Page 1 of 10 This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. Subject Editor: Shin-Jae Rhim Editor-in-Chief: Ilse Storch Accepted 4 October 2023 doi: 10.1002/wlb3.01133 2023: e01133 WILDLIFE BIOLOGY Wildlife Biology www.wildlifebiology.org © 2023 The Authors. Wildlife Biology published by John Wiley & Sons Ltd on behalf of Nordic Society Oikos Wild turkeys Meleagris gallopavo are diurnally active birds that spend the dark hours roosting in trees. We tested the hypothesis that multiple benefits exist for roost tree selection by wild turkeys, including thermoregulation, resource acquisition, and pro- tection from predators. We compared 48 roost trees used by eastern wild turkeys M. g. silvestris in Ontario, Canada to 48 non-roost trees sampled contemporaneously during 2017–2019 to determine roost site selection between seasons. Mean ( ± SE) roost tree height (21.4 ± 0.8 m) was taller than non-roost trees (18.2 ± 0.8 m), and roost trees were also larger in diameter at breast height (58.1 ± 5.5 vs 38.7 ± 3.1 cm). Using ibut- tons to collect microclimate temperatures at the tree, we found that mean temperature ( ± SE) of a deciduous roost (14.5 ± 0.1 ° C) was higher than temperature at either a coniferous roost (13.9 ± 0.1 ° C) or ambient temperature (13.2 ± 0.1 ° C) during the summer months. In winter however, we did not find any relationship between tem- perature and tree type. Roosts were closer to buildings (150.8 ± 26.0 m) in the winter compared to summer and year-round roosts, and winter roosts were also farther away from crops (395.2 ± 63.7 m) compared to roost sites used year-round. Summer roosts were closer to roads (143 ± 36.3 m) than the roosts in the winter and roosts used year- round. Our data suggest that thermoregulation is not the driving force behind roost selection; instead, predator avoidance appears to play the most important role, with some weaker evidence in support of proximity to resources. Keywords: Meleagris , roost, social, supplemental food, thermoregulation Introduction It is important for wild animals to dedicate time and energy to selecting high qual- ity sleeping locations because the suitability of those locations can influence fitness in many ways. There can be benefits when a location is close to resources, thermally suitable, or safe from predation risk (Paclík and Weidinger 2007, Wiens et al. 2014, O’Brien et al. 2021). Numerous factors affect the suitability of sleeping locations, and patterns of selection arise from the trade-offs between costs and benefits of these Seasonal roost selection of wild turkeys at their northern range edge Elizabeth A. Adey 1 , Jennifer E. Baici 1 and Jeff Bowman ✉ 2 1Trent University, Peterborough, ON, Canada 2Ontario Ministry of Natural Resources and Forestry, Peterborough, ON, Canada Correspondence: Jeff Bowman (Jeff.Bowman@ontario.ca) Research article Page 2 of 10 different locations. For example, an animal may benefit from the increased vigilance of sleeping in a group with conspecif- ics, but this may also come with a cost of increased compe- tition for food, or increased risk of disease spread (Buckley 1998, Laughlin et al. 2019). When animals sleep in a group, this is often referred to as communal roosting, which occurs when individuals aggre- gate at a roost location, often at night, in a specific area (Beauchamp 1999). Some long-term benefits of communal roosting can include information gathering and mate acquisi- tion (Blanco and Tella 1999), as well as short-term benefits like thermoregulation, predator avoidance, and establishing social status (Draulans and van Vessem 1986, Summers et al. 1987, Bishop and Groves 1991, Buckley 1998). Three hypotheses aim to explain how these benefits can lead to com- munal roosting. The information centre hypothesis (Ward and Zahavi 1973) states that roosts act as information shar- ing centers for food source locations. The recruitment centre hypothesis (Richner and Heeb 1996), states that recruiting new members benefits the flock as a whole through predator dilution and increased foraging success. The two strategies hypothesis tries to reconcile the multiple explanations for the evolution of roosting behavior by proposing that not all indi- viduals use a roost for the same reason (Weatherhead 1983); for example, communal roosting may evolve if individuals with a higher social status benefit from higher quality perch locations that provide predator avoidance and increased ther- moregulation, as a trade-off for informing the lower status individuals where the resource locations are (Weatherhead 1983). In this scenario, roost use is affected by different mechanisms depending on the status of individual birds. It has long been known that wild turkeys Meleagris gal- lopavo are highly social birds that roost communally year- round in trees (Fig. 1), with the exception of the period when females are nesting (Watts and Stokes 1971, Healy 1992). Numerous studies have been undertaken to assess roost selection by wild turkeys. For example, it has been proposed that turkeys choose roosts based on the benefits of being close to foraging sites, close to water, protection from predators, and because of thermoregulatory benefits (Boeker and Scott 1969, Kilpatrick et al. 1988, Rumble 1992, Chamberlain et al. 2000, Sasmal et al. 2018, Bakner et al. 2022, Gonnerman et al. 2023). Wild turkeys at northern latitudes are especially challenged to survive the winter because of restricted access to natural foods due to snow cover, longer nights, and colder tempera- tures (Haroldson et al. 1998, Nguyen et al. 2004, Niedzielski and Bowman 2015, Lavoie et al. 2017, Gonnerman et al. 2023). Minimizing nocturnal energy expenditure is critical to increase fasting endurance as lower temperatures require greater energy and food use to maintain thermoregulation (Paclík and Weidinger 2007). Consistent with the two strate- gies hypothesis, we consider that it is likely that wild tur- keys benefit from communal roosts in winter for multiple reasons, such as thermoregulation and resource acquisition. With respect to thermoregulation, energetic savings of com- munal roosting can be high compared to solitary roosting. For example, the acorn woodpecker Melanerpes formicivorus in California gains a 37% energetic saving by roosting com- munally (Du Plessis et al. 1994). We have also observed how- ever, that roosts tend to be close to supplemental food sources in winter. Most information on roosting behaviour for the eastern subspecies of the wild turkey M. g. sylvestris comes from stud- ies in the southern and central part of their range. A study in Rhode Island, USA, found roost site selection by wild turkeys included larger diameter trees, trees close to open water, and white pines Pinus strobus (Kilpatrick et al. 1988). Winter survival in Minnesota, has been linked to roost site selection near supplemental food sources, such as waste grain Figure 1. (A) Wild turkey preening in a roost tree. (B) One of the most frequently used roost trees in our study area. (C) A flock of turkeys perched in their roost. All photos were taken in our study site at a sugar maple Acer saccharum roost tree that was heavily used. 1903220x, 2024, 1, Downloaded from https://nsojournals.onlinelibrary.wiley.com/doi/10.1002/wlb3.01133 by University Of Florida, Wiley Online Library on [13/03/2025]. 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 Page 3 of 10 from agricultural activities (Kane et al. 2007). Similar results were found in Massachusetts during the winter where flocks restricted their movements by selecting roost sites that were close to pastures and fields spread with manure (Vander Haegen et al. 1989). There are few studies looking at roosting selection by eastern wild turkeys near their northern range edge, where thermal stress in winter is more likely to be a limiting factor. One study found that after an extralimital northern introduction, turkeys selected mostly conifers, and linked that selection to the likelihood of the tree’s ability to reduce impacts of wind and therefore heat loss to the birds (Nguyen et al. 2004). We were interested in the selection of roost trees by eastern wild turkeys at the northern edge of their range in Ontario, Canada. In this region, wild turkeys roost com- munally year-round, but this behavior is especially evident in winter, with large flocks forming in the autumn and per- sisting throughout the winter (Niedzielski and Bowman 2016, Baici and Bowman 2023). We assessed the charac- teristics of roosts selected, the relationship of the roost site to the resources around it, and the potential for seasonal changes in roost selection. Winter survival is a limiting fac- tor for sustainable populations of wild turkeys in northern latitudes (Wright et al. 1996, Kane et al. 2007, Niedzielski and Bowman 2015, Gonnerman et al. 2023). Mammalian predation has been identified as the main cause of death in studies from the region (Niedzielski and Bowman 2015). Understanding roost tree selection during the winter may help to test whether thermoregulation is a driver of roosting behavior in turkeys, and therefore may be linked to their survival and success as a reintroduced species in their north- ern range. Consistent with the two strategies hypothesis, and the idea that not every individual uses a roost for the same reason, we hypothesized that multiple benefits exist for roost selection by eastern wild turkeys, including ther- moregulatory, resource acquisition, and predator protection benefits. We predicted that turkeys should roost in large, tall trees, relative to available trees in the area, which provide refuge from many mammalian predators while allowing for numerous large individuals to roost in one location together. We also predicted that winter roosting locations would have better thermoregulatory properties than summer roost sites and that winter roost sites would be in close proximity to food resources. Material and methods Our study site, near Peterborough, Ontario (44.37N, − 78.28W), was close to the northern edge of the eastern wild turkey’s geographic range in Canada. Our study area consisted of mainly agricultural fields, mixed forests, wetlands, active farming properties, as well as deciduous and coniferous trees lining properties. In this region, the winters can be severe with low temperatures, and more snowfall than what turkeys experience in the core parts of their range. January, the cold- est month of the year, had a mean temperature of − 8.7 ° C in 2017, and − 14.3 ° C in 2018. The mean monthly snowfall in Peterborough during the winter months (November–April) was 24 cm (Government of Canada 2019). Wild turkeys were trapped using rocket nets (Grubb 1988) between January and March of 2017 and 2018, in open fields using bait piles consisting of corn. Locations to trap were based on driving surveys of the larger study area in Peterborough County where we located turkey flocks that were fairly consistent in their daily movements and flock size (Baici and Bowman 2023). We also needed landowner per- mission and a safe area to operate the rocket net. Trapping in the winter allowed us to target larger flocks and increase our chances of deploying GPS and VHF transmitters on an equal number of males and females (Fig. 2). From 2017 to 2018, we had 13 capture events at a total of five trapping sites, where one site had five capture events, two sites had three capture events, and the remaining capture events happened at the other two sites one time each. Capture and handling methods were in compliance with approved Trent University animal care protocol no. 24653. To track turkey movements, we used GPS transmitters (Lotek PinPoint VHF-3600L, Ontario, Canada) weighing 85 g ( < 3% mean mass of an adult hen, mean mass = 4.3 kg). Transmitters were attached using a backpack-style har- ness made of shock cord (Norman et al. 1997). The tags were programmed to collect GPS locations ( ± 19.9 m from our field trials) at various schedules. In 2017, locations were collected every 4 h and 15 min for 24 tags. In 2018 loca- tions were collected every hour between 6:30 and 10:30 h and once at 12:00 and again at 4:00 for 21 tags. Our total sample included 45 individuals: 24 adults, 21 juveniles, 22 males and 23 females. We located 48 roost trees during the day by ground truth- ing nocturnal GPS locations for each turkey and identify- ing roosts, where the roost could have been used for only one night, or multiple nights. The majority of our study area was private property, which influenced our ability to access and ground truth all possible roost locations. We were there- fore only able to actively collect roosting information from areas where we had landowner permission. We walked to the Figure 2. An eastern wild turkey Meleagris gallopavo sylvestris with a GPS transmitter. 1903220x, 2024, 1, Downloaded from https://nsojournals.onlinelibrary.wiley.com/doi/10.1002/wlb3.01133 by University Of Florida, Wiley Online Library on [13/03/2025]. 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 Page 4 of 10 GPS location, and then confirmed the roost tree by the pres- ence of fecal droppings, or feathers around the base of the tree (Hoffman 1968). Morphological measurements were collected at the tree, including tree height (m), and lowest branch to ground (m) using an electronic clinometer (model Haglof HCH, Haglof, Sweden). The tree’s diameter at breast height (DBH; taken at 1.3 m) was measured using diameter tape (C.F.E. Equipment Norfolk, Virginia) and the tree spe- cies was identified. Aspect and slope were collected based on previous research suggesting morning sun could affect roost selection (Boeker and Scott 1969, Donázar and Feijoo 2002). Aspect was transformed from degrees to radians for data analysis. We additionally took those same measurements at the nearest accessible tree greater than 20 cm DBH found at least 50 m north of the roost location. A 20 cm DBH was selected as a minimum size for an available tree as this was the smallest tree we had observed actively being used as a roost. Hereafter we refer to these systematically selected trees as non-roost trees. Using ArcGIS (ESRI 2011), we determined the distances to primary roads, water, crop, building, and livestock for each roost and non-roost location. In our study area, a predomi- nantly agricultural landscape, residential homes and farms were often connected to primary roads with secondary roads, including long unpaved driveways. In many cases, long drive- ways connecting buildings to primary roads were not classi- fied within the road layer. In our analysis, only primary roads were included. This included roads that existed as a path for conventional four-wheeled street legal vehicles and excluded paths constructed for off-road vehicles (e.g. ATV trails) and those related to forest and forest fire management (Land Information Ontario (LIO) 2019a). We used the Ontario Land Cover Database (Ontario Ministry of Natural Resources and Forestry 2014) for water and forest type and used Land Information Ontario base maps for primary roads, crops, and buildings (Provincial Mapping Unit 2017, Agriculture and Agri-Food Canada 2018, Graham 2019). Livestock locations were recorded and mapped using comprehensive road sur- veys and aerial imagery. We drove each road inside our study area and recorded location information for each livestock operation visible from public access roads. We then used a combination of satellite imagery (Land Information Ontario (LIO) 2019b) and qualitative observations recorded during ground-truthing to construct spatial polygons representing the approximate size and distribution of all livestock opera- tions within the study area (Baici and Bowman 2023). To compare temperature between a deciduous and a coniferous roost tree, we used 10 ibutton temperature sensors (model Maxim Integrated DS1922L, San Jose, CA, USA). We placed ibuttons in one tree of each type that occurred in close proximity to one another to gain insight into the temperature variation we might expect between coniferous and deciduous trees in the study area, and to assess varia- tion within each type of tree. Five ibuttons were placed in a known, well used coniferous roost tree (eastern white cedar, Thuja occidentalis ), and five ibuttons were placed in a known, well used deciduous roost tree (sugar maple, Acer saccharum ). The ibuttons were placed in locations where the turkeys might perch within the roost. One each was placed, close to the trunk, on the lowest branch, on a branch in the middle of the tree, and on a branch closest to the top of the tree. The fourth and fifth were placed at the farthest point on the eastern and western facing branches respectively regardless of the height at which this point occurred. The ibuttons were installed in July of 2018 and were collected in February of 2019, recording environmental temperature once every hour for 212 days. The ibutton temperature data were averaged between the 5 ibuttons placed in each tree, and then filtered for overnight (10:00–5:00 h) readings. Regional ambient temperature was taken from the closest weather station to our study site (Government of Canada 2019). Roosts were categorized into three classes: winter, summer, and year round. Winter occurred between 1 November and 30 April, and summer between 1 May to 31 October. The seasons were divided in this way to include the majority of low regional ambient temperatures in the winter and the high regional ambient temperatures in the summer, similar to a study done in Quebec (Lavoie et al. 2017). This division also aligns with turkey behaviour where birds flock up in large groups beginning in November, and then begin to separate into smaller groups for breeding by May (Kurzejeski et al. 1987, Badyaev et al. 1996, Humberg et al. 2009, Niedzielski and Bowman 2015). Roosts were categorized as a winter roost if 90% or more of the total GPS locations (pooled across all birds) occurred during the months of November–April. The same rule was applied to categorizing summer roosts for points at roosts during the months of May–October. A third category included roosts that were used all year round. During the study period, we also closely monitored wild tur- key nesting behaviour, and removed any GPS points associ- ated with incubation, as female wild turkey roost sites are known to vary substantially during the reproductive period (Streich et al. 2015, Wood et al. 2019). A Kruskal–Wallis one way analysis of variance test was used to compare roost trees to non-roost trees grouped by season of use, where there were three groups, summer roosts, winter roosts, and roosts used all year. For significant rela- tionships, Wilcoxon tests were then used to make post-hoc comparisons. A Kruskal–Wallis was also used to evaluate the effect of season on environmental temperature at the roost, using the ibutton temperature data. A post-hoc test (Dunn’s test) was then used to test the difference between the groups. All descriptive comparisons were conducted using RStudio ver. 1.2.1 (www.r-project.org). We used three binomial generalized linear mixed models in RStudio ver. 1.2.1 to test for differences in the character- istics of roost and non-roost trees with season as a categorical variable. We evaluated coefficients of the global model, a site model that considered site characteristics, and a tree model that considered tree characteristics to test our hypotheses and to gain insight into the selection of roost trees. We compared the three models with Akaike’s Information Criterion (AIC) to evaluate whether variables at the tree, site, or a mix of both best explained wild turkey roost selection. 1903220x, 2024, 1, Downloaded from https://nsojournals.onlinelibrary.wiley.com/doi/10.1002/wlb3.01133 by University Of Florida, Wiley Online Library on [13/03/2025]. 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 Page 5 of 10 Results We sampled 48 unique roost trees over the two-year study period. Eleven species of tree were used as roost sites, includ- ing American beech ( Fagus grandifolia ; 4%), American elm ( Ulnus americana ; 8%), basswood ( Tilia americana ; 8%), black ash ( Fraxinus nigra ; 10%), black locust ( Robinia pseu- doacacia ; 4%), eastern white cedar ( Thuja occidentalis ; 21%), large tooth aspen ( Populus grandidentata ; 6%), Norway maple ( Acer platanoides ; 2%), Manitoba maple ( Acer negundo ; 2%), sugar maple ( A. saccharum ; 29%), and trembling aspen ( Populus tremuloides ; 4%). In total, 44% (n = 21) of the roosts were used year-round, 19% (n = 9) of the roost sites were used only in winter, and 38% of roosts (n = 18) were used only in summer. Fifteen species of tree were used at random sites, including American elm (6%), black ash (17%), black locust (2%), blue spruce ( Picea pungens ; 2%), buckthorn ( Rhamnus cathartica ; 4%), crab apple ( Malus sylvestris ; 2%), eastern white cedar (29%), honey locust ( Gleditsia triacan- thos ; 2%), larch ( Larix decidua ; 2%), large tooth aspen (4%), Norway spruce ( Picea abies ; 2%), silver maple ( Acer sacchari- num ; 2%), sugar maple (15%), trembling aspen (7%), and white pine ( P. strobus ; 4%). The average distance between a roost tree and an associated random tree was 48.8 m. Distances between roosts and buildings, roads, and crops differed among seasons (Table 1). Roosts were on average, more than 100 m closer to buildings in the winter compared to roosts used year-round or roosts used in the summer (pair- wise Wilcoxon test, p < 0.01). Winter roosts were farther away from crops compared to year-round roosts (pairwise Wilcoxon test, p = 0.01). Summer roosts were closer to roads than the roosts in the winter and the roosts used year-round (pairwise Wilcoxon test, p = 0.01). There was no difference between roosts in the winter and the roosts used year-round in distance to roads. During summer, we observed temperature differences between the coniferous and deciduous trees and ambi- ent temperature (Table 2). In particular, it appeared that temperatures in the deciduous roost were higher than in either the coniferous tree or ambient temperature. There was a significant difference between all groups; ambi- ent temperature was cooler than the temperatures found at either the deciduous site or the coniferous site (Dunn test, z-value = 5.04 and 2.65 respectively, both p < 0.005, n = 882), and the coniferous site was cooler than the decid- uous site (Dunn test, z-value = 2.39, p < 0.01, n = 882). There was also a significant difference between temperature at the different trees and ambient temperature during the summer (Kruskal–Wallis comparison of rank means, df = 2, χ 2 = 25.39, p < 0.0001). There was no significant effect of location (tree type or ambient) on temperature during the winter months (Kruskal–Wallis comparison of means, df = 2, χ 2 = 0.19, p = 0.9). Comparing the binomial models using AIC (Table 3) showed that the best fit model was the tree model (Table 4, AIC = 126.01). Significant variables were consistent between the global model (Table 5, AIC = 137.07) and the tree model (Table 4), highlighting the height and DBH of roost trees being taller and larger than non-roost trees. The site model did not contain any significant variables (Table 6, AIC = 148.75). Closest branch to the ground, aspect, slope, distance to water, to road, to closest crop, to buildings, to livestock, and season did not differ significantly (all p > 0.1) between roost trees and the non-roost trees (Table 5–6). Discussion Wild turkey roosts in our study were predominantly influ- enced by tree characteristics, and not by site characteris- tics. Turkeys roosted in taller trees with larger diameters than nearby non-roost trees. We also found that there were some microclimate differences between types of trees, but this pattern was not what we expected. Summer tempera- tures were warmer in a deciduous tree than either a conif- erous tree or ambient temperature, whereas we predicted that trees might be selected for thermoregulatory advantage in winter. However, instead of seasonal differences due to tree type, we found that winter roosts were closer to build- ings, but farther from roads and crops than summer or Table 1. Comparison of 11 characteristics of eastern wild turkey Meleagris gallopavo silvestris summer roosts (n = 18), winter roosts (n = 9), and roosts used year-round (n = 21) in Peterborough, Ontario, as determined from Kruskal–Wallis rank sum test. Bold type indicates signifi- cant (p < 0.05) comparisons. Characteristic Summer All year round Winter p-value Mean SE Mean SE Mean SE Distance to buildings (m) 258.3 29.0 302.0 23.4 150.8 26.0 0.002 Distance to road (m) 143.0 36.3 218.0 25.1 235.2 33.7 0.008 Distance to crop (m) 262.4 55.2 179.0 33.4 395.2 63.7 0.045 Distance to water (m) 41.9 14.4 84.6 20.3 71.9 20.3 0.092 Tree ype (deciduous 0 coniferous 1) 0.3 0.1 0.1 0.1 0.3 0.2 0.229 Closest branch to the ground (m) 4.7 0.9 5.6 0.7 4.1 1.1 0.351 DBH (cm) 61.8 7.8 49.2 6.0 71.6 20.8 0.510 Distance to livestock (m) 908.4 163.0 970.0 164.2 591.5 155.7 0.674 Height (m) 21.2 0.2 21.1 0.9 22.0 2.3 0.921 Slope (%) − 7.4 2.6 − 6.8 2.3 − 9.9 5.0 0.944 Aspect (radians) 1.4 0.3 1.4 0.3 1.6 0.4 0.974 1903220x, 2024, 1, Downloaded from https://nsojournals.onlinelibrary.wiley.com/doi/10.1002/wlb3.01133 by University Of Florida, Wiley Online Library on [13/03/2025]. 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 Page 6 of 10 year-round roosts. These site-level differences in roost proximity to buildings, roads, and crops were evident in descriptive bivariate analyses, but were not important in the multivariate regression models. Overall, the height and DBH findings suggest that reduc- ing predation risk may be the most important benefit of roosts for wild turkeys in the study area. Large trees may also be important to accommodate large flocks of roosting birds. In contrast, we observed some limited evidence for proximity to buildings in winter, which appears to be associated with resource acquisition because of the importance of supple- mental food in winter, and the availability of this resource near buildings via bird feeders or livestock operations (Vander Haegen 1989, Thompson et al. 2009, Hill et al. 2023). This represents only weak evidence supporting the two strategies hypothesis, where the strategies are predator avoidance and resource acquisition. We had no evidence in support of ther- moregulation being an important basis of roost selection. Ward and Zahavi (1973) proposed that information shar- ing was the principal explanation for communal roosting in birds, but Weatherhead (1983) modified this idea, suggest- ing that individual birds may often have different reasons for roosting communally, depending on their status and role in a flock (i.e. the two strategies hypothesis). For example, subordinate birds gain access to food resources, while domi- nant individuals tolerate food sharing due to the benefits of a predator dilution effect (Weatherhead 1983). Beauchamp (1999) found that across bird species, increased foraging efficiency was a key factor in promoting communal roost- ing, although he noted that there were exceptions suggesting other processes can play a role. Our data provide only weak support for any of these ideas. In our study area, wild turkeys roost communally year-round, but the importance of tall, large diameter trees suggests that reducing predation risk is the simplest explanation for roost selection. Our study supports previous research (Boeker and Scott 1969, Kilpatrick et al. 1988, Rumble 1992) showing turkeys’ selection for roosts that are larger in diameter and height compared to non-roost trees. This selection, based on specific tree characteristics, aligns with the fact that these large-bodied birds require larger trees to accommodate larger groups and foster social interactions that take place during roosting. The height that they are selecting for also relates to the ability of that roost tree to distance them from ground predators (Beauchamp 1999). We speculate that there are hierarchies in perch locations within a roost, and that perches close to the ground are lower quality due to increased predation risk. Future research could evaluate this idea. Observations of roosting behaviour in our study suggests a lot of social interaction, including agonistic behaviour while birds estab- lish perch locations on a roost (McGowan et al. 2006, Evans and Sordhal 2009, Duerr et al. 2019). Agonistic behaviours are those which cause, threaten to cause, or seek to reduce physical damage and typically involve threats, aggression, and submission (McGlone 1986). Our results indicate that wild turkeys roosted farther from primary roads than expected during the winter months. However, because long driveways connecting buildings to primary roads were not classified within the road layer, different patterns may be observed when accounting for secondary roads. For example, many species use secondary roads to move through habitat (Whittington et al. 2019). Furthermore, wild turkeys may have a negative association with primary roads because birds closer to roads may be more likely to experience increased road mortality as a confound- ing effect of hunting pressure (Holbrook and Vaughan 1985) or increased predation (Thogmartin and Schaeffer 2000). When examining the microclimate characteristics of the roosting sites, we noticed contrasting seasonal variation between the coniferous and deciduous roost trees. We note that placing ibuttons in only two trees limits our inference about observed patterns, but we were surprised that the dif- ferences were more pronounced during the summer, whereas we initially anticipated greater variations between the tree types during winter. Previous research suggests coniferous roosts are selected based on their ability to shelter from win- ter weather better than deciduous trees which drop all their leaves and are more exposed to the elements (Schmitz 1991). Table 3. Comparison of three models (tree, landscape and global models) describing selection of roost trees by eastern wild turkeys , Meleagris gallopavo sylvestris . The tree model consisted of seven tree characteristics comparing roost trees with non-roost trees. The land- scape model consisted of seven coarser landscape-scale characteristics comparing roost and non-roost trees. The global model consisted of all (12) roost and non-roost characteristics. Bold type indicates the top model. K AICc Delta_AICc AICcWt Cum.Wt LL Tree model 9 126.01 0 1 1 − 52.96 Global model 14 137.07 11.06 0 1 − 51.94 Site model 8 148.75 22.74 0 1 − 65.55 Table 2. Comparison of ibutton temperature data ( ° C) located in a deciduous tree, coniferous tree, and ambient temperature data from a weather station close to sampled trees in Peterborough, Ontario. Data are expressed as means, and inter-quartile ranges (IQR). Comparisons were done using a Kruskal–Wallis one way analysis of variance. Temperature data from July 2018 to February 2019. Bold type indicates significant (p < 0.05) comparisons. Season Coniferous Deciduous Ambient p-value n mean SE IQR n mean SE IQR n mean SE IQR Winter 815 − 5.3 0.09 9.8 815 − 5.4 0.09 9.8 815 − 5.4 0.09 9.9 0.91 Summer 882 13.9 0.09 11.3 882 14.5 0.09 11.4 882 13.2 0.09 10.7 0.000003 1903220x, 2024, 1, Downloaded from https://nsojournals.onlinelibrary.wiley.com/doi/10.1002/wlb3.01133 by University Of Florida, Wiley Online Library on [13/03/2025]. 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 Page 7 of 10 However, we found no difference in temperature between coniferous trees and deciduous trees during the winter. One factor that is not recorded by ibuttons, is wind speed. Wind, in combination with temperature, affects the operative cool- ing of an organism, which likely plays an important role in their energy balance in the winter (Bakken 1976). The tur- keys may be favouring the coniferous sites because of the shelter they receive from wind, rather than temperature. The ibutton data did not show significant temperature differences between tree types in the winter, but we did find that decidu- ous roosts were warmer in the summer. Our winter roost sites were significantly closer to build- ings than roost sites selected during the summer. This under- scores the reliance that wild turkeys may have on humans during winter, and how we may be playing an important role in their survival during harsh winter conditions (Niedzielski and Bowman 2016). In addition to low temperatures (Oberlag et al. 1990), turkeys have been shown to be limited by snow depth in terms of their ability to forage for resources (Lavoie et al. 2017). If their roost sites are closer to build- ings, and therefore humans, it is likely they are receiving some sort of supplemental food sources, and the energetic cost of acquiring that food source might often be lower than foraging through deep snow for natural foods, which can be hard to access (Wright et al. 1996, Baici and Bowman 2023). Snow in agricultural landscapes is also often compacted by anthropogenic and livestock activities, reducing travel costs even further. The turkey’s ability to survive winter will likely increase if their selection of winter roost sites are closer to supplemental food sources, and therefore closely tied to the presence of humans. If summer roosting locations of eastern wild turkeys play a role in preventing overheating, expanding their range farther north may suggest colder weather is not the critical factor impeding the expansion of their range. For example, availabil- ity of supplemental food likely shifts the range limit north- ward (Niedzielski and Bowman 2016). A study modelling 50 bird species’ heat balance, showed the likelihood of southern range declines driven by climate changes, specifically water requirements for evaporative cooling, exacerbated in larger bodied birds (Riddell et al. 2019). Wild turkeys may be more affected by heat stress in the summer, rather than heat loss in the winter in relation to their selection of roosting sites. This idea is supported by physiological adaptations like their featherless heads (Buchholz 1996), and behavioural adapta- tions of avoiding heat stress (Rakowski et al. 2019). Turkeys may be selecting winter roots sites on the basis of other prop- erties not related to thermoregulation, and therefore there Table 4. The tree model. Comparison of seven characteristics of eastern wild turkey Meleagris gallopavo silvestris roost trees (n = 48) and non-roost trees (n = 48) in Ontario, Canada, as determined from binomial generalized linear model comparing characteristics of roost trees and non-roost trees (residual deviance = 105.92, null deviance = 133.08). The categorical season variable had three groups (summer, winter, year round) indicating when the roost was used. Bold type indicates significant coefficients (p < 0.05). Characteristic Roost Non-roost Z-value p-value Coefficient SE Mean SE Mean SE Height (m) 21.4 0.8 18.2 0.8 2.014 0.04 0.11 0.05 DBH (cm) 58.1 5.5 38.7 3.1 3.169 0.002 0.03 0.01 Closest branch to ground (m) 5.0 0.5 3.6 0.4 1.369 0.173 0.12 0.09 Aspect (radians) 1.9 0.3 1.1 0.2 − 0.421 0.674 − 0.09 0.22 Slope (%) − 7.6 1.6 − 5.0 1.4 − 0.880 0.379 − 0.04 0.04 Tree type (coniferous 1, deciduous 0) 0.2 0.06 0.4 0.07 − 0.997 0.319 − 0.60 0.60 Season summer 0.611 0.541 0.33 0.54 Season winter 1.296 0.195 0.94 0.72 Table 5. The global model. Comparison of 12 characteristics of eastern wild turkey Meleagris gallopavo silvestris roost trees (n = 48) and non-roost trees (n = 48) in Ontario, Canada, as determined from binomial generalized linear model comparing characteristics of roost trees and non-roost trees (residual deviance = 103.89, null deviance = 133.08). The categorical season variable had three groups (summer, winter, year round) indicating when the roost was used. Bold type indicates significant coefficients (p < 0.05). Characteristic Roost Non-roost Z-value p-value Coefficient SE Mean SE Mean SE Height (m) 21.4 0.8 18.2 0.8 2.169 0.03 0.12 0.05 DBH (cm) 58.1 5.5 38.7 3.1 3.220 0.001 0.03 0.01 Closest branch to ground (m) 5.0 0.5 3.6 0.4 1.458 0.145 0.14 0.09 Aspect (radians) 1.9 0.3 1.1 0.2 − 0.674 0.500 − 0.15 0.22 Slope (%) − 7.6 1.6 − 5.0 1.4 − 1.015 0.310 − 0.04 0.04 Distance to water (m) 66.2 11.3 72.0 11.8 − 0.004 0.997 − 0.001 0.003 Distance to road (m) 184.0 20.3 194.1 18.5 0.507 0.612 0.001 0.002 Tree type (coniferous 1, deciduous 0) 0.2 0.06 0.4 0.07 − 0.974 0.330 − 0.61 0.62 Distance to crop (m) 264.8 28.1 363.6 128.1 − 0.485 0.628 − 0.0003 0.0006 Distance to buildings (m) 252.3 16.2 265.6 18 − 1.021 0.307 − 0.002 0.002 Distance to livestock (m) 845.1 102.3 917.8 99.1 0.589 0.556 0.0002 0.0004 Season summer 0.789 0.430 0.47 0.60 Season winter 1.440 0.150 1.08 0.75 1903220x, 2024, 1, Downloaded from https://nsojournals.onlinelibrary.wiley.com/doi/10.1002/wlb3.01133 by University Of Florida, Wiley Online Library on [13/03/2025]. 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 Page 8 of 10 may be a stronger driver in the selection process of roost sites throughout the season. Habitat features around the roost, proximity to resources, and flock hierarchy structure, could, for example, play a more significant role in their selection of roost sites. The wild turkey is not only a game species in Ontario, but it is also a successful reintroduction story (Mosby 1959, 1973, 1975, Bailey 1980, Kennamer 1986, Bellamy 2001). Our findings could help to inform management decisions for population sustainability in Ontario, and elsewhere. Conserving stands with large coniferous and deciduous trees will be important for roosting habitat as wild turkeys switch tree types throughout the year and the roost tree selection may be linked to microclimate temperatures at the roost tree itself. In our study area, we observed a link between humans and wild turkeys, with a more direct link in the winter where the birds may be relying on waste grain and livestock for sup- plemental food resources. As our climate changes and we see a northern expansion of many species, using this information may also help to predict where we will see wild turkeys select- ing favorable roosting habitat as their range expands. Funding – Funding for this research was provided by the Ontario Ministry of Natural Resources and Forestry and by an NSERC Discovery Grant to JB. Permits – All research was carried out under a protocol approved by the Trent Univ. Animal Care Committee protocol no. 24653. Author contributions Elizabeth A. Adey : Conceptualization (equal); Data cura- tion (equal); Formal analysis (lead); Investigation (lead); Methodology (lead); Writing – original draft (lead). Jennifer E. Baici : Data curation (equal); Formal analysis (supporting); Investigation (equal); Methodology (supporting); Writing – revi