Altered natal dispersal at the range periphery: The role of behavior, resources, and maternal condition

Abstract Natal dispersal outcomes are an interplay between environmental conditions and individual phenotypes. Peripheral, isolated populations may experience altered environmental conditions and natal dispersal patterns that differ from populations in contiguous landscapes. We document nonphilopatric, sex‐biased natal dispersal in an endangered small mammal, the Mt. Graham red squirrel (Tamiasciurus hudsonicus grahamensis), restricted to a single mountain. Other North American red squirrel populations are shown to have sex‐unbiased, philopatric natal dispersal. We ask what environmental and intrinsic factors may be driving this atypical natal dispersal pattern. We test for the influence of proximate factors and ultimate drivers of natal dispersal: habitat fragmentation, local population density, individual behavior traits, inbreeding avoidance, competition for mates, and competition for resources, allowing us to better understand altered natal dispersal patterns at the periphery of a species’ range. A juvenile squirrel's body condition and its mother's mass in spring (a reflection of her intrinsic quality and territory quality) contribute to individual behavioral tendencies for movement and exploration. Resources, behavior, and body condition have the strongest influence on natal dispersal distance, but affect males and females differently. Male natal dispersal distance is positively influenced by its mother's spring body mass and individual tendency for movement; female natal dispersal distance is negatively influenced by its mother's spring body mass and positively influenced by individual tendency for movement. An apparent feedback between environmental variables and subsequent juvenile behavioral state contributes to an altered natal dispersal pattern in a peripheral population, highlighting the importance of studying ecological processes at the both range center and periphery of species’ distributions.

Natal dispersal patterns differ between birds and mammals, where dispersal in birds is often female-biased and male-biased in mammals (Greenwood, 1980). In birds and mammals, three underlying ecological processes are thought to ultimately drive natal dispersal and observed dispersal differences between sexes: inbreeding avoidance, competition for resources, and competition for mates (Gaines & McClenaghan, 1980). Testing for proximate and ultimate drivers of natal dispersal within a theoretical framework can elucidate important ecological influences, how these may vary among populations, and identify potential conservation implications, particularly in threatened populations.
In contrast to general mammalian dispersal patterns, natal dispersal in red squirrels is characterized as sex-unbiased and tends to be philopatric (Larsen, 1993;. Competition for resources best explained the observed patterns of sex-unbiased red squirrel dispersal, and the continuous occupied habitat in most areas where red squirrels occur likely explains philopatric settlement . Territory acquisition and associated resources, including conifer cone storage (Williams, Lane, Humphries, McAdam, & Boutin, 2014), are critical to the survival and reproduction of both male and female red squirrels (Kemp & Keith, 1970;Larsen & Boutin, 1994, 1998Rusch & Reeder, 1978), which may influence the tendency for both sexes to settle within or adjacent to their mother's territory (Berteaux & Boutin, 2000;Haughland & Larsen, 2004b;Kerr et al., 2007;Sun, 1997). While sex-unbiased, philopatric dispersal appears common throughout the red squirrel's range, no data exist for isolated, peripheral tree squirrel populations.
Peripheral, isolated populations could differ from range center populations due to environmental heterogeneity in availability of resources, landscape fragmentation, population dynamics, and local population density. Peripheral populations, in turn, may be influential in determining species distributions, and natal dispersal in these populations likely influences range expansion and contraction. Environmental heterogeneity may influence intrinsic characteristics of individuals within a population including body condition of mother and offspring (Bowler & Benton, 2005;Rémy, Le Galliard, Gundersen, Steen, & Andreassen, 2011) and individual personality (Cote et al., 2010), which can influence natal dispersal patterns.
Herein, we characterize natal dispersal in an isolated red squirrel subspecies occurring at the southern extent (trailing edge) of the species' range ( Figure 1) and compare dispersal in this isolated population to populations in the range center. We examine the influence of intrinsic and extrinsic factors on natal dispersal distance and the probability of nonphilopatric dispersal to include local population parameters (local male and female density), litter sex ratios, mother spring body mass, juvenile body condition, natal habitat patch size, and individual behavior traits. We develop a priori models to test support for three proximate factors and three ultimate ecological processes hypothesized to influence the probability of dispersing and dispersal distance: natal patch size, local density, individual behavior traits, inbreeding avoidance, competition for mates, and competition for resources (Greenwood, 1980;.

| Proximate hypotheses
Natal patch size: In highly fragmented landscapes, habitat patches might be smaller and farther apart. We examine the relationship between natal patch size and distance to the nearest patch and the probability of dispersing and dispersal distance. Local density: We test for positive and negative density dependence (e.g., Matthysen, 2005) based upon local density of occupied territories. Behavior: Personality traits may predispose some individuals to leave the natal area and disperse farther compared to others. We examine the relationship between individual behavior traits and the probability of dispersing and dispersal distance (Table 1).

| Ultimate hypotheses
Inbreeding avoidance: We examine the influence of local neighborhood and litter sex ratios on the probability of nonphilopatric dispersal and dispersal distance. Male competition for mates: We examine the influence of local male density and proportion of male littermates on the probability of nonphilopatric dispersal and dispersal distance. The promiscuous mating system in red squirrels and many other small mammals implies that there is likely little intrasexual competition for mates among females Lawson Handley & Perrin, 2007). Competition for food: We examine the influence of resource proxies (spring body mass of the squirrel's mother and juvenile body condition) on the probability of nonphilopatric dispersal and dispersal distance (Table 1).

| Study site and population
The Mt. Graham red squirrel (Tamiasciurus hudsonicus grahamensis, hereafter MGRS) is an endangered subspecies of red squirrel inhabiting the Pinaleño Mountains, in Arizona, USA, 32.7017°N, 109.8714°W, and is the southernmost population of red squirrels in North America (Sanderson & Koprowski, 2009; Figure 1). MGRS have been isolated for at least 10,000 years following post-Pleistocene glacial retreat (Harris 1990) and are morphologically, vocally, and genetically distinct from the nearest subspecies of red squirrel, T. h. mogollonensis, inhabiting the White Mountains of east central Arizona (Fitak, Koprowski, & Culver, 2013;Koprowski, Alanen, & Lynch, 2005 3,267 m in elevation. Interannual availability of food resources from conifer seeds and fungi can vary by an order of magnitude, and total resource abundance has decreased following recent disturbance events (King & Koprowski, 2009;Koprowski et al., 2005). The Pinaleños have experienced patchy forest damage at varying levels of severity, due to insect infestations

| Live trapping and quantifying individual behavior traits
Between May 2010 and February 2014, we trapped, radio-collared and tracked 94 juvenile (≤190 g) and four subadults (>190 g) MGRS through dispersal, settlement, and postsettlement. To capture juveniles, we monitored the location and reproductive condition of radiocollared adult females as part of a long-term study of MGRS space use (Koprowski, King, & Merrick, 2008 and fit radio collars (Koprowski et al., 2008). To reduce radio collar weight and allow for growth, we used a thin (3 mm) nylon zip-tie neck band with a 3 mm × 20 mm strip of thin, compressible foam mounting tape affixed to the inside of the neck band (total collar weight = 5 g; 3% of mean juvenile body mass, range: 2.5-5%). We recaptured individuals at least every 3 month to measure growth and check radio collar fit.
To characterize individual behavior traits that comprise personality, we performed two, 7.5-min behavior trials on 84 juveniles at the site

| Dispersal
We used digital receivers (Communication Specialists Inc. R-1000 receiver) and yagi 3-element directional antennae (Wildlife Materials Inc., Murphysboro IL, USA) to track juvenile MGRS movements from capture to settlement, locating each juvenile a minimum of 12 times monthly until settlement, death, or disappearance from our study area.
We monitored individuals for signs of settlement, which included conifer cone caching at a central midden (larderhoard) and territorial vocalizations (Larsen & Boutin, 1994). After settlement, we continued to monitor individual space use and survivorship. We measured straight-line dispersal distance from the natal nest to the territory center (midden) at which it settled. In addition to dispersal distances quantified in this study (2010-2013), we also had 11 records of dispersal distances for animals ear-tagged as juveniles in prior years (n = 8), and an early attempt to track natal dispersal in this population (n = 3; Kreighbaum & Van Pelt, 1996). We compiled published natal dispersal distances for red squirrels to identify range-wide mean dispersal distance for males and females and compared range-wide mean dispersal distances to dispersal distances in MGRS. Mean adult female 95% fixed kernel home range size in mixed conifer forest during fall (when juveniles settle) over 12 years was 1.7 ha: a territory diameter of 147.12 m, 73.56 m radius. We considered juveniles moving distances ≤150 m as settling within a territory contiguous with that of its mother (Larsen & Boutin, 1994), and juveniles moving distances >150 m as dispersers.

| Animal density
We determined occupancy of central larder hoards (middens) during quarterly censuses where we recorded signs of recent activity, including fresh conifer cone scales, digging, and cached cones and mushrooms (Koprowski & Snow, 2009) along with the age and sex of the resident. Between 2002 and 2015, mean adult female 95% fixed kernel home range size was 3.2 ha (range 1.1-7 ha), and we used this mean area to represent the local density that juveniles were exposed to prior to dispersal. We compiled occupancy of middens each quarter (December, March, June, and September) and determined local neighborhood density and sex ratios within a 100-m-radius (3.14 ha) buffers around natal nests by summarizing census occupancy records within each buffer. We used June census data to represent the density of occupied middens (where 1 occupied midden = 1 squirrel) and sex of residents within each 100-m-radius buffer, as summer is coincident with juvenile growth, development, and dispersal.

| Food availability
We quantified conifer cone availability in the natal area each fall via methods similar to Humphries and Boutin (2000) and Studd, Boutin, McAdam, Krebs, and Humphries (2014

| Natal patch size
To delineate patches of red squirrel habitat in the Pinaleños based on MGRS use, we developed a habitat suitability model based upon 9,424 MGRS juvenile lifetime telemetry locations relative to seven 25-m-resolution LiDAR-derived raster layers that included percent canopy cover, mean tree height, standard deviation in tree height, total basal area, live basal area, slope, and elevation (Appendix S1).
We followed Greco's (2007, 2009) patch morph algorithm in ArcGIS to create habitat patches with quality and marginal edge habitat delineated (Appendix S1). We used patch area in hectares, patch code (quality patch interior or edge) associated with each individual's natal and settlement location, and distance to the nearest patch as explanatory variables in subsequent natal dispersal models. MISPC4 Pearson's r = .14, p = .34); thus, we included these as intrinsic variables in subsequent models.

| Statistical analyses
We estimated the repeatability of individual behavior scores by selecting a subset of individuals (n = 13) to receive repeat OF and MIS trials. We compiled behavior data from trials 1 and 2 for each individual, and collapsed variables via principal component analysis as above. We calculated the intraclass correlation coefficient (ICC) and confidence intervals (α = 0.05) for principal components from OF and MIS trials, with animal ID as the subject, and n = 2 "raters" (trial 1 and trial 2), and specified a test for consistency between trials (model one, one way), where subject effects are random (Gamer, Lemon, Fellows, & Singh, 2015).

| Models for dispersal distance and probability of long-distance dispersal
We examined the influence of environmental and intrinsic factors on natal dispersal distance and probability of nonphilopatric dispersal within three model sets: sexes combined, female-only, and male-only models (Tables 3 and S2). Our candidate model sets contained five basic models: (1) Table S3 for descriptions). Female-and male-specific models included the same five basic models in addition to models developed to specifically test dispersal hypotheses (density, natal patch size, behavior, inbreeding avoidance, competition for mates, competition for resources; Tables 3 and S2) in either sex.
We identified 16 candidate generalized linear models a priori to test for the influence of intrinsic and extrinsic factors on log-transformed straight-line dispersal distance (Gaussian error structure), and on the probability of dispersing long distances (logit link, binomial error structure; Table S2), and compared models, fit with maximum-likelihood, within an information-theoretic model selection framework. For probability models, we specified nonphilopatric dispersal for males ≥150 m (>total diameter of mean adult female home range) and ≥100 m for females (as few females dispersed >150 m).

| Live trapping and individual behavior differences
Of the 98 radio-collared juvenile and subadult MGRS in our study (51 females and 47 males), 12 died prior to settlement (nine females and three males) and 24 had unknown fates (14 females and 10 males); of these, seven collars were found (four females and three males) and 17 went missing and were never relocated (nine females and seven males). Sixty-three individuals survived and were successfully tracked to settlement locations (29 females and 34 males). Combined with known dispersal distances from previous years (six males and five females), we were able to quantify dispersal distance for 74 MGRS.
We 3.2 | Dispersal, density, food, and natal patch size

| Dispersal
Natal dispersal in MGRS is male-biased with exaggerated dispersal distances compared to other red squirrel populations ( of males ≥150 m annually. The proportion of juveniles that are nonphilopatric and distances moved varied from year to year (proportion males 2 3df df = 6.45, p = .09; proportion females 2 3df = 3.99, p = .26; all individuals 2 3df = 7.56, p = .06), and this interannual variation in dispersal may be influenced by conifer seed crop availability (Table S4, Figure 3). For both sexes, the proportion of individuals dispersing and female dispersal distance was highest in 2011 (proportion males: 1.0, proportion females: 0.67, female mean dispersal distance: 915.1 m), a year of lowest food availability (Table S4, Figure 3).
During our study, local midden density within a 3.14-ha buffer surrounding natal nests ranged from 0.3 to 3.8 middens/ha (mean 2.1 ± 0.9), and local occupancy ranged from 0.0 to 1.9 occupied middens/ha (mean 0.8 ± 0.4) and this value did not vary significantly by year (one-way ANOVA F 1,58 = 0.28, p = .60). The mean proportion of occupied middens within 3.14-ha buffers was 0.37 ± 0.19.

| Natal patch size
Animals in our study were born in habitat patches ranging in size from 1.88 to 126.60 ha (mean 26.72 ± 26.00). The majority of individuals that dispersed and settled (66%) were born in small patches <30 ha in size, and 88% were born in patches designated as "quality patch interior" (90% of cells within a 50 m search radius of any focal cell are classified as suitable). Natal patches within our study areas were all <50 m of the next nearest patch; thus, we omitted distance to nearest patch from our models. distance and probability of nonphilopatric dispersal in both sexes included mother spring mass, juvenile body condition, and activity (female AIC weight = 0.84, male AIC weight = 0.83; Tables 3 and 4).

| Drivers and tests of mammalian dispersal hypotheses
In dispersal distance models, the relationship between dispersal distance, mother spring mass, and individual body condition is reversed for males and females, with female dispersal distance negatively influenced by increases in mother spring mass and individual body condition, whereas male dispersal distance is positively influenced by increases in both activity and mother spring mass (Figure 4, Table 4).
For females, the top probability of dispersal model (resources.locomotion; AIC weight 0.68; Table 3) was similar to the dispersal distance (resources.locomotion) model in both significance and sign of coefficients, but this was not the case for males (Table 4). For males, the coefficient for body condition is reversed in the top probability of dispersal model (resources.locomotion; AIC weight 0.74), but both body condition and mother spring mass had very little explanatory power, as this model is driven primarily by activity, whereby for every unit increase in movement PC score, males are over seven times more likely to disperse long distances (Table 4, Figure 4).
We found no support for local density and natal patch size as proximate explanatory factors or inbreeding avoidance as an ultimate driver of dispersal distance or probability of dispersal (Table S2). Evidence for an intrasexual effect of female density on female dispersal distance exists, whereby juvenile females dispersed farther with increasing local female density ( Figure 5), yet despite this relationship, female density was not a top model (Tables 3 and S2). Male dispersal distance was not influenced by local female density, further evidence against current inbreeding avoidance ( Figure 5). We found no relationship between mother spring mass and litter sex ratio (proportion male offspring) (t = 0.06, df = 47, p = .95) or an effect of year (F 3,54 = 0.99, p = .40). is the most supported ultimate hypothesis. We found little support for proximate influences of natal patch size, local density, litter sex ratios, or ultimate drivers competition for mates and inbreeding avoidance.

| The role of behavior and resources on dispersal distance
Individual behavior differences, or personalities, and their associated behavioral syndromes (correlated behavior traits) have been F I G U R E 3 Proportion of male and female juvenile Mt. Graham red squirrels (T. h. grahamensis) making long-distance dispersal movements (males dispersing ≤150 m; females dispersing ≤100 m) relative to an annual index of the current year's conifer cone availability (2010)(2011)(2012)(2013). Proportion of female long-distance dispersers for each year is indicated with gray bars, males with black bars. Mean ± standard deviation in dispersal distance for males and females for each year is shown above the bars T A B L E 3 Model descriptions and multimodel selection results for models developed a priori to explain dispersal distance and probability of long-distance dispersal in juvenile Mt. Graham red squirrels (Tamiasciurus hudsonicus grahamensis) between 2010 and 2013. Models with AICc weights >0 are shown. Models developed to test for dispersal hypotheses are indicated: DEN = local density, FRAG = habitat fragmentation, BEHAV = individual behavior differences, CFR = competition for resources, IA = inbreeding avoidance. See Table S3  documented in many taxa (Sih, Cote, Evans, Fogarty, & Pruitt, 2012), including red squirrels (Boon, Réale, & Boutin, 2008;Boon et al., 2007;Kelley, Humphries, McAdam, & Boutin, 2015), and are thought to be maintained within populations by differential fitness relative to highly variable resource availability and population densities in time and space (Cote et al., 2010;Duckworth, 2008;Wolf & Weissing, 2012). Positive correlations between behavior traits and dispersal distance have been documented in birds, mammals, lizards, and fishes (Clobert et al., 2009;Cote et al., 2010;Dingemanse et al., 2003;Duckworth, 2008) and may be important in population dynamics and maintenance of gene flow, especially for species threatened with habitat shifts or other disturbances (Massot, Clobert, & Ferrière, 2008;Sih et al., 2012). Natal dispersal distance in MGRS is correlated with an individual's tendency to actively explore a novel environment. Vagile behavior trait expression appears to be mediated by external cues from mothers and the surrounding environment related to resource availability, an example of condition dependence and phenotype dependence (Clobert et al., 2009;Cote et al., 2010).
The competition for resources hypothesis implies that sex-biased dispersal should occur only if resources are more important to one sex (the philopatric sex) than the other (Greenwood, 1980;. For most mammals, including Sciurids (but not observed in red squirrels), this dichotomy between the resource needs of females (competition for resources) and males' need for access to mates (competition for mates) has explained the primarily male-biased dispersal patterns observed in mammals (Clutton-Brock & Harvey, 1978;Greenwood, 1980). In red squirrels, acquiring a quality territory that supports the accumulation of food resources is critical for overwinter survival in both males and females (Kemp & Keith, 1970;Larsen & Boutin, 1994, 1998Rusch & Reeder, 1978), and as breeding does not occur until after a juvenile's first winter (Koprowski, 2005), it follows that natal dispersal and settlement decisions are driven by resource availability (competition for resources) rather than mates . Our models, along with evidence of increased dispersal in years of low conifer cone abundance, support the finding that resource availability is an important driver of natal dispersal in MGRS.
However, resources available to mothers prior to parturition, partially

| Maternal influences on natal dispersal in a highly variable world
Maternal influence on offspring phenotype is widespread in mammals (Maestripieri & Mateo, 2009), and maternal effects are shown to influence offspring behavior and propensity for dispersal in response to resource variability (Duckworth, 2009). In North American red squirrels, a female's body mass following winter is a reflection of her territory quality and the number of conifer cones she was able to collect from her territory and hoard the previous fall (Becker, Boutin, & Larsen, 1998). External influences such as resource availability and competition for resources affect maternal condition and subsequent behavioral and physiological phenotypes and sex ratios in a female's offspring (Love & Williams, 2008;Maestripieri & Mateo, 2009). In this study, we observed no offspring sex ratio differences and no relationship between mother's mass or year and offspring sex ratio, and provide evidence for maternal effects that may maximize the fitness of both sons and daughters within a highly variable environment. The positive associations between dispersal distance and mother spring body mass and dispersal distance and individual activity score in males provide some evidence that mothers in good condition tend to have active, exploratory sons that were long-distance dispersers. The negative relationship between dispersal distance and mother spring body mass in females provides some evidence that the same mothers in good condition tend to have philopatric daughters. In poor years, the majority of all offspring dispersed. In years of high-resource abundance, allowing daughters to settle adjacent to or within a quality territory increases overall fitness of both mother and daughter (Berteaux & Boutin, 2000). Such resource-and density-mediated adjustments F I G U R E 4 Linear relationships between variables included in our top model (resources.locomotion: mother spring mass, body condition index, and individual activity score, Tables 3 and 4)  in offspring sex ratios via natal dispersal could represent a flexible (rather than fixed) dispersal strategy that is adaptive in a highly variable environment.

| Insight from the range periphery
Ecological conditions such as extreme heterogeneity in resource availability characteristic of trailing edge peripheral populations can contribute to heterogeneous dispersal patterns among populations.
Intraspecific variation in dispersal distance has been documented in arvicoline rodents and suggests that while most studies report short dispersal distances, long-distance dispersal events, while infrequent, can occur (Le Galliard et al., 2012).
Peripheral populations represent microcosms of evolution, with distinct physical, physiological, and behavioral adaptations resulting from long-term isolation and environmental conditions different from the range center (Channell & Lomolino, 2000;Foster, 1999;Hampe & Petit, 2005). Compared to range center, peripheral populations exhibit decreased densities (Lomolino & Channell, 1995), expanded home range size (Koprowski et al., 2008), lower within-population genetic diversity (Fitak et al., 2013;Vucetich & Waite, 2003), variation in demographic parameters, and changes in the frequency of behaviors or shifts in behavior reaction norms (Foster, 1999 [2010][2011][2012][2013]. These relationships between dispersal distance and local conspecific and heterospecific density are of interest despite having less support within a multimodel selection framework as they demonstrate lack of support for the inbreeding avoidance hypothesis and suggest possible competition for resources among females