SEARCH

SEARCH BY CITATION

Keywords:

  • asymmetric gene flow;
  • asymmetric migration;
  • bird conservation;
  • coalescent theory;
  • detection probability;
  • dispersal asymmetry;
  • landscape ecology;
  • landscape genetics;
  • microsatellite DNA;
  • subtropical rainforest

Summary

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

1. Local extinctions in habitat patches and asymmetric dispersal between patches are key processes structuring animal populations in heterogeneous environments. Effective landscape conservation requires an understanding of how habitat loss and fragmentation influence demographic processes within populations and movement between populations.

2. We used patch occupancy surveys and molecular data for a rainforest bird, the logrunner (Orthonyx temminckii), to determine (i) the effects of landscape change and patch structure on local extinction; (ii) the asymmetry of emigration and immigration rates; (iii) the relative influence of local and between-population landscapes on asymmetric emigration and immigration; and (iv) the relative contributions of habitat loss and habitat fragmentation to asymmetric emigration and immigration.

3. Whether or not a patch was occupied by logrunners was primarily determined by the isolation of that patch. After controlling for patch isolation, patch occupancy declined in landscapes experiencing high levels of rainforest loss over the last 100 years. Habitat loss and fragmentation over the last century was more important than the current pattern of patch isolation alone, which suggested that immigration from neighbouring patches was unable to prevent local extinction in highly modified landscapes.

4. We discovered that dispersal between logrunner populations is highly asymmetric. Emigration rates were 39% lower when local landscapes were fragmented, but emigration was not limited by the structure of the between-population landscapes. In contrast, immigration was 37% greater when local landscapes were fragmented and was lower when the between-population landscapes were fragmented. Rainforest fragmentation influenced asymmetric dispersal to a greater extent than did rainforest loss, and a 60% reduction in mean patch area was capable of switching a population from being a net exporter to a net importer of dispersing logrunners.

5. The synergistic effects of landscape change on species occurrence and asymmetric dispersal have important implications for conservation. Conservation measures that maintain large patch sizes in the landscape may promote asymmetric dispersal from intact to fragmented landscapes and allow rainforest bird populations to persist in fragmented and degraded landscapes. These sink populations could form the kernel of source populations given sufficient habitat restoration. However, the success of this rescue effect will depend on the quality of the between-population landscapes.


Introduction

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

The local extinction of populations in habitat patches and dispersal between the patches are key processes structuring animal populations in heterogeneous environments (Andrewartha & Birch 1954; Day & Possingham 1995; Hanski 1998). Increasing rates of local extinction are symptomatic of regional population declines, and the fraction of habitat patches that are occupied is often related to metapopulation persistence (Lande 1987; Vos et al. 2001; but see Elkin & Possingham 2008). Dispersal is important for the recolonization of vacant habitat patches, regulation of local population dynamics and reduction of extinction risk in spatially structured populations (Bowler & Benton 2005).

Dispersal between populations in heterogeneous landscapes is one of the most important, yet least understood, ecological processes related to the persistence of animal populations (Bowler & Benton 2005). Although often assumed to be symmetric in spatial population models (Kleinhans & Jonsson 2011), dispersal is a complex trait regulated by independent processes operating during emigration, immigration and transition (Bowler & Benton 2005). Emigration and immigration are population specific vital rates, whereas transition can be considered the transfer of individuals between populations or interpatch movement (Bowler & Benton 2005). Not only are there multiple stages of dispersal, but the same proximate factor may have compensatory effects on the different stages of dispersal (Ims & Hjermann 2001). For example, high local population density often promotes emigration, but tends to inhibit immigration (Ims & Hjermann 2001). This imbalance between the emigration and immigration stages of dispersal is an important cause of asymmetry in dispersal rates (Kawecki & Holt 2002). There is accumulating evidence that asymmetric dispersal is the rule rather than the exception in a wide range of animal taxa (Senar, Conroy & Borras 2002; McIntire, Schultz & Crone 2007; Smith et al. 2008).

The identification of populations functioning as net exporters of dispersing individuals, because of asymmetric dispersal, is becoming increasingly important to biological conservation (Donovan et al. 1995). Asymmetric dispersal of individuals into populations that are likely to go locally extinct (Vuilleumier & Possingham 2006; Elkin & Possingham 2008) and net movement into recently modified landscapes (Remeš 2000; Battin 2004) are detrimental to long-term metapopulation persistence. However, asymmetric dispersal from source to sink populations may stabilize asynchronous population dynamics in modified landscapes (Doebeli 1995), prevent population declines (With, Schrott & King 2006) and allow populations to persist in landscapes degraded by habitat loss (Brown & Kodric-Brown 1977; McIntire, Schultz & Crone 2007). The effects of landscape change on the three stages of dispersal (emigration, immigration and transition) often have multiple consequences for animal populations (Bowler & Benton 2005). For example, habitat loss may limit emigration through within-population processes such as depressed reproduction and survival (Pulliam & Danielson 1991), but habitat loss may also inhibit transition through between-population processes such as increased dispersal mortality and reduced landscape permeability (Gustafson & Gardner 1996). Effective landscape conservation thus requires understanding of asymmetric dispersal in terms of within-population processes, such as emigration and immigration, as well as between-population processes, such as interpatch movement (Haynes et al. 2007; Revilla & Wiegand 2008). In addition, knowledge about the mechanisms of landscape change (Fahrig 2003; Fischer & Lindenmayer 2007) and the relative influence of habitat loss and habitat fragmentation on asymmetric dispersal are necessary for landscape management for wildlife conservation (Wiegand, Revilla & Moloney 2005; With, Schrott & King 2006).

Recently developed population genetic models allow the estimation of bidirectional migration rates over ecologically relevant time frames and can approximate biologically realistic scenarios such as asymmetric dispersal (Pearse & Crandall 2004). One such model uses coalescent theory (Kingman 1982) to approximate the genealogical history of genetic samples backward in time to estimate the number of immigrants per generation that originated in other sampled populations (Beerli & Felsenstein 2001; Wakeley 2001). This analytic approach has proved useful for identifying the dynamics of asymmetric dispersal and effective population sizes expected under source-sink population structure (Fraser et al. 2007).

In this paper, field survey and molecular data were used to evaluate a priori hypotheses for the effects of landscape change on patch occupancy and asymmetric migration in a spatially structured rainforest bird population. We estimated logrunner (Orthonyx temminckii, Ranzani 1822) occupancy rates for 46 rainforest patches in a regional study area and quantified bidirectional genetic migration rates among 11 logrunner populations in a portion of the study area. Our research objectives were to (i) investigate spatial variation in logrunner patch occupancy rates to determine the relative effects of landscape composition, landscape change and patch structure on local extinction; (ii) determine the extent to which emigration and immigration rates were asymmetric; (iii) evaluate spatial variation in emigration and immigration rates to discover the relative influence of local and between-population landscapes on asymmetric dispersal; and (iv) investigate spatial variation in emigration and immigration rates to determine the relative contributions of habitat loss and habitat fragmentation to asymmetric dispersal.

Materials and methods

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

Study Species

The logrunner was selected to investigate the effects of landscape change on patch occupancy and asymmetric dispersal because specialized life-history traits may predispose this species to local extinction and restricted dispersal (Sodhi, Liow & Bazzaz 2004; Fahrig 2007). Logrunners are ground-dwelling Passerines (Family Orthonychidae) endemic to subtropical rainforests of south-eastern Australia (Higgins & Peter 2002). The adults are sedentary and maintain year-round home ranges, and natal dispersal is the primary mode of dispersal between populations. Because logrunners possess short rounded wings and a partially atrophied sternum, they have a limited ability for sustained flight. In addition, this species has highly specialized morphological and behavioural adaptations for terrestrial foraging in the leaf litter of rainforests (Higgins & Peter 2002).

Previous research indicated the connectivity of logrunner populations was historically limited by the extent of dry eucalypt (Eucalyptus spp.) forest, but naturally heterogeneous landscapes with interdispersed patches of rainforest facilitated gene flow (Pavlacky et al. 2009). In addition, there was evidence for a contemporary shift in the pattern of dispersal, with deforestation having become the most important barrier to migration between logrunner populations. The generation time of logrunners was estimated to be 4 years, which suggested approximately 25 generations have elapsed since deforestation began in the early 1900s (Pavlacky et al. 2009).

Study Area

The regional study area for the investigation of logrunner site occupancy (8049 km2) was located within the South East Queensland Biogeographic Region, Australia (27°30′–28°23′S, 152°30′–153°38′E) (Fig. 1). This area included several World Heritage sites that are part of the Gondwana Rainforests of Australia, including Lamington, Moogerah Peaks, Mount Barney, Mount Chinghee, Tamborine and Springbrook National Parks, as well as rainforest remnants on private land. Eucalypt forest was the dominant vegetation type with rainforest on the mountain tops, and in alluvial fans and wet gullies. A reconstruction of historical forest structure (EPA 2004) within the regional study area showed rainforest cover had decreased by 52% since the early 1900s and occupied 4% of the land area at the time of this study (Pavlacky 2008).

image

Figure 1.  Map of the regional and dispersal study areas in South East Queensland, Australia. The map of the regional study area depicts the location of the occupancy transects and the distribution of rainforest vegetation. The map of the dispersal study area shows the contemporary landscape structure and local landscape buffers for the (1) Binna Burra, (2) Coomera River, (3) Fairview Mountain, (4) Green Mountains, (5) Mount Nimmel, (6) Nixon Creek, (7) Numinbah Valley, (8) Quamby Falls, (9) Sarabah, (10) Warrie and (11) Willowie logrunner populations.

Download figure to PowerPoint

The study area for the genetic analyses of dispersal covered a portion (415 km2) of the regional study area (28°07′19″–28°16′37″S, 153°05′59″–153°19′49″E) on the McPherson Range, including Lamington and Springbrook National Parks, and private land holdings (Fig. 1). The contemporary landscape was composed of 38% subtropical rainforest, 36% eucalypt forest, 21% deforested land and 4% second-growth forest (Pavlacky et al. 2009). A reconstruction of the historical landscape structure (EPA 2004) in the dispersal study area indicated 27% of the rainforest cover had been lost to deforestation over the previous 100 years (Pavlacky 2008).

Occupancy Sampling

A two-way random stratified sample design was used to select 46 rainforest patches from the regional study area (Pavlacky 2008). The patches were stratified into three rainforest types: (i) upland notophyll vine forest; (ii) lowland notophyll vine forest; and (iii) Araucarian (Araucaria cunninghamii) notophyll–microphyll vine forest (Webb, Tracey & Williams 1984). Upland rainforests occurred between 800 and 900 m, lowland forests below 700 m and Araucarian rainforests below 700 m on dry, well-drained slopes and foothills (Webb, Tracey & Williams 1984). For each rainforest type, nine small (2·5–16·7 ha), three intermediate (20·2–75·9 ha) and three large (>189·0 ha) rainforest patches were randomly selected. One additional intermediate-sized Araucarian rainforest patch was sampled. For the purpose of this study, the lowland and Araucarian rainforests occurring below 700 m were combined to represent lowland rainforests.

The logrunners were surveyed by sight and song by D. Pavlacky along a single transect within each of the 46 rainforest patches from 2 November 2004 to 21 January 2005 and from 29 October 2005 to 8 January 2006. The surveys occurred shortly after the logrunner breeding season (April–October) during the fledgling to independence period (Higgins & Peter 2002). Along each transect, logrunners were sampled at 4–8 unlimited-distance point count stations between 0·5 and 4 h after sunrise, for durations of 10 min at each station. The point counts were located at 100-m intervals along the line transects using a systematic design with a random start. The mean transect length was 605 m, and the mean number of points per transect was 6·7, which, based on the estimated population density and effective detection radius (not shown), intersected approximately three logrunner territories. Point-specific covariates for time of day, as well as survey-specific covariates for year and time of year were recorded to account for potential heterogeneity in logrunner detection probabilities (Pavlacky 2008).

Genetic Sampling

Between 2001 and 2006, 220 logrunners were captured and sampled by D. Pavlacky and D. Green at 11 rainforest sites using a systematic sampling design, with nine eastern locations separated by c. 5 km and two western sites separated by c. 10 km (Pavlacky et al. 2009). The sample locations were selected to ensure the local and between-population landscapes varied in the composition and configuration of forest types, as well as to ensure the landscapes varied in the degree of habitat loss and fragmentation (Fig. 1). The birds were lured into a 9-m-long, 2·5-m-high and 38-mm mesh mist-net using broadcasts of territorial vocalizations from two 4-W speakers arranged on either side of the net lane. For each bird captured, a 40–60 μL blood sample was obtained from the brachial vein, and the samples were stored in Queens’ lysis buffer solution.

DNA was extracted from the blood samples using an ammonium acetate salting-out method, and 10 microsatellite markers were amplified using the polymerase chain reaction (PCR) protocol of Nicholls et al. (2007). The primers were fluorescently labelled, and the PCR products were sized using capillary electrophoresis (MegaBACE 500; GE Healthcare, Chalfont St. Giles, Buckinghamshire, UK). The mean genotyping error per locus over the 10 loci for this study was 0·0063 (Pavlacky et al. 2009). The performance of the 10 microsatellite loci was evaluated, and two loci exhibited high frequencies of null alleles, and therefore, eight loci were used for the final analyses. In addition, Hardy–Weinberg and migration-drift equilibrium assumptions were evaluated, which were necessary for the population genetic analysis of allele frequency data (Beerli & Felsenstein 1999; Pearse & Crandall 2004). All populations except one (Sarabah) were within Hardy–Weinberg equilibrium. A coalescent-based, Markov chain Monte Carlo simulation indicated the logrunner populations were at migration-drift equilibrium. The logrunner populations demonstrated low but statistically significant genetic differentiation (FST = 0·015; SE = 0·005; 95% CI = 0·005, 0·025), and 33 of the 55 pair-wise estimates of FST were statistically different (Pavlacky et al. 2009).

Landscape Structure

The landscape structure immediately surrounding the sampled populations represented the ‘local landscapes’, and the landscapes separating the sampled populations represented the ‘between-population landscapes’. Little is known about the home range size and dispersal ecology of logrunners, but specialized life-history traits suggested this species is sedentary with limited dispersal ability (Higgins & Peter 2002). A 2 km radius was selected to characterize local landscapes because Australian woodland bird species with limited dispersal ability usually responded to landscape features within 2 km than rather than to landscapes at larger scales (Westphal et al. 2003). The structure of the local landscape surrounding each sample location was characterized using the preclearing and current vegetation cover data (regional ecosystems; EPA 2004) and ArcGIS version 9.1 (ESRI 2005). At each sample location, a 2-km-radius circular buffer was centred on the mean Universal Transverse Mercator (UTM) coordinates of the point count and capture locations (Pavlacky 2008). The land cover from the preclearing and current regional ecosystem data layers was clipped using the circular buffer. Within each of the historical and contemporary 12·6-km2 landscapes, the forest type of the sampled locations was recorded, and the total area of rainforest cover (km2) and the mean area of rainforest patches (km2) were measured using ArcGIS. The percentage of rainforest cleared and the reduction in mean patch area was measured by calculating the per cent change between the preclearing and current land cover data (Table 1). Mean patch isolation in the regional study area was measured as the mean distance (km) from the sampled patch to the edge of the nearest patch (>2·5 ha) in each of four quadrants delineated by the cardinal directions (Table 1).

Table 1.   Landscape variables describing the composition and modification of local (12·6 km2) landscapes for the dispersal and occupancy study areas, and the composition and configuration of between-population (0·5-km-wide) landscapes for the dispersal study area. The table includes the variable names, descriptions and means (ranges) for continuous variables or levels (frequencies) for categorical variables
Landscape variablesDescriptionsMeans (ranges) or levels (frequencies)
  1. aLandscape variables for the dispersal study area. bLandscape variables for the occupancy study area.

Local landscapes
 Forest typeRainforest type of the sampled locations within the landscapesUpland (3), Lowland (7)a Upland (15), Lowland (31)b
 Rainforest coverPer cent of rainforest cover within the contemporary landscapes41·6% (10·8–93·6%)a 27·3% (0·5–84·7%)b
 Rainforest cover lostPer cent change of rainforest cover for historical and contemporary landscapes24·8% (1·8–53·7%)a 36·8% (0·0–97·2%)b
 Patch area lostPer cent change of mean rainforest patch area for historical and contemporary landscapes62·0% (11·5–94·6%)a 64·2% (0·0–98·9%)b
Patch structure
 Patch areaThe natural log of the area of the rainforest patch31·9 km2 (<0·1–427·5 km2)b
 Mean patch isolationMean distance to the nearest rainforest patch in each quadrant of the four cardinal directions1·2 km (0·1–6·8 km)b
Between-population landscapes
 Rainforest coverPer cent of contemporary rainforest cover within the landscapes40·9% (7·7–98·9%)a
 Wet eucalypt forest coverPer cent of contemporary wet eucalypt forest cover within the landscapes24·8% (0·9–78·5%)a
 Deforested land coverPer cent of deforested land cover within the landscapes18·0% (0·0–50·3%)a
 Forest patch densityDensity of contemporary forest patches within the landscapes8·2 km−2 (1·0–16·5 km−2)a
 DistanceEuclidean distance between the populations9·4 km (3·2–18·7 km)a

For the between-population landscapes, 0·5-km-wide landscapes between the sampled logrunner populations were used to represent the vegetation encountered by dispersing individuals and to represent multiple dispersal pathways. The 0·5-km-wide buffer distance was approximately five times as wide as a typical logrunner territory. This buffer width was a compromise between minimizing the areas of overlap between the 0·5-km dispersal buffers for pairs of study populations and maximizing the coverage of the landscapes within the dispersal study area. The 0·5-km-wide landscapes effectively described the negative effect of deforestation and patch configuration on symmetric migration between logrunner populations (Pavlacky et al. 2009). The structure of these 0·5-km-wide between-population landscapes was characterized using the contemporary regional ecosystem data (EPA 2004) and ArcGIS (ESRI 2005). The 0·5-km-wide buffers were centred on straight-line vectors between each pair of sample locations, and the land cover was clipped from the contemporary regional ecosystem data layer to the extent of the buffers (Pavlacky et al. 2009). Within each 0·5-km-wide buffer, the per cent cover of rainforest, wet eucalypt forest, deforested land and forest patch density was measured (Table 1). The density of all distinct forest patches in the buffers was calculated as the number of patches per km2. The mean area of the buffers used to estimate the structure of the between-population landscapes was 4·9 km2 (SD = 2·1).

Occupancy Estimation

Logrunner detection (p) and occupancy (ψ) probabilities were estimated using a zero-inflated binomial model (MacKenzie et al. 2002; Tyre et al. 2003) implemented in the software programs mark (White & Burnham 1999) and presence (Hines 2006). mark was used for model construction, estimation and selection, and presence was used for assessing model fit. Detection probabilities were estimated from spatially replicated point count surveys within a single visit (MacKenzie et al. 2006). The model estimated the probability (pit) that the species was detected at replicated point count survey t, given presence at site i, and the probability (ψi) that the species was present at site i (MacKenzie et al. 2002). The detection (pit) and occupancy (ψi) parameters were modelled as functions of survey and/or site-specific covariates using the logit link function (MacKenzie et al. 2002). Categorical covariates were coded as indicator variables, and continuous covariates were standardized using the z-transformation. The occupancy models were constructed using all subsets of the three detection covariates (rainforest type, time of day and ordinal date) and six occupancy covariates (Table 1) with an upper limit of five covariates, which resulted in a candidate set of 130 models. The goodness-of-fit of the global model (number of parameters K = 9) was evaluated using the Pearson’s χ2 statistic and 10 000 parametric bootstrap iterations in program presence (MacKenzie & Bailey 2004).

Migration Estimation

The genetic estimates of migration within the expected dispersal distance of the species were used to infer dispersal success (Rousset 2001), which was defined as dispersal from a natal area to an area where breeding first takes place (Greenwood & Harvey 1982). The structured coalescent model in program migrate version 2.1.3 (P. Beerli, Tallahassee, Florida, USA) was used to estimate the migration rate per generation Mji = mji/μ and effective population size θi = 4Neμ, where mji was the number of immigrants per generation from subpopulation j into subpopulation i, Ne was the effective population size for subpopulation i, and μ was the locus specific mutation rate per generation (Beerli & Felsenstein 2001). The mutation rate μ was held constant across all eight loci and μ was modelled according to the step-wise mutation model (Pavlacky 2008). The migration rates Mji between all 110 population pairs were estimated within an unconstrained migration matrix. The unconstrained parameterization allowed the effective population size θi to vary by subpopulation and thus did not assume equal population sizes. The starting parameter values were calculated from the mean parameter estimates over four preliminary analyses. The model ran with a burn-in of 10 000 iterations, 10 short chains (1000 sampled genealogies), two long chains (10 000 sampled genealogies) and sampled one of every 20 constructed genealogies. The moving steps option was used to ensure a minimum tree acceptance of 200 genealogies for short chains and 2000 genealogies for long chains. Variation between independent model runs was accounted for by estimating the final parameters over 10 replicates of the model (Pavlacky 2008).

The assumptions of the structured coalescent model were (1) Wright-Fisher random mating within populations; (2) migration-drift equilibrium; and (3) constant effective population sizes and mutation rates through time (Beerli & Felsenstein 1999). Assumption (1) was satisfied for all populations except Sarabah (see Genetic Sampling section). Assumption (2) was verified using a coalescent simulation, which indicated sufficient time since habitat loss had elapsed for the pattern of migration and genetic drift to stabilize (Pavlacky et al. 2009). Assumption (3), which assumed effective population sizes had been constant over time, may be questionable in local landscapes experiencing as much as 54% habitat loss (Table 1). However, the estimates of migration and effective population size are long-term averages heavily weighted by coalescent events in the recent past (Beerli 2009). Because assumption (2) indicated sufficient time had elapsed for the pattern of migration to stabilize, the estimates of migration and effective population size likely reflect responses to contemporary landscape conditions post landscape change rather than partial non-equilibrium responses to ongoing habitat loss. In addition, high immigration rates are capable of arresting population declines in landscapes impacted by habitat loss (With, Schrott & King 2006). The lack of a correlation between effective population size and habitat loss (see below) provided little evidence for a relationship between current effective population size and past habitat loss.

The migration rates were grouped according to their respective emigration and immigration populations, and the data were stacked to represent 110 emigration and 110 immigration rates (= 220). A generalized linear model with the normal distribution and identity link function (Nelder & Wedderburn 1972; proc genmod, SAS Institute 2008) was used to investigate the asymmetry of logrunner migration rates. An analysis of variance parameterization and likelihood ratio test were used to investigate the difference between the emigration and immigration rates nested within the 11 populations. The 95% confidence intervals for the parameter estimates were calculated using likelihood profiles. The fit of the migration data to a normal distribution was evaluated using the Kolmogorov–Smirnov goodness-of-fit test (proc univariate, SAS Institute 2008).

The logrunner emigration (= 110) and immigration (= 110) rates were modelled as a function of local and between-population landscape features using a generalized linear mixed models with the normal distribution and identity link function (proc glimmix, SAS Institute 2008; Bolker et al. 2009). The emigration and immigration models included block covariance structures for the respective emigration and immigration populations, and population pairs. This within- and between-covariance structure permitted landscape covariates on emigration or immigration within populations as well as directional migration between populations, and appropriately accounted for the nonindependence of the clustered observations. All subsets of the covariance structures for the full fixed-effect models (number of parameters K = 8) were evaluated using restricted maximum pseudo-likelihood and likelihood ratio tests. After selecting an appropriate error structure for the models, the model parameters were estimated using the Laplace maximum likelihood approximation. The best covariance structure for the emigration data included random effects for the 11 emigration populations and 55 population pairs (inline image; = 0·020). The best covariance structure for the immigration data included random effects for the 11 immigration populations and 55 population pairs (inline image; = 0·001). After determining the appropriate covariance structure for the data, all one and two variable subsets of the fixed-effect covariates in Table 1 were evaluated, which resulted in candidate sets of 45 models for emigration and immigration. The 95% confidence intervals for the fixed-effect parameters and covariance parameters were calculated using the t distribution and likelihood profiles, respectively. The colinearity among the predictor variables (Table 1) was investigated using the Spearman’s rank-order correlation (proc corr, SAS Institute 2008), and with exception of rainforest lost and patch area lost (= 0·75), all pairs of variables exhibited correlations |r| < 0·7. In addition, correlations between logrunner effective population size (θi) and the amount of rainforest lost and mean patch area lost in local landscapes were investigated using the Spearman’s rank-order correlation (= 11), and no relationship between θi and the landscape modification variables was found (|r| < 0·08, > 0·81).

Hypotheses and Model Justification

Predictive models and the method of multiple working hypotheses (Chamberlain 1965) were used to evaluate a priori hypotheses for how landscape processes could influence the patch occupancy and asymmetric dispersal between logrunner populations. Objective (1) investigated spatial variation in logrunner patch occupancy rates to determine the relative influence of landscape and patch structure on local extinction. We hypothesized that patch occupancy was a function of (1·1) landscape composition; (1·2) habitat loss and habitat fragmentation; and (1·3) the area and isolation of rainforest patches. Three hypotheses for logrunner detection probabilities were also evaluated, which proposed that detection varied by rainforest type, time of year and time of day. We assumed that declining occupancy rates in response to landscape modification over time would be indicative of local extinction, while declining occupancy in response to increasing patch isolation would be symptomatic of restricted dispersal and reduced colonization of isolated patches (Hanski 1998). A species-oriented approach was adopted to represent landscape change (Fischer & Lindenmayer 2007), where habitat loss was represented by percentage of rainforest lost and habitat fragmentation was represented by the per cent change of mean rainforest patch area over time. Hypothesis (1·1) was represented by models containing the rainforest cover and forest type variables, hypothesis (1·2) by models containing the rainforest cover lost and patch area lost variables, and hypothesis (1·3) by the patch area and isolation variables (Table 1).

To address objective (2), emigration and immigration rates were investigated to determine the degree of asymmetric dispersal between logrunner populations. Hypothesis (2·1) proposed that dispersal rates were symmetric, and hypothesis (2·2) proposed that dispersal rates were asymmetric. Evidence for asymmetric dispersal was inferred when populations demonstrated net emigration or net immigration (Kawecki & Holt 2002).

After establishing the extent of asymmetric dispersal, objective (3) investigated spatial variation in emigration and immigration rates to determine how local and between-population landscapes influenced asymmetric dispersal between logrunner populations. We hypothesized that emigration and immigration were related to (3·1) the composition and modification of local landscapes and (3·2) the composition and configuration of the between-population landscapes. Our inference about the effects of landscape structure on emigration and immigration was based on the perspective that the composition and modification of local landscapes would influence asymmetric dispersal by within-population vital rates (Pulliam 1988; Revilla & Wiegand 2008), whereas the composition and configuration of between-population landscapes would influence asymmetric dispersal through dispersal mortality or landscape connectivity between populations (Gustafson & Gardner 1996; Haynes et al. 2007). Hypothesis (3·1) was represented by models containing the forest type, rainforest cover, rainforest cover lost and patch area lost variables measured in local landscapes, and hypothesis (3·2) was represented by the rainforest cover, wet eucalypt forest cover, deforested land cover and forest patch density variables measured in the between-population landscapes (Table 1). A null hypothesis that emigration and immigration were related to the distance between the populations was also evaluated.

Objective (4) investigated emigration and immigration rates to determine the relative contributions of habitat loss and habitat fragmentation to asymmetric dispersal between logrunner populations. We hypothesized that spatial variation in emigration and immigration rates was a function of (4·1) habitat loss and (4·2) habitat fragmentation in local landscapes. These two hypotheses reflected the perspective that spatial patterns of landscape change can influence the extent of asymmetric dispersal (With, Schrott & King 2006). Because large populations are often characterized by net emigration and small populations by net immigration (Stacey & Taper 1992), habitat loss and fragmentation were expected to reduce population size (Bender, Contreras & Fahrig 1998) and produce asymmetric migration between the modified and intact landscapes. Hypothesis (4·1) was represented by the loss of rainforest cover over time, and hypothesis (4·2) was represented by the reduction in mean patch area over time (Table 1).

Model Selection

Relative Kullback–Leibler Information lost for models used to approximate reality was estimated using the Akaike Information Criterion adjusted for sample size (AICc) (Burnham & Anderson 2002). The models were ranked by ΔAICc, the strength of evidence for alternate hypotheses was measured by AICc weights (wi), and the likelihood of the modelled hypotheses given the data was quantified by evidence ratios (wi/wj). The relative importance of the hypotheses and predictor variables were quantified by cumulative AICc weights [w+(j)]. Model averaged predictions, parameter estimates, unconditional standard errors and 95% confidence intervals were calculated for model sets defined by the 0·135 evidence ratio (AICc < 4). The strength of evidence for effect sizes was determined by evaluating the model averaged parameter estimates (inline image) with respect to zero using unconditional standard errors and 95% confidence intervals (Burnham & Anderson 2002).

Results

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

Patch Occupancy

Multimodel inference for the effects of landscape features on logrunner patch occupancy indicated slightly more support for the patch area and isolation hypothesis [w+(j) = 0·98] than the habitat loss and fragmentation hypothesis [w+(j) = 0·74]. The patch area and isolation hypothesis was 17 times more probable, and the habitat loss and fragmentation hypothesis was 12 times more probable than the landscape composition hypothesis [w+(j) = 0·06]. Cumulative AICc weights indicated patch isolation [w+(j) = 0·96], habitat loss [w+(j) = 0·45] and habitat fragmentation [w+(j) = 0·28] were the best predictors of logrunner patch occupancy. Rainforest forest type, remnant rainforest cover and patch area demonstrated much lower ability to predict logrunner occupancy [w+(j) < 0·04]. The habitat loss hypothesis was twice as probable as the habitat fragmentation hypothesis. Logrunner detection probabilities were best predicted by rainforest type [w+(j) = 0·91], followed by time of day [w+(j) = 0·25] and time of year [w+(j) = 0·04].

Logrunner patch occupancy was best predicted by a model containing mean patch isolation and habitat loss (Table 2). Occupancy declined with increasing mean patch isolation (inline image; SE = 1·19; CI = −5·16, −0·49) and the percentage of rainforest cover lost since settlement (Table 3; Fig. 2), and the 95% confidence intervals for these effects excluded zero (Table 3). There was nearly equal support for the second best model containing the effects of mean patch isolation and patch area lost (Table 2). Logrunner occupancy declined with increasing patch isolation and patch area lost (βPatch area lost = −2·56; SE = 1·36; CI = −5·23, 0·11) (Fig. 2). The 95% confidence interval for the effect of patch area lost narrowly covered zero, indicating a marginal effect size. There was also support for the third best model containing only the effect of mean patch isolation, but this model was two times less probable than the best approximating model (Table 2). The goodness-of-fit test indicated the full model fit the data reasonably well (χ2 = 537·0; = 0·199).

Table 2.   Model selection statistics for the effects of local landscapes and patch structure on logrunner patch occupancy, and the effects of local and between-population landscapes on emigration and immigration
ModelKLog (L)AICcΔAICc w i
Patch occupancy
 ψ (patch isolation + rainforest lost) p (forest type)5−160·46332·410·000·374
 ψ (patch isolation + patch area lost) p (forest type)5−160·88333·250·840·246
 ψ (patch isolation) p (forest type + time of day)5−161·29334·081·670·163
Emigration
 Patch area lost (local) + rainforest cover (local)6−305·49623·800·000·171
 Patch area lost (local)5−306·92624·420·620·125
 Patch area lost (local) + forest type (local)6−306·01624·831·030·102
 Patch area lost (local) + patch density (between)6−306·40625·611·810·069
 Patch area lost (local) + distance (between)6−306·53625·872·070·061
 Patch area lost (local) + rainforest lost (local)6−306·61626·032·230·056
 Patch area lost (local) + rainforest cover (between)6−306·92626·642·840·041
 Patch area lost (local) + deforested cover (between)6−306·91626·642·840·041
 Patch area lost (local) + wet eucalypt cover (between)6−306·92626·662·860·041
 Rainforest lost (local)5−308·18626·933·130·036
 Forest type (local)5−308·27627·113·310·033
 Rainforest lost (local) + forest type (local)6−307·49627·793·990·023
Immigration
 Patch area lost (local) + patch density (between)6−305·33623·470·000·241
 Rainforest lost (local) + patch density (between)6−306·44625·692·220·079
 Patch area lost (local) + rainforest cover (between)6−306·53625·872·400·073
 Patch area lost (local)5−308·17626·913·440·043
 Rainforest lost (local) + rainforest cover (between)6−307·14627·093·620·039
 Patch area lost (local) + wet eucalypt cover (between)6−307·15627·113·640·039
 Forest type (local) + patch density (between)6−307·20627·203·730·037
 Patch area lost (local) + deforested cover (between)6−307·33627·463·990·033
Table 3.   Parameter estimates, standard errors (SE), and lower (LCL) and upper (UCL) 95% confidence limits from the best approximating models for the effects of local landscapes and patch structure on logrunner patch occupancy, and the effects of local and between-population landscape features on emigration and immigration
ParametersEstimateSELCLUCL
Patch occupancy
 ψ (intercept)2·3840·9240·5734·196
 ψ (patch isolation)−2·6831·049−4·738−0·627
 ψ (rainforest lost)−1·7010·770−3·210−0·192
 p (intercept)−0·6880·188−1·056−0·320
 p (upland forest)0·9170·2830·3621·472
Emigration
 Intercept12·8201·14210·18515·454
 Patch area lost (local)−0·0500·014−0·079−0·022
 Rainforest cover (local)0·0290·016−0·0030·061
 Random effect of emigration population0·0550·9330·0003·292
 Random effect of population pair3·1662·9190·0009·424
 Residual error12·2353·3967·44918·586
Immigration
 Intercept10·3481·6356·65014·047
 Patch area lost (local)0·0560·0200·0160·097
 Patch density (between)−0·3530·148−0·651−0·056
 Random effect of immigration population2·0561·5020·1047·707
 Random effect of population pair5·1332·2970·77510·540
 Residual error9·5102·1336·32415·307
image

Figure 2.  Predicted logrunner patch occupancy as a function of mean patch isolation, rainforest lost and mean patch area lost from the best approximating models. The bold trend lines are model averaged probabilities of occupancy, and the filled areas are unconditional 95% confidence intervals.

Download figure to PowerPoint

Asymmetric Migration

The emigration and immigration rates for the 11 logrunner populations were asymmetric (inline image; < 0·001). The 95% confidence intervals of the effects with respect to zero indicated emigration exceeded immigration for the Binna Burra (βBinna Burra = 4·9; SE = 1·2; CI = 1·5, 8·3) and Sarabah (βSarabah = 5·5; SE = 1·2; CI = 2·1, 8·9) populations, while emigration was less than immigration for the Fairview (βFairview = −4·0; SE = 1·2; CI = −7·4, −0·6), Numinbah (βNuminbah = −3·8; SE = 1·2; CI = −7·2, −0·4) and Warrie (βWarrie = −5·1; SE = 1·2; CI = −8·5, −1·7) populations. The migration data fit the normal distribution (= 0·07; > 0·150).

Do Local or Between-Population Landscapes Influence Emigration and Immigration?

Multimodel inference for emigration away from logrunner populations indicated the local landscape hypothesis [w+(j) = 0·97] was three times more probable than the between-population landscape hypothesis [w+(j) = 0·33]. Cumulative AICc weights indicated mean patch area lost in local landscapes [w+(j) = 0·71] was the best predictor of logrunner emigration, followed by forest type [w+(j) = 0·23], rainforest cover [w+(j) = 0·21] and rainforest lost [w+(j) = 0·20] in local landscapes. The effect of patch area lost on emigration was seven times more probable than the effect of distance between the populations [w+(j) = 0·10], and rainforest cover and rainforest lost were two times more probable than distance. The between-population landscape variables, patch density, deforested land, rainforest cover and wet eucalypt cover, exhibited much lower predictive ability [w+(j) < 0·12].

Emigration away from logrunner populations was best predicted by mean patch area lost and the amount of remnant rainforest cover in local landscapes (Table 2). Emigration declined with increasing mean patch area lost (inline image SE = 0·017; CI = −0·083, −0·014) and increased with increasing remnant rainforest cover in local landscapes (Table 3; Fig. 3). The 95% confidence interval for the effect of mean patch area lost did not cover zero, and the interval for the effect of remnant rainforest cover narrowly covered zero (Table 3), indicating large and marginal effect sizes, respectively. There was nearly equal support for the second best model containing only the effect of mean patch area lost (Table 2). The addition of the other predictor variables did not appreciably increase model fit as measured by the log likelihood (Table 2), and the 95% confidence intervals for these effects substantially covered zero. For example, there was little evidence for an effect of distance on emigration in the fifth best model (βDistance = −0·095; SE = 0·107; CI = −0·310, 0·121).

image

Figure 3.  Predicted logrunner emigration as a function of mean patch area lost and remnant rainforest cover in local landscapes and predicted logrunner immigration as a function of mean patch area lost in local landscapes and forest patch density in between-population landscapes. The bold trend lines are model averaged emigration and immigration rates, and the filled areas are unconditional 95% confidence intervals. The open circles are mean emigration and immigration rates for each population, and the error bars are one standard error of the mean. The open circles without error bars are the immigration rates for the Binna Burra population.

Download figure to PowerPoint

Multimodel inference for immigration into logrunner populations showed equal support for the local landscape [w+(j) = 0·82] and between-population landscape [w+(j) = 0·80] hypotheses. Cumulative AICc weights [w+(j)] indicated mean patch area lost in local landscapes [w+(j) = 0·51] and forest patch density in the between-population landscapes [w+(j) = 0·46] were the best predictors of logrunner immigration, followed by rainforest lost in local landscapes [w+(j) = 0·21] and rainforest cover in between-population landscapes [w+(j) = 0·17]. The effects of local patch area lost and intervening forest patch density on immigration were six times more probable than the effect of distance between the populations [w+(j) = 0·08]. Wet eucalypt forest and deforested cover in the between-population landscapes, and forest type and rainforest cover in local landscapes exhibited low predictive ability [w+(j) < 0·13].

Immigration into logrunner populations was best predicted by mean patch area lost in local landscapes and forest patch density in between-population landscapes (Table 2). Immigration increased with increasing mean patch area lost in local landscapes (inline image; SE = 0·020; CI = 0·012, 0·093) and declined with increasing forest patch density in the between-population landscapes (inline image; SE = 0·148; CI = −0·640, −0·059) (Table 3; Fig. 3). The 95% confidence intervals for the effects of reduced mean patch area and forest patch density did not cover zero, indicating large effect sizes for these variables. As with emigration, there was little evidence for an effect of distance on immigration.

Does Habitat Loss or Fragmentation Influence Emigration and Immigration?

Multimodel inference for emigration away from logrunner populations indicated the habitat fragmentation hypothesis [w+(j) = 0·71] was four times more probable than the habitat loss hypothesis [w+(j) = 0·20]. As presented in the previous section, logrunner emigration was best predicted by a model containing the negative effect of mean patch area lost (Tables 2 and 3; Fig. 3). There was support (AICc < 4) for models containing the negative effects of rainforest cover lost (inline image; SE = 0·032; CI = −0·122, 0·006), but these models were five times less probable than the best approximating model containing the fragmentation effect (Table 2).

Multimodel inference for immigration into logrunner populations indicated the habitat fragmentation hypothesis [w+(j) = 0·51] was two times more probable than the habitat loss hypothesis [w+(j) = 0·21]. As presented earlier, logrunner immigration was best predicted by a model containing the positive effect of mean patch area lost (Tables 2 and 3; Fig. 3). There was support (AICc < 4) for the second best model containing the positive effect of rainforest cover lost (inline image; SE = 0·033; CI = 0·005, 0·138), but this model was three times less probable than the best approximating model containing the fragmentation effect (Table 2).

Discussion

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

The occurrence of logrunner populations in our study region was primarily influenced by rainforest patch structure and anthropogenic landscape change. Patch isolation was the best predictor of logrunner patch occupancy, with smaller effects of rainforest loss and fragmentation over time. Patch area, remnant rainforest cover and rainforest type had very little influence on logrunner occupancy rates. The large effect of patch isolation suggested dispersal and colonization from neighbouring habitat patches were important processes for maintaining regional patch occupancy (Hanski 1998). However, after controlling for patch isolation, the loss and fragmentation of rainforest over time influenced patch occupancy rates to a greater extent than patch isolation alone. This suggested that immigration from adjacent patches was unable to prevent local extinction in landscapes that have experienced high levels of habitat loss and fragmentation. Moreover, the occurrence of logrunner populations was primarily influenced by habitat loss over the past 100 years since European settlement, while reduction in mean patch area over time and associated edge effects were less important. This result provided little support for the hypothesis that habitat fragmentation was more important than habitat loss for explaining local extinction of a highly specialized rainforest species (Bender, Contreras & Fahrig 1998; Fahrig 2003).

The patch occupancy study indicated that dispersal from surrounding rainforest patches and anthropogenic landscape change played important roles for the occurrence of logrunner populations. We investigated spatial variation in emigration and immigration rates to gain a mechanistic understanding of how landscape composition and modification influences dispersal between logrunner populations. Our results suggested that the effects of landscape change on asymmetric dispersal played an important role in local extinction of logrunner populations.

However, the differences in the landscape structure of the regional and dispersal study areas may have limited our ability to make inferences about the relationship between asymmetric dispersal and local extinction. The occupancy study occurred in a large region with 4% rainforest cover and 52% habitat loss, while the dispersal study occurred in a portion of the region with 38% rainforest cover and 21% habitat loss (Fig. 1). The local landscapes for the occupancy study contained on average 27% rainforest cover and 37% habitat loss, whereas the landscapes for the dispersal study contained on average 42% rainforest cover and 25% habitat loss (Table 1). The large degree of asymmetric dispersal at moderate levels of landscape modification suggested the effects of landscape change on dispersal between logrunner populations may be larger in the regional study area than reported for the dispersal study area. Nevertheless, the regional occupancy study controlled for variation in patch isolation, making the landscape effects comparable in the regional and dispersal study areas. Finally, although the sizes of the local and between-population landscapes were based on logrunner life history, the scales of the landscape measurements may have limited the ability to evaluate the relative effects of habitat loss and fragmentation on occupancy and asymmetric dispersal (Smith, Fahrig & Francis 2011).

The comparison of emigration with immigration rates indicated a high degree of asymmetric dispersal between the 11 populations. The migration rates reported in this study can reflect effective dispersal, or natal dispersal to an area where successful breeding occurred. The estimates of effective dispersal represented long-term averages of the number migrants per generation and thus were useful for approximating the dispersal dynamics of logrunners since European landscape change began in the early 1900s. The Binna Burra and Sarabah populations in contiguous upland and lowland forests of Lamington National Park (427 km2) were net exporters of dispersing logrunners, while the Fairview and Warrie populations in contiguous lowland forests of Springbrook National Park (37 km2), as well as the Numinbah population in a small isolated lowland patch (1 km2), were net importers of dispersing birds. Relatively intact landscapes were net exporters, and landscapes experiencing large reductions in mean rainforest patch size over time were net importers of dispersing logrunners. In fact, a >60% reduction in mean patch size over time was capable of switching populations from net exporters to net importers of dispersing logrunners. For example, estimated emigration was 39% greater than immigration in landscapes with an 11% reduction in mean patch area, whereas estimated immigration was 37% greater than emigration in landscapes with a 95% reduction in mean patch area (Fig. 3). The effect of between-population forest patch density on immigration also influenced the extent of asymmetric migration (Fig. 3).

The capacity of logrunner populations to produce emigrants that went on to reproduce in other populations was primarily related to the composition and modification of local landscapes. The structure of the between-population landscapes had very little influence on emigration. This indicated that processes occurring within local landscapes were more important than the connectivity of between-population landscapes for maintaining emigration and suggested that emigration was limited by low reproduction and high mortality in modified local landscapes (Revilla & Wiegand 2008). Local landscape structure was expected to cause variation in population vital rates (Pulliam 1988; Revilla & Wiegand 2008) and population size (Stacey & Taper 1992), which would have important implications for the generation of surplus individuals that are available to emigrate. On the other hand, between-population processes, such as landscape permeability, boundary effects and mortality during transition (Bowler & Benton 2005), appeared less important for successful emigration. This result contrasts with the findings of Haynes et al. (2007) who found the composition of between-population landscapes was more important than local habitat quality for emigration in an experimental invertebrate system.

The capacity of logrunner populations to accept immigrants that originated in other landscapes was influenced by the structure of both local and between-population landscapes. Local and between-population landscapes were equally important for successful immigration, suggesting that immigration was influenced by high mortality in modified landscapes, as well as by dispersal mortality and/or landscape connectivity between populations (Revilla & Wiegand 2008). We found that between-population landscapes composed of small numbers of large forest patches facilitated logrunner immigration, while landscapes with large numbers of small forest patches inhibited immigration. Our results are similar to those of Revilla & Wiegand (2008) who found that adding dispersal habitat and augmenting patch sizes removed the asymmetry in dispersal and decreased the variability of immigration into source populations.

The fragmentation of between-population landscapes can produce patterns of asymmetric dispersal by funnelling potential immigrants away from habitat patches, preventing emigration because of hard-edge boundaries or by increasing dispersal mortality (Gustafson & Gardner 1996; Revilla & Wiegand 2008). We used a linear mixed model to account for the nonindependence of the emigration and immigration rates, and this approach was useful for evaluating the importance of the between-population landscapes while controlling for the effects of local landscapes. Although between-population landscapes were important for successful immigration, the structure of between–between landscapes was less important for effective emigration. The negative effect of landscape heterogeneity on immigration was sufficient to explain asymmetric dispersal between logrunner populations. These findings suggested that dispersal mortality and/or boundary effects, while animals are actively moving between habitat patches (transition stage of dispersal) may be particularly important for successful immigration. Conversely, we found little evidence for dispersal mortality or reluctance to emigrate in response to landscape structure as logrunners dispersed from natal to breeding populations. A previous study demonstrated that symmetric migration between logrunner populations was influenced by the deforestation and heterogeneity of between-population landscapes (Pavlacky et al. 2009). The current study evaluated the relative importance of local and between-population landscapes and found that reduced connectivity in the landscape was determined by rainforest fragmentation and a complex pattern of asymmetric dispersal.

Similar to the findings of With & King (2001), we found that habitat fragmentation was more important than habitat loss for promoting asymmetric dispersal between logrunner populations. This result supported the hypothesis that species from tropical or subtropical systems are particularly vulnerable to habitat fragmentation (Fahrig 2003; Sodhi, Liow & Bazzaz 2004). While habitat loss appeared to be more important for explaining local extinctions in the regional logrunner population, habitat fragmentation had more impact on the direction and extent of dispersal between logrunner populations. Large population sizes and high reproductive rates within intact landscapes were expected to promote emigration, while reduced population sizes and survival in modified landscapes were expected to facilitate immigration (Donovan et al. 1995; Revilla & Wiegand 2008). In addition, high immigration rates into landscapes experiencing population declines may have reduced the availability of high-quality nesting territories and depressed reproduction and survival rates (Lande 1987; Pulliam & Danielson 1991; Kawecki & Holt 2002). This may have further reduced the capacity of populations to produce surplus emigrants. Patch area and edge effects (With & King 2001; Laurance et al. 2002) appeared to play a bigger role than habitat loss in determining which populations were net exporters and net importers of dispersing logrunners. Although we expected substantial population declines in response to habitat loss and fragmentation (Bender, Contreras & Fahrig 1998), there was little evidence for relationships between the effective size of logrunner populations and habitat loss or habitat fragmentation. High net immigration in modified landscapes with no apparent decline in effective population sizes suggested a high level of population turnover, and the population declines appeared to be offset by high immigration rates from intact landscapes (With, Schrott & King 2006).

Conservation Implications

Our results supported simulation studies, indicating asymmetric dispersal from landscapes functioning as net exporters of dispersing individuals is important for the conservation of forest dwelling bird populations (Donovan et al. 1995; Baillie et al. 2000; With, Schrott & King 2006). Conservation measures that maintain large patch sizes in the landscape may promote asymmetric dispersal from intact to fragmented landscapes and allow bird populations to persist in landscapes degraded by habitat loss. Habitat fragmentation had a larger influence on the ability of populations to function as net exporters of dispersing birds than net habitat loss, which suggested the negative effects of habitat loss can to some extent be mitigated by conservation measures that maintain large patch sizes in the landscape (Kareiva & Wennergen 1995; With & King 2001). However, immigration may be unable to prevent local extinction in landscapes experiencing high levels of habitat loss. Fragmented sink landscapes may allow populations to persist until habitat restoration can re-establish more extensive forests (Turner & Corlett 1996). However, small forest patches in fragmented landscapes may prevent the successful immigration of dispersing individuals (Gustafson & Gardner 1996; Haynes et al. 2007). Clearly, if animals are unable to move between the habitat patches, the ecological benefits of immigration and the rescue effect will not be realized (Howe, Davis & Mosca 1991).

Acknowledgements

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

We thank the Norman Wettenhall Foundation, The Ecology Centre, Centre for Applied Environmental Decision Analysis, Queensland Ornithological Society, Birds Australia, ARC Federation fellowship to H. Possingham and ARC Discovery Grant DP0210350 to D. Green for financial support. J. Austin was instrumental in developing the microsatellite markers. J. Nicholls optimized the microsatellite markers, genotyped many of the samples and provided guidance for field collections. D. Putland supplied high-quality recordings of logrunner vocalizations. SAPAC, University of Adelaide, and M. Watts, The Ecology Centre, University of Queensland, provided computer resources for running the coalescent models. S. Blomberg contributed statistical advice. We are grateful to A. du Plessis, M. Hall, and A. and V. Quirk, Queensland Parks and Wildlife Service, and L. Fairall and P. Watts, Gold Coast City Council for logistic support. Thank you to I. and N. Gaskins, P. Kennedy, R. and J. Panitz, and P. Yaun and brothers for allowing us to capture and release logrunners on their properties.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  • Andrewartha, H.G. & Birch, L.C. (1954) The Distribution and Abundance of Animals. University of Chicago Press, Chicago, Illinois, USA.
  • Baillie, S.R., Sutherland, W.J., Freeman, S.N., Gregory, R.D. & Paradis, E. (2000) Consequences of large-scale processes for the conservation of bird populations. Journal of Applied Ecology, 37, 88102.
  • Battin, J. (2004) When good animals love bad habitats: ecological traps and the conservation of animal populations. Conservation Biology, 18, 14821491.
  • Beerli, P. (2009) How to use MIGRATE or why are Markov chain Monte Carlo programs difficult to use? Population Genetics for Animal Conservation (eds G. Bertorelle, M.W. Bruford, H.C. Hauffe, A. Rizzoli & C. Vernesi), pp. 4279. Cambridge University Press, Cambridge, UK.
  • Beerli, P. & Felsenstein, J. (1999) Maximum-likelihood estimation of migration rates and effective population numbers in two populations using a coalescent approach. Genetics, 152, 763773.
  • Beerli, P. & Felsenstein, J. (2001) Maximum likelihood estimation of a migration matrix and effective population sizes in n subpopulations by using a coalescent approach. Proceedings of the National Academy of Sciences of the United States of America, 98, 45634568.
  • Bender, D.J., Contreras, T.A. & Fahrig, L. (1998) Habitat loss and population decline: a meta-analysis of the patch size effect. Ecology, 79, 517533.
  • Bolker, B.M., Brooks, M.E., Clark, C.J., Geange, S.W., Poulsen, J.R., Stevens, M.H.H. & White, J.S.S. (2009) Generalized linear mixed models: a practical guide for ecology and evolution. Trends in Ecology & Evolution, 24, 127135.
  • Bowler, D.E. & Benton, T.G. (2005) Causes and consequences of animal dispersal strategies: relating individual behaviour to spatial dynamics. Biological Reviews, 80, 205225.
  • Brown, J.H. & Kodric-Brown, A. (1977) Turnover rates in insular biogeography: effect of immigration on extinction. Ecology, 58, 445449.
  • Burnham, K.P. & Anderson, D.R. (2002) Model Selection and Multimodel Inference: A Practical Information-Theoretic Approach. Springer-Verlag, New York, New York, USA.
  • Chamberlain, T.C. (1965) The method of multiple working hypotheses. Science, 148, 754759.
  • Day, J.R. & Possingham, H.P. (1995) A stochastic metapopulation model with variability in patch size and position. Theoretical Population Biology, 48, 333360.
  • Doebeli, M. (1995) Dispersal and dynamics. Theoretical Population Biology, 47, 82106.
  • Donovan, T.M., Lamberson, R.H., Kimber, A., Thompson III, F.R. & Faaborg, J. (1995) Modeling the effects of habitat fragmentation on source and sink demography of Neotropical migrant birds. Conservation Biology, 9, 13961407.
  • Elkin, C.M. & Possingham, H. (2008) The role of landscape-dependent disturbance and dispersal in metapopulation persistence. American Naturalist, 172, 563575.
  • Environmental Protection Agency (EPA) (2004) Survey and Mapping of 2001 Remnant Vegetation Communities and Regional Ecosystems of Queensland: Incorporating Version 4.0 and, in part, Version 4.1. Queensland Herbarium and Environmental Protection Agency, Toowong, Queensland, Australia.
  • Environmental Systems Research Institute (ESRI) (2005) ArcGIS, Version 9.1. Environmental Systems Research Institute, Inc., Redlands, California, USA.
  • Fahrig, L. (2003) Effects of habitat fragmentation on biodiversity. Annual Review of Ecology Evolution and Systematics, 34, 487515.
  • Fahrig, L. (2007) Non-optimal animal movement in human-altered landscapes. Functional Ecology, 21, 10031015.
  • Fischer, J. & Lindenmayer, D.B. (2007) Landscape modification and habitat fragmentation: a synthesis. Global Ecology and Biogeography, 16, 265280.
  • Fraser, D.J., Hansen, M.M., Østergaard, S., Tessier, N., Legault, M. & Bernatchez, L. (2007) Comparative estimation of effective population sizes and temporal gene flow in two contrasting population systems. Molecular Ecology, 16, 38663889.
  • Greenwood, P.J. & Harvey, P.H. (1982) The natal and breeding dispersal of birds. Annual Review of Ecology and Systematics, 13, 121.
  • Gustafson, E.J. & Gardner, R.H. (1996) The effect of landscape heterogeneity on the probability of patch colonization. Ecology, 77, 94107.
  • Hanski, I. (1998) Metapopulation dynamics. Nature, 396, 4149.
  • Haynes, K.J., Dillemuth, F.P., Anderson, B.J., Hakes, A.S., Jackson, H.B., Jackson, S.E. & Cronin, J.T. (2007) Landscape context outweighs local habitat quality in its effects on herbivore dispersal and distribution. Oecologia, 151, 431441.
  • Higgins, P.J. & Peter, J.M. (2002) Handbook of Australian, New Zealand and Antarctic Birds. Volume 6, Pardalotes to Spangled Drongo. Oxford University Press, Melbourne, Victoria, Australia.
  • Hines, J.E. (2006) PRESENCE2: Software to Estimate Patch Occupancy and Related Parameters. US Geological Survey, Patuxent Wildlife Research Center, Laurel, Maryland, USA.
  • Howe, R.W., Davis, G.J. & Mosca, V. (1991) The demographic significance of ‘sink’ populations. Biological Conservation, 57, 239255.
  • Ims, R.A. & Hjermann, D.Ø. (2001) Condition-dependent dispersal. Dispersal (eds J. Clobert, E. Danchin, A.A. Dhondt & J.D. Nichols), pp. 203216. Oxford University Press, Oxford, UK.
  • Kareiva, P. & Wennergen, U. (1995) Connecting landscape patterns to ecosystem and population process. Nature, 373, 299302.
  • Kawecki, T.J. & Holt, R.D. (2002) Evolutionary consequences of asymmetric dispersal rates. American Naturalist, 160, 333347.
  • Kingman, J.F.C. (1982) On the genealogy of large populations. Journal of Applied Probability, 19A, 2743.
  • Kleinhans, D. & Jonsson, P.R. (2011) On the impact of dispersal asymmetry on metapopulation persistence. Journal of Theoretical Biology, 290, 3745.
  • Lande, R. (1987) Extinction thresholds in demographic models of territorial populations. American Naturalist, 130, 624635.
  • Laurance, W.F., Lovejoy, T.E., Vasconcelos, H.L., Bruna, E.M., Didham, R.K., Stouffer, P.C., Gascon, C., Bierregaard, R.O., Laurance, S.G. & Sampaio, E. (2002) Ecosystem decay of Amazonian forest fragments: a 22-year investigation. Conservation Biology, 16, 605618.
  • MacKenzie, D.I. & Bailey, L.L. (2004) Assessing the fit of site-occupancy models. Journal of Agricultural Biological and Environmental Statistics, 9, 300318.
  • MacKenzie, D.I., Nichols, J.D., Lachman, G.B., Droege, S., Royle, J.A. & Langtimm, C.A. (2002) Estimating site occupancy rates when detection probabilities are less than one. Ecology, 83, 22482255.
  • MacKenzie, D.I., Nichols, J.D., Royle, J.A., Pollock, K.H., Bailey, L.L. & Hines, J.E. (2006) Occupancy Estimation and Modeling: Inferring Patterns and Dynamics of Species Occurrence. Elsevier, Burlington, Massachusetts, USA.
  • McIntire, E.J.B., Schultz, C.B. & Crone, E.E. (2007) Designing a network for butterfly habitat restoration: where individuals, populations and landscapes interact. Journal of Applied Ecology, 44, 725736.
  • Nelder, J.A. & Wedderburn, R.W.M. (1972) Generalized linear models. Journal of the Royal Statistical Society. Series A, 135, 370384.
  • Nicholls, J.A., Austin, J.J., Pavlacky, D.C. & Green, D.J. (2007) Characterization of polymorphic microsatellites in the logrunner Orthonyx temminckii (Aves: Orthonychidae). Molecular Ecology Notes, 7, 11171119.
  • Pavlacky Jr, D.C. (2008) Avian patch occupancy and landscape genetics of logrunners (Orthonyx temminckii) in fragmented subtropical rainforests of South East Queensland. PhD thesis, University of Queensland, Brisbane, Queensland, Australia.
  • Pavlacky Jr, D.C., Goldizen, A.W., Prentis, P.J., Nicholls, J.A. & Lowe, A.J. (2009) A landscape genetics approach for quantifying the relative influence of historic and contemporary habitat heterogeneity on the genetic connectivity of a rainforest bird. Molecular Ecology, 18, 29452960.
  • Pearse, D.E. & Crandall, K.A. (2004) Beyond FST: analysis of population genetic data for conservation. Conservation Genetics, 5, 585602.
  • Pulliam, H.R. (1988) Sources, sinks, and population regulation. American Naturalist, 132, 652661.
  • Pulliam, H.R. & Danielson, B.J. (1991) Sources, sinks, and habitat selection: a landscape perspective on population dynamics. American Naturalist, 137, S50S66.
  • Remeš, V. (2000) How can maladaptive habitat choice generate source-sink population dynamics? Oikos, 91, 579582.
  • Revilla, E. & Wiegand, T. (2008) Individual movement behavior, matrix heterogeneity, and the dynamics of spatially structured populations. Proceedings of the National Academy of Sciences of the United States of America, 105, 1912019125.
  • Rousset, F. (2001) Genetic approaches to the estimation of dispersal rates. Dispersal (eds J. Clobert, E. Danchin, A.A. Dhondt & J.D. Nichols), pp. 1828. Oxford University Press, Oxford, UK.
  • SAS Institute (2008) SAS/STAT. Version 9.2. SAS Institute Inc, Cary, North Carolina, USA.
  • Senar, J.C., Conroy, M.J. & Borras, A. (2002) Asymmetric exchange between populations differing in habitat quality: a metapopulation study on the citril finch. Journal of Applied Statistics, 29, 425441.
  • Smith, A.C., Fahrig, L. & Francis, C.M. (2011) Landscape size affects the relative importance of habitat amount, habitat fragmentation, and matrix quality on forest birds. Ecography, 34, 103113.
  • Smith, S.J., Leslie Jr, D.M., Hamilton, M.J., Lack, J.B. & Van Den Bussche, R.A. (2008) Subspecific affinities and conservation genetics of western big-eared bats (Corynorhinus townsendii pallescens) at the edge of their distributional range. Journal of Mammalogy, 89, 799814.
  • Sodhi, N.S., Liow, L.H. & Bazzaz, F.A. (2004) Avian extinctions from tropical and subtropical forests. Annual Review of Ecology Evolution and Systematics, 35, 323345.
  • Stacey, P.B. & Taper, M. (1992) Environmental variation and the persistence of small populations. Ecological Applications, 2, 1829.
  • Turner, I.M. & Corlett, R.T. (1996) The conservation value of small, isolated fragments of lowland tropical rain forest. Trends in Ecology & Evolution, 11, 330333.
  • Tyre, A.J., Tenhumberg, B., Field, S.A., Niejalke, D., Parris, K. & Possingham, H.P. (2003) Improving precision and reducing bias in biological surveys: estimating false-negative error rates. Ecological Applications, 13, 17901801.
  • Vos, C.C., Verboom, J., Opdam, P.F.M. & ter Braak, C.J.F. (2001) Toward ecologically scaled landscape indices. American Naturalist, 157, 2441.
  • Vuilleumier, S. & Possingham, H.P. (2006) Does colonization asymmetry matter in metapopulations? Proceedings of the Royal Society B-Biological Sciences, 273, 16371642.
  • Wakeley, J. (2001) The coalescent in an island model of population subdivision with variation among demes. Theoretical Population Biology, 59, 133144.
  • Webb, L.J., Tracey, J.G. & Williams, W.T. (1984) A floristic framework of Australian rainforests. Australian Journal of Ecology, 9, 169198.
  • Westphal, M.I., Field, S.A., Tyre, A.J., Paton, D. & Possingham, H.P. (2003) Effects of landscape pattern on bird species distribution in the Mt. Lofty Ranges, South Australia. Landscape Ecology, 18, 413426.
  • White, G.C. & Burnham, K.P. (1999) Program MARK: survival estimation from populations of marked animals. Bird Study, 46, 120139.
  • Wiegand, T., Revilla, E. & Moloney, K.A. (2005) Effects of habitat loss and fragmentation on population dynamics. Conservation Biology, 19, 108121.
  • With, K.A. & King, A.W. (2001) Analysis of landscape sources and sinks: the effect of spatial pattern on avian demography. Biological Conservation, 100, 7588.
  • With, K.A., Schrott, G.R. & King, A.W. (2006) The implications of metalandscape connectivity for population viability in migratory songbirds. Landscape Ecology, 21, 157167.