Received: 28 August 2023 | Revised: 23 January 2024 | Accepted: 28 January 2024 DOI: 10.1002/jwmg.22567 R E S E A R C H A R T I C L E Sex ‐ specific resource use by wild turkeys in response to hunting activity Alaina P. Roth 1 | Patrick H. Wightman 2 | Nicholas M. Masto 3 | Jay R. Cantrell 4 | Charles Ruth 4 | Bradley S. Cohen 3 | Michael J. Chamberlain 2 | Bret A. Collier 5 1 School of Renewable Natural Resources, Louisiana State University, Baton Rouge, LA 70803, USA 2 Warnell School of Forestry and Natural Resources, University of Georgia, Athens, GA 30602, USA 3 College of Arts and Science, Tennessee Technological University, Cookeville, TN 38505, USA 4 South Carolina Department of Natural Resources, Columbia, SC 29202, USA 5 School of Renewable Natural Resources, Louisiana State University, Agricultural Center, Baton Rouge, LA 70803, USA Correspondence Bret A. Collier, School of Renewable Natural Resources, Louisiana State University Agricultural Center, Baton Rouge, LA 70803, USA. Email: bret@lsu.edu Present address Alaina P. Roth, Wisconsin Department of Natural Resources, 101 S. Webster Street, Madison, WI 53703, USA. Funding information South Carolina Department of Natural Resources Abstract Hunting causes direct mortality and potentially disrupts normal activities of game and non ‐ game species. As spatial (i.e., selection of hunting areas) and temporal (i.e., only diurnally present) patterns of hunters can become predictable, hunted species may respond accordingly. Risk becomes more dynamic and complex for species that are hunted concurrent with their breeding cycle, and growing literature has noted that wild turkey ( Meleagris gallopavo spp.) behaviors can be altered by hunting activity. We allocated global positioning system (GPS) units to 1,500 wild turkey hunters and affixed GPS transmitters to 175 wild turkeys during 2014 – 2018 on the Webb Wildlife Management Area Complex in South Carolina, USA. We evaluated whether wild turkeys shifted resource selection as a function of hunter resource selection during the progression of hunting seasons. Male wild turkeys avoided areas where stationary hunting bouts occurred during the early hunting season (and selected for these areas before hunting began), whereas females were more likely to select those areas avoided by males by the end of the hunting season. For every 15% increase in predicted probability of an area being hunted, male wild turkeys were 4.16 times less likely to select that area, whereas female wild turkeys were 1.08 times more likely to select that area relative to pre ‐ season periods when hunters were not on the landscape. Hunting activity induced immediate responses by male wild turkeys as they sought refuge away from hunted areas. Coupled with recent research suggesting Journal of Wildlife Management 2024;88:e22567. wileyonlinelibrary.com/journal/jwmg © 2024 The Wildlife Society | 1 of 17 https://doi.org/10.1002/jwmg.22567 hunters are more influential than natural predators in reducing the frequency of male vocalizations and eliciting fleeing and avoidance behavior, our results indicate hunting activity could affect distribution and courtship behaviors of male wild turkeys during their breeding season. K E Y W O R D S anti ‐ predator behaviors, hunting, landscape of fear, movement ecology, predation risk, resource selection function, wild turkey Independent of direct mortality, predators can influence prey by inducing antipredator behavioral responses (Lima and Dill 1990, Lima 1998, Wirsing and Ripple 2011). Behavioral responses entail varying levels of vigilance (Lima and Dill 1990, Brown 1999), moving to safer locations (Caldwell 1986, Heithaus and Dill 2002), or altering diel activity (Creel and Christianson 2008, Crosmary et al. 2012, Smith et al. 2019), but these responses accrue costs to the individual by reducing time for behaviors such as foraging, vocalizations, or mate acquisition (Creel and Winnie 2005, Smith et al. 2017, Wakefield et al. 2020, Van Beeck Calkoen et al. 2022). Consequently, non ‐ consumptive effects of predators on prey demography can be equal to or greater than consumptive effects (Schmitz et al. 1997, Nelson et al. 2004, Pangle et al. 2007). In addition to natural predators, human presence can induce mortality (e.g., road accidents, hunting; Madsen 1995, Tolon et al. 2009, Isrealsen et al. 2020) and non ‐ consumptive effects (Gaynor et al. 2018, Fattebert et al. 2019). Furthermore, human activity is likely analogous to predation risk (Frid and Dill 2002, Beale and Monaghan 2004 a , b ) and can cause animals to adjust behaviors to avoid human contact (Doherty et al. 2021). Hunting is an extreme type of human activity that affects wildlife because it causes direct mortality (Fleskes et al. 2007, Zbinden et al. 2018) and disrupts normal activities such as activity rhythms (Crosmary et al. 2012), flight initiation distances (Madsen and Fox 1995), vigilance (Jayakody et al. 2008, Ciuti et al. 2012, Paton et al. 2017), or habitat use patterns (Davis and Afton 2010, Ciuti et al. 2012, Yetter et al. 2018). Thus, the spatiotemporal distribution of hunting can influence animal behavior and distributions at both local and regional scales. Hunted animals encounter a perplexing situation as hunters change the predation risk landscape daily and seasonally. Predation risk becomes more dynamic and complex for species that are hunted concurrent with their breeding cycle. Thus, animals hunted during their reproductive season must adjust movements to seek needed resources (e.g., mates) while also avoiding encounters with hunters (Reiss 1987). The predation risk allocation hypothesis suggests that predictability and intensity of predation risk across a landscape is the primary factor influencing prey behavioral responses. Thus, in situations where risk is spatiotemporally unpredictable but intense, anti ‐ predator behaviors will be magnified and should result in pronounced responses (Lima and Bednekoff 1999). Hunter behaviors are generally predictable spatially and temporally, as they spend substantive time near roads and trails (Lyon and Burcham 1998, Diefenbach et al. 2005, Karns et al. 2012) and are constrained to hunting during diurnal periods, but hunted species face variation in hunter risk due to restricted seasons, specific days, or times of day when hunting is permitted (Toïgo et al. 2008, Tolon et al. 2009). Therefore, as spatial and temporal patterns of hunters become predictable, prey may respond accordingly. Eastern wild turkeys ( Meleagris gallopavo silvestris ) have declined in abundance and productivity across much of their range (Byrne et al. 2015, Eriksen et al. 2015, Isabelle et al. 2018, Chamberlain et al. 2022). Unlike hunting seasons for most other North American galliforms, spring wild turkey season occurs concurrent with reproductive efforts (Chamberlain et al. 2018, Wightman et al. 2019, Wakefield et al. 2020). Spring harvest is the primary form of mortality for male wild turkeys (Vangilder 1992, Chamberlain et al. 2012, Wightman et al. 2023 c ). Recent work also indicates hunting negatively influences gobbling activity (Chamberlain et al. 2018; Wightman et al. 2019, 2023 b ) 2 of 17 | ROTH ET AL 19372817, 2024, 4, Downloaded from https://wildlife.onlinelibrary.wiley.com/doi/10.1002/jwmg.22567 by University Of Florida, Wiley Online Library on [03/12/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License and could affect reproduction (Chamberlain et al. 2018, Wakefield et al. 2020). Likewise, spring wild turkey hunting seasons often occur concurrent with female egg laying and nesting periods, so hunting ‐ induced shifts in female habitat selection could indirectly affect fitness (Isabelle et al. 2018). Unlike animals that can avoid hunters by shifting to more nocturnal movements (Gaynor et al. 2018), wild turkeys are diurnal and must access resources and engage in reproductive activities while hunters are present on the landscape. Several studies have evaluated wild turkey responses to hunting activity (Gross et al. 2015, Collier et al. 2017, Wakefield et al. 2020), but detailing how hunting pressure is distributed across the landscape and how hunter resource selection affects wild turkey behavior would provide a useful case study of anti ‐ predator responses (Gerrits et al. 2020, Wightman et al. 2023 a ). Our objective was to evaluate if eastern wild turkey resource selection changed with the onset of hunting season. We evaluated resource selection by hunters and modeled how male and female wild turkeys' selection for land cover types and areas with hunting activity changed before and during hunting seasons on a public hunting area in South Carolina, USA, during 2014 – 2018. We hypothesized wild turkeys would exhibit changes in resource selection with the onset of hunting. Specifically, we predicted turkeys of both sexes would avoid areas selected by hunters immediately at the onset of the hunting season. Given that female wild turkeys could not be legally harvested and hunters would not actively pursue them, we predicted that female response to hunters would be less pronounced than males. Lastly, we predicted that land cover types selected by hunters would be avoided by both sexes of wild turkeys during hunting season. STUDY AREA We conducted research on the Webb Wildlife Management Area (WMA) Complex from 2014 – 2018 (Figure 1). The Webb WMA Complex was a conglomerate of 3 contiguous WMAs (Webb, Palachacola, and Hamilton Ridge) in Hampton and Jasper counties owned and managed by the South Carolina Department of Natural Resources (SCDNR). The Webb WMA Complex was 10,483 ha and consisted of bottomland and upland hardwood stands along drainages (~4,673 ha) and approximately 65 km of roads available for vehicle access. Planted and managed upland pines, primarily loblolly ( Pinus taeda ) and longleaf pine ( P. palustris ), comprised approximately 3,346 ha. The remaining 2,464 ha were composed of mixed ‐ pine hardwoods, wildlife openings, and bottomland hardwoods. Management activities included prescribed fire, timber management, and fallow field management focused on wildlife species such as white ‐ tailed deer ( Odocoileus virginianus ), wild turkey, red ‐ cockaded woodpeckers ( Picoides borealis ), and northern bobwhite ( Colinus virginianus ). During our study, the regular wild turkey season opened on 1 April and closed 1 May in 2014 and 2015. In 2016 – 2018 the regular wild turkey season opened 1 April and closed 5 May. Youth hunts occurred annually and were the Saturday before the regular season opened (Table 1). Hunting was permitted all day with legal hunting hours 30 minutes prior to sunrise until 30 minutes post sunset. Hunting was not permitted on Sundays (Wightman et al. 2019, Gerrits et al. 2020). METHO DS Turkey capture We captured male and female wild turkeys during January – March 2014 – 2018 using rocket nets at sites baited with cracked corn. We classified each individual by sex and aged them based on barring of the ninth and tenth primary feathers (Pelham and Dickson 1992). We banded each individual with a uniquely numbered aluminum rivet band (National Band and Tag Company, Newport, KY, USA) and radio ‐ tagged them with a backpack ‐ style global positioning system (GPS) ‐ very high frequency (VHF) combination backpack (Biotrack, Wareham, Dorset, United HUNTING AFFECTS WILD TURKEY RESOURCE SELECTION | 3 of 17 19372817, 2024, 4, Downloaded from https://wildlife.onlinelibrary.wiley.com/doi/10.1002/jwmg.22567 by University Of Florida, Wiley Online Library on [03/12/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Kingdom; Guthrie et al. 2011). We released all turkeys at the capture site immediately after processing. We programmed GPS transmitters placed on females to collect hourly locations between 0500 and 2000 daily and one location nightly (23:58:58; Cohen et al. 2018). We programmed GPS transmitters placed on males similarly, except that during 1 March ‒ 30 June locations were recorded every 30 minutes between 0500 and 2000 (Collier et al. 2017, Wightman et al. 2019, Gerrits et al. 2020). Throughout our study, we located turkeys ≥ 3 times weekly via VHF telemetry to monitor general location, survival, and nesting activity (Wood et al. 2019). F I G U R E 1 Webb Wildlife Management Area (WMA) Complex in South Carolina, USA, 2014 – 2018, including all boundaries, primary roads (yellow), and secondary roads (red) mapped during a study of wild turkey resource selection in response to hunter activity. T A B L E 1 Spring wild turkey hunting season dates (2014 – 2018), mean numbers of hunters allocated global positioning system (GPS) units daily, and hunters allocated GPS units annually on the Webb Wildlife Management Area (WMA) Complex, South Carolina, USA. Year Youth hunt dates Season dates Mean GPS hunters/day GPS hunters each year 2014 28 Mar 1 Apr – 1 May 7 186 2015 28 Mar 1 Apr – 1 May 10 304 2016 26 Mar 1 Apr – 5 May 10 367 2017 25 Mar 1 Apr – 5 May 7 274 2018 31 Mar 2 Apr – 5 May 10 369 4 of 17 | ROTH ET AL 19372817, 2024, 4, Downloaded from https://wildlife.onlinelibrary.wiley.com/doi/10.1002/jwmg.22567 by University Of Florida, Wiley Online Library on [03/12/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Hunter locations Hunters were required to complete a self ‐ clearing hunter data card before entering the WMA and return the card upon exiting. As detailed in Gerrits et al. (2020), we stationed personnel at road access points near self ‐ clearing stations approximately 2 hours before until 30 minutes after sunrise to meet with hunters and allocate GPS units (i ‐ gotU GT ‐ 120 GPS; Mobile Action, New Taipei City, Taiwan) to individuals who volunteered to participate. We scheduled GPS units to record a location every 30 seconds, and if hunters planned to group hunt (e.g., parent and child), we only allocated one unit per group. We asked hunters to carry the GPS units in a front pocket or in the top of a backpack to ensure data acquisition, and they could deposit GPS units at self ‐ clearing permit stations at the end of their hunt. We did not collect any identifying information from hunters. We collected GPS units twice daily and, after downloading hunter locations, erased data and recharged the battery on each unit for redeployment the following day (see Gerrits et al. 2020 for additional details). Hunting season timing varied slightly by year (Table 1), but the only day hunting occurred during our study that had no hunter GPS units deployed was on 28 March 2014 during the youth hunt before the regular season opening. Before analyzing hunter locations, we considered any GPS locations falling within 150 m of infrastructure such as hunter check stations and campgrounds or in a parking area as non ‐ hunting points and removed these locations from analysis. Next, as wild turkey hunters have specific search behaviors when hunting (Hazel et al. 1990) that typically include periods of stationary hunting activity interspersed with active periods where a hunter is moving, we separated stationary from active hunting points (Gerrits et al. 2020) and used dynamic Brownian Bridge movement models (dBBMM) in program R version 4.0.3 (R Core Team 2020) package move (Kraunstauber et al. 2020) to create 45% utilization distributions (UDs) for each hunting bout (Gerrits et al. 2020). We set the locational error at 20 m with a window size of 7 locations and margin size of 3 locations (Byrne et al. 2014). We classified each location within the UD as stationary and locations outside the UD as either active locations as hunters moved through the landscape or abandoned points (e.g., points with no preceding or subsequent locations), which were primarily errant locations tied to stationary points not captured within the 45% UD. We then grouped all the hunter bout points together and classified the location closest to the UD center as the hunter centroid (Gerrits et al. 2020). We used hunter centroids for subsequent analyses as we sought to identify areas where hunters focused their hunting activity for a period of time (Gerrits et al. 2020). As detailed below and in Gerrits et al. (2020), we considered secondary roads to be representative of areas where hunters actively moved across the landscape to engage in stationary hunting bouts. Landscape classification We used 30 ‐ m resolution landscape classifications from the 2016 National Land Cover Database (NLCD; Jin et al. 2019, Wickham et al. 2021) to map distinct land cover types. We classified land cover into 3 types: pine ‐ dominated, bottomland hardwood, and open area. Pine ‐ dominated areas were identified by NLCD as being predominately evergreen vegetation, and on our site represented upland pine stands. Bottomland hardwood was identified by NLCD as woody wetland and hardwood areas, and on our site generally represented bottomland hardwoods and areas influenced by the Savannah River. Lastly, open areas were identified by NLCD as being hay, pasture, and cultivated crops, which on our site were early successional fields and openings planted for wildlife. These 3 categories represented the predominate land cover types across our study site. The Webb WMA Complex contained 2 general road types, which we defined as primary and secondary. We categorized roads as primary if they were paved or graveled and vehicle access was not limited, whereas secondary roads were unpaved gravel and logging roads or fire breaks where vehicle use was prohibited (Gerrits et al. 2020). Secondary roads were typically dirt or early successional vegetation maintained using herbicide applications and mowing to allow foot travel. Primary roads generally represented access areas where hunters parked and entered HUNTING AFFECTS WILD TURKEY RESOURCE SELECTION | 5 of 17 19372817, 2024, 4, Downloaded from https://wildlife.onlinelibrary.wiley.com/doi/10.1002/jwmg.22567 by University Of Florida, Wiley Online Library on [03/12/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License secondary roads or other areas to initiate hunting activities. Hunters typically walk secondary roads (Gross et al. 2015) and initiate stationary hunting bouts after walking some distance from secondary roads (Gerrits et al. 2020). We calculated availability at the second order by generating 3 paired random locations for each known hunter and wild turkey location to examine resource selection of both hunters and wild turkeys. Because we were specifically interested in resource selection within areas that could be hunted on the WMA by hunters, we required random locations to fall within the WMA boundary. We generated 5, 30 × 30 ‐ m Euclidean distance raster grids, bounded by the WMA boundaries, where each pixel represented distance (km) to a land cover (open, pine, bottomland) or road type (primary or secondary). We intersected all known and paired random hunter and wild turkey locations (1:3 known to random locations, respectively; Fieberg et al. 2021) with distance rasters and extracted the distance (km) for each location to the nearest land cover and road type (Connor et al. 2003, Benson 2013). Resource selection functions Hunting bout selection patterns We used resource selection functions (RSFs) to examine where hunters selected to engage in hunting (i.e., stationary hunting bouts). We specified stationary hunting bouts as a binomial response of known hunter centroid locations and random hunter locations (0 and 1, respectively). We fitted a generalized linear model with a logit link function using glmmTMB in R (Brooks et al. 2017) to model hunter selection as a function of distance ‐ based main effects that included pine, bottomland hardwood stands, open areas, primary roads, and secondary roads. We considered known centroid locations to be the selected area of a hunting bout and the 3 randomly generated locations to represent available but unused hunting areas. Negative RSF coefficients were associated with selection (i.e., individuals were predicted to be closer to the covariate) and positive RSF coefficients were associated with avoidance (i.e., individuals were predicted to be farther from the covariate; Connor et al. 2003). We then predicted the spatial probability (0 – 1) of a hunter engaging in a hunting bout relative to land cover covariates across the 30 × 30 ‐ m grid using our fitted RSF for hunters (Morris et al. 2016). In other words, each 30 × 30 ‐ m pixel received a relative intensity of use by hunters ranging from 0 to 1 ( λ hunt ). We subsequently intersected and extracted λ hunt (i.e., relative hunter use) values to each known and paired random male and female wild turkey location, which served as an additional covariate that directly measured hunting risk for wild turkeys (hunting bout probability). Wild turkey selection patterns We examined how wild turkey selection changed across the hunting season because we expected hunter activity to induce short ‐ and long ‐ term risk aversion strategies. Specifically, we examined how selection for land cover types (pine, open, bottomland) and risk variables (primary roads, secondary roads, λ hunt ) changed beginning before the hunting season (pre ‐ season), to the youth season (youth), to the first 2 weeks of the general season (early season), and to the last 2 weeks of the general season (late season). We assigned all known and paired wild turkey locations a season based on the date of the location. We specified season as a factor variable for analysis. We fitted generalized linear mixed effects models with logit link functions using the glmmTMB package in R version 4.2.2 (Brooks et al. 2017). We fitted separate RSF models for male and female wild turkeys with identical parameterization. Main effects included season (pre ‐ , youth, early, and late season), land cover types (open, bottomland hardwood, and pine [km]), and risk variables (primary roads [km], secondary roads [km], and relative hunter risk probability [0 – 1]). We included interactions among the categorical season variable and all other land cover and risk predictors because we were interested in whether wild turkeys altered selection patterns through 6 of 17 | ROTH ET AL 19372817, 2024, 4, Downloaded from https://wildlife.onlinelibrary.wiley.com/doi/10.1002/jwmg.22567 by University Of Florida, Wiley Online Library on [03/12/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License time with varying levels of risk. We calculated Pearson correlations between pairs of covariates prior to model fitting and removed one of a pair of correlated variables ( | r | ≥ 0.7 Dormann et al. 2013). Relative hunting bout probability was correlated with secondary roads ( r = − 0.87 for male and female wild turkeys). Therefore, we omitted secondary roads from wild turkey RSFs because hunting bout probability was a more direct measure of hunting risk. We standardized all variables to mean = 0 and standard deviation = 1 to facilitate model convergence and interpretation (Schielzeth 2010). Last, we specified a random intercept of wild turkey individual ID to account for repeated sampling and temporal autocorrelation for each individual. We used fitted male and female wild turkey RSFs to predict relative intensities of use ( λ male or λ female ) across the hunting season as a function of land cover and risk variables. We illustrate results graphically because interactions are most easily interpreted visually (Smith et al. 2023). We used the conditional fixed ‐ effects model (i.e., average across individual IDs) to generate predictions of focal variables at the population level and set non ‐ focal variables to their mean values. For all RSFs, we fitted only a single model (hunter, male wild turkey, and female wild turkey = 3 RSFs) because we were interested in inferences for all predictor variables (Fieberg and Johnson 2015). We report hunter and wild turkey RSF coefficients ( β ), standard errors (SE), test statistics ( Z ), P ‐ values, odds ratios, and associated 95% confidence intervals for all covariates. We considered covariates to be statistically significant if 95% confidence intervals around coefficient estimates did not overlap zero and therefore concluded they influenced selection or avoidance patterns. R E S U L T S We monitored 175 GPS ‐ tagged wild turkeys during 2014 – 2018, including 101 females (69 adults, 32 juveniles) and 74 males (47 adults, 27 juveniles). Of these, 1 adult female was poached and 14 adult males were harvested. We collected 68,332 ( x ̅ = 719 ± 3 05 [SD]) and 69,467 ( x ̅ = 978 ± 687) locations on female and male wild turkeys, respectively. We allocated GPS units to 1,500 hunters and estimated that 37.9% of hunters on the Webb WMA Complex carried GPS units (Table 1). We identified 6,823 unique hunting bouts. Hunting bouts were closer to pine stands and secondary roads but farther from primary roads (Table 2). For every 1 km farther from secondary roads, hunting bouts were 38.9 (95% CI = 30.3 – 50.0) times less likely. Likewise, for every 1 km farther from pine stands, T A B L E 2 The estimated resource selection coefficients ( β ) on the log ‐ odd scale, standard errors (SE), test statistic, P ‐ values, and 95% confidence intervals for the hunter resource selection functions, where hunters selected or avoided open, bottomland hardwood (bottomland), or pine ‐ dominated (pine) cover types and primary and secondary roads. Coefficients are distance ‐ based effects and are interpreted as selection for negative coefficients and avoidance for positive coefficients. The 95% confidence intervals that do not overlap zero suggest a statistically significant effect. 95% CI Variable Coefficient ( β ) SE Z P Lower Upper Intercept 484.84 45.93 10.56 <0.001 394.82 574.86 Open − 0.02 0.05 − 0.42 0.67 − 0.13 0.08 Bottomland − 0.15 0.18 − 0.87 0.39 − 0.50 0.19 Pine − 0.74 0.05 − 13.61 <0.001 − 0.85 − 0.64 Primary roads 0.44 0.06 7.50 <0.001 0.32 0.55 Secondary roads − 3.66 0.13 − 28.58 <0.001 − 3.91 − 3.41 HUNTING AFFECTS WILD TURKEY RESOURCE SELECTION | 7 of 17 19372817, 2024, 4, Downloaded from https://wildlife.onlinelibrary.wiley.com/doi/10.1002/jwmg.22567 by University Of Florida, Wiley Online Library on [03/12/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License hunting bouts were 2.1 (95% CI = 1.9 – 2.3) times less likely. Hunters were 1.5 (95% CI = 1.4 – 1.7) times more likely to engage in a hunting bout for every 1 km farther from a primary road (Table 2). Male and female wild turkeys selected areas closer to pines and bottomland hardwoods and farther from open areas, but selection strength varied across the hunting season periods (Figure 2; Table S1, available in Supporting Information). Females also selected areas closer to pines, but selection strength declined when hunting started and females began selecting areas farther from pines relative to the pre ‐ season period (Figures 3A and 4). Males and females selected locations farther from open areas; however, females showed reduced avoidance and began selecting locations closer to open areas when hunting season started (Figures 3 and 4). Conversely, males selected locations farther from open areas as hunting progressed (Figures 3A and 4). Males showed consistent selection patterns for areas closer to bottomland hardwoods, whereas females selected to be nearer bottomland hardwoods during pre ‐ , youth, and early season but selected areas farther from (i.e., avoided) bottomland hardwoods during late hunting season (Figures 3A and 4). Male and female selection of areas near primary roads depended on season and diverged as hunting season progressed (Figure 3B; Table S1). Both sexes selected areas farther from primary roads during pre ‐ season (Figure 5); however, females showed selection during youth, early, and late seasons for areas closer to primary roads, while males continued to select areas farther from primary roads across all seasons (i.e., avoidance; Figures 3B and 5). Notably, males selected for areas farthest from primary roads during the early hunting season when most hunters were present on the WMA (Figure 5). Specifically, males were 1.04 (95% CI = 1.02 ‐ 1.07) times more likely to select areas for every 3.5 km farther from primary roads during early hunting season compared to pre ‐ season (Figure 3B; Table S2, available in Supporting Information). Females were 1.06 (95% CI = 1.03 ‐ 1.09), 1.05 (95% CI = 1.03 ‐ 1.07), and 1.13 (95% CI = 1.10 ‐ 1.16) times more likely to select areas closer to primary roads during youth, early, and late hunting seasons relative to pre ‐ season (Figure 3B; Table S1 and S2). Female wild turkeys avoided areas where hunters were more likely to engage in hunting bouts compared to males, which generally selected areas where hunters were more likely to hunt (Figures 3B and 6). Males and females showed opposite relative intensities of use during pre ‐ season where males were more likely to select areas where hunters would engage in hunting bouts and females increasingly avoided these areas (Figure 6). During youth season, selection and avoidance by males and females decelerated and in early hunting season, both male and female selection changed to avoid areas with increased hunting bout probability (Figure 6). For females, relative intensity of use remained constant during late hunting season; however, males returned to pre ‐ season levels of selection for areas with greater probability of hunting bouts (Figure 6). Specifically, for every 15% increase in hunting bout probability, avoidance by females for these areas declined by 1.13 (95% CI = 1.09 – 1.16), 1.18 (95% CI = 1.15 – 1.21), and 1.27 (95% CI = 1.24 – 1.30) during youth, early season, and late season. Conversely, males selected areas where hunters were more likely to engage in hunting bouts during pre ‐ season. Selection for these areas became weaker as the season progressed declining by 1.16 (95% CI = 1.13 – 1.2) and 1.31 (95% CI = 1.28 – 1.34) times for every 15% increase in hunting bout probability during youth and early season, respectively compared to pre ‐ season. Male selection trends during late hunting season returned to pre ‐ season levels. D I S C U S S I O N Hunters can exert substantive influences on prey through direct harvest (Fleskes et al. 2007, Zbinden et al. 2018) and indirect, non ‐ lethal interactions (Crosmary et al. 2012). Prey can respond to hunting by altering space use, choosing less risky areas, or adjusting activity patterns to avoid predation risk (Heithaus and Dill 2002, Wirsing and Ripple 2011, Smith et al. 2019). We tested the risk allocation hypothesis using a data set containing male and female wild turkey and turkey hunter locational data collected over 5 years. Under the predation risk allocation hypothesis, we expected a strong and immediate anti ‐ predator response by male and female wild turkeys, manifested by spatial avoidance of areas selected by hunters. There was general agreement for the hypothesis for 8 of 17 | ROTH ET AL 19372817, 2024, 4, Downloaded from https://wildlife.onlinelibrary.wiley.com/doi/10.1002/jwmg.22567 by University Of Florida, Wiley Online Library on [03/12/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License F I G U R E 2 Fitted resource selection coefficients for male (blue) and female (red) eastern wild turkeys on Webb Wildlife Management Area Complex in South Carolina, USA, 2014 – 2018. Points are estimated means with associated 95% confidence intervals. The vertical line at zero indicates no effect. All coefficients were standardized (μ = 0, σ = 1) to illustrate the magnitude and direction of effect. Negativity is selection for distance ‐ based variables (vegetation and primary roads) and avoidance for non ‐ distance variables (hunters). HUNTING AFFECTS WILD TURKEY RESOURCE SELECTION | 9 of 17 19372817, 2024, 4, Downloaded from https://wildlife.onlinelibrary.wiley.com/doi/10.1002/jwmg.22567 by University Of Florida, Wiley Online Library on [03/12/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License males, as they avoided areas associated with hunting activity in the early hunting season, whereas females showed limited response to hunting activity. Our results contribute to the growing body of literature demonstrating that game species perceive hunters as predators and shift resource selection to avoid predation risk (Ordiz et al. 2012, Gross et al. 2015, McGrath et al. 2018, Wightman et al. 2023 a ). We specifically investigated how male resource selection changed prior to and within the hunting season in response to hunter distribution. Although hunting pressure varies across the hunting season with most effort and harvest occurring within the first 2 weeks of the season (Hubbard and Vangilder 2005, Gerrits et al. 2020), avoidance of areas where hunting bouts were more likely to occur suggests that anti ‐ predator behaviors exhibited by males are immediate and durable during times of peak hunting activity. Male wild turkeys tended to avoid primary roads after the onset of hunting as these roads provided access points for hunters on our study site. Our data parallel contemporary works noting that male wild turkeys move away from areas associated with public access and increase the frequency of using courtship vocalizations farther from public access during and after hunting (Wightman et al. 2023 a , b ). Male turkeys returned to similar pre ‐ hunting selection patterns during the late hunting season, suggesting behavioral plasticity and habituation to hunting activity (Montgomery et al. 2020) or lower densities of hunters later in the season may have reduced the strength of anti ‐ predator responses (Little et al. 2016). F I G U R E 3 Fitted resource selection function (RSF) coefficients for interactions between hunting season and vegetation communities (A) and risk variables (B) for male (blue) and female (red) eastern wild turkeys on Webb Wildlife Management Area Complex in South Carolina, USA, 2014 – 2018. Points are estimated means with associated 95% confidence intervals. The dashed horizontal line at zero indicates no effect in comparison to pre ‐ season resource selection. All resource selection coefficients are standardized (μ = 0, σ = 1) to illustrate the relative magnitude and direction of effect. Negativity denotes selection for distance ‐ based variables (vegetation and primary roads [prim rds]) and avoidance for non ‐ distance variables (hunters) relative to pre ‐ hunting season. 10 of 17 | ROTH ET AL 19372817, 2024, 4, Downloaded from https://wildlife.onlinelibrary.wiley.com/doi/10.1002/jwmg.22567 by University Of Florida, Wiley Online Library on [03/12/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Resource selection by female wild turkeys was seemingly unaffected by hunting activity. Female wild turkeys avoided areas predicted to have a greater probability of hunting activity in the pre ‐ season, but the relationship weakened as hunting season progressed. Females tend to select upland pine forests and roads when they are reproductively active (Badyaev 1995, Moore et al. 2010, Kilburg et al. 2014, Yeldell et al. 2017) and although females on our study site selected to be closer to primary roads, the strength of selection for primary roads was weak and inconsistent. Risk ‐ reward theory suggests that animals assess predation risk and balance benefits of increased fitness and reproduction with the negative tradeoff of being at risk of predation (Lima 1998, Réale et al. 2007). For example, brown bears ( Ursus arctos ) exposed to predictable human disturbance in space and time only exhibited spatial shifts if alternative food resources of equal quality were <1 km away (Rode et al. 2007). We suspect that although roads were associated with increased hunter presence on our study site, the perceived risk of hunters was likely not strong enough to deter females from using areas near roads during nesting and brooding (Badyaev 1995, Yeldell et al. 2017, Wood et al. 2019). Likewise, the lethality hypothesis states that prey will alter space use to avoid predators that represent the greatest source of lethality (Morosinotto et al. 2010, Kohl et al. 2019, Montgomery et al. 2020). The lack of lethality of hunters to females versus the high lethality of hunters to males may play an important role in inducing anti ‐ predator responses in wild turkeys, and likely accounts for the sex ‐ specific variation we observed. Regardless, our results suggest that hunting activity did not significantly influence resource selection by female wild turkeys during the hunting season. F I G U R E 4 Predicted relative intensity of use for male (dotted) and female (solid) eastern wild turkeys on Webb Wildlife Management Area Complex in South Carolina, USA, 2014 – 2018, relative to distance (km) to open (red), pine (green), and bottomland hardwood (blue) vegetation communities. Predictions for the focal variable were generated by holding other non ‐ focal variables at their mean value and were based on the fixed ‐ effects conditional model (i.e., averaged across random effect of individual). The 95% confidence intervals are omitted to better illustrate resource selection trendlines. HUNTING AFFECTS WILD TURKEY RESOURCE SELECTION | 11 of 17 19372817, 2024, 4, Downloaded from https://wildlife.onlinelibrary.wiley.com/doi/10.1002/jwmg.22567 by University Of Florida, Wiley Online Library on [03/12/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Hunter resource selection was primarily influenced by availability and distribution of areas where they could park and access hunting areas. Hunters chose to be farther from primary roads and closer to secondary roads because hunters parked vehicles along primary roads and used secondary roads to access hunting areas. Our results indicating that hunters extensively relied on secondary roads supports previous studies focused on turkey hunters and hunters targeting other species (Lyon and Burcham 1998, Diefenbach et al. 2005, Karns et al. 2012, Gross et al. 2015). For example, Diefenbach et al. (2005) concluded that 87% of white ‐ tailed deer ( Odocoileus virginianus ) hunters stayed within 500 m of a road or trail, whereas Lebel et al. (2012) reported that 80% of all deer hunter movements were within 100 m of a road or trail. Gross et al. (2015) reported that 75% of all turkey hunter locations were within 150 m of a trail, and 40% of hunter locations were within 25 m of the nearest trail on our study site (Gerrit