Introduction

Individuals of wild animal populations often disperse from natal territories to breeding territories (natal dispersal) and can subsequently disperse again between breeding attempts (breeding dispersal) (Greenwood 1980; Clarke et al. 1997). By facilitating spatial and temporal gene flow (Clobert 2001; Ronce 2007), this dispersal behavior of individuals has long-term consequences on the structure and genetic diversity of populations. Thus, there has been considerable effort by conservation biologists and behavioral and population ecologists to understand the drivers, patterns, and consequences of dispersal (Koenig et al. 1996).

Dispersal, especially long-distance movement through unfamiliar habitat, increases the risk of mortality through predation and can reduce post-dispersal fitness due to high energy requirements of travel (Greenwood 1980; Greenwood and Harvey 1982; Pusey and Wolf 1996). Nevertheless, the potential benefits of dispersal must outweigh its considerable risks; otherwise, this behavior would disappear (Perrin and Mazalov 1999). Competition for resources and mates, spatial and temporal heterogeneity of the environment, and the combination of these factors drive dispersal (Greenwood 1980; Greenwood and Harvey 1982; Dobson and Jones 1985; Pusey and Wolf 1996). As the sexes are affected differently by these factors, a sex bias in dispersal frequency and/or distance can occur (Greenwood 1980; Waser and Jones 1983; Pusey 1987). In mate defense systems, typical in mammals, males are more likely to disperse as they invest little in their offspring (Greenwood 1980; Mabry et al. 2013). In these systems, males are generally interested in maintaining access to females until fertilization, after which time, ongoing investment ceases. The female remains defending the territory and raising the offspring, while the male departs (Greenwood 1980; Mabry et al. 2013). In contrast, in resource defense systems, typical in birds, males invest heavily in their offspring. By remaining close to their natal territory, they can take advantage of their familiarity with local resources in order to attract a mate and breed successfully (Greenwood 1980; Clarke et al. 1997). In addition, sex-biased dispersal may reduce the risk of inbreeding by facilitating spatial separation of close relatives (Greenwood 1980; Pusey 1987; Pusey and Wolf 1996). This is especially important in small and isolated populations that are at risk of inbreeding depression through increased levels of homozygosity caused by low gene flow (Charlesworth et al. 1993). However, inbreeding avoidance alone cannot predict the direction of the dispersal bias (Perrin and Mazalov 1999; Ronce 2007). Dispersal patterns ultimately depend on the relative cost of dispersal for each sex (Moore and Ali 1984).

The main challenge to studying dispersal in open populations is the inability to distinguish emigration from mortality, which leads to underestimation of dispersal distance and the proportion of individuals dispersing, as well as obfuscation of sex bias (Koenig et al. 1996). Additionally, in open populations of species where sex cannot be determined in nestlings or juveniles and is only determined in adults by non-molecular methods, measurement of sex bias in natal dispersal is limited to distance dispersed. The proportion dispersing of each sex cannot be measured as sex is unknown for birds that leave the study site or die. Some researchers assume a 50:50 sex ratio in offspring, but this may introduce bias because sex allocation can be manipulated by parents as a result of population structure or environmental variables (Komdeur 1996). Without knowledge of sex, and therefore proportion dispersing, it is impossible to understand how sex-biased dispersal may be influencing evolutionary processes (e.g., gene flow) and whether sex-biased dispersal limits inbreeding. Closed populations of rare species where no immigration occurs are most vulnerable to inbreeding, and thus, studying sex-biased dispersal in these species as a mechanism to avoid inbreeding is particularly important. Moreover, studies on island species in a closed environment where no immigration or emigration occurs offer a unique opportunity to distinguish between emigration and mortality. Closed populations that have been studied for multiple years, where all individuals are uniquely marked and where breeding system and life history traits are known, allow for clear, a priori predictions regarding sex bias, providing a comprehensive way to explore the evolution of dispersal patterns (Goudet et al. 2002).

In this study, we investigated dispersal patterns in the Chatham Island black robin (Petroica traversi), a small (22–25 g), insectivorous passerine endemic to the Chatham Island archipelago 800 km east of mainland New Zealand (Higgins and Peters 2002). The black robin passed through a severe population bottleneck in 1980 when the entire population was reduced to five individuals, including only a single breeding pair (Butler and Merton 1992; Massaro et al. 2013a), but as of January 2016, the population has since recovered to 289 adult birds. The entire species is restricted to Mangere and Rangatira/Hokorereora Islands, two small islands free of exotic mammalian predators. As there is no dispersal between the two islands, it is an ideal, simplified system for testing possible causes of dispersal. Specifically, we addressed the following questions: (1) do black robins use a resource defense mating system, (2) is there a sex bias in natal dispersal, (3) what are the potential factors driving natal dispersal, (4) what are the effects of natal dispersal on breeding success, (5) is there a sex bias in breeding dispersal, (6) what are the potential factors driving breeding dispersal, and (7) what are the effects of breeding dispersal on reproductive success?

Methods

Study site

Data were collected from the Rangatira/Hokorereora Island (218 ha, 44° 20′ 50″ S 176° 10′ 30″ W) population during the 2007 to 2011 breeding seasons (October to January). Since 2007, black robins have been banded with a unique color band combination. Each year, individually marked adults were re-sighted, and since the island is relatively small, all alive birds have generally been seen within the 3-month field season. To ensure that unique band combinations of individual birds were re-sighted correctly, all birds were triple checked by at least two different observers. It was not possible to record data blind because our study involved focal animals in the field. At the end of the 2011 season, the black robin population numbered 239 adults, 200 on Rangatira and 39 on Mangere (Massaro et al. 2013b). Each year, during the breeding season, locations of all nests were recorded by GPS (Garmin GPSMAP 60CSx, <10 m) and nestlings were banded shortly before fledging. Adults were sexed by behavior (i.e., males sing, display frequent territorial behaviors, and courtship feed their partners; females beg for food from their partners, incubate eggs, and brood chicks). To ensure that birds were sexed correctly, behaviors that indicated sex were recorded three times by different observers.

Spatial analyses

Spatial analysis was completed in ArcGIS 10.2 (ESRI 2013). Territory maps for each breeding season were produced by generating Thiessen polygons in ArcMap. Thiessen polygons, also known as Dirichlet (Voroni) tessellations, use Delaunay triangulation to assign to a given point all area that is closer to that point than to any other point (Fortin and Dale 2005; ESRI 2013). This method is appropriate for identifying territories where nests are irregularly spaced (Adams 2001; Valcu and Kempenaers 2008). A vegetation map of the island was created through manual interpretation of satellite imagery where vegetation was classed as forest, scrubby vegetation, or bare ground (A. Chick, unpublished data). The resulting territory polygons were clipped to the vegetated areas (forest and scrubby vegetation) of the island. A territory size metric analogous to diameter (Valcu and Kempenaers 2008) was estimated for each territory polygon. We measured the Euclidean distance in meters from the nest location to the territory polygon edge 120 times (once every 3° rotation) and averaged the results. From these results, we calculated the mean territory diameter for each breeding season. As both adult and juvenile black robins were associated with a nest location (natal or breeding) for every breeding season, we calculated density based on nest coordinates for each bird. With these coordinates, 1-m resolution density maps of birds per hectare were produced using the ArcGIS Kernel Density tool with the search radius parameter set to the average territory diameter of each breeding season.

We measured the Euclidean distance (in meters) from natal nest to that of first breeding (natal dispersal) or between nests from breeding season to breeding season (breeding dispersal) for both males and females to test whether one sex moves farther than the other. To determine if this movement qualifies as dispersal, we analyzed territory boundaries. In sedentary and spatially constrained species in a heterogeneous environment where densities fluctuate between breeding seasons, distance moved may not be useful for determining dispersal status (Greenwood and Harvey 1982). Moving further distances may not reduce kin competition or potential inbreeding if an individual’s territory still shares borders with that of a close relative. Therefore, for natal dispersal, we classified movement as dispersal if there was at least one territory in between the natal and breeding territories as measured in a straight line from natal to breeding nest. A bird was classed as philopatric when breeding and natal territories intersected or were adjacent. For classifying breeding dispersal, we examined whether the territories intersected from season to season (Valcu and Kempenaers 2008). If they intersected, the birds were classed as non-dispersers.

Statistical analyses

All statistical analyses were performed in R version 3.2.3 (R Core Team 2016) using the lme4 R package version 1.1–12 (Bates et al. 2016) for linear and generalized linear (mixed) effect models. For all analyses, model fit was determined by obtaining p values through likelihood ratio tests of the full model, which included the fixed effect being tested, against the model without the effect in question (Tables S1, S2, and S3). In all instances, visual inspection of residual plots did not reveal any obvious deviations from normality or homoscedasticity.

To investigate whether territory or mate retention was higher for black robins from breeding season to breeding season, we evaluated pair data where both members of the pair survived consecutive breeding seasons (i.e., pairs where one member died from 1 year to the next were eliminated from this dataset). Birds were classed as deceased if they were not resighted for two or more consecutive years. Only birds that occupied a territory and were paired for consecutive seasons were considered. We tested each sex independently. To test for the likelihood of mate or territory change, we fitted generalized linear mixed effect models (GLMMs) in the binomial family with change in status (yes/no) as the response variable and change type (mate/territory) as a fixed effect. As each bird had two lines of data per year, one for each change type (mate/territory), the unique bird band number was fitted as a random effect (see Table S1 for model selection).

We tested for sex bias in natal dispersal in both the proportion of birds dispersing (no. of individuals that dispersed) and the distance moved (in m). To test for sex bias in the proportion of birds dispersing, we fitted generalized linear models (GLMs) in the binomial family with dispersal status (philopatric/dispersed) as the response variable. We included sex and bird density (birds/ha) in the natal year as fixed effects (see Table S2 for model selection). To test for sex bias in natal dispersal distance, we fitted linear models (LM) with distance moved (in meters) as the response variable. Both dispersing and philopatric birds were included. Sex was fitted as a fixed effect (see Table S3 for model selection).

As a measure of the effectiveness of natal dispersal, we investigated post-dispersal breeding success. Breeding success was calculated as the ratio of chicks fledged to the number of eggs laid to account for differences in clutch sizes. As causes of dispersal differ between the sexes, dispersal costs and benefits may also differ; therefore, we tested each sex independently. We fitted GLMs with breeding success as the response variable. The final model included dispersal class and delayed first breeding (females only, see Table S2 for model selection).

Changing mates, whether through divorce or death, may impact dispersal. Therefore, we investigated whether the reason for mate change (divorced/widowed) affected either the likelihood of dispersal or the distance moved in order to determine if they could be combined in the breeding dispersal analyses. We fitted mixed effect models to test for a difference in the likelihood of dispersal (GLMM) and distance moved (LMM) due to either divorce or death of a mate. Territory change and meters moved were the response variables, respectively. In order to account for annual records for each bird, the unique bird band number was included as a random effect. The fixed effects included were sex, the reason for mate change, and their interaction. Sex was only included to test for an interaction effect. As there was no difference in either the likelihood of dispersal or the distance moved by divorced or widowed birds, we combined the two reasons for mate change in subsequent analyses (Table S1). We fitted mixed effect models to test for sex bias in breeding dispersal in both proportion of birds dispersing (GLMM) and the distance moved (LMM). For both tests, bird band number was included as a random effect. Fixed effects retained in the final models were sex and mate (see Table S1 for model selection).

As a measure of the effectiveness of breeding dispersal, we investigated post-dispersal breeding success. Each sex was tested independently. We fitted GLMMs with breeding success (chicks fledged/eggs laid) as the response variable. Bird band number was included as a random effect. The original models included the fixed effects of territory change, distance moved, mate change, and the two-way interactions of the fixed effects.

Results

Most black robin males (88 %) did not change their mate, nor their territory between breeding seasons (Table 1). If they did change either mate (10 %) or territory (0.93 %), it was significantly more likely that they would change mates than territories (χ 2 (1) = 4.75, p < 0.05). Neither mate nor territory change was significantly more likely for females (Table 1).

Table 1 Territory and mate change events [n (%)] between consecutive breeding seasons by sex from 2007 to 2011 (excludes birds that were not alive or paired in consecutive years)
Full size table

Natal dispersal in black robins was significantly sex-biased in both proportion of birds dispersing and the distance moved (Fig. 1, Table 2). Females were more likely to disperse than males (χ 2 (1) = 22.46, p < 0.001), and they moved an average of 212 m ± 47 SE further than males (R 2 = 0.24, F 1,64 = 19.94, p < 0.001) (Fig. 1). Black robin density in the natal year correlated with the likelihood of dispersal for both sexes (χ 2 (1) = 4.75, p < 0.05) (Fig. 2) but not the distance moved (R 2 = 0.25, F 2,63 = 10.42, p = 0.34).

Fig. 1
figure 1

Natal dispersal for 2007–2011 for a males (n = 38) and b females (n = 28). Philopatric birds remained in their natal or adjacent territories for breeding (orange lines); for dispersing birds, there was at least one territory separating the natal and breeding territories (blue lines). The proportion of females dispersing from their natal site was significantly higher than that of males (82.1 % of females dispersed compared to 26.3 % of males). Females also moved further from their natal site (mean = 331 m, range = 58–1177 m, CV = 72) than males (mean = 119 m, range = 4–803 m, CV = 122)

Full size image
Table 2 The number of natal dispersal events [n (%)] for males and females from 2007 to 2011 and the distance moved in meters [mean (range, CV)] for all individuals from natal nest to first breeding nest
Full size table
Fig. 2
figure 2

Predicted increase in proportion of black robins dispersing from their natal territories due to an increase in black robin population density (birds/ha) in the natal year for both males (solid line) and females (dashed line), (χ 2 (1) = 4.75, p < 0.05). Gray ribbons represent 95 % confidence intervals

Full size image

For females, breeding success in the first breeding season was lower for dispersers than those that remained philopatric (χ 2 (1) = 7.04, p < 0.01). Females that delayed their first breeding attempt had relatively higher success regardless of whether they dispersed or remained philopatric (χ 2 (1) = 6.73, p < 0.01) (Fig. 3). There was no significant difference in breeding success between dispersing and philopatric males (χ 2 (1) = 0.056, p = 0.81).

Fig. 3
figure 3

Predicted breeding success (ratio of chicks fledged to eggs laid) of female black robins for their first breeding season after natal dispersal. Circles show mean values and whiskers indicate 95 % confidence intervals. Dispersing females had reduced breeding success (p < 0.01). Females that delayed (solid circles) their first attempt increased their breeding success compared to those that did not delay (open circles) regardless of dispersal status (p < 0.01)

Full size image

Breeding dispersal of black robins was significantly sex-biased in the proportion of birds dispersing but not in the distance moved (Table 3). Females were more likely to disperse between breeding seasons than males (χ 2 (1) = 4.80, p < 0.05). Black robins that changed mates from breeding season to breeding season were significantly more likely to disperse (χ 2 (1) = 29.1, p < 0.001) (Fig. 4). While breeding dispersal distances were not sex-biased (χ 2 (1) = 1.68, p = 0.20), they, too, were significantly affected by change in mate. There was a significant interaction between mate status and sex. Divorced or widowed females moved an average of ~99 m (±12 SE) further than those that retained their mate (χ 2 (1) = 60.33, p < 0.001), while males only moved ~20 m (χ 2 (1) = 16.25, p < 0.001). Poor breeding success was neither a factor in the likelihood of dispersal (χ 2 (2) = 3.49, p = 0.18), nor the distance moved (χ 2 (2) = 1.16, p = 0.56). Breeding success was not significantly different between males that changed territories between seasons and those that retained territories (χ 2 (1) = 0.99, p = 0.32). We also did not find any differences in breeding success between females that dispersed between seasons and those that retained territories (χ 2 (1) = 3.43, p = 0.06).

Table 3 The number of breeding dispersal events [n (%)] for males and females from 2007 to 11 and the distance moved in meters [mean (range, CV)] for all individuals between breeding seasons
Full size table
Fig. 4
figure 4

Proportion of male and female black robins dispersing between breeding seasons. Boxes show the median, upper, and lower quartiles of data. Whiskers indicate values outside the middle 50 %, excluding outliers. Both male and female black robins that changed mates (filled boxes) were more likely to disperse between breeding seasons than those that retained their mates (open boxes) (χ 2 (1) = 29.1, p < 0.001). Females that changed their mate were more likely to disperse than males (χ 2 (1) = 4.80, p < 0.05)

Full size image

Discussion

In this study, we found that black robins practice a resource defense mating system, whereby both adult males and females are sedentary and socially monogamous. Black robins are unlikely to disperse between breeding seasons unless they lose a mate through either death or divorce. Natal dispersal was strongly sex-biased, whereby females were more likely to disperse and to disperse further than males from their natal territory to their first breeding site. This suggests that there may be a clear advantage for young males to establish territories close to their natal territory, while there are less costs associated with dispersal for females. However, when we investigated first-time breeding success of males and females, we found no difference in success between dispersing and philopatric males, but unexpectedly, dispersing females had lower breeding success than those that were philopatric. If natal dispersal is costly for females in respect to their first breeding attempt, there may be other benefits associated with dispersal.

The female bias in natal dispersal may reduce inbreeding among close relatives, which can lead to a decrease in genetic diversity (Lewin 1989). The risk of inbreeding is particularly high in small and isolated populations of island endemic species and species in fragmented landscapes because there are few or no opportunities for gene flow to introduce novel alleles from other populations (Charlesworth et al. 1993). In large, outbred populations with high gene flow among unrelated individuals, offspring are expected to have dissimilar (heterozygous) alleles at any given location (locus) on a gene (Keller and Waller 2002). However, in inbred populations, offspring are more likely to have identical (homozygous) alleles at a given locus, which reduces genetic diversity and, thus, fitness (Keller and Waller 2002). This reduced fitness due to inbreeding (inbreeding depression) can include reduced survival, reduced fecundity, or increased susceptibility to disease (Ardern and Lambert 1997). Additionally, rare alleles within small populations may be lost through genetic drift, further decreasing genetic diversity and fitness (Frankham et al. 1999). Evidence of inbreeding depression in wild populations is now well established (Keller and Waller 2002; Clobert et al. 2004; O’Grady et al. 2006). Recent studies analyzing black robin pedigree data from the intensive management and monitoring period (1982–1998) have found a high level of inbreeding in the species. The median level of inbreeding, F = 0.34, was found to be much higher than the inbreeding level of offspring produced by full siblings in outbred populations, F = 0.25 (Weiser et al. 2016). The effects of this inbreeding in combination with genetic drift have been shown to lead to the expression of a deleterious, maladaptive trait in the black robin (Massaro et al. 2013a). Furthermore, there is no evidence of genetic purging, a process whereby deleterious alleles are removed via natural selection, in this species (Kennedy et al. 2014). However, as the black robin population has substantially increased since the severe bottleneck in 1980, it is now possible for black robins to avoid mating with a parent or sibling.

There are two mechanisms that may facilitate inbreeding avoidance within populations: (1) sex-biased dispersal, which results in close kin of the opposite sex not being in the same area for breeding (Pusey 1987), and (2) selective mating behaviors that allow individuals to recognize close kin and avoid them as mates (Komdeur and Hatchwell 1999; Hauber and Sherman 2001). Through associative learning, closely related individuals of some species become familiar with the cues identifying their kin within their natal group or territory and apply those cues to other locations in order to discriminate between close relatives and unrelated individuals (Komdeur and Hatchwell 1999). However, for individuals not raised in family groups, such as brood parasites (cuckoos, cowbirds), or cooperative breeders that have unrelated helpers, another method is needed to recognize close kin (Hauber and Sherman 2001). In such systems, discriminating close relatives from other individuals may consist of determining which animals look most like oneself, self-referent phenotype matching (Hauber and Sherman 2001). However, a study on two island endemics, the New Zealand robin (Petroica australis), which is closely related to the black robin, and the New Zealand saddleback (Philesturnus carunculatus), found no link between kin recognition and inbreeding avoidance (Jamieson et al. 2009). This result is consistent with several other studies of semi-isolated bird species and populations where kin recognition did not facilitate inbreeding avoidance, despite the risk and adverse consequences of inbreeding being high (e.g., Keller and Arcese 1998; Wheelwright and Mauck 1998; Hansson et al. 2007; Eikenaar et al. 2008). Hence, it is possible that sex-biased dispersal is a default mechanism that avoids pairings between close relatives in those species, as well as the black robin. A study of natal dispersal and inbreeding avoidance in the black grouse (Tetrao tetrix) found that female-biased natal dispersal alone was enough to avoid inbreeding (Lebigre et al. 2010). Similarly, in a study on red-cockaded woodpeckers (Picoides borealis), female-biased natal dispersal, and not kin recognition, was the mechanism that reduced inbreeding (Daniels and Walters 2000). In the black robin, post-natal dispersal breeding success is reduced in dispersing females compared to those that remain philopatric, suggesting that the cost of inbreeding is higher than the cost of dispersal. Therefore, female-biased natal dispersal may be a mechanism to reduce inbreeding in the black robin; however, further genetic work is needed to confirm this.

We found that higher densities of black robins during the natal year increased the frequency of dispersal for both male and female birds. Density has been found to affect dispersal patterns (Matthysen 2005). In an island population of song sparrows (Melospiza melodia) in British Colombia, intrasexual competition for food increased dispersal, but when supplemental food was provided, dispersal frequency decreased (Arcese 1989). In black robins, both sex-biased and density-dependent dispersal may explain much of the habitat occupancy structure (Aars and Ims 2000; Clobert et al. 2009). Currently, black robin density is highest in lowland bush and few black robins are found in higher elevation areas. It is unclear if this is due to weather or exposure, food, or nest site availability. However, since 2010, more black robins have nested in higher elevation areas. This may be a factor of density, as the population has steadily increased. Males are significantly less likely to disperse than females; thus, they are more likely to remain in lowland areas. As in other species (sensu Arlt and Pärt 2008), even though females do disperse and may disperse a great distance, their breeding territory is ultimately determined by locating a male with a territory. However, as density increases, both females and males are more likely to disperse. This cycle of sex-biased and density-dependent dispersal is incremental, as the measureable outcome of dispersal—breeding success—is reduced in dispersing female black robins and is generally low in the overall population (clutch size ± std. err. 2.02 ± 0.03, number of chicks fledged ± std. err. 1.03 ± 0.06 (Massaro et al. 2013a)). However, this cycle may accelerate if a positive density-dependent fitness increase (i.e., Allee effect) also occurs (Allee and Bowen 1932; Stephens and Sutherland 1999).

While there are few island endemic bird species that are restricted to a single island, studying species in island systems, including habitat fragments, allows for explicit testing of clear, a priori, predictions regarding aspects of population biology and behavioral ecology that would not be possible in open populations of more complex systems (Goudet et al. 2002).