Carnivore Occupancy within the Early Successional Habitat of New England Cottontails Free

Understanding spatial patterns of carnivore distributions is an important aspect of wildlife management. Carnivores influence managed prey species by creating top–down effects on lower trophic-level organisms within complex local food webs (Baum and Worm 2009; Gervasi et al. 2012; Sandom et al. 2013; Winnie and Creel 2017). Top–down effects of predators include direct killing of prey and “fear”-induced trophic cascades (Suraci et al. 2016). Prey may respond to predation pressure with demographic and behavioral changes, such as shifts in foraging habitat and reduced foraging time, all of which can result in decreased reproduction (Winnie and Creel 2017). Carnivores may also have important intraguild interactions with other predators, such as interference and exploitative competition that ultimately shape both predator and prey communities (St. Pierre et al. 2006; Thompson and Gese 2007).

Mammalian carnivore communities in the northeastern United States changed dramatically following European settlement. Habitat loss and persecution resulted in the loss of several large carnivores, including mountain lions Puma concolor and gray wolves Canis lupus, and in their absence, the smaller coyotes Canis latrans expanded into the Northeast (Ray 2000). Several species experienced range loss (e.g., American marten Martes americana, Canada lynx canadensis, and black bear Ursus americanus), while others found ways to thrive in human-dominated landscapes by exploiting new food resources and shelters. Carnivores that remain in the Northeast have continued to expand due to improved habitat suitability, suspended trapping seasons (e.g., bobcat Lynx rufus), natural recolonizations (e.g., black bear), and reintroductions (e.g., fisher Pekania pennanti; Ray 2000; Litvaitis et al. 2006b). Increasing carnivore populations has the potential to increase predation pressure on prey species, such as New England cottontails Sylvilagus transitionalis, that are habitat specialists or that have experienced historical declines and have yet to fully recover (Ray et al. 2005; Reddy et al. 2019).

New England cottontails are habitat specialists in early successional habitat (Barbour and Litvaitis 1993; Oehler at al. 2006) and were historically found throughout the Northeastern United States but declined significantly following the widespread abandonment of agricultural land and subsequent succession of land into mature forest that occurred throughout the 20th century (Litvaitis 1993). This decline of early successional habitat has left the remaining New England cottontails isolated into five geographically distinct groups, two of which are in Connecticut. One of these groups is in western Connecticut in adjacent areas of southwestern Massachusetts and eastern New York. The second population is east of the Connecticut River, extending into Rhode Island. The other three geographic populations are in southwestern Maine and coastal New Hampshire, in the Merrimack River Valley of southern New Hampshire, and in Cape Cod, Massachusetts (Litvaitis et al. 2006a; Fenderson et al. 2011). Early successional habitats consisting of young forests and shrublands made up 15–20% of the landscape in the coastal states of the Northeast during presettlement times but now cover only 1–5% of the landscape in states such as Connecticut (Oehler et al. 2006).

New England cottontails and at least 50 other species of wildlife associated with early successional habitats are now listed as species of greatest conservation need in Connecticut (CT DEEP 2015). New England cottontails are at risk from habitat loss and the limited amount of available habitat and also must compete for space and resources with nonnative eastern cottontails Sylvilagus floridanus (Cheeseman et al. 2018). Predation is also a significant documented source of mortality in New England cottontails and is primarily associated with four mammalian carnivores, including bobcats, coyotes, red fox Vulpes vulpes, and gray fox Urocyon cinereoargenteus (Barbour and Litvaitis 1993; Litvaitis and Villafuerte 1996; Litvaitis 2001; Gilbart 2012). Predation pressure by generalist carnivores in early successional habitat may be associated with the availability of prey and the ability to successfully exploit prey resources there (Litvaitis et al. 2006b). Of the four species, bobcats and red fox show a stronger dietary preference toward rabbits (Story et al. 1982; Hockman and Capman 1983; Litvaitis 2001). Smith and Litvaitis (1999) also reported mortality from weasels Mustela spp., fisher, and several other avian predators. Several of these carnivores have experienced significant changes in their abundances and distributions in Connecticut and have the potential to affect the recovery of New England cottontails (CT DEEP 2015). Bobcats were once restricted to the northwestern part of the state but have increased their extent of occurrence in the last two decades (J. Hawley, CT DEEP, personal communication). Likewise, coyotes only appeared in Connecticut in the mid-1950s (CT DEEP 2015) but now are reported from every town in the state. Other important predators that use early successional habitat, such as red fox and gray fox, are considered widespread in Connecticut (CT DEEP 2015). However, both species of fox have declined from historic levels, and their decline may be associated with land use changes or interspecific competition and predation (Major and Sherburne 1987; Sargeant and Allen 1987) from coyotes following their emergence over the last half century (CT DEEP 2015). Interactions among the primary predator species are likely to include competition, particularly between coyote and red fox, but differences in activity levels and habitat selection may limit this (Chamberlain 1999). The use of early successional habitat by fishers is poorly understood in Connecticut but their distribution is associated with historical declines and recovery through natural recolonizations and reintroductions in the western part of the state (Hapeman et al. 2011). Thus, to begin to understand the potential impacts of carnivores on the recovery of New England cottontails, it is important to have a better understanding of the occupancy patterns of widely distributed carnivores in early successional habitat in Connecticut. Studying patterns in local carnivore occupancy patterns provides distributional data while simultaneously estimating probabilities of detection and site use by each target species and the influence of covariates on each species. These outcomes are particularly important when focused within areas of conservation concern, such as the restricted range of New England cottontails.

Here, we use an occupancy modeling framework to examine detection data from New England cottontails, eastern cottontails, and their four primary predators (coyote, bobcat, red fox, and gray fox) collected from automatically triggered camera surveys in early successional habitat (MacKenzie et al. 2018). We viewed both Sylvilagus species (hereafter referred to collectively as cottontails) as ecologically similar enough in their interactions with predators to group them together for analyses. We add to the limited body of information on carnivore communities associated with early successional habitat in the range of New England cottontails in Connecticut (O'Connor and Rittenhouse 2017). The objectives of our study associated with cottontails and their four primary predators were to 1) identify regional differences in occupancy and detectability in early successional habitat, 2) develop species-specific models of occupancy and detectability, and 3) identify interactions between predators and prey and among the primary predators. We predicted that of the top four mammalian predators, bobcats would have the highest occupancy in early successional habitat because they preferentially prey on cottontails in the Northeastern United States (Litvaitis 2001). We also predicted that regional differences in occupancy would only be evident for bobcats and coyotes and that occupancy would be higher in western Connecticut than in eastern Connecticut because both species were thought to be more abundant in northwestern Connecticut during the last two decades (CT DEEP 2020). We did not predict regional differences for red fox because no pattern was evident in the historical records for this species apart from indication that residents increased in number in the Northeast after the mid-1800s and were supplemented by natural recolonization from southern Canada and introductions from Europe (Kasprowicz 2016).

Between September 2014 and July 2015, we collected detection data from motion-triggered cameras (Moultrie P-150) at sites (n = 34) located in early successional habitats in Connecticut within the expected range of the two populations of New England cottontails east (n = 17) and west (n = 17) of the Connecticut River (Figure 1). These early successional habitats broadly consisted of old fields, shrublands (wet and dry), young open-canopy woodlands, regenerating clear cuts, and areas of previous forest thinning. The actual composition of vegetation at sites often depended on the site's history (Oehler et al. 2006). Most of our sites (n = 24) were located on Connecticut Department of Energy and Environmental Protection (CT DEEP) lands and consisted of state forests, parks, and wildlife management areas, with the remainder of the sites on private land (n = 10). All sites, except for three, were also within Connecticut's New England cottontail focus areas, specially designated areas with habitat suitability and current and historic use by New England cottontails (Fuller and Tur 2012). The three sites not in these focus areas were within the same habitat and in close proximity to the focus areas. CT DEEP, land trusts, and private landowners granted permission to access the sites before commencing all camera surveys.

We based the camera methods for this study on a 7-mo pilot study where we used the following methods and assessed their efficacy for detecting the target species while minimizing bias. We determined camera placement within early successional habitat in our study area by using randomly generated points in ArcMap 10.1 (Environmental Systems Research Institute) with the additional criteria that we spaced camera sites at least 5 km apart to avoid spatial autocorrelation and to maximize the number of independent individuals contributing to the dataset (Long et al. 2008). At each site, we placed cameras on trees approximately 0.5 m above ground level, but we adjusted them between surveys if snow cover appeared to interfere with expected operation. We set cameras to three-shot pulse and to be active for 24 h per day. We stapled a scent-soaked (Hawbaker's Wildcat 2) cloth wick to the top of a 0.6-m wooden post covered with a rain guard (modified plastic cup) that we placed 3 to 5 m in front of the camera (following Mahard 2014). We used half of an aluminum pie plate with a goose feather hanging below as a visual attractant and strung it to a branch above the camera's motion detection zone. Survey periods for each camera lasted 2 wk (14 trap days [TD]). After each survey period, we checked the camera batteries and replaced any if necessary. We removed photographs by swapping in a new Secure Digital memory card, and we reapplied scent lure to the cloth wick on the wood post. The number of surveys that we completed at each site ranged between two and seven (28–98 TD) and depended on when we received approved research permits, travel logistics, weather, and the number of cameras available at a given time (Figure S1, Supplemental Material).

We collected and measured habitat-associated covariates at each camera site that, based on the primary literature, we believed could potentially influence both detection and occupancy probabilities of carnivores (Table 1). To assess the extent to which species have spread from historical ranges throughout the state, we divided sites into two groups, accounting for whether the site was east or west of the Connecticut River. Other site-specific covariates included the distance to nearest water, road density, and the amount of early successional vegetation (Table 1). We derived values for habitat-associated covariates from the most recent available geographic information system layers in ArcMap 10.1. We made site covariate measurements at two different spatial buffers (1 km² and 5 km²) around cameras to accommodate the smaller patchy and irregular nature of early successional habitat within the landscape. We evaluated covariates for collinearity using a Pearson correlation, and we did not include those with Pearson r values of >0.50 in the same model (Wilson and Schmidt 2015).

We were also interested in the interactions of carnivores and New England cottontails using camera data. However, we were unable to separate New England cottontails from eastern cottontails that also appeared at camera sites. We quantified the number of hours during each 2-wk survey that at least one cottontail was detected on camera and used that as a survey-specific covariate on each predator's detectability. We created a site-specific covariate for each predator's occupancy based on the intensity of cottontail activity in front of the camera by averaging the number of hours with cottontail detections over all surveys at a particular site (Figure S2, Supplemental Material). These two covariates provided a way to account for the possible effects that localized cottontail activity could have on each predator's occupancy or detectability.

We recorded detection data for target species (0 or 1) on a per survey basis and combined information over surveys to create a detection history for each species (e.g., 0010). We used detection data from the first four survey visits at each site to estimate occupancy (Ψ) and detectability (p) for cottontails, each of the four principal cottontail predators, and a combined multicarnivore species group by using the program MARK version 9.0 (White and Burnham 1999; Cooch and White 2019). The combination of low detection rates and unequal survey effort at sites led to excessive variation in our occupancy models when using the entire data set. We found that using data from the average number of surveys (n = 4) minimized information loss and provided results that were informative and produced stable results with standard errors that allowed interpretations of the data. Therefore, species-specific occupancy analysis only included detection/nondetection data from the first four survey periods (28 to 56 TD at each site). We assumed that sites were either occupied or unoccupied by a particular species for all survey periods. We sampled some units during the breeding season with the possibility that a previously unoccupied unit could have potentially been colonized by a juvenile, thus violating the closure assumption. However, given that movements of the species in and out of sampling units could be considered random, we believe our occupancy estimates are unbiased and regard our results as the proportion of sites used by our target species (Mackenzie et al. 2018).

For cottontails and the combined carnivore group, we did not model the effects of any covariates on our estimates of occupancy and detectability. Individually, for each of our target carnivore species, we took a multistep approach to modeling factors affecting occupancy and detectability parameters (Doherty et al. 2012). This approach allowed us to identify an optimal predictive model for each species (Barbieri and Berger 2004) from covariates that were strongly supported by cumulative model weights of >0.5 (Doherty et al. 2012). In the first round of modeling (“step 1”), we built models using all the covariates predicted to influence detectability while keeping occupancy constant; we created a balanced subset of all possible additive linear combinations of covariates that used each detection covariate an equal number of times throughout the entire model set. To reduce overparameterization given the number of sites and surveys, we included a maximum of four covariates in any single model (two intercepts and up to two covariates). We ran models and evaluated them using corrected Akaike information criterion (AICc) weights (Burnham and Anderson 2002) based on their relative importance in the balanced set of models (see Table S3, Supplemental Material, for the complete list of models). We then calculated the strength of each covariate in the model set by summing the weights of all the models that included each covariate (Barbieri and Berger 2004). Therefore, covariates that were consistently in the top models had the highest cumulative weights (Burnham and Anderson 2002; Barbieri and Berger 2004; Doherty et al. 2012). We then selected covariates with cumulative weights greater than or equal to 0.5 for inclusion in the final round of modeling. Next, we used the same process to develop our models for occupancy, keeping detectability constant and running models using covariates an equal number of times. We brought occupancy covariates with cumulative weights of 0.5 and above into the final round of modeling. If no models accumulated enough weight to make it to step 2, we used the base model as the top model for parameter estimates.

In the final round (“step 2”) of modeling, we combined the covariates from models of detectability and occupancy with weights greater than or equal to 0.5 from the earlier round (step 1). We simultaneously modeled detectability and occupancy using these covariates in all possible combinations. We determined the relative importance of each covariate in this step by summing the AICc weights of the models that contained a particular covariate. We considered covariates with cumulative weights of 0.5 and above the strongest predictors, and we used them to build a species' optimal predictive model after the second step was completed (Barbieri and Berger 2004; Doherty et al. 2012; Martin et al. 2020). This approach emphasizes the relative strength of individual covariates in a balanced set of model combinations while reducing the potential effects of pretending variables that may be included in otherwise high-ranking models when relying solely on a model's AICc ranking for evaluation. We completed goodness of fit tests using the most general model for each species (Cooch and White 2019).

We also attempted to use our four-survey dataset to investigate interactions between species. We paired detection data for coyote and bobcat, coyote and red fox, coyote and cottontail, bobcat and red fox, bobcat and cottontail, and red fox and cottontail, coding detection histories for each possible outcome (MacKenzie et al. 2018). We used the program PRESENCE (Hines 2006) to run single-season, two-species occupancy models without covariates to estimate how likely each pair of species was to cooccur compared with what one would expect if each species was distributed independently (MacKenzie et al. 2018).

We monitored 34 camera sites for a total of 1,947 TD of effort from September 2014 to July 2015. Logistically, it was not possible to survey every site with equal effort, but we surveyed every site for at least 28 TD (two surveys), and the average over all sites was 56 TD (four surveys). Latencies to first detection for bobcat, coyote, and red fox were 28, 28, and 12 d, respectively (Table S1, Supplemental Material). We detected gray fox at only one single site and did not analyze further with occupancy modeling. In addition to our target species, we identified 16 additional mammal and seven avian species during this study (Table S2, Supplemental Material).

In all, we detected carnivores thought to hunt New England cottontails at 25/34 sites (0.74 naive occupancy). Estimated detectability of at least one species during a survey was 0.52 (SE = 0.06), and we estimated that 0.81 (SE = 0.09) of the sites were used by predators (Table 2). We detected coyotes 23 times at 17 sites (Figure 2) during the four surveys, making them the predator most often encountered during the study (Table 2). Bobcats were the second most often encountered predator that we observed during surveys, with 15 detections at 11 sites throughout the survey area. We detected red fox eight times at only five sites, but we detected them the quickest where they were present (Table S1, Supplemental Material).

The estimated probability of detection in a 2-wk survey period was similar between bobcat (p = 0.25, SE = 0.09) and coyote (p = 0.26, SE = 0.07) but higher for red fox (p = 0.46, SE = 0.16). Overall estimated occupancy among the target species differed; occupancy was greatest for coyotes (Ψ = 0.81, SE = 0.21), followed by bobcats (Ψ = 0.53, SE = 0.19) and red fox (Ψ = 0.17, SE = 0.07; Table 2). The multistep approach to modeling for each species provided estimates of the effects of covariates on occupancy and detectability parameters. For coyotes, the intensity of cottontail detections at a site (RabInt) had the greatest cumulative AICc weight for occupancy, while detectability was constant across surveys and sites. Coyote occupancy in eastern Connecticut (Ψ_East = 0.83, SE = 0.26) was slightly greater than in western Connecticut (Ψ_West = 0.78, SE = 0.24), but regional differences were not supported in the model set (Table 3).

Bobcat detectability and occupancy were most influenced by two covariates. The only two covariates to accumulate enough support in step 2 included region for occupancy (0.50 AICc weight) and RabInt for detectability (0.71 AICc weight; Table 3). These two covariates make up our optimal predictive model for bobcats in our area. Bobcat occupancy differed between focus area regions (β = 3.53, SE = 3.20), with greater occupancy west of the Connecticut River (Ψ_West = 0.89, SE = 0.32; Ψ_East = 0.20, SE = 0.14). While there was sufficient support for the regional parameter using cumulative AICc weight, the 95% confidence interval of the β estimate for the regional parameter estimate was large and overlapped zero. Greater intensity of rabbit activity at a site (RabInt) significantly increased the likelihood of bobcat detections (β = 0.45, SE = 0.14).

Increasing road densities within the 5-km² buffer around cameras was the best predictive covariate for red fox detectability (Table 3). Road densities had a small positive, but not significant, effect on red fox detectability (β = 0.001, SE = 0.001). The best overall predictive model for red fox did not include any covariates for occupancy; consequently, the best detectability model was also the best overall model and performed better than the base model (Table 4). While our top model suggests that road densities are important to the fox detection process, the β value was not significant. Therefore, we cannot conclude anything other than detection and occupancy are constant among sites.

We failed to find any interactions between species using two-species occupancy models in PRESENCE. Excessively large standard errors in our results prevented us from drawing any conclusions. We believe that we did not have sufficient power in the dataset to make meaningful interpretations about species interactions and that additional sampling is required to evaluate this hypothesis.

We detected fisher at 3/34 sites (8% naive occupancy), and all sites were located east of the Connecticut River. We detected weasel species at 4/34 sites (11% naive occupancy) east and west of the Connecticut River. The low detection rates and the relatively few sites of detection prevented us from modeling their species-specific occupancy and detectability in early successional habitat. We detected cottontails 34 times at 16 sites (47% naive occupancy, Ψ = 0.50, SE = 0.09) and had higher detectability than any other species measured in the study (p = 0.60, SE = 0.07; Table 4). We found that 12 of 16 sites (75%) with cottontails had at least one predator species present during sampling.

We detected cottontails and their key predators throughout our study area in early successional habitat. However, our estimates of occupancy and detectability were not homogeneous and varied between species and sample sites. The limited number of detections of gray fox, fisher, and weasels in early successional habitat suggests that most of the potential predation pressure on cottontails is associated with coyote, bobcat, and red fox. We only found a positive association between the intensity of cottontail detections and occupancy for bobcats; therefore, predation pressure from bobcats may be greater than other species. Also, predation from bobcats may be greater in the western part of the state. To our knowledge, this is the first study to systematically survey for carnivores throughout New England cottontail focus areas in Connecticut, and we suspect that similar patterns of occupancy exist in early successional habitat outside of these focus areas.

Our detection data and results from our occupancy models suggest that coyotes have successfully expanded their area of occupancy throughout Connecticut since their first observation in the 1950s and confirm public sighting data that indicates coyotes are now found in many suburban communities throughout the state (CT DEEP 2015). Over the last century, coyotes have naturally expanded their range from the Midwestern United States into the Northeast despite efforts to eradicate the species as a livestock threat (Gompper 2002). We predicted that coyotes would occupy a higher proportion of sites in western Connecticut, but the data did not support our hypothesis. Our prediction was based on historical reports of their occurrence in the western part of the state and the potential for the Connecticut River to serve as a dispersal barrier. Coyotes are vagile carnivores and are known to use roads and bridges as travel corridors (Way 2009). Their presence in the eastern part of the state may reflect their ability to use existing river crossings combined with range expansion from the north that began in the late 1800s as coyotes moved into the Eastern United States and Canada following the decline of larger predators (Fener et al. 2005).

Coyotes may pose a minimal risk to cottontails in early successional habitat in Connecticut. Coyotes were not originally a threat to New England cottontails in the region, as they were not present in Connecticut at the time of European settlement (DeGraaf and Yamasaki 2001). Studies of coyotes outside the region confirmed that rabbits are part of the coyote diet (Latine and Giuliano 2017; Santana and Armstrong 2017). Coyote occupancy was negatively associated with the localized activity of cottontails. Coyotes have a wide dietary breadth (Colborn et al. 2020) and may not be specifically looking for cottontails in early successional habitat or may be avoiding aggressive confrontations with bobcats that compete for cottontails. In Connecticut, white-tailed deer Odocoileus virginianus seem to be the preferred prey and were found in 43% of road-killed and harvested coyote stomachs collected from 1993 through 2006 that contained food. Cottontail species were not detected in these samples, but the data may also reflect the relatively low abundance of cottontails compared with other small mammals (P. Rego and J. Hawley, CT DEEP, unpublished data).

Detection data supported our prediction that bobcats are more likely to occur in the early successional habitat of western Connecticut based on both recent and historical occurrence records (P. Rego and J. Hawley, CT DEEP, unpublished data). In Connecticut, bobcats experienced a significant range reduction during the 19th and early 20th century and persisted in small numbers in the northwestern part of the state (J. Hawley, CT DEEP, personal communication). Bobcat populations began to recover when agricultural land use declined in the early 20th century, and abandoned fields reverted to early successional habitats with diverse and abundant prey (Litvaitis 1993, 2001; Litvaitis et al. 2006b; P. Rego and J. Hawley, CT DEEP, unpublished data). Increased trapping rates throughout the region (Pollack 1949; Litvaitis 1993, 2001; Litvaitis et al. 2006b; P. Rego and J. Hawley, CT DEEP, unpublished data) and bounties paid by local governments in Connecticut suggest that bobcats made a strong recovery during this time. Bounty records between 1939 and 1948 from towns in Massachusetts that border Connecticut suggest that recovery was much stronger in the western portion of the state (Pollack 1949). However, forest maturation (habitat degradation) and expanding human development (habitat loss) throughout the 1960s and 1970s led bobcats in Connecticut and most of New England into decline once again (Litvaitis et al. 2001, 2006b; P. Rego and J. Hawley, CT DEEP, unpublished data) and resulted in the classification of bobcats as a protected furbearer species in Connecticut. Bobcats have benefited from their protected status, and sighting reports submitted to CT DEEP indicate that bobcats are throughout the state; however, most of the bobcat sightings were reported from western Connecticut (2009–2016). This is consistent with the greater number of detections in that half of the state in this study (P. Rego and J. Hawley, CT DEEP, unpublished data). We acknowledge that there was seasonal variation in the timing of surveys across sites and in the number of surveys per camera site that could potentially influence our detection rates (Figure S1, Supplemental Material). However, we attempted to account for seasonal variation in our sampling by including an average monthly temperature covariate in our models. The covariate was not significant in our models; therefore, we believe our detection rates are reasonable.

Findings from other studies show that bobcats tend to avoid or are less likely to occur in areas with higher road densities (Lesmeister et al. 2015; Litvaitis et al. 2015). We found limited support of this, despite a road density covariate making it into step 2 of the modeling process. We detected more bobcats at camera sites where the intensity of cottontail activity was greater. Our finding suggests that where bobcats are already present, they are more likely to use an area in the near term if there are high levels of recent cottontail activity. One plausible explanation for this is that bobcats are allocating more time to areas around the cameras where important prey species (cottontails) are more likely to be located.

We detected more red fox at occupied sites with higher road densities within a 5-km² buffer around camera sites. Red fox may be taking advantage of higher concentrations of small mammal prey in developed areas. Alternatively, red fox may use early successional habitat in developed areas as a strategy to avoid less developed areas where other intraguild predators and competitors may spend more time (Samuel and Nelson 1982; Dekker 1983; Melville et al. 2015). Previous studies reported home range overlap of the four focal predators (Major and Sherbure 1987; Litvaitis and Harrison 1989; Chamberlain and Leopold 2005), while others showed that coyotes displace or express aggression toward fox (Schmidt 1986; Sargeant et al. 1987; Harrison et al. 1989). CT DEEP's 2015 Wildlife Action Plan noted that both fox species may be declining due to the loss of early successional habitats and competition with coyotes. Considering that this study was conducted in what is thought to be a prey-dense habitat, there was an expectation that fox would be detected at a greater number of locations, but the rather limited number of fox detections could be the result of high occupancy of coyotes, displacing fox in early successional habitat. Our species interaction models did not provide any information that would help address competitive interactions between carnivores in our study. It is noteworthy that we detected gray fox only once during our pilot study and once during the present study. Gray fox may simply be less abundant in Connecticut than the other focal predators. Coyotes are also thought to negatively affect bobcat populations in the Northeastern United States (Litvaitis and Harrison 1989; Moruzzi et al. 2002).

Anecdotally, the presence of both coyotes and bobcats at many of the same camera sites suggests that coexistence between the two species may be possible through some form of temporal partitioning of activity patterns. Across all sites, bobcats tended to be more active between 0600 and 2100 hours, with highest activity in early afternoon (Figure S3, Supplemental Material). This period of higher activity for bobcats coincided with the time coyotes seemed to be least active in our study. Coyote detections occurred throughout the day, peaking midmorning when bobcat activity was at its lowest (Figure S4, Supplemental Material).

Predators of New England cottontails in Connecticut experienced dramatic changes following European settlement. In Connecticut, a historical period of decline preceded a more recent widespread recovery through natural recolonization. As predators have increased their numbers and distribution, their potential to influence the persistence of important prey species through predation pressure has also increased (Litvaitis and Villafuerte 1996; Litvaitis 2001). The New England cottontail is a habitat specialist, is found in only five genetically and geographically isolated populations in the Northeastern United States, and is dependent on habitat for foraging food and eluding predators (Fenderson et al. 2011). The results of this study highlight the importance of understanding local distributions of carnivores within habitats of interest. Regional differences in occupancy for bobcat may be a concern for recovery of New England cottontails in western Connecticut compared with eastern Connecticut if predator occupancy is associated with cottontail mortality. It is thought that bobcats selectively hunt for cottontails more than other predators (i.e., coyotes and fisher), and we observed that more sites were occupied by bobcats in the western part of the state and that bobcats were targeting areas where there was more cottontail activity. We did not see the same association between cottontail detections and occupancy in red fox and coyote. Therefore, they may not be specifically searching for cottontails. Additional data on the abundance of carnivores in early successional habitat would be beneficial. Habitat restoration efforts that promote larger patches of cottontail habitat and connectivity between patches will benefit cottontails because predation risk and localized extinctions are more likely in small, isolated patches (Barbour and Litvaitis 1993). Larger and more connected tracts of habitat will provide more cover and allow for larger populations of New England cottontails to persist in the presence of carnivores that are also recovering from historical declines. Wildlife and habitat managers have continued to prioritize early successional habitat restoration, and while populations of some early successional obligate species have been increasing (CT DEEP 2015), New England cottontails have continued to decline (Rittenhouse and Kovach 2020).

Our findings demonstrate that early successional habitat is not used uniformly by members of the local carnivore guild. Rather, the probability of use differed between species and, in some cases, by site. Differences in occupancy and detectability for some target species may be influenced by the natural landscape, human development, past historical events, movement rates, and local abundance. Over time, the factors that influence carnivore use (e.g., habitat suitability) could have implications for how fast New England cottontail populations recover. Future studies could be directed toward surveying a greater number of habitat types to better understand how carnivores use the overall landscape in Connecticut and to resample early successional habitat with the addition of more sites, sampled consistently as they become accessible. We also recommend that sampling during colder portions of the year, when prey availability is lower, may increase detection rates, leading to more efficient sampling. We believe that having a limited number of cameras to deploy at any one time (8–10) combined with delays in access permits were the greatest factors affecting when sites could be surveyed during the project. Genetic methods that differentiate between New England and eastern cottontails could also be combined with camera surveys to improve our understanding of predator–cottontail relationships in early successional habitat.

Please note: The Journal of Fish and Wildlife Management is not responsible for the content or functionality of any supplemental material. Queries should be directed to the corresponding author for the article.

Table S1. Carnivore detections, observed occupancy, and latency to first detection (LTD); the average number of trap days (TD) to first detection of a species at a site. Values are reflective of all detection data that we collected during camera surveys between September 2014 and July 2015 to identify distribution, occupancy, and detectability of the four primary mammalian predators of New England cottontails Sylvilagus transitionalis in Connecticut.

Available: https://doi.org/10.3996/JFWM-21-049.S1 (833 KB DOCX)

Table S2. Nontarget species that we identified at 34 sites during camera surveys completed between September 2014 and July 2015 in the New England cottontail Sylvilagus transitionalis focus areas of Connecticut. We performed camera surveys to identify distribution, occupancy, and detectability of the four primary mammalian predators of New England cottontails in the area.

Available: https://doi.org/10.3996/JFWM-21-049.S1 (833 KB DOCX)

Table S3. List of models that we ran in MARK (White and Burnham 1999; Cooch and White 2019) derived from detection data collected between September 2014 and July 2015 in New England cottontail Sylvilagus transitionalis focus areas to estimate occupancy (Ψ) and detectability (p) of the four primary mammalian predators of New England cottontails in Connecticut. We ran models during step 1 of model evaluation for bobcat Lynx rufus, coyote Canis latrans, and red fox Vulpes vulpes. We used the same list of models for each species, except that while we ran bobcat and coyote as two groups (east and west), we ran red fox as a single group. We ran 35 models for detectability, and we ran 27 models for occupancy.

Available: https://doi.org/10.3996/JFWM-21-049.S1 (833 KB DOCX)

Figure S1. Distribution of camera effort between eastern Connecticut and western Connecticut survey regions and across the survey season from September 2014 through July 2015. We attempted to apply similar camera effort at sites in each region throughout the project. Black bars indicate the approximate date and duration of camera deployment at each site.

Available: https://doi.org/10.3996/JFWM-21-049.S1 (833 KB DOCX)

Figure S2. Diagram outlining the process of calculating cottontail Sylvilagus activity-related covariates from camera detections for use in modeling carnivore occupancy and detectability. This diagram uses hypothetical data from three surveys at five sites. We used the number of detections during an individual survey as a survey-specific covariate for the level of cottontail activity near the camera during that survey. We then averaged those survey-specific numbers for each indiviudal site to use as a site-specific covariate for the level of overall cottontail activity at the site. We conducted camera surveys in Connecticut between September 2014 and July 2015.

Available: https://doi.org/10.3996/JFWM-21-049.S1 (833 KB DOCX)

Figure S3. Histogram depicting the frequency of detection events split into 3-h time blocks for coyotes Canis latrans and bobcats Lynx rufus. We collected detections in Connecticut over an 11-mo period, September 2014 to July 2015, within early successional habitat using triggered cameras with visual and scent lures. Peak activity time for bobcats coincided with the lowest activity time for coyotes, supporting the possibility of temporal resource partitioning.

Available: https://doi.org/10.3996/JFWM-21-049.S1 (833 KB DOCX)

Figure S4. A bobcat Lynx rufus is detected by a triggered camera while it rubs against a wooden post that is coated with a scent lure. We used camera detections of bobcats and other carnivores to estimate species occupancy and detectability within early successional habitat throughout Connecticut between September 2014 and July 2015.

Available: https://doi.org/10.3996/JFWM-21-049.S1 (833 KB DOCX)

Reference S1.[CT DEEP] Connecticut Department of Energy and Environmental Protection. 2020. Wildlife fact sheet: bobcat Lynx rufus.

Available: https://doi.org/10.3996/JFWM-21-049.S2 (280 KB PDF) and https://portal.ct.gov/DEEP/Wildlife/Learn-About-Wildlife/Wildlife-Fact-Sheets

Reference S2. Connecticut Department of Energy and Environmental Protection. 2020. Wildlife fact sheet: eastern coyote Canis latrans.

Available: https://doi.org/10.3996/JFWM-21-049.S3 (4.562 MB PDF) and https://portal.ct.gov/DEEP/Wildlife/Learn-About-Wildlife/Wildlife-Fact-Sheets

Reference S3. Fuller S, Tur A. 2012. Conservation strategy for the New England cottontail Sylvilagus transitionalis. U.S. Fish & Wildlife Publications 320.

Available: https://doi.org/10.3996/JFWM-21-049.S4 (4.562 MB PDF) and https://digitalcommons.unl.edu/usfwspubs/320/

Reference S4. Gilbart M. 2012. Under cover: wildlife of shrublands and young forest. Cabot, Vermont: Wildlife Management Institute.

Available: https://doi.org/10.3996/JFWM-21-049.S5 (9.633 MB PDF) and https://youngforest.org/sites/default/files/Under_Cover-010412_FINAL.pdf

Funding for the study was provided by the Norcross Wildlife Foundation-AV Stout Fund and Central Connecticut State University. We greatly appreciate the Farmington Land Trust, Aspetuck Land Trust, Redding Land Trust, New Hartford Land Trust, Joshua's Tract, CT DEEP, and other landowners for allowing us to access specific sites and to biologists Paul Rego and Jason Hawley for sharing their thoughts, time, and unpublished historical data. We would also like to thank the Journal reviewers and Editors for the considerable time and effort in helping us to improve this work. Finally, we would like to K.J. and P.D. for the technical support from 2017 to 2019.

Any use of trade, product, website, or firm names in this publication is for descriptive purposes only and does not imply endorsement by the U.S. Government.