Landscape context affects use of restored grasslands by mammals in a dynamic agroecosystem.
Habitat loss due to agriculture is one of the leading causes of endangerment for terrestrial vertebrates (Czech el al, 2000; Kerr and Cihlar, 2004). In the Midwest region of the U.S.A., the landscape is now dominated by intensive row crop agriculture (Mankin and Warner, 1997). For instance, 99.9% of tallgrass prairie in Illinois has been lost due to this land conversion (Samson and Knopf, 1994). Programs exist to ameliorate this habitat loss by providing financial incentives for landowners to create habitat for native species. For example, the Conservation Reserve Program (CRP) and the newer State Acres for Wildlife Enhancement (SAFE) have created grasslands on former croplands by encouraging farmers to enroll for 10-15-y intervals. Due to the dominance of agriculture and nature of voluntary enrollment, the landscape is a patchwork of habitat embedded in a matrix of cropland.
Mammals are ecologically important organisms in grassland ecosystems because of their significant roles in trophic interactions, both as predators and prey. Herbivores also can influence vegetative communities through their foraging (Olofsson et al., 2008; Batzli and Dejaco, 2013) and may affect restoration trajectories. Some mammals are also valued by the public as game species. The agricultural matrix could affect use of restored grasslands by mammals in two main ways. First, crop fields could serve as supplemental habitats providing food (Dunning et al, 1992), especially during the growing season (Nixon et al, 1991; Beasley et al., 2006). Second, the agricultural matrix could provide cover and affect movements (Grovenburg et al., 2011) to restored grasslands from other landscape elements (e.g., forests). This connectivity provided by agricultural fields is temporally dynamic due to the annual planting and harvesting of crops (Cosentino et al., 2011).
Other landscape variables also may affect use of restored grasslands by mammals. Some restoration sites may be too small to support populations of species dependent on grassland habitat, and their occurrence may require movements among grassland patches. Use of restored grasslands by other mammal species may be contingent on proximity to critical habitats such as forest and water, and the ability of the species to move through the agricultural matrix. Hence, landscape context should be considered when evaluating programs like CRP and SAFE.
However, landscape context is rarely included in assessments of ecological restoration (but see Mulligan et al., 2013; Cosentino et al, 2014). For instance, in Brudvig's (2011) extensive review of restoration studies, which included plants and animals, only 10% of the studies investigated how landscape factors (e.g., surrounding landscape, connectivity) affected site-level biodiversity, whereas 78% of the studies tested how site-level factors affected biodiversity.
Despite the integral role of medium to large mammals in grasslands, past monitoring of ecological restorations has focused on small mammals (Stone, 2007; Mulligan et al., 2013; DeGolier et al., 2015) because population-scale processes can occur within the typical size of restored grasslands, and small mammals are easier to sample via live trapping. Now, noninvasive survey methods such as camera trapping provide an effective way to monitor habitat use of larger mammals (Rowcliffe and Carbone, 2008; Long et al., 2011; Burton et al., 2015). Data from camera traps can be integrated with occupancy modeling to estimate use of sites by species despite imperfect detection (Cove et al., 2012; Kalle et al, 2014; Robinson et al., 2014).
We employed camera trapping to evaluate how medium and large mammals used restored grasslands during summer and winter in a dynamic agroecosystem in Illinois. Our focus was on effects of patch size, vegetation cover, and landscape context because these variables can be managed through selective enrollment when creating new grassland habitats. We frame our predictions regarding seasonal differences in grassland use in terms of crop harvesting and changes to the agricultural matrix because this alteration is so extensive (i.e., ~80% of the landscape becomes bare fields), but we consider alternative explanations for observed patterns. We documented all mammals using the grasslands and had adequate data to conduct statistical analysis for four species: raccoon (Procyon lotor), eastern cottontail (Sylvilagus floridanus), coyote (Canis latrans), and white-tailed deer (Odocoileus virginianus).
Raccoons are generalist mesocarnivores and important nest predators (Heske et al., 1999; Rodewald and Kearns, 2011; Friesen et al., 2013) including of grassland birds (Renfrew and Ribic, 2003; Lyons et al., 2015). Hence, factors that affect use of restored grasslands by raccoons could be consequential to nest predation rates for declining grassland bird species (Schmidt, 2003). Lagomorphs such as eastern cottontails are herbivores that can affect vegetation composition due to preferential foraging (Crawley, 1990; Olofsson et al., 2008). Eastern cottontails are a prey species for many predators including coyotes (Morey et al., 2007), and they prefer thick hiding cover (Althoff et al., 1997; Bock, 2006). Cottontails are also a game species that declined in Illinois coincident with intensification of agriculture (Mankin and Warner, 1999a). Coyotes have expanded their geographic range over time and are now the top predators in many regions (Gompper, 2002; Prugh et al., 2009). Coyotes can displace smaller canids such as red foxes (Vulpes vulpes; Gosselink et al., 2003) due to intraguild predation (Gosselink et al., 2006; Robinson et al., 2014). The white-tailed deer is the largest herbivore in Midwestern grassland ecosystems where it can affect plant species composition and the rate of succession (Batzli and Dejaco, 2013) while being an economically important game species (Illinois Department of Natural Resources, 2015).
We tested the following predictions for our four focal species (Table 1): (1) Raccoons use diverse habitats but are typically associated with forest (Newbury and Nelson, 2007; Beasley et al, 2006, 2011), and water sources (Gehrt and Fritzell, 1998; Newbury and Nelson, 2007; Beasley and Rhodes, 2010). Therefore, we predicted use of grassland restorations by raccoons would be positively associated with their proximity to forest, ponds, and streams. (2) Because crops provide cover that should encourage movements across crop fields, we expected use of grasslands by raccoons to be greater during summer. (3) If areas with extensive grasslands promote movement through the agricultural matrix, especially when crops are absent, we expected greater use of sites near other grasslands by raccoons during winter. (4) Our cool season grasslands were often dominated by smooth brome (Bromus inermis) that provides denser cover than sites dominated by warm season grasses (Eggebo et al., 2003). If habitat selection by cottontails is driven by a preference for dense hiding cover that reduces predation risk, then there should be a positive association between the probability of site occupancy and dominance of cool season grasses. (5) We expect that cottontails perceive the open agricultural matrix as risky during winter, whereas restored grasslands cotild serve as refuges after crops are removed. If cottontails constrict their use of agricultural fields after crop harvest, occupancy probabilities on grasslands should increase from summer to winter. (6) Cottontails will select farmsteads during winter because of the protection provided by small buildings and sheds (Mankin and Warner, 1999b). Therefore, we predicted greater occupancy probabilities during winter for sites near human structures. (7) If coyotes seek patches where their lagomorph prey are abundant (Arias-Del Razo et al., 2012), then we expected a pattern of co-occurrence among sites for coyotes and cottontails and increased use of restored grasslands by coyotes in winter. (8) Coyotes generally avoid interactions with humans (Gosselink et al., 2003; Magle et al., 2014). Therefore, we expected a positive relationship between site occupancy for coyotes and distance to nearest human structure. (9) Because coyotes use forested habitat extensively for diurnal cover in agricultural landscapes (Atwood et al., 2004), we predicted grassland occupancy would be related positively to proximity of nearest forest. (10) White-tailed deer in the Midwest favor early successional upland forest (Nixon et al., 1991); therefore, we predicted that grassland use by deer should be positively related to proximity to nearest forest. (11) Because deer fawns are at risk of predation by coyotes and other predators (Nelson and Woolf, 1987), denser hiding cover should be preferred during summer when fawns are present. As a result, we expected higher use during summer for sites dominated by cool season grasses.
Our 30 study sites were grasslands located in central Illinois (within 148 km of Decatur; 39[degrees]50'25"N, 88[degrees]57'17"W) in the Grand Prairie and Southern Till Plain Natural Divisions (Fig. 1). These grasslands were a mixture of parcels in the State Acres for Wildlife Enhancement (SAFE) Program (see Mulligan et al., 2013) and the Conservation Reserve Program (CRP). The sites were representative of these habitat programs in the region in terms of variation in size (1-256 ha, median = 29 ha), vegetation cover, and landscape context. The sites were all established grasslands seeded [greater than or equal to] 3 y prior to sampling.
Central Illinois is dominated by intensive agricultural production (Mankin and Warner, 1997). For instance, about 80% of the land in the Grand Prairie Region is planted with annual row crops (corn or soybean; Schooley et al., 2012). Our grassland sites were embedded in this agricultural landscape (Fig. 1).
Our SAFE and CRP sites were seeded with mixes dominated by grasses we classified into three categories: cool season grasses (10 sites), warm season grasses (12 sites), or a mix (eight sites). We classified sites by visually estimating the dominant grass species (>50% cover) at each of four camera quadrants (see below) as either cool or warm season. If three or more quadrants were dominated by cool or warm season grasses, then the site received that designation. Otherwise, sites were considered as mixed. The common cool season species were smooth brome, Virginia wild rye (Elymus virginicus), and Canada wild rye (E. canadensis). The common warm season species were big bluestem (Andropogon gerardi), little bluestem (Schizachyrium scoparium), switchgrass (Panicum virgatum), side oats (Bouteloua curtipendula), eastern gamagrass (Tripsacum dactyloides), and Indian grass (Sorghastrum nutans).
CAMERA TRAP SAMPLING
We sampled mammals with camera traps on the 30 sites during two seasons, "summer" (Jun. 2014-0ct. 2014) and "winter" (Nov. 2014-Mar. 2015), which corresponded to the growing and nongrowing seasons for agricultural crops (Mankin and Warner, 1999b).
Within each season, we sampled each site for four continuous weeks (Magle et al., 2014). We could sample 10 sites at one time; therefore, sites were placed into three groups for sampling within a season based on logistical considerations.
We sampled each site with four trail cameras (Bushnell Trophy Cam, Model 119436c) after splitting sites into four quadrants and placing a camera near the center of each (Fig. 1). The cameras have a passive infrared motion sensor and were set to take bursts of three shots with a 10-s trigger delay after activation. The cameras were placed 150 cm above the ground by mounting them on fence posts. Cameras were pointed down toward a bait post (3.4 m away) at an angle of 45[degrees] to reduce false triggers due to moving vegetation in the background. We cleared vegetation from a 6.3-[m.sup.2] area centered on the bait post to reduce false triggers further. The bait post was placed north of the camera to avoid solar interference. Bait posts consisted of a punctured can of cat food (salmon flavor) and a fatty acid tablet (Wildlife Control Supplies) in wire mesh. The bait was intended to attract nearby mesocarnivores, but herbivores might have been curious about the strong scent. We stored photographs from the cameras according to the protocol by Harris et al. (2010) and categorized photographs following Sanderson and Harris (2013). The resulting output from each site included an occupancy matrix plus the number of independent photographs (>60 min apart) for our focal species.
We measured covariates that could be relevant either for detection or occupancy for the focal species (MacKenzie et al., 2006). Patch size and sampling effort were deemed potentially important detection covariates for all species. Patch size (ha) was measured as the continuous area of grassland delimited by maintained breaks, roads, or other habitat. Because each site had four cameras (one in each quadrant) regardless of its total area, larger sites might be less thoroughly sampled than smaller sites, and species might have a greater probability of being missed. Patch size was ln-transformed prior to analysis. Sampling effort equaled the number of days that cameras were active during a sampling season. The maximum was 28 d, but some cameras were active for less time due to failure (summer: mean = 25.98 d, se = 0.36; winter: mean = 24.25 d, se = 0.47). Although variation in sampling effort among sites was small, sites where cameras were operating for a greater number of days might have had a higher probability of detecting less common species.
Occupancy covariates included patch size, vegetation type, distance to nearest other grassland (measure of patch isolation), and distances to nearest forest, stream, pond, and human structure. Vegetation was an ordinal variable indicating the dominant vegetation at a site was (1) cool season grasses, (2) mixed, or (3) warm season grasses. The other occupancy covariates assessed landscape context and were measured as linear distances (m) from edges of focal grasslands to edges of other features. We measured these landscape covariates from digital orthophotos with ArcMap (ESRI, 2015) and Google Earth (ver. 184.108.40.2067). Distances to forest stands (0-1500 m, median = 92 m), stream (0-2260 m, median = 102 m), pond (02009 m, median = 193 m), and human structure (0-1139 m, median = 80 m) were ln-transformed. Human structures included buildings (minimum of 112 m2) with an associated footprint of mowed lawns and gravel areas (minimum footprint of 4500 m2). We did not measure distance to nearest agricultural field because there was minimal variation among sites (0 m for 27 of 30 sites). Because the covariates "patch size" and "distance to nearest grassland" exhibited multicollinearity (r = 0.66), we used Principal Components Analysis (PCA; PRINCOMP Procedure, SAS Institute Inc., 2013) to create orthogonal principal components. The first axis (Size_Iso) accounted for 83% of the variation and was positively correlated with patch size (r = 0.91) and distance to grassland (r = 0.91). Size_Iso represented a gradient from small, connected grasslands to large, isolated grasslands.
We also wanted to determine if presence of coyote (Predator) affected presence of eastern cottontail (Prey), and the reverse, beyond effects of other occupancy covariates. To do this, we used the conditional occupancy probability from the most supported occupancy model based on habitat and landscape covariates for each species (see next section).
Occupancy modeling is a widely used method for using presence-absence data to assess habitat relationships while accounting for false absences (MacKenzie et al., 2006; Duggan et al, 2011; Long et al, 2011). We conducted occupancy modeling for three mammal species with adequate data: raccoons, eastern cottontails, and coyotes (Appendix 1). We used single-season models in program PRESENCE 9.5 (Hines, 2006) to examine relationships separately for summer and winter. Multiseason models for evaluating turnover across seasons (MacKenzie et al., 2006) did not converge probably due to a moderate sample size (30 sites) and data sparseness. We used a design in which each of the four camera stations at a site was treated as a spatial replicate and sampling was conducted without replacement (Guillera-Arroita, 2011). Hence, there were four surveys per site during each season.
We used a two-step approach to evaluate occupancy models (Duggan et al., 2011; Cove et al., 2012; Robinson et al., 2014). First, we determined the most supported model for detection (p) while holding occupancy (psi) constant. The candidate set for detection included the intercept-only model [p(.)], p(Size), p(Effort), and p(Size, Effort). Second, after identifying the best detection model, we evaluated a candidate set of occupancy covariates to test our predictions (Table 1). Vegetation and Size_Iso were selected as occupancy covariates for all species. For raccoons, we also included distances to forest, pond, and stream. For cottontail rabbits, we also included Predator and distances to human structures. For coyotes, we also included Prey and distances to forest and human structures. These covariates combinations were limited to two to avoid overparameterization and increase the likelihood of model convergence. However, not all models converged (see Berry, 2016 for full candidate sets).
We evaluated models at both stages (detection and occupancy) using an information theoretic approach (Burnham and Anderson, 2002). Models with a AAIC value < 2.0 were considered as competitive. We also considered model fit of competitive models when making inferences to avoid support of uninformative variables (Arnold, 2010).
ACTIVITY OF WHITE-TAILED DEER
Because naive occupancy of white-tailed deer was high, especially in summer, occupancy modeling would be uninformative. Instead, we modeled activity of deer using the number of independent photographs (>60 min apart) as the response variable. We used a negative binomial model with a log link function and included sampling effort as an offset variable (GENMOD Procedure, SAS Institute Inc., 2013). Separate models were evaluated for each season using an information theoretic approach (Burnham and Anderson, 2002). Each candidate set included the intercept-only model plus models with Vegetation, Size_Iso, Forest, and Predator singly and in combinations with two predictors.
We evaluated two response variables to determine if overall use of the created grasslands by species differed between summer and winter. First, we used conditional occupancy probability for each site derived from the most supported model for each species. Second, we used photo- activity rate (DaVanon et al., 2016) that equaled the number of independent photos of each species divided by sampling effort for each site. We tested for differences in occupancy probability and photo-activity rate between seasons with the Wilcoxon Signed Rank Test (UNIVARIATE Procedure, SAS Institute Inc., 2013).
Eight species of mammals were encountered at varying frequencies (greatest to least): white-tailed deer, eastern cottontail, coyote, raccoon, striped skunk (Mephitis mephitis), Virginia opossum (Didelphis virginiana), American mink (Neovison vison), and long-tailed weasel (Mustela frenata) (Appendix 1). Of those, four species (skunk, opossum, mink, and weasel) were detected too infrequently for occupancy modeling. Despite cameras operating for a total of 5638 nights across the two seasons, we did not capture any photographs of red fox on our restored grassland sites.
For raccoons, the best detection model in summer was p(Size). A second model, p(Size, Effort), was competitive but did not substantially improve model fit (Table 2). Detection of raccoons decreased with patch size ([beta] = -1.048, SE = 0.464). Per-survey detection probabilities for sites averaged 0.26 (SE = 0.04). For winter, there was no strong support for any detection covariates for raccoons as the intercept-only model was ranked first (Table 3). The average per-survey detection probability for winter was 0.32 (SE = 0.12). During summer occupancy of raccoons was best explained by a model that included distance to nearest forest (Table 2). Raccoon occupancy was low when forested habitat was >400 m away (Fig. 2A). The second ranked model, which also included distance to nearest pond, did not improve model fit substantially (Table 2). During winter, there was less model uncertainty and raccoon occupancy was explained by Forest and Size_Iso (Table 3). Only grassland sites near forests had high occupancy probabilities (Fig. 2B), and occupancy was higher for smaller connected sites than for larger isolated sites (Fig. 2C).
For cottontails, the best model for detection during summer was p(Size), and the best model during winter was p(Size, Effort) (Tables 2, 3). Detection of cottontails was related negatively to patch size in summer ([beta] = -0.759, SE = 0.277) and winter ([beta] = -0.477, SE = 0.237). During winter, cottontail detection also was related positively to effort ([beta] = 0.113, SE = 0.061). Per-survey detection probabilities averaged 0.31 (SE = 0.03) for summer and 0.37 (se = 0.02) for winter. For occupancy the best model during summer was psi(Size_Iso) (Table 2). Site occupancy for cottontails was greater for larger isolated grasslands than for smaller connected grasslands (Fig. 2D). The intercept-only model was competitive, but the addition of Size_Iso improved model fit (Table 2). The other competitive models included Size_Iso plus one additional predictor that did not improve model fit (Table 2). For winter, the best occupancy model was psi(Vegetation). The intercept-only model was again competitive, but inclusion of Vegetation improved model fit (Table 3). Occupancy probability was 0.832 (se = 0.173) for sites dominated by cool season grasses, 0.643 (se = 0.156) for mixed sites, and 0.395 (se = 0.173) for sites dominated by warm season grasses.
For coyotes, we did not find support for any detection covariates; the top model for detection in both seasons was the intercept-only model (Tables 2, 3). The average per-survey detection probability was 0.12 (se = 0.08) for summer and 0.25 (SE = 0.10) for winter. For occupancy the best model during summer was psi (Human, Prey) (Table 2). Occupancy for coyotes was related negatively to distance to nearest human structure ([beta] = -1.34, SE = 0.833) and to conditional occupancy of cottontail rabbits ([beta] = -3.55, SE = 2.68) (Fig. 2E, F). For winter, the top model of coyote occupancy was the intercept-only model (Table 3). Other competitive models did not substantially improve model fit.
ACTIVITY OF WHITE-TAILED DEER
During summer activity of white-tailed deer based on photographic rate was best explained by a model that included distance to nearest forest (Table 4). A second-ranked model that also included Sizejso was competitive but did not substantially improve model fit. Activity of deer was related negatively to distance to forest ([beta] = --0.001, SE = 0.0003). During winter the intercept-only model was ranked first (Table 4). Two models that included either Forest or Predator (coyote) were competitive but did not substantially improve model fit.
Based on estimates of conditional site occupancy for our most supported models (Tables 2, 3), occupancy by raccoons changed between seasons (S = -155, P < 0.001, N = 30), decreasing from summer to winter (Fig. 3). However, site occupancy for cottontails (S = 48.5, P = 0.198, N = 30) and coyotes (S = -5.5, P = 0.892, N = 30) did not differ between seasons. Based on photograph rate, activity of raccoons (S =-12, P = 0.384, N = 30) and coyotes (S = 5, P = 0.821, N = 30) did not vary between seasons. In contrast activity of cottontails (S = 56, P = 0.013, N = 30) and white-tailed deer (5= 119, P = 0.012, N = 30) increased on restored grasslands during the winter (Fig. 3).
We detected eight native species of medium and large mammals using restored grasslands in Midwestern landscapes dominated by agriculture. Thus, grasslands created under the CRP and SAFE programs provide habitat for these species as well as for more intensively studied vertebrates such as small mammals (Mulligan et al, 2013) and birds (McCoy et al, 1999; Eggebo et al, 2003; Andrews et al., 2015). Grassland use differed between summer and winter for three of the four species examined in detail (raccoons, cottontails, deer), which suggests the dynamic nature of the agricultural matrix was consequential. Landscape context also had strong effects on restoration outcomes. Mulligan et al (2013) likewise found landscape context affected colonization rates of newly created grasslands by small mammals in the region.
Landscape context had clear influences on site occupancy by raccoons. The most supported models for raccoon occupancy in both seasons included distance to nearest forest. This result was expected as raccoons favor forested habitats (Beasley and Rhodes, 2010; Beasley et al., 2011), especially for denning (Henner et al, 2004). However, there were seasonal nuances to this main effect likely related to the harvesting of crops in the matrix. During summer, occupancy of grasslands by raccoons was higher, and occupancy probabilities remained moderate when sites were <400 m from forest (Fig. 2A). During winter, when the matrix was bare of crops, there was a much sharper reduction in occupancy probability with distance to forest. Moreover, raccoons were more likely to use sites that were well connected to other grasslands in winter (Fig. 2C), presumably because the cover encouraged movements among nearby grasslands and from nearby forest. Because Size_Iso is a synthetic variable that reflects both patch isolation and patch size, raccoons could also have used the smaller grasslands more often because they had more edge habitat, which raccoons prefer (Renfrew and Ribic, 2003; Renfrew et al, 2005; Barding and Nelson, 2008). Reduced use of grassland sites during winter by raccoons could also just reflect reduced activity and movements during winter. However, home range sizes of raccoons in our region do not vary seasonally (Nixon et al., 2009; Beasley et al, 2007), and movement distances and rates do not differ between seasons (Newbury and Nelson, 2007).
Our hypothesis that herbivorous prey species (cottontails and deer) would prefer cool season grasslands that provide denser hiding cover (Eggebo et al., 2003) over warm season grasslands received only limited support. A positive association was found between site occupancy for cottontails and dominance of cool season grasses during winter only. Evidently, vegetation type is not as important during summer when crops can provide extra cover and there is more standing cover in nonagricultural habitats. For white-tailed deer, which should primarily be concerned with fawn predation, vegetation type did not affect photo-activity rate in either season, and only distance to the nearest forest (Nixon et al., 1991) predicted activity during summer. Overall, we did not detect any negative consequences of vegetation dominated by cool season grasses, typically smooth brome, instead of warm season grasses. This outcome is consistent with research for small mammals on newly created SAFE grasslands (Mulligan et al., 2013). Likewise, Duggan et al (2011) found smooth brome grasslands could function as suitable habitat for Franklin's ground squirrel (Poliocitellus franklinii), often considered a prairie-obligate species. Clearly, seeding with native warm season species remains preferable from a floristics standpoint, and other animal taxa may be more discriminating and prefer warm season grasslands.
Both cottontails and white-tailed deer substantially increased their activity on our grassland sites during winter based on photo-activity rate (Fig. 3). These species likely spent more time in the restored grasslands due in part to the drastic reduction in vegetative cover and food resources in the surrounding agricultural matrix after harvest. Cottontails in central Illinois, and other regions with intensive farming, reduce their home ranges substantially after crop harvest, but seasonal shifts are uncommon in less modified regions (Mankin and Warner, 1999b). In contrast, white-tailed deer in central Illinois have larger home ranges during winter (Nixon et al., 1991), which could contribute to their greater use of restored grasslands during winter.
Although prey may be concentrated on our created grasslands during winter, we found no evidence of coyotes spatially tracking one of their key prey species, eastern cottontails. The occupancy probability and photo-activity rate of coyotes on grasslands did not change between seasons despite coyotes increasing their home range size during winter in our region (Gosselink et al., 2003). Moreover, even though "Prey" was in the top model for coyote occupancy during summer, the relationship was negative, indicating coyotes were less likely to occupy sites with cottontails. This result was consistent with patterns of habitat use for coyotes and desert lagomorphs (Arias-Del Razo et al., 2012). One explanation for this discordance is predators might spend more time hunting in areas where prey is less common but more vulnerable (AriasDel Razo et al., 2012). Coyote occupancy during summer also was higher when restoration sites were closer to human structures, which was opposite to our hypothesis that coyotes avoid humans (Gosselink et al., 2003; Magle et al., 2014). Additional factors not included in our models, such as abundance of small mammal prey, could have produced the positive association with farmsteads. For winter the intercept-only model was most supported for coyote occupancy. In general, the factors affecting the spatial distribution of coyotes within our patchy landscapes were not well resolved (see also Cove et al., 2012).
One unexpected result was we did not detect red foxes on any of the restoration sites despite >5000 camera nights of sampling. Our camera trapping was intensive enough to record secretive mustelids like long-tailed weasels and American mink. Red foxes are a species associated with open habitat, and although we anticipated a negative pattern of occurrence between red foxes and coyotes due to competitive exclusion and intraguild predation (Lavin et al., 2003; Gosselink et al., 2006), we still expected to detect foxes on a subset of the sites. These results suggest red foxes are now quite rare in rural areas of central Illinois. Foxes may occur more often in habitats near people (Goad et al., 2014; Lesmeister et al., 2015). For instance, red foxes found refuge from coyotes near farmsteads in rural areas of central Illinois (Gosselink et al., 2003). In our study, however, an avoidance of farmsteads by coyotes was not evident. Based on trends from annual harvest adjusted for effort using pelt prices, red fox declined in Illinois from 1987 to 1996 during a period when coyotes increased substantially (Gosselink et al., 2003). This apparent decline in red fox has leveled off in the past decade (A. Ahlers, Kansas State University, unpubl. data). The role of interspecific interactions in limiting the distribution of red foxes in the Midwest region requires additional investigation.
Our results demonstrate landscape context strongly affects occupancy of restored grasslands by mammals in a highly modified landscape. Species occurrences and activity were associated with grassland size and isolation, distances to human structures, and distances to nearest forest. These outcomes can assist managers in selecting land parcels to enroll in habitat restoration programs to either promote use by games species or to reduce undesirable predation. For instance, the increased use of restored grasslands by raccoons during summer coincides with nesting of grassland bird species. Many of these species have declined in North America (Knopf, 1994; Herkert, 1995), and new grasslands created under federal programs may benefit some species (Herkert, 2009). Given the potential importance of raccoons as nest predators (Heske et al, 1999; Schmidt, 2003; Lyons et al, 2015), we recommend that managers concerned with habitat restoration for grassland birds should focus on sites >400 m from forests. Currently, enrollment in the SAFE program is restricted to designated SAFE areas and includes basic eligibility requirements (USDA, 2008), but the location of potential SAFE grasslands relative to forests is not a consideration for enrollment.
Acknowledgments.--We thank E. Pritchard, J. Nawrocki, H. Crank, and L. Chow for field assistance; W. Louis and S. Simpson for help in identifying field sites; and the landowners who allowed research on their property. E. Heske, J. Jacquot, and two anonymous reviewers provided helpful comments on the manuscript. This project was funded by the Illinois Department of Natural Resources through the Federal Aid in Wildlife Restoration Program and by the University of Illinois.
ALTHOFF, D. P., G. L. STORM, AND D. R. DEWALLE. 1997. Daytime habitat selection by cottontails in central Pennsylvania. J. Wildl. Manage., 61:450-459.
ANDREWS, J. E., J. D. BRAWN, AND M. P. WARD. 2015. When to use social cues: conspecific attraction at newly created grasslands. Condor: Ornithol. Appl., 117:297-305.
ARIAS-DEL RAZO, I., L. HERNANDEZ, J. W. LAUNDRE, AND L. VELASCO-VAZQUEZ. 2012. The landscape of fear: habitat use by a predator (Canis latrans) and its main prey (Lepus californicus and Sylvilagus audubonii). Can. J. Zool., 90:683-693.
ARNOLD, T. W. 2010. Uninformative parameters and model selection using Akaike's Information Criterion. J Wildl. Manage., 74:1175-1178.
ATWOOD, T. C., H. P. WEF.KS, AND T. M. GEHRINC;. 2004. Spatial ecology of coyotes along a suburban-to-rural gradient. J. Wildl. Manage., 68:1000-1009.
BARDING, E. E. AND T. A. NELSON. 2008. Raccoons use habitat edge in northern Illinois. Am. Midi. Nat., 159:394-402.
BATZLI, G. O. AND C. E. DEJACO. 2013. White-tailed Deer (Odocoileus virginianus) facilitate the development of nonnative grasslands in central Illinois. Am. Midi. Nat., 170:323-334.
BEASLEY, J. C., T. L. DEVAULT, M. I. RETAMOSA, AND O. E. RHODES Jr. 2006. A hierarchical analysis of habitat selection by raccoons in Northern Indiana. J. Wildl. Manage., 71:1125-1133.
--. --, AND O. E. RHODES JR. 2007. Home-range attributes of raccoons in a fragmented agricultural region of northern Indiana. J. Wildl. Manage., 71:844-850.
--, Z. H. OLSON, G. DHARMARAJAN, T. S. EAGAN II, AND O. E. RHODES JR. 2011. Spatiotemporal variation in the demographic attributes of a generalist mesopredator. Landscape Ecoi, 26:937-950.
--AND O. E. RHODES. 2010. Influence of patch- and landscape-level attributes on the movement behavior of raccoons in agriculturally fragmented landscapes. Can. J. Zool., 88:161-169.
BERRY, W. B. 2016. Use of restored grasslands by mammals in a dynamic agroecosystem: insights from camera trapping. MS Thesis, University of Illinois, Urbana-Champaign.
BOCK, C. E., Z. F.JONES, AND J. H. BOCK. 2006. Abundance of cottontails (Sylvilagus) in an exurbanizing southwestern savanna. Southwestern Nat., 51:352-357.
BRUDVIG, L. A. 2011. The restoration of biodiversity: where has research been and where does it need to go? Am. J. Bot., 98:549-558.
BURNHAM, K. P. AND D. R. ANDERSON. 2002. Model selection and multimodel inference: a practical information-theoretic approach. 2nd ed. Springer, New York. 488 p.
BURTON, A. C., E NEILSON, D. MOREIRA, A. LADLE, R. STEENWEG, J. T. FISHER, E. BAYNE, AND S. BOUTIN. 2015. Wildlife camera trapping: a review and recommendations for linking surveys to ecological processes. J. Appl. Ecol., 52:675-685.
COSENTINO, B. J., R. L. SCHOOLEY, AND C. A. PHILLIPS. 2011. Connectivity of agroecosystems: dispersal costs can vary among crops. Landscape Ecol., 26:371-379.
--, R. L. SCHOOLEY, B. T. BESTELMEVER, J. F. KELLY, .AND J. M. COFEMAN. 2014. Constraints and time lags for recovery of a keystone species (Dipodomys spectabilis) after landscape restoration. Landscape Ecol., 29:665-675.
COVE, M. V., B. M.JONES, A. BOSSERT, D. R. CLEVER, R. K. DUNWOODY, B. C. WHITE, AND V. L. JACKSON. 2012. Use of camera traps to examine the mesopredator release hypothesis in a fragmented midwestern landscape. Am. Midi. Nat., 168:456-465.
CRAWLEY, M. J. 1990. Rabbit grazing, plant competition and seedling recruitment in acid grassland. J. Appl. EcoL, 27:803-820.
CZECH, B., P. R. KRAUSMAN, AND P. K. DEVERS. 2000. Economic associations among causes of species endangerment in the United States. Bioscience, 50:592-601.
DAVANON, K. A., L. K HOWARD, K. E. MABRY, R. L. SCHOOLEY, AND B. T. BESTELMEYER. 2016. Effects of exurban development on trophic interactions in a desert landscape. Landscape Ecol. In press.
DEGOLIER, T., J. PORT, AND S. P. SCHOTTLER. 2015. Small mammal habitat preferences in a patchwork of adjacent reconstructed grasslands subject to semiannual burns. Ecol. Restor., 33:388-394.
DUGGAN, J. M., R. L. SCHOOLEY, AND E.J. HESKE. 2011. Modeling occupancy dynamics of a rare species, Franklin's ground squirrel, with limited data: are simple connectivity metrics adequate? Landscape Ecol., 26:1477-1490.
DUNNING, J. B., B.J. DANIELSON, AND H. R. PULLIAM. 1992. Ecological processes that affect populations in complex landscapes. Oikos, 65:69-175.
EGGEBO, S. L., K. F. HIGGINS, D. E. NAUGLE, AND F. R. QAUMEN. 2003. Effects of CRP field age and cover type on ring-necked pheasants in eastern South Dakota. Wildl. Soc. Bull, 31:779-785.
ESRI. 2015. ArcMap, version 10.3. ESRI, Redlands, California.
FRIESEN, L. E., G. CASBOURN, V. MARTIN, AND R. J. MAGKAY. 2013. Nest predation in an anthropogenic landscape. Wilson J. Ornithol., 125:562-569.
GEHRT, S. D. AND E. K FRITZELL. 1998. Resource distribution, female home range dispersion and male spatial interactions: group structure in a solitary carnivore. Anim. Behav., 55:1211-1227.
GOAD, E. H., L. PEJCHAR, S. E. REED, AND R. L. KNIGHT. 2014. Habitat use by mammals varies along an exurban development gradient in northern Colorado. Biol. Consent., 176:172-182.
GOMPPER, M. E. 2002. Top carnivores in the suburbs? Ecological and conservation issues raised by colonization of north-eastern North America by coyotes. Bioscience, 52:185-190.
GOOGLE INC. 2015. Google Earth, version 220.127.116.117. Google Inc, Mountain View, California.
GOSSELINK, T. E., T. R. VAN DEELEN, R. E. WARNER, AND M. G. JOSELYN. 2003. Temporal habitat partitioning and spatial use of coyotes and red foxes in east-central Illinois. J. Wildl. Manage., 67:90-103.
--, --, --, AND P. C. MANKIN. 2006. Survival and cause-specific mortality of red foxes in agricultural and urban areas of Illinois. J. Wildl. Manage., 71:1862-1873.
GROVENBURG, T. W., C. N.JACQUES, AND J. A. JENKS. 2011. Drought effect on selection of Conservation Reserve Program grasslands by white-tailed deer on the Northern Great Plains. Am. Midi. Nat., 166:147-162.
GUILLERA-ARROITA, G. 2011. Impact of sampling with replacement in occupancy studies with spatial replication. Methods Ecol. Evol., 2:401-406.
HARRIS, G., R. THOMPSON, J. L. CHILDS, ANDJ. G. SANDERSON. 2010. Automatic storage and analysis of camera trap data. Ecol. Soc. Am. Bull, 91:352-360.
HENNER, C. M., M. J. CHAMBERLAIN, B. D. LEOPOLD, AND L. W. BURGER JR. 2004. A multi- resolution assessment of raccoon den selection. J. Wildl. Manage., 68:179-187.
HERKERT, J. R. 1995. An analysis of Midwestern breeding bird population trends: 1966-1993. Am. Midi. Nat., 134:41-50.
--. 2009. Response of bird populations to farmland set-aside programs. Conseru. Biol., 23:1036-1040.
HESKE, E. J., S. K. ROBINSON, AND J. D. BRAWN. 1999. Predator activity and predation on songbird nests on forest-field edges in east-central Illinois. Landscape Ecol., 14:345-354.
HINES, J. 2006. PRESENCE--software to estimate patch occupancy and related parameters. USGS-PWRC.
ILLINOIS DEPARTMENT OF NATURAL RESOURCES. 2015. Summary of 2014-2015 Illinois deer seasons, https:// www.dnr.illinois.gov/hunting/deer/Pages/AnnualDeerHarvestReports.aspx
KALLE, R., T. RAMESH, Q. QURESHI, AND K. SANKAR. 2014. Estimating seasonal abundance and habitat use of small carnivores in the Western Ghats using an occupancy approach. J. Trap. Ecol., 30:469-480.
KERR, J. T. AND J. CIHLAR. 2004. Patterns and causes of species endangerment in Canada. Ecol. Appl, 14:743-753.
KNOPF, F. L. 1994. Avian assemblages on altered grasslands. Stud. Avian Biol., 15:247-257.
LAVIN, S. R., T. R. VAN DEELEN, P. W. BROWN, R. E. WARNER, AND S. H. AMBROSE. 2003. Prey use by red foxes (Vulpes vulpes) in urban and rural areas of Illinois. Can. J. Zool., 81:1070-1082.
LESMEISTER, D. B., C. K. NIELSEN, E. M. SCHAUBER, AND E. C. HELLGREN. 2015. Spatial and temporal structure of a mesocarnivore guild in Midwestern North America. Wildl. Monogr., 191:1-61.
LONG, R. A., T. M. DONOVAN, P. MACKAY, W.J. ZIELINSKI, AND J. S. BUZAS. 2011. Predicting carnivore occurrence with noninvasive surveys and occupancy modeling. Landscape EcoL, 26:327-340.
LYONS, T. P., J. R. MILLER, D. M. DEBINSH, AND D. M. ENGLE. 2015. Predator identity influences the effect of habitat management on nest predation. Ecol. Appl., 25:1596-1605.
MACKENZIE, D., J. NICHOLS, J. ROYLE, K. POLLOCK, L. BAILEY, AND J. HINES. 2006. Occupancy estimation and modeling: inferring patterns and dynamics of species occurrence. Academic Press, Burlington. 324 p.
MAGLE, S. B" L. S. SIMONI, E. W. LEHRER, AND J. S. BROWN. 2014. Urban predator-prey association: coyote and deer distributions in the Chicago metropolitan area. Urban Ecosyst., 17:875-891.
MANKIN, P. C. AND R. E. WARNER. 1997. Mammals of Illinois and the Midwest: ecological and conservation issues for human-dominated landscapes, p. 135-153. In: M. W. Schwartz (ed). Conservation in highly fragmented landscapes. Chapman and Hall, New York.
--. 1999a. A regional model of the eastern cottontail and land-use changes in Illinois. J. Wildl. Manage., 63:956-963.
--. 1999b. Responses of eastern cottontails to intensive row-crop farming. J. Mammal., 80:940-949.
MCCOY, T. D., M. R. RYAN, E. W. KURZEJESKI, AND L. W. BURGER JR. 1999. Conservation reserve program: source or sink habitat for grassland birds in Missouri? J. Wildl. Manage., 63:530-538.
MOREY, P. S., E. M. GESE, AND S. GEHRT. 2007. Spatial and temporal variation in the diet of coyotes in the Chicago metropolitan area. Am. Midi. Nat., 158:147-161.
MULLIGAN, M. P., R. L. SCHOOLEY, AND M. P. WARD. 2013. Effects of connectivity and regional dynamics on restoration of small mammal communities in Midwestern grasslands. Restor. Ecol., 21:678-685.
NELSON, T. A. AND A. WOOLF. 1987. Mortality of white-tailed deer fawns in southern Illinois. J. Wildl. Manage., 51:326-329.
NEWBURY, R. K. AND T. A. NELSON. 2007. Habitat selection and movements of raccoons on a grassland reserve managed for imperiled birds. J. Mammal., 88:1082-1089.
NIXON, C. M., L. P. HANSEN, P. A. BREWER, AND J. E. CHELSVIG. 1991. Ecology of white-tailed deer in an intensively farmed region of Illinois. Wildl. Monogr., 18:3-77.
--, P. C. MANKIN, T. L. ESKER, J. B. SULLIVAN, A. A. ROTHERING, R. KOERKENMEIER, AND R. D. BLUETT. 2009. Aspects of raccoon life history in west-central Illinois. Trans. Illinois State Acad. Sci., 102:117-129.
OLOFSSON, J., C. DE MANZANCOURT, AND M. J. CRAWLEY. 2008. Spatial heterogeneity and plant species richness at different spatial scales under rabbit grazing. Oecologia, 156:825-834.
PRUGH, L. R., C.J. STONER, C. W. EPPS, W. T. BEAN, W.J. RIPPLE, A. S. LALIBERTE, AND J. S. BRASHARES. 2009. The rise of the mesopredator. Bioscience, 59:779-791.
RENFREW, R. B. AND C. A. RIBIC. 2003. Grassland passerine nest predators near pasture edges identified on videotape. Auk, 120:371-383.
--,--, AND J. L. NACK. 2005. Edge avoidance by nesting grassland birds: a futile strategy in a fragmented landscape. Auk, 122:618-636.
ROBINSON, Q. H., D. BUSTOS, AND G. W. ROEMER. 2014. The application of occupancy modeling to evaluate intraguild predation in a model carnivore system. Ecology, 95:3112-3123.
RODEWALD, A. D. AND L.J. KEARNS. 2011 Shifts in dominant nest predators along a rural-tour-ban landscape gradient. Condor, 113:899-906.
ROWCLIFFE, J. M. AND C. CARBONE. 2008. Surveys using camera traps: are we looking to a brighter future? Anim. Conserv., 11:185-186.
SAMSON, F. AND F. KNOPF. 1994. Prairie conservation in North America. Bioscience, 44:418-421.
SANDERSON, J. AND G. HARRIS. 2013. Automatic data organization, storage, and analysis of camera trap pictures. J. Indonesian Nat. Hist., 1:6-14.
SAS INSTITUTE INC. 2013. SAS[R] 9.4--SAS Institute Inc, Cary, North Carolina, U.S.A.
SCHMIDT, K A. 2003. Nest predation and population declines in Illinois songbirds: a case study for mesopredator effects. Conserv. Biol., 7:1141-1150.
SCHOOLEY, R. L., L. A. COTNER, A. A. AHLERS, E. J. HESKE, AND J. M. LEVENGOOD. 2012. Monitoring site occupancy for American mink in its native range. J. Wildl. Manage., 76:824-831.
STONE, E. R. 2007. Measuring impacts of restoration on small mammals in a mixed-grass Colorado prairie. Ecol. Restor., 25:183-190.
USDA. 2008. Conservation Reserve Program State Acres for Wildlife Enhancement (SAFE). Farm Service Agency, United States Department of Agriculture. https://www.fsa.usda.gov/Internet/FSA_ File/safe08.pdf.
SUBMITTED 7 MARCH 2016
ACCEPTED 13 DECEMBER 2016
APPENDIX 1.--Naive occupancy (no. sites with detections-30 total sites) and number of independent photographs (>60 min apart) of all mammals detected by camera traps in central Illinois, 2014-2015 Summer Species Naive occupancy No. photos White-tailed deer 0.97 117 (Odocoileus virginianus) Eastern cottontail 0.30 131 (Sylvilagus floridanus) Coyote (Canis latrans) 0.33 18 Raccoon (Procyon lotor) 0.27 48 Striped skunk (Mephitis mephitis) 0.23 9 Virginia opossum 0.10 3 (Didelphis virginiana) American mink (Neovison vison) 0.10 2 Long-tailed weasel (Mustela frenata) 0.00 0 Winter Species Naive occupancy No. photos White-tailed deer 0.77 203 (Odocoileus virginianus) Eastern cottontail 0.47 360 (Sylvilagus floridanus) Coyote (Canis latrans) 0.33 16 Raccoon (Procyon lotor) 0.20 17 Striped skunk (Mephitis mephitis) 0.20 22 Virginia opossum 0.20 11 (Didelphis virginiana) American mink (Neovison vison) 0.00 0 | Long-tailed weasel (Mustela frenata) 0.03 2 j
BRIAN BERRY, ROBERT L. SCHOOLEY (1) and MICHAEL P. WARD
Department of Natural Resources and Environmental Science, University of Illinois, Urbana 61801
(1) Corresponding author: e-mail: email@example.com
Caption: Fig. 1.--(A) Distribution of grassland study sites in central Illinois where mammals were sampled with camera traps. (B) Example grassland site showing placement of four camera traps. The dark band running through the site is a grassy waterway. The site is surrounded by row crop agriculture
Caption: Fig. 2.--Factors affecting occupancy of restored grasslands by mammals in central Illinois, 2014--2015. Size_Iso is a synthetic variable that ranges from small and connected grasslands to large and isolated grasslands. For raccoons in winter and coyotes in summer, the predicted occupancy probability is shown for each covariate with the other covariate held to its mean value
Caption: Fig. 3.--Seasonal use of restored grasslands by mammals in central Illinois, 2014-2015. The left panel is conditional occupancy; the right panel is photo-activity rate (no. independent photographs/sampling effort). Bars are means (+1 SE). Significant differences in use between summer and winter are indicated (* for P < 0.05; *** for P < 0.001)
TABLE 1.--Predictions for how focal species of mammals will use grasslands created under the Conservation Reserve Program (CRP) and State Acres for Wildlife Enhancement (SAFE) Program in central Illinois. Full hypotheses and predictions are presented in the Introduction Focal species Prediction Supported? (a) Raccoon 1. Grassland use associated positively Yes (b) with proximity to forest, ponds, and streams. 2. Greater use of grasslands during Yes summer when crop cover promotes movements. 3. Greater use of sites near other Yes grasslands during winter when crops are absent. Eastern cottontail 4. Grassland use positively associated Yes (c) with dominance of cool season grasses that provide thick hiding cover. 5. Grassland occupancy should increase Yes from summer to winter as grasslands become refuges after crop harvest. 6. Grassland use associated negatively No with distance to human structure. Coyote 7. Increased use of grasslands in winter No if coyotes track their cottontail prey. 8. Positive relationship between No grassland occupancy and distance to nearest human structure. 9. Grassland occupancy associated No positively with proximity to forest. White-tailed deer 10. Grassland use should be related Yes (d) positively to proximity to nearest forest. 11. Higher use during summer for cool No season grasslands that provide thick hiding cover for fawns. (a) "Yes" indicates prediction was supported by occupancy modeling, photo- activity rate, or both (b) Distance to forest supported. Distances to ponds and streams not supported (c) Supported during winter only (d) Supported during summer only TABLE 2.--Model selection statistics for summer detection and occupancy for raccoons, cottontails, and coyotes during 2014 in central Illinois. Occupancy models presented include those with [DELTA]AIC [less than or equal to] 2 plus the intercept-only model. Covariates are defined in the text. [DELTA]AIC = difference between model AIC and AIC for top model, [w.sub.i] = Akaike weight, K = no. estimable parameters, and -2LogLike = twice the negative log-likelihood Species Parameter Model [DELTA]AIC [w.sub.i] K Raccoon Detection psi(.), p(Size) 0 0.493 3 psi(.), p(Size, 0.91 0.313 4 Effort) psi(-). P(-) 2.76 0.124 2 psi(.), 3.88 0.071 3 p(Effort) Occupancy psi(Forest), 0 0.233 4 p(Size) psi(Forest, 1.26 0.124 5 Pond), p(Size) psi(.), p(Size) 1.33 0.120 3 psi (Forest, 1.92 0.089 5 Stream), p(Size) psi(.), p(.) 4.09 0.0293 2 Cottontail Detection psi(.), p(Size) 0 0.532 3 psi(.), p(Size, 1.52 0.249 4 Effort) psi(.), p(.) 2.97 0.121 2 psi(.), 3.37 0.099 3 p(Effort) Occupancy psi(Size_Iso), 0 0.253 4 p(Size) psi(.), p(Size) 0.88 0.163 3 psi(Size_Iso, 1.91 0.097 5 Vegetation), p(Size) psi(Size_Iso, 1.97 0.095 5 Predator), p(Size) psi(Size_Iso, 2 0.093 5 Human), p(Size) psi(.), p(.) 3.85 0.036 2 Coyote Detection3 psi(.), p(.) 0 0.584 2 psi(.), p(Size) 0.68 0.416 3 Occupancy psi(Human, 0 0.301 4 Prey), p(.) psi(Human), 0.73 0.209 3 p(.) psi(.), p(.) 2.39 0.091 2 Species Parameter Model -2 * LogLike Raccoon Detection psi(.), p(Size) 70.82 psi(.), p(Size, 69.73 Effort) psi(-). P(-) 75.58 psi(.), 74.70 p(Effort) Occupancy psi(Forest), 67.49 p(Size) psi(Forest, 66.75 Pond), p(Size) psi(.), p(Size) 70.82 psi (Forest, 67.41 Stream), p(Size) psi(.), p(.) 75.58 Cottontail Detection psi(.), p(Size) 77.61 psi(.), p(Size, 77.13 Effort) psi(.), p(.) 82.58 psi(.), 80.98 p(Effort) Occupancy psi(Size_Iso), 74.73 p(Size) psi(.), p(Size) 77.61 psi(Size_Iso, 74.64 Vegetation), p(Size) psi(Size_Iso, 74.70 Predator), p(Size) psi(Size_Iso, 74.73 Human), p(Size) psi(.), p(.) 82.58 Coyote Detection3 psi(.), p(.) 77.73 psi(.), p(Size) 76.41 Occupancy psi(Human, 71.34 Prey), p(.) psi(Human), 74.07 p(.) psi(.), p(.) 77.73 (a) Detection models psi(.), p(Effort) and psi(.), p(Size, Effort) did not converge TABLE 3.--Model selection statistics for winter detection and occupancy for raccoons, cottontails, and coyotes during 2014-2015 in central Illinois. Occupancy models presented include those with [DELTA]AIC [less than or equal to] 2 plus the intercept-only model. Covariates are defined in the text. [DELTA]AIC = difference between model AIC and AIC for top model, [w.sub.i] = Akaike weight, K = no. estimable parameters, and -2LogLike = twice the 623 negative log-likelihood Species Parameter Model [DELTA]AIC [w.sub.i] K Raccoon Detection psi(.), p(.) 0 0.368 2 psi(.), 0.2 0.333 3 p(Size) psi(.), 1.77 0.152 4 p(Size, Effort) psi(.), 1.83 0.147 3 p(Effort) Occupancy psi(Size_Iso, 0 0.506 4 Forest), p(.) psi(.), p(.) 5.53 0.032 2 Cottontail Detection psi(.), 0 0.470 4 p(Size, Effort) psi (.), p 1.11 0.270 3 (Effort) psi(.), 2.28 0.150 3 p(Size) psi(.), p(.) 2.91 0.110 2 Occupancy psi 0 0.245 5 (Vegetation), p(Size, Effort) psi(.), 0.49 0.192 4 p(Size, Effort) psi(Human), 1.32 0.127 5 p(Size, Effort) psi 1.88 0.096 6 (Vegetation, Predator), p(Size, Effort) psi 1.9 0.095 5 (Predator), p(Size, Effort) psi(-), p(.) 3.4 0.043 2 Coyote Detection psi(.), p(.) 0 0.427 2 psi(.), 0.7 0.301 3 p(Size) psi(.), p 1.94 0.162 3 (Effort) psi (.), 2.7 0.111 4 p(Size, Effort) Occupancy psi(.), p(.) 0 0.204 2 psi(Forest), 1.47 0.098 3 p(.) psi(Size_Iso), 1.7 0.087 3 p(.) psi(Human), 1.7 0.087 3 p(.) psi (Prey), 1.79 0.083 3 p(.) psi 1.92 0.078 3 (Vegetation), p(.) Species Parameter Model -2 * LogLike Raccoon Detection psi(.), p(.) 59.58 psi(.), 57.78 p(Size) psi(.), 57.35 p(Size, Effort) psi(.), 59.41 p(Effort) Occupancy psi(Size_Iso, 50.05 Forest), p(.) psi(.), p(.) 59.58 Cottontail Detection psi(.), 100.06 p(Size, Effort) psi (.), p 103.17 (Effort) psi(.), 104.34 p(Size) psi(.), p(.) 106.97 Occupancy psi 97.57 (Vegetation), p(Size, Effort) psi(.), 100.06 p(Size, Effort) psi(Human), 98.89 p(Size, Effort) psi 97.45 (Vegetation, Predator), p(Size, Effort) psi 99.47 (Predator), p(Size, Effort) psi(-), p(.) 106.97 Coyote Detection psi(.), p(.) 80.17 psi(.), 78.87 p(Size) psi(.), p 80.11 (Effort) psi (.), 78.87 p(Size, Effort) Occupancy psi(.), p(.) 80.17 psi(Forest), 79.64 p(.) psi(Size_Iso), 79.87 p(.) psi(Human), 79.87 p(.) psi (Prey), 79.96 p(.) psi 80.09 (Vegetation), p(.) TABLE 4.--Model selection statistics for negative binomial models for activity of white-tailed deer (photographic rate) in central Illinois, 2014-2015. Models presented include those with [DELTA]AIC [less than or equal to] 3 plus the intercept-only model. Covariates are defined in the text. [DELTA]AICc = difference between model AICc and 630 AICc for top model, [w.sub.i], = Akaike weight, K=no. estimable parameters, and LL = log-likelihood Season Model [[DELTA]AIC.sub.c] [w.sub.i] K Summer Forest 0 0.483 3 Forest, Size_Iso 1.767 0.200 4 Forest, Predator 2.267 0.155 4 Forest, Vegetation 2.547 0.135 4 Intercept-only 7.666 0.010 2 Winter Intercept-only 0 0.252 2 Forest 0.512 0.195 3 Predator 1.575 0.115 3 Size_Iso 2.2 0.084 3 Vegetation 2.395 0.076 3 Forest, Predator 2.804 0.062 4 Season Model LL Summer Forest 54.19 Forest, Size_Iso 54.60 Forest, Predator 54.20 Forest, Vegetation 54.21 Intercept-only 49.07 Winter Intercept-only 242.13 Forest 243.11 Predator 242.58 Size_Iso 242.47 Vegetation 242.17 Forest, Predator 243.51
|Printer friendly Cite/link Email Feedback|
|Author:||Berry, Brian; Schooley, Robert L.; Ward, Michael P.|
|Publication:||The American Midland Naturalist|
|Date:||Apr 1, 2017|
|Previous Article:||Discovery of an exotic population of Thereuonema tuberculata (Chilopoda: Scutigeromorpha), the Japanese House Centipede, in Ohio, U.S.A.|
|Next Article:||Range-wide microsatellite analysis of the genetic population structure of Prairie voles (Microtus ochrogaster).|