Landscape-Abundance Relationships of Male Eastern Wild Turkeys Meleagris gallopavo silvestris in Mississippi, USA Authors: Davis, Annie, Wang, Guiming, Martin, James, Belant, Jerrold, Butler, Adam, et al. Source: Acta Ornithologica, 52(2) : 127-139 Published By: Museum and Institute of Zoology, Polish Academy of Sciences URL: https://doi.org/10.3161/00016454AO2017.52.2.001 BioOne Complete (complete.BioOne.org) is a full-text database of 200 subscribed and open-access titles in the biological, ecological, and environmental sciences published by nonprofit societies, associations, museums, institutions, and presses. Your use of this PDF, the BioOne Complete website, and all posted and associated content indicates your acceptance of BioOne’s Terms of Use, available at www.bioone.org/terms-of-use. Usage of BioOne Complete content is strictly limited to personal, educational, and non - commercial use. Commercial inquiries or rights and permissions requests should be directed to the individual publisher as copyright holder. BioOne sees sustainable scholarly publishing as an inherently collaborative enterprise connecting authors, nonprofit publishers, academic institutions, research libraries, and research funders in the common goal of maximizing access to critical research. Downloaded From: https://bioone.org/journals/Acta-Ornithologica on 30 Jan 2025 Terms of Use: https://bioone.org/terms-of-use Access provided by University of Florida Landscape-abundance relationships of male Eastern Wild Turkeys Meleagris gallopavo silvestris in Mississippi, USA Annie D AVIS 1 , Guiming W ANG 1 *, James M ARTIN 2 , Jerrold B ELANT 1 , Adam B UTLER 3 , Scott R USH 1 & Dave G ODWIN 4 1 Department of Wildlife, Fisheries and Aquaculture, Mail Stop 9690, Mississippi State University, Mississippi State, Mississippi 39762, USA 2 Warnell School of Forestry and Natural Resources & Savannah River Ecology Laboratory, University of Georgia, Athens, GA 30602, USA 3 The Mississippi Department of Wildlife, Fisheries and Parks, 1505 Eastover Drive, Jackson, MS 39211, USA 4 Mississippi Forestry Association, 620 North State Street, Suite 201, Jackson, MS 39202, USA *Corresponding author, e-mail: guiming.wang@msstate.edu Davis A., Wang G., Martin J., Belant J., Butler A., Rush S., Godwin D. 2017. Landscape-abundance relationships of male Eastern Wild Turkeys Meleagris gallopavo silvestris in Mississippi, USA. Acta Ornithol. 52: 127–139. DOI 10.3161/00016454AO2017.52.2.001 Abstract. The effects of landscape alteration on wildlife have drawn increasing attention from ecologists as landscapes have rapidly changed worldwide. Large-bodied birds are highly sensitive to habitat loss and fragmentation. We aimed to test two hypotheses concerning landscape-abundance relationships of male Wild Turkeys Meleagris gallopavo silvestris , the largest galliform in North America. The landscape fragmentation hypothesis posits that animal populations would decline linearly with habitat fragmentation, independent of the effects of habitat loss. The landscape composition hypothesis states habitat quantity, measured at spatial scales larger than the individual territory, would influence species abundance. Using the National Land Cover Database, we extracted the metrics of landscape fragmentation and composition at turkey annual dispersal (i.e., a buffer of 16-km radius) and annual home range (i.e., a buffer of 2.15-km radius) scales, respectively, for 28 wildlife management areas across Mississippi, USA. We used the mean annual num- ber of harvested adult males per hunter day during annual spring hunting seasons (mid-March to mid-April) from 2001 to 2005 as a relative abundance index. Male Wild Turkey relative abundance increased with increasing fragmentation of hardwood forest (i.e., deciduous trees as the dominant form of vegetation) at both annual home range and dispersal scales, contrary to the prediction of the landscape fragmentation hypothesis. Male Wild Turkey abundance also increased with increasing landscape interspersion and juxtaposition index (i.e., an index of landscape spatial configu- ration) at the annual dispersal scale. We found nonlinear relationships between the proportion of hardwood forest and male turkey abundance, supporting the landscape composition hypothesis. Male turkey relative abundance peaked at 29% hardwood forest within the annual dispersal distance. Diverse landscapes are beneficial to Wild Turkeys at large spatial scales. Key words: abundance, landscape composition, landscape fragmentation, nonlinear response, Meleagris gallopavo silvestris , population, management, hardwood forest Received — Dec. 2016, accepted — Dec. 2017 ACTA ORNITHOLOGICA Vol. 52 (2017) No. 2 INTRODUCTION The effects of habitat distribution and fragmenta- tion on species have drawn increasing attention from ecologists and wildlife managers as land- scapes have rapidly changed worldwide (Fahrig 2003, Ewers & Didham 2006, Lande et al. 2014). Landscape structure and heterogeneity shape avian community structure and population dynamics substantially, creating the landscape- abundance relationship (Pimm et al. 1995, Boulinier et al. 1998, Villard et al. 1999, Lande et al. 2014). For instance, Betts et al. (2006) found that the composition (e.g., amount and proportion of habitat or land cover) of landscape affected occur- rence of Ovenbirds Seiurus aurocapilla and Blackburnian Warblers Dendroica fusca on a spatial scale larger than individual home ranges. Likewise, proportion of forest cover was a good predictor of the occurrence of Cantabrian Capercaillie Tetrao urogallus cantabricus in the Cantabrian Mountains of Spain (Quevedo et al. Downloaded From: https://bioone.org/journals/Acta-Ornithologica on 30 Jan 2025 Terms of Use: https://bioone.org/terms-of-use Access provided by University of Florida 2006). Patch shape and isolation were the best pre- dictors of the effect of landscape fragmentation on forest bird abundance in southern Chile (Magrach et al. 2011). Landscape-scale habitat evaluations can be beneficial to understanding animal-habitat relationships and wildlife habitat management (Glennon & Porter 1999, Goetz & Porter 2005). Habitat loss and fragmentation is currently one of the greatest threats to biodiversity (Pimm et al. 1995, Ewers & Didham 2006). Habitat loss reduces the amount and composition of habitat, whereas habitat fragmentation alters the spatial configura- tion and size of habitat patches (Fahrig 2003, Sauder & Rachlow 2014). Relative importance of habitat loss and fragmentation is critical for understanding the roles of landscape in shaping animal population dynamics. The effects of habi- tat loss and fragmentation are often confounding and habitat fragmentation may occur during habi- tat loss (Fahrig 2003). Habitat fragmentation may possibly take place in different times as habitat loss (Mazgajski et al. 2010). Therefore, landscape composition and spatial configuration (e.g., patch juxtaposition and isolation) are two major factors concurrently influencing habitat suitability and population dynamics of wildlife (Ewers & Didham 2006, Fahrig et al. 2011). The landscape composition hypothesis posits that habitat quantity on spatial scales larger than the individual territory would influence species abundance (Betts et al. 2006). For example, Lande et al. (2014) found positive relationships between adult density and amount of habitat of Black Grouse Tetrao tetrix and Capercaillie Tetrao urogal- lus in Norway. These two species depend on dif- ferent forest types throughout their life cycles, fur- ther emphasizing that habitat amount and diver- sity are important for population abundance of habitat generalists (Fahrig 2003, Ewers & Didham 2006). The landscape fragmentation hypothesis sug- gests that increases in habitat fragmentation and edge amount would reduce wildlife population abundance linearly due to increases in predation risk, reduction in food, and (or) decreases in dis- persal and ecological connectivity, independent of the effects of habitat loss (Betts et al. 2006). Detrimental edge effects on avian demographics may result in declines in population size (Lampila et al. 2005). Nevertheless, empirical evidence shows that habitat fragmentation may increase or decrease animal population abundance, partially dependent on life history strategies (Fahrig 2003, Ewers & Didham 2006). 128 A. Davis et al. Large-bodied birds, in particular, can be highly sensitive to habitat loss and fragmentation (Boulinier et al. 1998, Thornton et al. 2012). Thornton et al. (2012) found that Ocellated Turkeys Meleagris ocellata in Guatemala were often rare or absent in remnant forest patches, and were more vulnerable to habitat loss. The Eastern Wild Turkey ( M. gallopavo silvestris ; hereafter, Wild Turkey), the largest galliform native to North America, has experienced landscape fragmenta- tion and habitat loss across much of its range (Tapley et al. 2005). Although Wild Turkeys use a variety of habitats (e.g., forest, agriculture, and shrubs), they are primarily forest dwellers (Porter 1992, Boulinier et al. 1998, Rioux et al. 2009). Therefore, forest fragmentation may affect Wild Turkey dispersal and survival substantially (Boulinier et al. 1998, Thogmartin & Schaeffer 2000). Landscape fragmentation metrics, such as patch density, edge density, and patch isolation (e.g., distances to nearest patch), could potentially be related to Wild Turkey population abundance (Glennon & Porter 1999, Rioux et al. 2009). Male Wild Turkey abundance had a quadratic relation- ship (i.e., increasing, peaking and then declining) with the amount of forests in the northern part of the species’ range (Rioux et al. 2009). Recently, Bowling et al. (2016) found that the productivity of Wild Turkeys was related to interposition and juxtaposition index (IJI) of landscapes in a quad- ratic manner in New York, USA. Abundance of the species in the southeastern United States (US) has declined since the 1990s (Byrne et al. 2016); how- ever, the role of the changing landscape (e.g., habitat distribution and landscape fragmentation) on this decline is unclear. Our objective was to determine the landscape- abundance relationship of male Wild Turkeys in the state of Mississippi, USA. We tested the land- scape composition hypothesis, which predicts that habitat quantity and composition on spatial scales larger than the individual territory influ- ence male Wild Turkey abundance (Betts et al. 2006) and the landscape fragmentation hypothe- sis, which predicts that increases in habitat frag- mentation would reduce male Wild Turkey popu- lations linearly. The landscape composition hypothesis also predicts nonlinear relationships between lanscape composition and animal abun- dance (Betts et al. 2006). Because of the Wild Turkeys affinity for forests, particularly hardwood forests i.e., deciduous trees as the dominant form of vegetation, we predicted that abundance of male Wild Turkeys would increase with increasing Downloaded From: https://bioone.org/journals/Acta-Ornithologica on 30 Jan 2025 Terms of Use: https://bioone.org/terms-of-use Access provided by University of Florida forest. The loblolly-shortleaf pine systems con- tribute more than 50% of the commercial trees produced in this forest type. The oak-pine mixed forests consist of oaks, hickories, Sweetgum, Blackgum Nyssa sylvatica , Loblolly Pines, and Shortleaf Pines. The Southwest region vegetation consists of primarily oak-hickory forests (60%) along the loess hills bordering the Mississippi River flood- plain. A narrow oak-pine forest adjacent to oak- hickory forests in the loess hills transitions to loblolly-shortleaf pine forests. Vegetation in the Southeast region is dominat- ed by longleaf and slash pine forests (59%), including the pine flat woods on the coast and the rolling pine hills in the remaining part of the region. Longleaf Pines Pinus palustris and Slash Pines P. elliottii make up the majority of the forest community. The remainder of the forest contains dogwood Cornus spp., beech, Magnolia Magnolia grandiflora , and Sweetbay Magnolia M. virginiana along water, and Blackjack Oak Quercus marilan- dica , Post Oak Q. stellate , and Southern Red Oak ( Q. falcata ) in the uplands. We used 28 wildlife management areas (WMAs) across Mississippi that had four or more years of Wild Turkey harvest records, from 2001 to 2005 (Fig. 1). The 28 WMAs ranged from 73.5 km 2 to 922 km 2 in size and were located within the five Wild Turkey management regions: Northeast (n = 6), Delta (5), East-Central (6), Southwest (3), and Southeast (8) regions (Table 1, Fig. 1). Spring Wild Turkey hunting seasons of Mississippi lasted from mid-March to May 1, about 33 days. Each turkey hunter was required to purchase a hunting permit from MDWFP with a bag limit of three adult males. It was mandatory that turkey hunters completed and returned permit cards to self-serv- ice stations, often located at the entrance of a WMA, to report the number of adult males har- vested (Lint et al. 1995). Each hunter trip afield was considered a hunter-day. Relative Abundance Index of adult male Wild Turkeys Mississippi Department of Wildlife, Fisheries and Parks maintained the database of annual spring Wild Turkey hunting statistics, including the total number of adult males harvested by hunters and the total number of hunter days, for each WMA having spring Wild Turkey seasons. The number of harvested adult male Wild Turkeys per hunter day during spring hunting season, a capture per unit effort (CPUE), provides a reasonable relative Landscape-wild turkey abundance relationships 129 the percentage of hardwood forest, percentage of forest cover, and (or) landscape heterogeneity, but would decrease with increasing patch density (PD), edge density (ED), and Euclidean nearest neighbor (ENN) of forest patches. Habitat amounts may affect animal abundance nonlinearly (Fahrig 2003, Betts et al. 2006, Rioux et al. 2009); thus, we also tested whether the abundance of Wild Turkeys was nonlinearly related to percentage of hardwood forests. This work evaluated the effects of heterogeneous landscapes on Wild Turkey abundance on a large spatial scale to provide insight into Wild Turkey habitat management for maintaining or increasing harvestable Wild Turkey population abundance. METHODS Study area The Mississippi Department of Wildlife, Fisheries and Parks (MDWFP) established five Wild Turkey management regions statewide, based primarily on vegetation and physiographic regions: North- east, Delta, East-Central, Southwest, and South- east (Fig. 1). Vegetation of the Northeast region is primarily composed of oak-hickory and oak-pine forests (44%). The oak-hickory forest is dominated by oak trees Quercus spp. and hickories Carya spp., as well as American Beech Fagus grandifolia , Black Cherry Prunus serotina , and Yellow Poplar Liriodendron tulipifera . The oak-pine forest consists of about half upland oaks, mixed with Loblolly Pinus taeda and Shortleaf Pines P. echinata. The Delta region is covered primarily by oak- gum-cypress forests (19%) and agricultural lands. This area can be divided into two main areas, the batture land, which is the area west of the Mississippi River levee, and the interior floodplain and can be subjected to extensive flooding in the spring. The batture land vegetation consists of primarily Sugarberry Celtis laevigata , American Sycamore Platanus occidentalis , Eastern Cottonwood Populus deltoides , Sweetgum Liquidambar styraci- flua , elms Ulmus spp., and cypress Taxodium spp., whereas the interior Delta consists of extensive agricultural areas with fragmented interspersed hardwood forests comprising Overcup Oak Quercus lyrata , Nuttall Oak Q. nuttallii , Water Oak Q. nigra , Green Ash Fraxinus pennsylvanica , Sweetgum, and Eastern Cottonwood. Vegetation in the East-Central region is composed of two major forest types (60%), loblolly-shortleaf pine forest and oak-pine mixed Downloaded From: https://bioone.org/journals/Acta-Ornithologica on 30 Jan 2025 Terms of Use: https://bioone.org/terms-of-use Access provided by University of Florida 130 A. Davis et al. Fig. 1. Geographic locations and boundaries of 28 wildlife management areas in Mississippi, USA. The 28 wildlife management areas were distributed in the five Wild Turkey management regions (Delta, East-Central, Northeast, Northwest, Southeast, and Southwest) created by the Mississippi Department of Wildlife, Fisheries, and Parks. Wildlife Management Areas 1 - Leroy Percy 2 - O’Keefe 3 - Shipland 4 - Stoneville 5 - Sunflower 6 - Choctaw 7 - Bienville 8 - Caney Creek 9 - John W. Starr 10 - Okatibbee 11 - Tallahala 12 - Calhoun County 13 - Chickasaw 14 - Divide Section 15 - John Bell Williams 16 - Malmaison 17 - Upper Sardis 18 - Chickasawhay 19 - Leaf River 20 - Little Biloxi 27 - Pearl River 21 - Marion County 22 - Old River 23 - Pascagoula 24 - Red Creek 25 - Wolf River 26 - Copiah 28 - Sandy Creek 0 25 50 100 Kilometers Northeast East-Central Southwest Southeast 17 15 14 13 12 16 4 Delta 1 6 9 5 3 27 7 8 11 10 26 18 21 25 19 23 24 20 22 28 2 N Downloaded From: https://bioone.org/journals/Acta-Ornithologica on 30 Jan 2025 Terms of Use: https://bioone.org/terms-of-use Access provided by University of Florida Landscape-wild turkey abundance relationships 131 abundance index (RAI) of Wild Turkeys in central Mississippi (Lint et al. 1995). The CPUE-based RAI was correlated with capture-recapture estimates over six years (correlation coefficient r = 0.66, p = 0.05, n = 6) (Lint et al. 1995). Therefore, we calculated the RAI as the number of adult males harvested divided by the number of the total hunter days during the spring hunting season for each of the 28 WMAs. We used the average over five years to reduce the effects of annual variation in Wild Turkey harvests (Goetz & Porter 2005). Landscape variables We derived landscape variables using 2001 and 2006 National Land Cover Database (NLCD) satel- lite imagery classified by the Multi-Resolution Land Characteristics Consortium (http://www. mrlc.gov/). The 2001 NLCD was developed based on LandSat images from 2000, and the 2006 NLCD was based on LandSat images collected between 2000 and 2006, both at a 30-m spatial resolution (Homer et al. 2007, Fry et al. 2011). The land cover and land use (LCLU) layers were classified into 15 classes: open water, developed open space, devel- oped low intensity, developed medium intensity, developed high intensity, barren land, hardwood forest, pine forest, mixed forest, shrub/scrub, grassland/herbaceous, pasture/hay, cultivated crops, woody wetlands, and emergent herbaceous wet- lands (Homer et al. 2007, Fry et al. 2011). The 2001 NLCD overall classification accurancy for region 7, including Mississippi, was 74%, and the 2006 NLCD accuracy was 72% (Wickham et al. 2010, 2013). We derived landscape variables on two spatial scales: the annual dispersal distance and the annual home range of Wild Turkeys. Because the annual maximum dispersal distance of Wild Table 1. Area, management region, dominant vegetation type, and years of data for wildlife management areas (WMAs), Mississippi, USA. Region and WMA Area (ha) Vegetation Type Years of Data Delta Leroy Percy 7 350 Crop 2000–2005 O'Keefe 25 780 Crop 2000–2005 Shipland 19 650 Crop 2000–2005 Stoneville 10 660 Crop 2000–2004 Sunflower 25 220 Crop 2000–2005 East-Central Choctaw 86 910 Hardwood forest 2000–2005 Bienville 10 830 Pine forest 2000–2005 Caney Creek 11 510 Pine forest 2000–2005 John W. Starr 33 460 Woody wetland 2000–2005 Okatibbee 26 950 Hardwood forest 2000–2005 Tallahala 11 140 Pine forest 2000–2005 Northeast Calhoun County 35 510 Hardwood forest 2000–-2005 Chickasaw 10 840 Pasture 2000–2005 Divide Section 65 340 Hardwood forest 2000–2005 John Bell Williams 10 730 Hardwood forest 2000–2005 Malmaison 34 540 Crop 2000–2005 Upper Sardis 19 320 Hardwood forest 2000–2005 Southeast Chickasawhay 11 750 Pine forest 2000–2005 Leaf River 16 760 Pine forest 2000–2005 Little Biloxi 58 340 Pine forest 2000–2005 Marion County 29 710 Pine forest 2000–2005 Old River 59 780 Woody wetland 2000–2005 Pascagoula 15 190 Woody wetland 2000–2005 Red Creek 92 200 Pine forest 2000–2005 Wolf River 41 690 Pine forest 2000–2005 Southwest Copiah 26 670 Pine forest 2000–2004 Pearl River 28 220 Pasture 2001–2005 Sandy Creek 68 730 Hardwood forest 2000–2005 Downloaded From: https://bioone.org/journals/Acta-Ornithologica on 30 Jan 2025 Terms of Use: https://bioone.org/terms-of-use Access provided by University of Florida Turkeys is about 16 km (Porter 1977, Hopkins et al. 1982), we created a 16-km circular buffer centered on the centroid of each WMA using ArcGIS v10.2, a geographical information system (GIS, Environmental Systems Research Institute, Redlands, California, USA). We also created five 2.15-km circular buffers for each WMA, each being equivalent to the average annual home range size for Wild Turkeys in Mississippi (1,460 ha) (Chamberlain 1995, Godwin et al. 1995, Miller & Conner 2007, Marable et al. 2012). The 2.15-km buffers were placed on the centroid and the mid- dle points of each cardinal direction of the 16-km buffer (Fig. 2). Using the clip tool in ArcGIS, we clipped the NLCD raster images to one 16-km and five 2.15-km buffers for each WMA. We used FRAGSTATS 4.2 to calculate class- and landscape-level land cover metrics (McGarigal et al. 2012). Class level metrics measured the frag- mentation and spatial configuration of all forests (i.e., pine forest, hardwood forest, pine-hardwood mixed forest, and woody wetland combined) or hardwood forests (i.e., hardwood forest and woody wetland combined). Wild Turkeys are habitat generalists, requiring different land covers throughout their annual life cycle. Thus, land- scape-level metrics were calculated with all 15 LCLU classes to measure the overall fragmenta- tion, diversity, and heterogeneity of the landscape (McGarigal et al. 2001). For each WMA, we com- puted patch density (PD), edge density (ED), and Euclidean nearest neighbor distance (ENN) to measure the level of fragmentation for hardwood forests, forests (i.e., hardwood, pine, and hard- wood-pine forests combined), and landscapes (i.e., all land cover types combined, Table 2). We did not include any boundary/background as edge (using the default setting: 0%). We used the interspersion and juxtaposition index (IJI) to measure spatial configuration of forests and landscapes (Table 2). The IJI measures percent of the observed interspersion (or adjacency to one another) of land cover types over the maximum possible interspersion for the given number of patch types within a boundary (Table 2). Greater values of landscape IJI represent more even spatial configuration or arrangement of different patch types in a landscape. Glennon & Porter (1999) found that Wild Turkey relative abundance was related to PD, ED and IJI. We used the percentage of forests and hardwood forests (PLAND) to measure the composition and relative amount of habitat because Wild Turkeys primarily select hardwood or forest covers in Mississippi (Miller et al. 1999, Chamberlain & Leopold 2000, McKinney 2013). We calculated the mean of each 132 A. Davis et al. Fig. 2. The outline of landscape analyses. Inner circles — five home range (2.15-km radius) scale buffers for a wildlife management area (WMA). The outer circle is a 16-km buffer representing the annual dispersal radius of Wild Turkeys. The polygon is the boundary of a WMA. Table 2. Landscape variables included in the analysis of landscape-abundance relationships of Wild Turkeys. * — Maximum interspersion and juxtaposition is 100 when all patch types are equally adjacent to all other patch types. Variable name Initial Definition Patch density PD The number of patches of a landcover type or all land cover types per 100 ha Edge density ED The sum of the lengths (m) of all edge segments per ha of a land cover type (e.g., hardwood or forest) or all types Euclidean nearest neighbor distance ENN Mean distance (m) to the nearest neighboring patch of the same type, based on shortest edge-to-edge distance Interspersion and juxtaposition index IJI Percent of the observed interspersion over the maximum possible inter- spersion* for the given number of patch types within a boundary Percentage of land cover type PLAND Percentage (%) of a land cover type within a 16-km buffer or within a 2.15-km buffer Downloaded From: https://bioone.org/journals/Acta-Ornithologica on 30 Jan 2025 Terms of Use: https://bioone.org/terms-of-use Access provided by University of Florida Landscape-wild turkey abundance relationships 133 metric over five 2.15-km buffers as the home range scale landscape metrics for each WMA. We averaged landscape metrics over NLCD 2001 and NLCD 2006 by WMA in the analysis of landscape-abundance relationships of Wild Turkeys. Statistical analysis We built linear models to examine relationships between Wild Turkey RAI and landscape metrics. Many landscape metrics generated by FRAGSTATS are highly correlated with each other (Fahrig 2003). To remove multicollinearity between cov- ariates, two variables with a Pearson’s product- moment correlation coefficient | r | > 0.70 were not included simultaneously in a model. We used a two-stage model selection approach. At the first stage, we built models for evaluating the effects of landscape fragmentation, land- scape configuration, and landscape composition, respectively. We used a single-covariate LM for the effects of landscape configuration with IJI. We incorporated ED, ENN, and PD to quantify the effects of landscape fragmentation. We built single-covariate LMs for each of the landscape metrics, and then used forward stepwise varia- ble selection to build complex LMs with two or more uncorrelated covariates for the landscape fragmentation model. We also used forward stepwise selection for the landscape composition models with hardwood PLAND and total forest PLAND. We used the R 2 value to measure the relative importance of landscape fragmentation, configuration, and composition. At the second stage, we used forward stepwise selection to build complex LMs to incorporate landscape fragmentation, configuration, and composition. The entry level of covariates in the forward stepwise selection was 0.1. The two stage model selection was repeated for the home-range level and annual dispersal level, respectively, because the landscape composition hypothesis predicts Wild Turkey abundance would be related to land- scape variables beyond home range levels. The landscape composition hypothesis pre- dicts a nonlinear relationship between percentage of forests and Wild Turkey abundance. Thus, we built generalized additive models (GAMs) to test for nonlinear landscape-abundance relationships using the R package mgcv (Hastie & Tibshirani 1986, Wood & Augustin 2002). We established sta- tistical significance at α = 0.05 and performed all statistical analyses in the R 3.0.2 environment (R Development Core Team 2013). RESULTS Factors affecting male turkey abundance in dis- persal distance scale Means of annual hunter days over 2001 to 2004 were not related to areas of WMAs (t = -0.21, df = 26, p = 0.83), nor were mean CPUEs related to WMA areas (t = -1.12, df = 26, p = 0.27). At the annual dispersal distance, with all land cover classes combined (i.e., the landscape level), PD, ED, ENN and IJI were highly correlated with one another (| r | > 0.7). The model including IJI was the only significant model (coefficient = 0.0008, 95% confidence interval [CI] = [0.00004, 0.0016]; p = 0.04, R 2 = 0.15; Fig. 3, Table 3). At the forest class level, the model with an IJI term was the only significant model (coefficient 0.0008, 95% CI = [0.0003, 0.0013]; p ≤ 0.01, R 2 = 0.39; Table 4). Relative abundance was not related to the per- centage of forests (p = 0.36, Table 4). With only hardwood forests considered, Wild Turkey rela- tive abundance was positively related to ED (coefficient = 0.005, p= 0.01, R 2 = 0.25) and PD (coefficient = 0.005, p = 0.047, R 2 = 0.14, Fig. 4) but not the ENN between hardwood patches (p = 0.12, Table 5). Neither ED (t = 1.18, df= 25, p = 0.25) nor PLAND (t = 1.46, df =25, p = 0.16) was significant when the two variables were included in the same LM. The hardwood forest PLAND had a significant nonlinear relationship with Wild Turkey relative abundance in GAMs, Fig. 3. Positive relationships between wild turkey catch per unit effort (number of harvested adult males per hunter day) and landscape interspersion and juxtaposition index (IJI) at annual dispersal distance scale of 28 wildlife management areas, Mississippi, USA. Interspersion and juxtaposition index Number of harvested birds/hunter day 40 50 60 70 0.02 0.04 0.06 0.08 Downloaded From: https://bioone.org/journals/Acta-Ornithologica on 30 Jan 2025 Terms of Use: https://bioone.org/terms-of-use Access provided by University of Florida with turkey relative abundance peaking at an optimal hardwood forest proportion of 0.29 (F = 3.98, edf = 2.24, p = 0.02, Fig. 5a). Factors affecting male turkey abundance at home range scale At the annual home-range scale, no relationships were found between landscape metrics and rela- tive abundance index with all 15 land cover and land use layers being included. At the forest class level, only the model including IJI was significant (p = 0.02, R 2 = 0.21). Wild Turkey relative abun- dance was positively related to either hardwood forest PLAND (p = 0.01) and PD (p = 0.01, R 2 = 0.40) or hardwood forest PLAND (p = 0.02) and ED (p ≤ 0.01, R 2 = 0.46). The percentage of hardwood forest had a nonlinear relationship with Wild Turkey abundance at the home range scale in GAMs, with turkey relative abundance peaking at an optimal hardwood forest propor- tion of 0.39 (F = 3.82, edf = 3.16, p = 0.01, Fig. 5b). We did not have any LMs with > 1 significant covariate of coefficient p values < 0.05 at the dis- persal distance and home-range levels in this study. DISCUSSION Landscape spatial configuration is an important factor influencing the distribution and relative abundance of animals, including forest birds (Andren 1994, Lampila et al. 2005, Fahrig 2013, Villard & Metzger 2014). Our findings demon- strate that Wild Turkey relative abundance was affected by landscape fragmentation on both large and small spatial scales. Heterogeneous land- scapes with diverse forest covers in close proximi- ty enhanced Wild Turkey relative abundance. Additionally, Wild Turkey relative abundance var- ied with increasing hardwood forest cover on the landscape, supporting the landscape composition hypothesis. The occurrence and persistence of the Black Grouse were affected by the landscape com- position from an area greater than an individual Black Grouse home range (Ludwig et al. 2009). Conversely, our findings were inconsistent with the prediction of the landscape fragmentation hypothesis; we found a positive relationship between the relative abundance and hardwood edge density at large and small spatial scales. Our results suggest that landscape composition and fragmentation jointly influence Wild Turkey abun- dance (Rioux et al. 2009). Landscape heterogeneity influences animal species diversity and abundance (MacArthur 1964, Fahrig et al. 2011). Relative male Wild Turkey abundance was related to the IJI of forest covers at home range scales. Bowling et al. (2016) found that habitat selection during brooding in Wild Turkeys was related to landscape IJI at the 1-km scale but not at the 20-km. Female Wild Turkeys 134 A. Davis et al. Table 3. Linear models of landscape-Wild Turkey abundance relationships on landscape levels at two spatial scales on 28 wildlife management areas, Mississippi, USA. Each row was for a single-covariate model. Variable initials in the column Model were defined in Table 2. Model Annual dispersal distance scale Annual home range scale R 2 Coefficient p-value R 2 Coefficient p-value PD 0.04 0.001 0.302 0.03 0.001 0.417 ED 0.02 0.000 0.482 0.01 0.000 0.607 ENN 0.02 0.000 0.436 0.02 0.000 0.542 IJI 0.15 0.0008 0.040 0.11 0.001 0.087 Table 4. Linear models of forest class landscape-Wild Turkey abundance relationships at two spatial scales on 28 wildlife manage- ment areas in Mississippi, USA. Each row was for a single-covariate model. Variable initials in the column Model were defined in Table 2. Model Annual dispersal distance scale Annual home range scale R 2 Coefficient p-value R 2 Coefficient p-value PD 0.07 0.001 0.181 0.04 0.001 0.313 ED 0.02 0.000 0.481 0.01 0.000 0.556 ENN 0.02 0.000 0.440 0.01 0.000 0.555 IJI 0.39 0.0008 0.0005 0.21 0.0006 0.015 PLAND 0.03 0.000 0.355 0.02 0.000 0.531 Downloaded From: https://bioone.org/journals/Acta-Ornithologica on 30 Jan 2025 Terms of Use: https://bioone.org/terms-of-use Access provided by University of Florida exhibited seasonal variation in selection of differ- ent types of forests in burned Longleaf Pine Savanna in Georgia, USA (Little et al. 2016). Greater values of landscape IJI represent more even interspersion of patch types in a landscape (McGarigal et al. 2012). Positive correlation between the landscape (or forest) IJI and Wild Turkey relative abundance is consistent with known habits of Wild Turkeys whereby they require different land covers throughout their annual life cycle (Dunning et al. 1992, Porter 1992). Wild Turkeys use uplands or dense shrubs for nesting (Seiss et al. 1990, Miller 1997); forests, fields, thinned or burned pine stands, or stream- side ripirians for brooding (Phalen 1986, Godwin et al. 1992, Miller & Conner 2005); and a variety of habitat types such as upland forests, intensively managed pine forests, bottomland forests, and pastures and agricultural fields during non- nesting seasons (Godwin et al. 1992, Miller et al. 1999, Chamberlain et al. 2000, McKinney 2013). Therefore, interspersion of diverse land covers Landscape-wild turkey abundance relationships 135 within the annual dispersal range may be likely critical in meeting the demands of the annual cycle. The relative amount of hardwood forests appeared to affect the relative abundance of Wild Turkeys in a nonlinear manner. Turkey abundance plateaus at around 29% hardwood abundance. Hardwoods may be a limiting factor to turkey abundance, until they reach about 29% coverage within the landscape, and then beyond that turkey abundance does not increase with increas- ing hardwood coverage (Fig. 5). Rioux et al. (2009) also found male turkey density peaked at the intermediate amount of forests (0.2–0.5) after a mild winter. However, male turkey density had a positive relationship with forest cover after a severe winter. Translocated Wild Turkeys in north- central Mississippi and northwestern Louisiana selected hardwood forest at the home range level (McKinney 2013, Cohen et al. 2016). Therefore, landscapes of an intermediate-level of hardwood forests intermixed with other land covers in close Fig. 4. Positive relationships between Wild Turkey catch per unit effort (number of harvested adult males per hunter day) and hardwood forest patch density (A) and hardwood forest edge density (B) at annual dispersal distance scale for 28 wildlife management areas, Mississippi, USA. Number of harvested birds per hunter day Number of harvested birds per hunter day 0.02 0.04 0.06 0.08 1 2 3 4 5 2 6 0 4 8 10 A B Patch density (number/100 ha) Edge density (m/ha) Table 5. Linear models of hardwood class landscape Wild Turkey abundance relationships at two spatial scales on 28 wildlife management areas in Mississippi, USA. Each row was for a single-covariate model. Variable initials in the column Model were defined in Table 2. Model Annual dispersal distance scale Annual home range scale R 2 Coefficient value p-value R 2 Coefficient value p-value PD 0.14 0.005 0.047 0.10 0.003 0.116 ED 0.25 0.003 0.008 0.33 0.003 0.001 ENN 0.09 -0.002 0.120 0.01 0.000 0.702 PLAND 0.27 0.001 0.005 0.23 0.0009 0.011 0.02 0.04 0.06 0.08 Downloaded From: https://bioone.org/journals/Acta-Ornithologica on 30 Jan 2025 Terms of Use: https://bioone.org/terms-of-use Access provided by University of Florida proximity may benefit Wild Turkeys as a habitat generalist species. Types of habitats comprising a landscape or their fragmentation may impose either positive or negative impacts on animal abundance (Pimm & Askins 1995, Fahrig 2003, Ewers & Didham 2006). Wild Turkey abundance was positively related to hardwood forest fragmentation but was unrelated to overall landscape or forest fragmentation. The positive correlation we observed between Wild Turkey abundance and hardwood forest fragmen- tation is contrary to the prediction of the land- scape fragmentation hypothesis. Glennon & Porter (1999) also found Wild Turkey abundance was positively related to habitat fragmentation. Nearly 50% of earlier studies of landscape-animal abun- dance relationships have found positive effects of habitat fragmentation on animal abundance (Fahrig 2003). Animals with different life history traits (e.g., high vs. low mobility and habitat generalist vs. specialist) may exhibit different responses (positive or negative) to habitat fragmentation. Wild Turkeys are mobile habitat generalists (Marable et al. 2012, McKinney 2013). Habitat fragmentation or increases in edge density within annual dispersal ranges does not likely impede dispersal. In New York, Fleming & Porter (2005) suggest that habitat edges may facilitate dispersal of Wild Turkeys. We used the CPUE index to measure the rela- tive abundance of male Wild Turkeys. Lint et al. (1995) found that CPUE indices of male turkeys were significantly correlated with the capture- recapture estimates of Wild Turkey population sizes in central Mississippi. The relationship be - tween hardwood proportion and male turkey abundance of our study was similar to the rela- tionship between forest proportion and Wild Turkey density estimated using repeated surveys in Canada (Rioux et al. 2009). Harvest data (e.g., number of animals harvested by hunters) have been used to investigate wildlife population ecol- ogy for a long time (Elton & Nicholson 1942, Williams 1985, Henden et al. 2009). The main caveat of the use of harvest data is that harvest statistics as relative abundance indices may not be representative of population abundance and trends, and may bias estimation of population trends (Ranta et al. 2008, Simard et al. 2012). How- ever, comparisons between bag size (or CPUE) and independent population count are inconsis- tent between species and between studies. Bag size is a reasonable relative abundance index of Red Grouse Lagopus lagopus scoticus populations, being positively related to count and of variability indistinguishable with that of count (Cattadori et al. 2003). On the contrary, CPUE did not represent the abundance and population variability of Willow Grouse Lagopus lagopus (Willebrand et al. 2011). Therefore, the usefulness of bag size and CPUE as an index of relative abundance should be species or even region dependent. Wild Turkeys demonstrated scale-dependent or hierarchical habitat selection (Little et al. 2016, Pollentier et al. 2017). Likewise, Graf et al. (2007) found that landscape variables of multiple spatial scales are needed to study Capercaillie-habitat relationships. Our findings have important 136 A. Davis et al. 10 20 30 40 50 Partial residual Percentage of hardwood forest Partial residual Percentage of hardwood forest 10 20 30 45 25 15 5 -0.04 0.00 0.02 0.04 Fig. 5. Nonlinear relationship between Wild Turkey catch per unit effort (number of harvested adult males per hunter day) and proportion of hardwood forest at the annual dispersal distance scale (A) and at the annual home range scale (B) for 28 wildlife management areas, Mississippi, USA. Partial residual is the contribution of a covariate to model fit at a value of the covariate.