environment Australia ). In the mid‐1800s, extensive clearing of native vegetation for anthropogenic purposes such as agriculture and mining began in earnest. This clearing continued, such that ~14% of native habitat now remains in this region (Environment Conservation Council ). In southern parts of its range, the Grey‐crowned Babbler is now restricted to roadside or riparian vegetation, small adjacent remnant woodland patches within farmland (<0.5 ha), and habitat edges of the few remaining larger conservation reserves (>5 ha; Robinson ).

Because of their complex social structures and mating systems, cooperatively breeding bird species can be particularly vulnerable to habitat loss and fragmentation (Blackmore et al. , Harrisson et al. ). Grey‐crowned Babblers typically live in groups of up to 12 individuals (avg. ~5 individuals; Stevens et al. ) and occupy territories of between 2 and 53 ha in size (Higgins and Peter , Blackmore and Heinsohn ). Groups usually consist of a dominant breeding pair and past offspring that delay dispersal from natal territories for up to three years to help raise young (Blackmore et al. ). High levels of genetic relatedness across local neighborhoods suggest most dispersal occurs over relatively short distances (<2 km; Koenig et al. , Blackmore et al. , Stevens et al. ).

Sampling

The study region encompassed an area of ~22, 250 km² in north‐central and northeast Victoria, Australia (Fig. ). Potential sites were identified based on long‐term survey records (Robinson , Tzaros , , Davidson and Robinson ; N. Lacey, unpublished data; D. Robinson, unpublished data). Call playback confirmed the presence and size of a Grey‐crowned Babbler family group at each potential site. Territory occupancy was verified from nesting activity, and site locations were recorded using a Geographic Positioning System. An on‐ground search using call playback was conducted in areas of habitat within a 2 km radius of each study territory to determine distances to adjacent groups. An average Euclidean distance of 979.8 m separated sampled groups and their closest neighboring Grey‐crowned Babbler group, measured between group centroids (usually a nest). Structural connectivity distances between sampled groups were estimated using the distance‐measuring tool available in Google Earth satellite data 2015 to calculate the cumulative straight‐line distance total between pairs of groups through visibly connected habitat areas of tree cover. Sampling was undertaken at 39 sites selected from three geographic regions: west (n = 15, Kerang; Boort), southeast (n = 12, Violet Town; Lurg), and northeast (n = 12, Peechelba; Rutherglen; Chiltern; Fig. ). The sampling period incorporated two annual breeding seasons between June 2010 and April 2012. Birds were lured into mist nets using call playback. Each individual was banded with a metal leg band provided by the Australian Bird and Bat Banding Scheme and a unique combination of three colored plastic leg bands for identification in the field. Individuals were measured and a blood sample (~70 μL) collected from the brachial vein using a VITREX capillary tube. Blood was transferred to a Whatman FTA Card and stored at room temperature in paper envelopes. We sampled 135 Grey‐crowned Babbler individuals from 39 discrete family groups.

Molecular sexing, genotyping, and genetic marker behavior

All Grey‐crowned Babbler individuals were screened by polymerase chain reaction (PCR) and sexed using a standard molecular protocol (Griffiths et al. ). DNA isolates were subsequently genotyped for 13 Grey‐crowned Babbler microsatellite loci by the Australian Genomic Research Facility on an AB3730 capillary sequencer and analyzed using GeneMapper 3.7 (Applied Biosystems, Foster City, California, USA). Extraction protocol, primer sequences, PCR conditions and protocols, and appropriate genetic marker behavior checks are described in Stevens et al. ().

Since many population genetic analyses assume independence of individuals, in cooperatively breeding systems inclusion of close relatives has the potential to introduce some bias and contribute to patterns of population genetic structure. However, the potential influence of including close relatives should be greater on analyses conducted at the site level (<0.5 km; Stevens et al. ) than at the subpopulation or regional level (which pools multiple sites). Previously we found that removing closely related individuals did not substantially alter inferences (e.g., patterns of diversity, genetic structure, relatedness; Stevens et al. ), so we chose to retain all individuals to retain maximum power.

Analyses were based on six subpopulations: (1) Kerang north; (2) Kerang south and Boort; (3) Violet Town south; (4) Lurg, Violet Town north, and Peechelba; (5) Rutherglen; and (6) Chiltern (Fig. ). Subpopulations were defined according to geography and genetic substructure previously described in Stevens et al. (). Given Kerang north and Rutherglen (previously identified as sharing membership to the same genetic cluster) are separated by a very large geographic distance (>100 km), we chose to treat them as separate subpopulations here in order to be able to measure the extent of gene flow between them (Fig. ). Similarly, Violet Town south and Kerang south/Boort (which share membership to the same genetic cluster in the TESS analysis; Stevens et al. ) are also separated by a very large geographic distance (>100 km), and so were treated as separate subpopulations in gene flow analyses (Fig. ).

Contemporary gene flow and migration among subpopulations

Contemporary (previous 2–3 generations) levels of gene flow between all subpopulation pairs (n = 36 possible pairwise comparisons) were assessed using BayesAss v 3.1.1 (Wilson and Rannala ). As the average reproductive lifespan of the Grey‐crowned Babbler is five years and they exhibit overlapping generations (Counsilman and King ), we presumed contemporary gene flow levels to represent the 10–15 yr prior to sampling and therefore reflect genetic processes following extensive habitat fragmentation in the study area which have occurred within the past 200 yr (Bradshaw ).

BayesAss uses a Bayesian method with Markov chain Monte Carlo (MCMC) simulations to provide estimates of the mean and 95% confidence intervals (CI). BayesAss assumes both linkage equilibrium and that migration and genetic drift do not change subpopulation allele frequencies over the previous 2–3 generations, and relaxes assumptions of Hardy–Weinberg equilibrium (HWE) within populations (Wilson and Rannala ). Research has shown that BayesAss analyses may result in incorrect estimations of migration rates which arise from bimodality of the inference that models produce, as well as the effects of weak population structure (Meirmans ). Stevens et al. (), however, reported strong genetic structure across our study area. To increase the statistical power and inference reliability of BayesAss output, we followed Meirmans () further suggestions and ran over 30 repeats and did not average results, instead reporting the most biological meaningful and repeated results. These methods assist parameter optimization and ensure convergence. Furthermore, in instances where model assumptions may be violated, such as cooperatively breeding species, accurate estimates can still be obtained if migration rates are low (Faubet et al. ). Markov chain Monte Carlo mixing parameter values for migration rates (gene flow), allele frequencies, and inbreeding coefficients were adjusted to 0.50, 0.95, and 0.50, respectively, to achieve recommended acceptance rates (Wilson and Rannala ). We performed 3 × 10⁷ MCMC iterations with 10⁶ iterations to discard as burn‐in. Each run was initialized with different starting‐seed values to achieve consistency of mean parameter estimates between runs (Wilson and Rannala ).

Initial identification of putative first‐generation immigrants and their inferred origins also used BayesAss. To validate BayesAss ancestry assignments, we conducted a second method implemented in Geneclass2 (Piry et al. ) using a Bayesian approach (Rannala and Mountain ) with a Monte Carlo resampling algorithm (Paetkau et al. ). We tested 10,000 simulated individuals with a type I error threshold of 0.05 and used a likelihood ratio L_home/L_max. This ratio is computed from the likelihood of the population from which the individual was sampled (L_home) over the highest likelihood value among all population samples (L_max), including the population of the individual (Piry et al. ). The likelihood ratio of L_home/L_max has more statistical power to identify non‐resident individuals among populations than using only L_home (Piry et al. ). Both assignment methods assume all possible source populations have been sampled. Although some disparate Grey‐crowned Babbler groups exist between our populations (Robinson, unpublished data), ancestry analyses allowed us to identify general pathways of dispersal and to make direct comparisons of ancestry assignments between methods.

Detecting temporal gene flow and gene flow patterns between east and west regions

Common methods for enabling direct comparisons of temporal gene flow levels include comparing BayesAss and Migrate‐n estimates. Recent studies have shown that such comparisons may not always reflect biological reality (Faubet et al. , Meirmans , Samarasin et al. ). For instance, Samarasin et al. () suggest that in scenarios where there has been a recent decline in migration, Migrate‐n will underestimate historical migration rates (i.e., Migrate‐n will be biased to recent parts of the 4N_e time period). Furthermore, in the same situation, BayesAss will overestimate recent migration rates (Samarasin et al. ). Therefore, we undertook a qualitative investigation into long‐term and contemporary gene flow (connectivity) occurrence between regions. We also looked for any pattern variation in potential gene flow occurrence over time. We ensured greater robustness in the results by pooling our data which reduced the number of group comparisons to two regions rather than all possible pairs of subpopulations, while increasing the number of individuals within a group (east: n = 84; west: n = 51; Meirmans et al. 2014; Fig. ).

We used Migrate‐n to estimate mutation‐scaled, long‐term gene flow rates between the two regions. To reduce the number of potential parameters relative to the number of loci and improve statistical power (Kuhner ), we set parameters to include symmetrical gene flow. We used the Brownian motion model with F_st calculations of θ and M as starting parameters, and Metropolis‐Hastings sampling and uniform prior distributions to estimate θ (range, 0–100; delta, 10) and M (range, 0–500; delta, 50). The Markov chain settings recorded 10⁴ steps from 1 long chain of 10⁶ sampled steps, and a search strategy following a static heating scheme using four temperatures (1.0, 1.5, 3.0, and 1,000.0) to examine the genealogical space more effectively (Beerli , ). Runs were replicated twice to ensure posterior probabilities stabilized.

We used the commonly used method for estimating unscaled long‐term gene flow rates (Chiucchi and Gibbs , Dutta et al. , Wood et al. ) by multiplying the mutation‐scaled long‐term gene flow rates (M) generated in Migrate‐n with a typical vertebrate microsatellite mutation rate (0.001; Ellegren , Schlötterer ). Meanwhile, estimates of contemporary gene flow rates were obtained with BayesAss using the same methods as described above, but for east and west regions. We present means and CIs for Migrate‐n and BayesAss in our results.

Long‐term and contemporary effective population sizes of east and west regions

We derived the long‐term effective population sizes from θ values produced in Migrate‐n for east and west regions. To obtain a measure of contemporary effective population size (N_e), we estimated the effective number of breeders (N_eD; related to inbreeding and reflecting the parental generation) using the single‐sample linkage disequilibrium‐based method implemented in LDN_e (Waples and Do ). A recent study suggests that LDN_e analysis can be unreliable for sample sizes <30 (Tallmon et al. ). To increase the robustness of estimates of effective population size (Tallmon et al. ), we ran analyses for our pooled data set of east and west regions. We estimated N_eD using three different rates for the inclusion of rare alleles (pcrit: 0.05; 0.02; and 0.01), which allowed for comparisons of consistency across results. We report estimates from the criterion ≥0.05 as these provide a reasonable balance between maximum precision and minimal bias with polymorphic loci such as microsatellites (Waples and Do ).

Modeling population history

The genealogical history of the six subpopulations was investigated to estimate whether drift was more important than immigration in shaping contemporary population structure (Ciofi et al. ). Two models of population history, drift vs. immigration–drift equilibrium (gene flow), were assessed in 2‐Mod v 0.2 following the methods of Ciofi et al. (). Both models are based on population allele frequencies. The drift model computes allele frequencies as a product of pure drift with little evidence of gene flow between populations. The gene flow model works on an equilibrium principle between immigration and genetic drift to evaluate allele frequency within populations. The likelihood of each model's fit to the data is estimated using MCMC methods which compare estimates between models and provide probabilities of the goodness of fit for each (Ciofi et al. ). Simulations of MCMC were run for 10⁵ iterations, discarding the initial 10% of results as burn‐in to avoid possible bias from start conditions. The analysis was repeated three times to validate results.

Signature of bottlenecks within subpopulations

To investigate whether the six subpopulations had experienced genetic bottlenecks, we ran a two‐phase mutation model (TPM) in Bottleneck v 1.2.02 (Cornuet and Luikart ). This method investigates whether observed heterozygosity within each subpopulation was higher than would be expected for populations in mutation–drift equilibrium and can be used to detect bottlenecks over the last 2–4N_e generations (Cornuet and Luikart , Luikart et al. ). The proportion of stepwise mutation model in the TPM was set to 70%.

Results

Contemporary gene flow and migration among subpopulations

Very low levels of contemporary gene flow (i.e., the proportion of individuals within a subpopulation that are immigrants) per generation were recorded between six subpopulations of Grey‐crowned Babblers over the previous 2–3 generations using BayesAss (Table ). Estimates ranged from 0.01 to 0.19, with most rates being ≤0.03 (Table ). Two population pairs showed strong evidence of gene flow (CIs did not include zero) via immigration: Kerang south/Boort to Violet Town south and Chiltern to Rutherglen.

Estimates of recent (previous 2–3 generations) mean gene flow rates per generation among six Grey‐crowned Babbler subpopulations

Destination of gene flow	Origin of gene flow
Kn	KsB	Vs	LVP	Rg	Ch
Kn	0.91 (0.84 to 0.98)	0.02 (−0.02 to 0.05)	0.02 (−0.02 to 0.05)	0.02 (−0.02 to 0.05)	0.02 (−0.02 to 0.06)	0.02 (−0.02 to 0.05)
KsB	0.01 (−0.01 to 0.03)	0.96 (0.92 to 0.99)	0.01 (−0.01 to 0.02)	0.01 (−0.01 to 0.02)	0.01 (−0.01 to 0.03)	0.01 (−0.01 to 0.02)
Vs	0.03 (−0.02 to 0.08)	0.19 (0.11 to 0.27)	0.69 (0.65 to 0.74)	0.04 (−0.01 to 0.10)	0.02 (−0.02 to 0.06)	0.03 (−0.02 to 0.07)
LVP	0.01 (−0.01 to 0.03)	0.02 (−0.01 to 0.05)	0.01 (−0.01 to 0.02)	0.94 (0.90 to 0.99)	0.01 (−0.01 to 0.03)	0.01 (−0.01 to 0.02)
Rg	0.03 (−0.02 to 0.07)	0.04 (−0.003 to 0.09)	0.01 (−0.01 to 0.04)	0.02 (−0.01 to 0.05)	0.83 (0.75 to 0.90)	0.07 (0.01 to 0.13)
Ch	0.02 (−0.01 to 0.04)	0.02 (−0.02 to 0.06)	0.02 (−0.01 to 0.04)	0.02 (−0.02 to 0.06)	0.05 (−0.004 to 0.10)	0.88 (0.80 to 0.95)

Notes

Values indicate the mean proportion of individuals within subpopulations in rows (Destination of gene flow) that are immigrants from subpopulations in columns (Origin of gene flow). The 95% confidence intervals of gene flow rates are in parentheses, and immigration CI values that do not cross zero are in bold type. Proportions of non‐migrants are on the diagonal. Subpopulations are Kerang north (Kn); Kerang south/Boort (KsB); Violet Town south (Vs); Lurg/Violet Town north/Peechelba (LVP); Rutherglen (Rg); and Chiltern (Ch). Values were calculated using BayesAss (Wilson and Rannala ).

Initial ancestry assignments from BayesAss identified 10 individuals as likely first‐generation immigrants. Eight out of the 10 individuals were adult birds (≥2nd‐year bird), and this cohort was male‐biased (n = 7/1 sex ratio). The two remaining birds, one male and one female, were first‐year birds.

The Geneclass2 ancestry assignment method identified nine possible first‐generation immigrants (P < 0.05). Seven of the nine birds identified were adults, and two were immature. Three out of the nine birds were also identified as likely migrants with BayesAss (one adult; two immature). Geneclass2 results also supported a male bias among adult immigrants (n = 6/1 sex ratio; Table ).

First‐generation immigrants identified among six genetic subpopulations of the Grey‐crowned Babbler in southern parts of its range

Individual	Sample location	Origin of ancestry	Probability of ancestry	Log(L) of ancestry	Sex	Age class	Approximate Euclidean distance (km)
VB085	Vs	KsB	0.979		Male	Adult	170
VB086	Vs	KsB	0.956		Male	Adult	170
VA135	Vs	KsB	0.950		Male	Adult	175
VB087	Vs	KsB	0.946		Male	Adult	170
VB088	Vs	KsB	0.934		Male	Adult	170
VA092	Vs	KsB	0.910		Male	Adult	175
VA134	Vs	KsB	0.892		Female	Adult	175
CH006,	Ch	Rg	0.872	16.804	Male	Immature	15
VA091,	Vs	LVP	0.794	19.156	Female	Immature	20
RH035,	Rg	KsB	0.485	17.706	Male	Adult	220
CK042	LVP	Ch		21.958	Male	Adult	37
RE131	Kn	Rg		20.761	Male	Adult	215
RR040	LVP	Rg		20.036	Female	Adult	12
CH004	Rg	Ch		18.517	Male	Adult	15
VT068	KsB	LVP		16.254	Male	Adult	190
RI008	Ch	Rg		14.204	Male	Adult	12

Notes

Values indicate the probability (BayesAss; Wilson and Rannala ) and/or the log likelihood (log(L); Geneclass2; Piry et al. ) of an individual being a first‐generation immigrant. Euclidean distances are approximations and measured from the individual's sampling location to the closest sampled family group associated with the putative subpopulation of origin. Subpopulations are Kerang north (Kn); Kerang south/Boort (KsB); Violet Town south (Vs); Lurg/Violet Town north/Peechelba (LVP); Rutherglen (Rg); and Chiltern (Ch). All log(L) and probability values were below the significance threshold (P < 0.05). Results are shown in descending order based on probability, then log(L), values.

⁵Individual identified using BayesAss.

⁶Individual identified using Geneclass2.

Detection of temporal gene flow and pattern variation between east and west regions

We found evidence of symmetrical long‐term gene flow between the east and west regions (Fig. , Table ). Contemporary gene flow occurrence was evident in the direction of the east region to the west, but no evidence for contemporary gene flow occurring from the west to the east region (CIs included zero; Fig. , Table ).

Estimates of long‐term and contemporary gene flow per generation and effective population size estimates between two geographically separated regions of Grey‐crowned Babbler in southern parts of their distribution

Route	n	Long‐term m	Contemporary m	Recipient θ	Recipient NeD
East to West	51	0.015 (0.01–0.02)	0.066 (0.03 to 0.10)	4.29 (2.07–6.47)	17.0 (13.90–20.90)
West to East	84	0.015 (0.01–0.02)	0.018 (−0.02 to 0.04)	5.89 (3.53–8.20)	19.70 (16.50–23.50)

Notes

Shown are the sample size (n); long‐term (Migrate‐n; Beerli ) and contemporary (BayesAss; Wilson and Rannala ) levels of gene flow per generation (m); and mutation‐scaled, long‐term effective population size (θ; Migrate‐n) and contemporary effective number of breeders (N_eD; related to inbreeding and reflecting the parental generation; LDN_e; Waples and Do ) for the recipient population. Values shown in parentheses are 95% CIs, and values for m that do not cross zero are in bold type.

Long‐term and contemporary effective population sizes

Mutation‐scaled, long‐term effective population size estimates (θ) were higher in the east (5.89) than in the west (4.29; Table ). Contemporary LD‐based effective population size estimates (N_eD) were also higher in the east than in the west region and were smaller than their respective sample size (n; east: n = 83, N_eD = 19.7; west n = 51, N_eD = 17.0; Table ).

Demographic history of subpopulations

The pure drift model was identified as the most plausible model given the genetic history of subpopulations (probability, drift = 0.70; gene flow = 0.30). This result suggests that levels of gene flow among these subpopulations are not sufficient to counteract genetic drift.

Bottleneck signatures within subpopulations

Under the TPM, four of the six populations showed evidence of genetic bottlenecks: Kerang south/Boort; Violet Town south; Lurg/Violet Town north/Peechelba; and Rutherglen (Table ).

Models of bottleneck signatures for Grey‐crowned Babbler individuals sampled from six subpopulations

Subpopulation	n	k	TPM
Kn	25.85	5.46	0.66
KsB	76.00	7.23	<0.01
Vs	20.00	4.69	<0.01
LVP	73.54	7.08	<0.01
Rg	41.85	6.77	0.05
Ch	32.00	5.77	0.25

Notes

Subpopulations are Kerang north (Kn); Kerang south/Boort (KsB); Violet Town south (Vs); Lurg/Violet Town north/Peechelba (LVP); Rutherglen (Rg); and Chiltern (Ch). Values are mean number of individuals sampled per locus (n); mean observed number of alleles (k); and significant values (P < 0.05) for the two‐phase mutation model (TPM). Computations were calculated in Bottleneck v 1.2.02 (Cornuet and Luikart ).

Discussion

Our study demonstrates that the contemporary functional connectivity of landscapes used by the Grey‐crowned Babbler in the southern parts of its range is likely compromised relative to historical levels. The change in gene flow pattern over time shows that contemporary migration of individuals from the west to the east region has decreased to a level that provides no evidence of its occurrence. Demographic history models indicated that genetic drift was a greater influence on the species than gene flow across the study region, and most subpopulations show signatures of bottlenecks. Effective population size estimates of less than 100 for the regions are now well below what is required for long‐term population viability (Frankham ).

Gene flow decline despite evidence of long‐distance dispersal

Although evidence was found for continuing contemporary gene flow in an east‐to‐west direction, the few long‐distance dispersal events observed from the west to the east did not support evidence for continuing occurrences in this latter direction. In fact, overall contemporary gene flow levels remain very low or non‐existent between the east and west regions (<2 effective migrants per generation). We suggest that the evidence of gene flow from the west to the east found in contemporary immigration rates between subpopulations Kerang south/Boort to Violet Town south (Table ) may be a remnant of historical connectivity. Additionally, the highest number of samples for any of the genetic clusters found in our study area was evident in the Kerang south/Boort subpopulation (n = 38; Stevens et al. ). These data could potentially skew our results, indicating that the birds in Violet Town south (n = 4) from the same genetic cluster are from the western Kerang south/Boort subpopulation. Some evidence of the same genetic cluster was also recorded in the Chiltern subpopulation (n = 2), and being in the east region and geographically closer, the Violet Town south birds may have originated from Chiltern. In these northeastern areas, greater levels of structural connectivity, that is, more available tree cover, are provided by dispersed tree cover (Stevens et al. ), and roadside and riparian corridors (Fig. ; K. Stevens, personal observation).

The variation in contemporary patterns of gene flow between the east and west, and the (potential) change in gene flow patterns over time, may also be a consequence of higher levels of available habitat in the east region. Grey‐crowned Babblers in this region may exhibit increased fitness and greater mobility and be capable of flying further or more often. Higher population levels in the east also require more available habitat, and fitter birds in this region could utilize the higher levels of functional habitat connectivity to move west. By contrast, birds in the west region may not be as mobile due to a lack of habitat and lower levels of functional connectivity across their region, potentially producing negative effects on their fitness and movement between habitat patches. Less mobile species often rely on corridors as conduits for dispersal, and these types of habitat linkages can be crucial to animal movement through fragmented landscapes, particulary in agricultural systems (van der Ree and Bennett , Gillies and St. Clair , Vergara et al. ). Ongoing gene flow may be better facilitated by the presence of both corridors and dispersed (stepping‐stone) habitat connectivity in fragmented systems.

If our estimates of contemporary gene flow levels between east and west regions are overestimated as a result of recent declines in migration rates as studies suggest (Samarasin et al. ), this could mean gene flow between east and west regions is potentially occurring at even lower levels than our estimates show (Fig. , Table ). Under such a scenario, there is an even more pressing need to instigate targeted conservation management efforts for these birds in our study area. Similar studies on metapopulations that are reliant on relatively stable sources of habitat have shown that habitat loss and fragmentation are associated with decreased wildlife immigration and survival (Catlin et al. ). Populations that experience high levels of habitat disturbance can become demographically and genetically isolated as a result of reduced dispersal and gene flow. Although our study showed evidence for some long‐distance (<220 km) emigration from the west to the east region, potentially facilitated by extant riparian habitat connectivity between major rivers in the area (e.g., Murray and Goulburn rivers; Fig. ), the overall rate of observed gene flow may be insufficient to mitigate the detrimental effects of small population sizes on the long‐term genetic viability of these subpopulations (Weeks et al. , Segelbacher et al. ).

Signatures of genetic bottlenecks and small effective population sizes

Signatures of genetic bottlenecks likely reflect declines in population size and/or reduced gene flow (Cornuet and Luikart , Broquet et al. ). Detectable signatures of bottlenecks generally become apparent when high levels of population decline have occurred or numbers of breeding individuals are reduced to unsustainable levels (i.e., N_e < 100 individuals; Peery et al. ). Strong evidence of longer‐term signatures of bottlenecks in most subpopulations supports the small N_e estimates and evidence of drift. Our results are consistent with those of other studies on species experiencing major population declines resulting from recent isolation and/or population collapse as a consequence of habitat loss and fragmentation (Bender et al. , Fahrig , Radford et al. ).

Small N_e and severe reductions in N_e can lead to a loss of fitness through inbreeding depression and reduced evolutionary potential (Frankham et al. ). For species of conservation concern, identifying populations which have small N_e and that show evidence of recent bottlenecks is crucial for effective conservation decisions (McCusker et al. ). Long‐term and contemporary estimates of effective population sizes were higher for the east region than for the west, but were well below the level predicted to limit loss of fitness to ≤10% over five generations (Frankham et al. ). The census population in the southern extent of the species’ range is estimated at ≤2000 individuals (Davidson and Robinson ). Samples used in this study were collected within the same census population, and hence, our results may reflect a concerning trend across the entire population, and which is below the number required for the future genetic viability of these populations (i.e., N_e > 1000; Frankham et al. ).

Influences of drift rather than migration shaping contemporary population structure

Despite evidence for dispersal over large geographic distances, the higher probability of genetic drift influencing Grey‐crowned Babbler population structure in the study area will likely outweigh the level of migration required for mutation–drift equilibrium (Luikart et al. ). This finding is consistent with earlier studies indicating that habitat fragmentation implications include disrupted dispersal of the Grey‐crowned Babbler (Environment Australia ). Other studies investigating the effects of habitat modification on species’ population genetic structure and functional connectivity report similar detrimental effects (Dutta et al. , Harrisson et al. , McCusker et al. ). Declines in genetic exchange between small populations are likely to be associated with increased levels of inbreeding and elevated risk of local extinction as subpopulations lose genetic diversity (Sunnucks ). Analyses indicated that Kerang south/Boort was no longer receiving gene flow from other subpopulations, which suggests a decrease in genetic exchange from this subpopulation. Long‐term census records have shown population decline and extirpation of Grey‐crowned Babbler groups from habitat patches in these areas particularly (Tzaros , , Stevens et al. ). The lack of immigration from Kerang south/Boort, population decline, and local extinctions is a concerning trend. This concern is further compounded given the drift model estimated there was a 70% probability that drift had occurred. Such evidence strongly indicates Kerang south/Boort is exposed to an increasing threat of inbreeding and drift, and its long‐term viability is questionable without intervention (Volpe et al. , Weeks et al. ). As such, we identify the Kerang south/Boort population as a management priority within our study area.

Conclusion and Recommendations

An understanding of the role of landscape connectivity among spatially structured and declining populations is required to inform effective conservation measures that promote genetic variation and population demographic viability (Amos et al. ). Differences in gene flow patterns over time that were observed here suggest that these regions are now, or are becoming, isolated, and are consistent with a loss of functional connectivity resulting from large‐scale habitat loss and fragmentation since the mid‐1800s in this area (Fig. , Table ). Given a lack of functional landscape connectivity is a likely driver in this threatening process, there is potential to reverse this decline in gene flow. Across the Lurg area for instance, long‐term (>22 yr) and large‐scale (>1500 ha) habitat restoration has led to a substantial increase in woodland bird species diversity and richness, including the Grey‐crowned Babbler (Thomas , Vesk et al. ). Ongoing research into the long‐term effects of habitat restoration for the Grey‐crowned Babbler in these areas demonstrates an increase in population size (2001–2008, mean = 59; 2009–2015, mean = 106) with the average group size increasing by 0.8 birds (Thomas , Vesk et al. ; Lacey, unpublished data, Moylan, unpublished data). Although substantial areas of revegetated habitat support population increases in woodland fauna within the Lurg area (Vesk et al. ), this is a localized phenomenon within our study region. There remain large gaps in structural connectivity and a lack of habitat availability between subpopulations elsewhere, which may explain the low levels of contemporary gene flow between them. With similar habitat restoration effort within targeted areas, woodland species could experience an increase in gene flow levels.

Our study suggests that loss of functional connectivity of landscapes has had negative consequences for the future genetic viability of the Grey‐crowned Babbler in the southern part of its range. The current status of the species in the study area is symptomatic of faunal declines in fragmented systems (Radford et al. ). In our focal area, there are a suite of other woodland birds that are likely threatened by the same or similar processes (Amos et al. ). The Grey‐crowned Babbler is an exemplar in this context as its cooperatively breeding behavior makes it especially susceptible to the influences of habitat fragmentation owing to a substantially reduced N_e (relative to total population size; Sunnucks ). Under these circumstances, actions to promote/enhance gene flow for the fragmentation‐sensitive Grey‐crowned Babbler are likely to also have benefits for other threatened species, including species with less sensitive breeding strategies such as pair breeders.

Efforts that promote species genetic viability, such as conservation translocations and habitat connectivity enhancement, require information about functional connectivity and genetic variability of populations (Weeks et al. ). The data we have presented are highly relevant for targeting revegetation programs between subpopulations that have become disconnected, but could also be used to inform carefully managed translocation programs (Weeks et al. , Volpe et al. ). Translocations for genetic rescue/restoration purposes are increasingly being considered as a potentially powerful management strategy for boosting fitness and genetic diversity of small, isolated populations (Hoffmann et al. , Weeks et al. , Whiteley et al. ). Arguments warning against translocations often suggest that mixing genes between previously genetically isolated populations will lead to outbreeding depression (Storfer ). However, evidence of historical genetic connectivity across the study region indicates that efforts to increase functional connectivity would be highly unlikely to result in negative fitness consequences for the Grey‐crowned Babbler (Frankham et al. , Frankham ). Intervention programs, such as human‐assisted translocations, could potentially be implemented across the southern parts of the Grey‐crowned Babbler's range as an interim measure until habitat revegetation can provide functional landscape connectivity in these areas (Clarke et al. ). Such management interventions may be necessary to avoid localized extinctions as have been observed in other highly fragmented parts of the species range (e.g., south‐coastal Victoria, southeast South Australia; Barrett , Department of Environment and Heritage , Department of Land, Water, Environment and Planning ). Increasing structural landscape connectivity to facilitate gene flow for Grey‐crowned Babblers is also likely to provide long‐term benefits for other woodland bird species that are affected by loss of habitat in the same areas (Clarke and Oldland ).

Subpopulations in this fragmented landscape present a model for species that persist at the extremes of their range. But perhaps more importantly, here they also present a transferable model with broad applicability for many declining bird species. This study has detailed how genetic approaches can be used to drive intervention‐orientated conservation programs that aim to facilitate long‐term gene flow in a contemporary landscape.

Acknowledgments

Data on the location of babbler territories were kindly provided by D. Robinson, C. Tzaros, and N. Lacey. We thank many enthusiastic field assistants. Funding was provided by a Stuart Leslie Bird Research Award (BirdLife Australia); a Professor Allen Keast Research Award (BirdLife Australia); the Holsworth Wildlife Research Endowment; and a Jill Landsberg Trust Fund Scholarship (Ecological Society of Australia), for which we are most grateful. This research was conducted under Deakin University Animal Welfare Committee approval A66‐2009; Australian Bird and Bat Banding Scheme authority 1762; and Department of Sustainability and Environment (Victoria) bird banding and research permit 10005380. No authors of this paper have a conflict of interest to declare.