Risky movements? Natal dispersal does not decrease survival of a large herbivore

Abstract Natal dispersal is assumed to be a particularly risky movement behavior as individuals transfer, often long distances, from birth site to site of potential first reproduction. Though, because this behavior persists in populations, it is assumed that dispersal increases the fitness of individuals despite the potential for increased risk of mortality. The extent of dispersal risk, however, has rarely been tested, especially for large mammals. Therefore, we aimed to test the relationship between dispersal and survival for both males and females in a large herbivore. Using a radio‐transmittered sample of 398 juvenile male and 276 juvenile female white‐tailed deer (Odocoileus virginianus), we compared survival rates of dispersers and nondispersers. We predicted that dispersing deer would experience greater overall mortality than philopatric deer due to direct transfer‐related risks (e.g., vehicular collision), indirect immigration‐related mortality attributable to colonization of unfamiliar habitat, and increased overwinter mortality associated with energetic costs of movement and unfamiliarity with recently colonized habitat. For both male and female yearlings, survival rates of dispersers (male = 49.9%, female = 64.0%) did not differ from nondispersers (male = 51.6%, female = 70.7%). Only two individuals (both female) were killed by vehicular collision during transfer, and overwinter survival patterns were similar between the two groups. Although dispersal movement likely incurs energetic costs on dispersers, these costs do not necessarily translate to decreased survival. In many species, including white‐tailed deer, dispersal is likely condition‐dependent, such that larger and healthier individuals are more likely to disperse; therefore, costs associated with dispersal are more likely to be borne successfully by those individuals that do disperse. Whether low‐risk dispersal of large mammals is the rule or the exception will require additional research. Further, future research is needed to evaluate nonsurvival fitness‐related costs and benefits of dispersal (e.g., increased reproductive opportunities for dispersers).


| INTRODUC TI ON
Animal movements represent a cost-benefit trade-off, such that fitness may be increased (e.g., competition avoidance, resource acquisition) or decreased (e.g., energy expenditure, exposure to predators) by movement behavior (Fahrig, 2007;Lima & Dill, 1990;Ronce, 2007). Natal dispersal, or movement of individuals away from birth site to location of first potential reproduction (Howard, 1960), is a critical life-history strategy affecting many ecological and evolutionary processes (Bullock et al., 2002;Clobert et al., 2012).
Fitness consequences of dispersal remain relatively poorly understood in many systems. Despite substantial differences in dispersal patterns (e.g., active vs. passive dispersal), conceptual similarities exist such that three phases of natal dispersal are often recognized: emigration (i.e., departure from natal range), transfer (i.e., movement from natal range to a distinct, nonoverlapping adult range), and immigration (i.e., establishment of adult range; Andreassen et al., 2002;Jongejans et al., 2015). Although successful long-distance dispersers often demonstrate increased fitness relative to philopatric conspecifics (Bowler & Benton, 2005;Larsen & Boutin, 1994;Stephens et al., 2005), dispersers may incur substantial costs in each dispersal phase (Bonte et al., 2012), for instance, fitness benefits of dispersal trade-off with energetic costs of predeparture exploratory movements (Debeffe et al., 2013), ambulatory costs during transfer (Benoit et al., 2020), opportunity costs associated with lost familiarity of natal habitat (Part, 1991), and direct mortality risks of predation (Bonnet et al., 1999) or vehicular collision (Real & Mañosa, 2001). Further, estimating the balance of cause-specific costs and benefits can be complicated by the observation that dispersal within a population may have multiple ultimate causes Stenseth & Lidicker, 1992), and highly modified habitats may yield suboptimal movement patterns (Fahrig, 2007;Shaw et al., 2014).
For actively dispersing vertebrates, costs tend to be most pronounced in the transfer and immigration phases of dispersal (Bonte et al., 2012). Specifically, long-distance dispersal increases direct mortality risks during transfer and increases energy expenditure, which may decrease postdispersal survival. For instance, Massemin et al. (1998) demonstrated increased mortality of barn owls (Tyto alba) in France due to vehicular collision during dispersal, and Bonnet et al. (1999) similarly attributed young-of-the-year snake mortality in France to dispersal movements. Further, using animal-borne biologgers in France, Benoit et al. (2020)  White-tailed deer (Odocoileus virginianus) are large mammals whose dispersal patterns have been relatively well-studied and may serve as a model for large-mammal dispersal. For both sexes, dispersal primarily occurs in yearlings (i.e., 1-year-olds), and like most mammals, dispersal is male-biased, with yearling male emigration rates of approximately 50%-80% . Dispersal in females is less common, with yearling female dispersal rates of only 3%-50% . Greenwood (1980) attributed sex-biased dispersal of vertebrates to sex-specific ultimate causes, as we have suggested for white-tailed deer Lutz et al., 2015).
Thus, fitness consequences of potentially risky dispersal likely affect sexes differently.
Although causes of dispersal are becoming clearer, costs associated with dispersal of large mammals remain poorly understood (Bonte et al., 2012). Generally, costs are assumed to be high due to energetic expenditure and direct threat of transfer-related mortality (e.g., vehicular collision, exposure to predators); however, few studies of large mammals have documented dispersal-related risks. In whitetailed deer, dispersal has been predicted to increase mortality (McCoy et al., 2005;Nelson & Mech, 1986;Roseberry & Klimstra, 1974), but this prediction has rarely been tested. To our knowledge, only two prior studies of large mammals (both white-tailed deer) have compared mortality of dispersers and nondispersers: Nixon et al. (2001), working in highly fragmented agricultural region of Illinois (USA), suggested that dispersal increased mortality risk, but Haus et al. (2019), working in Delaware (USA), suggested that dispersal status was not a significant predictor of mortality for yearling male white-tailed deer.
Here, we present dispersal-related mortality for both sexes of white-tailed deer through all phases of dispersal. To test for mortality risks associated with dispersal, we compared time-specific survival of dispersing and philopatric yearling male and female white-tailed deer from multiple populations. We predicted that dispersing deer would experience greater overall mortality than philopatric deer due to (a) direct transfer-related risks such as vehicular collision, (b) indirect immigration-related mortality attributable to recent colonization of unfamiliar habitat, and (c) increased winter mortality associated with energetic costs of movement and unfamiliarity with recently colonized habitat. Female white-tailed deer transfer movements are longer and more tortuous than male movements (Lutz et al., 2016;Nixon et al., 2001); therefore, we predicted direct transfer-related risks would be greater in female dispersers. Further, yearling female dispersal is generally completed by early summer , whereas many yearling males disperse in fall just prior to or concurrent with hunting seasons Peterson et al., 2017); therefore, we predicted that male dispersers would bear greater indirect risks associated with recent colonization. and 4D). All study areas experience a temperate climate; elevations do not exceed 820 m; and deer within these areas do not migrate to winter ranges Lutz et al., 2015). Hunting of both male and female deer is common in all areas, with limited hunting beginning the first week of October and most hunting beginning the third week of November and ending by the third week of January (Wallingford et al., 2017). To our knowledge, natal and adult ranges of all deer included in this study comprised areas open to hunting.

| Study areas
Further, across the study areas, fawns are commonly depredated by coyotes (Canis latrans), black bears (Ursus americanus), and bobcats (Lynx rufus; Vreeland et al., 2004); however, predation of yearling and older deer is uncommon (Wallingford et al., 2017). Winter mortality is relatively uncommon, typically occurring Feb-Mar, and annual survival of yearling and adult deer outside hunting is >90% (Wallingford et al., 2017). Capture occurred during winter through early spring, ending by 23

| Deer capture and monitoring
Apr for males and 26 Mar for females. At time of capture, juvenile deer were approximately 7-10 months old. For both male and female white-tailed deer, natal dispersal prior to 11 months of age is rare; thus, capture between December and April decreased the likelihood of capturing juveniles that had already dispersed (Marchinton & Hirth, 1984;Vreeland et al., 2004). Deer were captured January-April using Clover traps, drop-nets, rocket-nets, and net-gun from helicopter. Additional capture, monitoring, and study area details are provided in Long (2005) and Lutz (2015).

| Analytical methods
To test for survival differences between dispersers and nondispersers, we estimated survival for males and females using known-fate models in Program MARK v. 6.2 (White & Burnham, 1999). We modeled males and females in separate analyses because data collection for the sexes did not overlap in time or, with the exception of WMU 2D, in space. Due to estimability and sample size issues, we pooled study areas and years within sex. Thus, for each sex, we developed two candidate models of annual survival: a null model that combined dispersers and nondispersers and a group model that estimated weekly survival separately for dispersers and nondispersers.
Individuals were entered into the analysis after the completion of all capture efforts, and weekly survival was estimated through 22 Apr of the following year. Because capture of females ended 1 month earlier than capture of males, the null postcapture model generated 52 and 56 weekly survival estimates for males and females, respectively (Table 1). Additionally, to investigate potential relationship of dispersal distance on mortality, dispersal distance was included as an individual covariate in all models. Dispersal distance was calculated as straight-line distance between median x and y predispersal and postdispersal coordinates (Haus et Peterson et al., 2017). We used Akaike's information criterion corrected for sample size (AIC c ) to select the most parsimonious model of survival (Burnham & Anderson, 1998).
Because dispersal movements were not observed before mid-April, deer were censored from analysis if they died before this period or if we were unable to determine their fate. We investigated all mortalities to determine cause of death, which included vehicular collision, starvation, disease, predation, legally killed, and illegally killed (Wallingford et al., 2017 In addition to testing an annual (i.e., postcapture) model of survival, we also performed two subsequent analyses over shorter timescales. In the first, to investigate potential mortality differences independent of actual transfer movements, we began survival analyses after most dispersal-related movements had ceased. For yearlings, 95% of natal dispersal was completed by 19 Nov for males and 2 Jul for females Lutz et al., 2015). Therefore, to control for any transfer-related mortality and to compare postset-    (Table 1). Thus, regardless of whether analyses included dispersal movements, began after establishment of adult range, or isolated the period of peak mortality (i.e., fall hunting), no differences in survival were detected between dispersing and nondispersing yearling deer.

| D ISCUSS I ON
Despite a large sample of male and female white-tailed deer, we failed to find evidence of direct, survival-related risks associated with dispersal for either sex, as dispersers and nondispersers survived at similar rates. This finding does not necessarily indicate costfree or even low-cost dispersal, as fitness-related costs may extend beyond measures of survival. For instance, even after successful immigration, dispersers could potentially experience reduced reproductive opportunities (Cant et al., 2001;Part, 1991). However, contrary to predictions, dispersal in white-tailed deer does appear to be a relatively low-risk behavior.
As in many other vertebrates, dispersal in white-tailed deer is strongly sex-biased, and we observed greater dispersal of males than females. But if dispersal is not particularly risky for females, why did so few disperse? Although the survival consequences of dispersal are similar, the ultimate causes of dispersal in white-tailed deer differ greatly between the sexes and seem to relate to reproductive, rather than survival, components of fitness. Male white-tailed deer disperse to reduce the probability of inbreeding when abundance of adult females is great and to reduce intrasexual competition when abundance of adult males is great Shaw et al., 2006). Thus, especially if male dispersal is a low-risk behavior, benefits of dispersal likely frequently outweigh costs, which would ultimately select for dispersal. Consistent with this, male dispersal movements in our study were quick (often <12 hr), direct, and relatively short (median dispersal distance in our system = 5.9 km; Long, 2005;Long et al., 2010), although dispersal movements in more agricultural landscapes are often longer in duration and distance (Anderson et al., 2015;Springer, 2017). These findings suggest that males have adapted efficient dispersal strategies, and like our study, Haus et al. (2019) found no effect of dispersal on yearling male white-tailed deer survival in Delaware, USA.
Despite being relatively low risk, female dispersal movements of white-tailed deer remained relatively rare. As Greenwood (1980) suggested, inbreeding is minimized if one sex disperses and the other remains philopatric; thus, inbreeding avoidance likely does not ultimately cause female deer dispersal and may select for philopatry through increased reproductive success. Dispersal that has been observed in females has been attributed to density-dependent limitation of parturition sites ; thus, female dispersal is likely a conditional strategy that is more rarely adopted than male dispersal. Female white-tailed deer dispersal movements seem to be longer and more tortuous than male movements (Lutz et al., 2016;Nelson & Mech, 1992;Nixon et al., 2001). For example, median yearling female straight-line dispersal distance in our system (15.3 km) was more than 2.5 times greater than median male dispersal distances (Lutz, 2015), and actual distance traveled across the landscape during transfer was certainly longer, especially for females.
Although dispersal distance was not a strongly predictive covariate for survival in either sex, the ratio of female to male beta estimates for dispersal as a covariate was 3.84:1, suggesting that increasing dispersal distance may increase risk for females more than males.
Determining whether the circuitous, longer-distance dispersal pattern typical of female white-tailed deer represents an inefficient search strategy resulting from low selection pressure for dispersal, or a fundamentally different transfer strategy than males' direct paths, will require additional research. Further, with GPS transmitters capable of more frequent fixes, future studies could incorporate path analysis of dispersers to investigate whether path length and tortuosity of transfer movements influence survival.
In addition to dispersal-related transfer movements, male and female white-tailed deer have also been shown to demonstrate F I G U R E 2 Weekly known-fate survival estimates (±95% CIs) for 276 yearling (1-2 years old) female (a) and 398 yearling male (b) whitetailed deer in Pennsylvania, USA. Models begin postcapture (26 Mar for females, 23 Apr for males) and continue through 22 Apr of the following year. For both females and males, no difference was detected in survival between dispersing (black lines) and nondispersing (gray lines) deer temporary, round-trip excursive forays, which could also potentially increase mortality risk (Jacobsen, 2017;Lutz et al., 2016).
These forays are often of short duration but may last 2 weeks or more (Springer, 2017). As such, some mortalities that occur during forays outside natal home ranges may be incorrectly attributed to dispersal-related mortality, and if animals die during movement without establishing a distinct adult range, distinguishing between forays and dispersal-related transfers becomes impossible. However, as we have shown, transfer-related mortality prior to establishing an adult range was low in our system. Further, of deer that did establish adult ranges distinct from natal ranges, >92% of males and >96% of females were recorded in adult ranges for >14 days prior to mortality, suggesting little potential for the inadvertent inclusion of excursive deer in our models of dispersal-related survival.
Further, although we detected no effect of dispersal on survival for females, due to the rarity of female dispersal we were able to observe only 27 dispersal events. The only transfer-related vehicularcollision mortalities were observed in females, and annual point estimates for survival were 6.7% lower for dispersers than nondispersers. Because females, on average, disperse longer distances than males in the same landscape, female dispersal may incur relatively greater costs and expose females to greater risk, but direct risks associated with dispersal seemed to be rare in both sexes. Further, although road-crossings do frequently occur during dispersal, both males and females seem actively to reduce risk by terminating movements on the near side of major roadways and other potentially dangerous barriers (Long et al., 2010;Lutz et al., 2016).
In addition to direct risks associated with transfer movements, we continued survival analyses through the following April to examine potential delayed costs and indirect risk of dispersal. In addition to hunting, winter mortality of white-tailed deer can be high (DelGiudice et al., 2002), and energetically costly movements prior to winter could increase winter mortality. For instance, male deer are known to increase their movement rates in late fall and early winter as they search for mates (Long et al., 2013;Whitman et al., 2018), and these activities can increase mortality after the breeding season (Ditchkoff et al., 2001). Benoit et al. (2020) showed that transferrelated movements of roe deer in France resulted in 22% more energy expenditure for dispersers relative to nondispersers, but any increased energetic costs of dispersal did not apparently translate to decreased survival in our study. Similarly, Whitman et al. (2018) did not see increased white-tailed deer mortality with increased breeding-related movements in New York, USA.
However, white-tailed deer are habitat generalists, and because dispersal patterns differ across landscapes Lutz et al., 2015;Nixon et al., 2007), dispersal risk may vary as well. In a highly modified and fragmented agricultural landscape with sparse, patchily distributed forest covering 1.6%-20% of the landscape, Nixon et al. (2001) found reduced survival in dispersing white-tailed deer in Illinois, USA. Nixon et al. (1991Nixon et al. ( , 2001 observed greater female dispersal rates, lower male dispersal rates, and greater dispersal distances than other studies of white-tailed deer from less fragmented forested habitats. Decreased survival of dispersers was attributed to unfamiliarity with newly settled habitat, described as heavily hunted "dispersal sinks," wherein survival beyond winter was unlikely (Nixon et al., 1991(Nixon et al., , 2001. If adult habitat comprised less cover than the forested patches where capture was concentrated, movement to subprime habitat likely increased mortality because forest patches serve as important refuges from hunting in agricultural landscapes (Foster et al., 1997). Further, much of the natal area was located inside reserves with no hunting, and increased mortality was associated with any movements outside refuges (Nixon et al., 1991(Nixon et al., , 2001. Similarly, Rosenberry et al. (1999) found that emigration from an area with restricted hunting increased yearling male white-tailed deer mortality in Maryland, USA. Therefore, dispersal from refuges likely increases risk.
In other mammals, such as red fox (Vulpes vulpes) in England,  , 1988;Johnson et al., 2009;Jones, 1988 Dispersal patterns of European roe deer have been well-studied and offer an interesting contrast. Unlike white-tailed deer, dispersal rates and distances do not vary by sex (Debeffe et al., 2012). Like white-tailed deer, dispersal distance distributions are right-skewed but vary by habitat (Debeffe et al., 2012;Diefenbach et al., 2008), and dispersal appears to be condition-dependent, such that larger and less parasitized roe deer are more likely to disperse (Debeffe et al., 2012(Debeffe et al., , 2014Wahlström, 1994). Therefore, in both species those individuals most likely to bear successfully the energetic costs of dispersal are those that disperse. But, the equal sex ratios of dispersal and similar dispersal distance distributions suggest that ultimate causes and consequences of roe deer dispersal may differ from white-tailed deer.
Animal movement incurs cost, and these costs are typically assumed to increase mortality risk, especially for natal dispersal (Bonte et al., 2012;Ronce, 2007). Although there are certainly costs associated with all phases of dispersal, this study suggests that within some populations and systems, animals can disperse with relatively low direct risk to survival. However, few studies have related natal dispersal to direct mortality risks, especially for large mammals, and it remains to be seen whether low-risk dispersal is widespread or rare.

CO N FLI C T O F I NTE R E S T
None declared. Writing-review & editing (equal).

DATA AVA I L A B I L I T Y S TAT E M E N T
The original data input files are available from the Dryad Digital