Is conservation based on best available science creating an ecological trap for an imperiled lagomorph?

Abstract Habitat quality regulates fitness and population density, making it a key driver of population size. Hence, increasing habitat quality is often a primary goal of species conservation. Yet, assessments of fitness and density are difficult and costly to obtain. Therefore, species conservation often uses “best available science,” extending inferences across taxa, space, or time, and inferring habitat quality from studies of habitat selection. However, there are scenarios where habitat selection is not reflective of habitat quality, and this can lead to maladaptive management strategies. The New England cottontail (Sylvilagus transitionalis) is an imperiled shrubland obligate lagomorph whose successful recovery hinges on creation of suitable habitat. Recovery of this species is also negatively impacted by the non‐native eastern cottontail (Sylvilagus floridanus), which can competitively exclude New England cottontails from preferred habitat. Herein, we evaluate habitat quality for adult and juvenile New England and eastern cottontails using survival and density as indicators. Our findings did not support selection following an ideal free distribution by New England cottontails. Instead, selected resources, which are a target of habitat management, were associated with low survival and density and pointed to a complex trade‐off between density, survival, habitat, and the presence of eastern cottontails. Further, movement distance was inversely correlated with survival in both species, suggesting that habitat fragmentation limits the ability of cottontails to freely distribute based on habitat quality. While habitat did not directly regulate survival of juvenile cottontails, tick burden had a strong negative impact on juvenile cottontails in poor body condition. Given the complex interactions among New England cottontails, eastern cottontails, and habitat, directly assessing and accounting for factors that limit New England cottontail habitat quality in management plans is vital to their recovery. Our study demonstrates an example of management for possible ecological trap conditions via the application of incomplete knowledge.


| INTRODUC TI ON
It is broadly recognized that the availability and quality of habitat influences wildlife density, survival, and reproduction and, as a result, strongly regulates population size (Heinrichs et al., 2016;Johnson, 2007). In fact, habitat loss and degradation are frequently cited as the leading causes of species declines globally (Butchart et al., 2006;Gibbons et al., 2000;Tilman et al., 2017). Thus, management to create habitat and improve habitat quality is often a primary focus of conservation efforts. These efforts have had notable success in recovering declining wildlife populations by improving survival and reproductive rates and increasing population size (Catlin et al., 2011;Innes et al., 1999;Warren, 1991;Whitehead et al., 2008). However, there are also many instances wherein management efforts were ineffective or even detrimental to target species, so much so that ecological restoration is considered among the top three anthropogenic causes of ecological traps (Armstrong et al., 2007;Ausden et al., 2001;Robertson et al., 2013). These failures are generally attributed to overlooking key factors that alter habitat quality, leading to ineffective conservation measures and a waste of limited conservation funding.
In order to enact effective habitat management, a robust understanding of how survival, reproduction, and density vary among patches of differing quality is recommended (Johnson, 2007).
However, measures of survival, reproduction, or density are often prohibitively difficult and costly to obtain and management decisions must proceed with the "best available science" (Doremus, 2004). This is particularly true for rare or threatened species given legal mandates to use "best available science" under the U.S. Endangered Species Act and considering their scarcity and low population densities often preclude the effective application of traditional research methods (Murphy & Weiland, 2016). Conservation and management strategies using "best available science" often apply existing information from similar systems or rely on ecological theory to make broader inferences from correlative studies-the latter is commonly the case with habitat selection studies.
Selection for high-quality habitat evolves due to its effects on individual fitness (Fretwell, 1969), and strong selection for habitat features is evident across diverse taxa (Binckley, 2017;Crosby et al., 2019;Gorosito et al., 2018;Sawyer & Brashares, 2013;White et al., 2019). When species are at equilibrium and can select habitat according to an ideal free distribution, selection is indicative of habitat quality (Jones, 2001;Pulliam & Danielson, 1991). Thus, practitioners often use patterns in species occupancy or resource selection to infer habitat quality and produce management recommendations (Johnson, 2007). Although this approach can be effective, habitat selection may not always be a reliable indicator of fitness and density (Van Horne, 1983). The relationship between survival, reproduction, and density and habitat selection can become decoupled in altered or disturbed systems, resulting in weak relationships between selected resources and fitness or in ecological traps where fitness is negatively impacted by preferentially selected resources (Battin, 2004;Robertson & Hutto, 2006). Inferences obtained from decoupled systems can lead to prioritizing conservation for poor-quality habitat, which is at best inefficient or at worst can exacerbate declines of imperiled target species (Hale & Swearer, 2017).
Moreover, preferential selection for poor-quality habitat is not uncommon in wildlife populations. For example, the incorrect use of cues to assess habitat quality can lead to selection for resources that are attractive but detrimental to fitness (Nawrocki et al., 2019;Titeux et al., 2019). Numerous processes can result in decoupled systems, for example, historic competitive interactions (Morris et al., 2000), despotic selection where dominant individuals exclude subordinate individuals to high-density low-quality habitat (Fretwell, 1969), or when populations are regulated through density-dependent mechanisms such as predation (Morris, 2005). Conflicting patterns of apparent habitat selection and fitness can also result where systems are not able to reach equilibrium (Shochat et al., 2002), such as when habitat fragmentation limits the movement capacity of individuals and thus their ability to distribute among patches on the basis of habitat quality (Stamps et al., 2005). These phenomena can result from the introduction of novel resources (Remeš, 2003), competitors (Cole et al., 2005), or predators (Igual et al., 2007) and occur more frequently in human-disturbed systems, such as in anthropogenically managed shrublands and grasslands (Bock & Jones, 2004;Robertson et al., 2013;Shochat et al., 2005).
Given the limited availability of conservation funding, it is critical that enacted management is effective at meeting targets. Hale and Swearer (2017) recently highlighted how management can create ecological traps for target species when these actions do not link habitat selection to habitat quality, but they note this possibility has largely been ignored in wildlife research. One species of intense management focus is the New England cottontail (Sylvilagus transitionalis, F I G U R E 1 New England cottontail (Sylvilagus transitionalis) resting in native-dominated late successional shrubland Figure 1), a shrubland obligate lagomorph that has experienced dramatic population declines since the mid-1900s .
It is listed as vulnerable by the IUCN and was considered for federal listing under the U.S. Endangered Species Act, but denied status as ongoing efforts were deemed sufficient for its recovery (Litvaitis & Lanier, 2019;USFWS, 2015). Declines of New England cottontails are primarily attributed to habitat loss and fragmentation as a result of anthropogenic development and natural forest succession, scramble competition with introduced eastern cottontails over unoccupied areas (Sylvilagus floridanus), and the modification of habitat by invasive shrub species (Cheeseman et al., 2018;Litvaitis et al., 2008).
These declines have gained regional conservation attention, and recovery plans hinge on the successful creation of 20,000 hectares of habitat (Fuller & Tur, 2012). However, despite considerable conservation effort, the area of occupancy for New England cottontails has declined by an additional 50% in the last decade (Rittenhouse & Kovach, in press), raising concerns about the efficacy of existing management within these highly impacted ecosystems.
Past research addressing habitat quality for New England cottontails has largely focused on studies of habitat use and selection (Brubaker et al., 2014;Buffum et al., 2015;Cheeseman et al., 2018Kilpatrick et al., 2013;Litvaitis et al., 2003Litvaitis et al., , 2006Shea et al., 2019;Tash & Litvaitis, 2007). Demographic studies related to habitat quality have been mostly limited to assessments of survival at their northern range boundary (Barbour & Litvaitis, 1993;Brown & Litvaitis, 1995) where eastern cottontails are still scarce, emphasizing the importance of vegetative cover with high stem densities and the influence of large habitat patches on survival (Barbour & Litvaitis, 1993;Brown & Litvaitis, 1995). As a result, conservation management planning for this species has focused on creating early successional shrublands with a high density of native shrubs and low overstory canopy closure to facilitate a denser understory (New England Cottontail Regional Technical Committee, 2017).
Unfortunately, management plans do not currently consider the multifaceted relationships among habitat quality, habitat selection, survival, and density nor how the present landscape context, including possible competitive exclusion by an introduced competitor, habitat modification by invasive shrubs, or dispersal costs in highly fragmented landscapes, might affect these relationships. More recent resource selection studies have supported recommendations to manage for early successional shrubland only where eastern cottontails are not prevalent, demonstrating that New England cottontails select for features typically associated with early successional shrublands such as low canopy closure (Cheeseman et al., 2018).
However, these authors also found that where eastern cottontails are prevalent, they displace New England cottontails from early successional shrublands. In areas where eastern cottontails are prevalent, New England cottontails instead select for areas of high canopy closure and moderate to high densities of the invasive shrub Japanese barberry (Berberis thunbergii), which are themselves indicative of mid-to late successional shrublands (Cheeseman et al., 2018). These findings raise concerns over the quality of later successional shrublands for New England cottontails and the suitability of early successional shrublands where eastern cottontails are sympatric. In this study, we examine how resources and landscape features, cottontail health, and competitor prevalence influence habitat quality for New England and introduced eastern cottontails. We examine the survival and density of each species as indicators of habitat quality and assess the impact of eastern cottontail abundance on New England cottontail density across patches with varying resources.

| Study area
We conducted our study at 15 sites in the Hudson Valley region of New York, USA, from December 2013 through April 2017. The climate is temperate, average annual rainfall for this period was 51.8 cm (SD = 36.5 cm), and winter snow depth was highly variable within and between years, varying from 0 to 63.5 cm (www.ncdc.noaa.gov).
Sites were characterized by different-aged successional shrublands, persistent forested ericaceous shrublands, and/or persistent forested shrub wetlands. Successional shrubland was classified as early, mid, late, and persistent based on the density and height of shrubs, degree of overstory canopy closure, and dominant shrub species in the case of persistent shrublands following Cheeseman et al. (2018).
Successional shrublands were often dominated by invasive multiflora rose (Rosa multiflora), Japanese barberry, and Oriental bittersweet (Celastrus orbiculatus); and ericaceous shrublands contained Vaccinium spp. and mountain laurel (Kalmia latifolia) understory and an oak (Quercus spp.), hemlock (Tsuga canadensis), and/or white pine (Picea glauca) overstory. Sweet pepperbush (Clethra alnifolia) and swamp azalea (Rhododendron viscosum) were common in forested shrub wetlands. Overstory tree species composition within forested wetlands was variable and representative of the general tree communities at the site.

| Field methods
From December 2013 to October 2016, New England and eastern cottontails were live-trapped using box traps baited with apple.
Traps were set along adaptive transects spanning site boundaries at approximately 25-m intervals. Captured cottontails were weighed, and morphological measurements were obtained to determine species identity and age. Each individual was examined for parasites, and the bone protrusion, rump, and loin muscles were examined to assess body condition and scored following Bonanno et al. (2008).
For analyses, we collapsed categories into "good" and "poor" body condition where individuals scoring in the lowest category in at least two bodily locations were considered to have "poor" body condition, and all others were considered in "good" body condition. For newly captured individuals, a tissue biopsy was obtained from the ear to confirm species identity through genetic analyses and a blood sample was collected in microhematocrit tubes for a hematocrit assessment. Tissue samples were stored in 100% ethanol for later genetic analysis. DNA from preserved tissues was extracted using the Qiagen DNeasy Blood and Tissue Kit (Qiagen Inc.) and rabbit species identified by PCR amplifying the D-loop region of the mitochondrial DNA, and subsequently analyzing this region using two restriction fragment length polymorphism assays (Kovach et al., 2003;Litvaitis et al., 1997;Ryan et al., 2016).
Microhematocrit tubes were centrifuged for 5 min within 8 hr of collection on a LW Scientific LWS-M24 Microhematocrit Centrifuge machine at 14,800 g. Readings were taken visually and recorded to the nearest percent (Ryan et al., 2016).
Cottontails under 800 g were affixed with a 1.1-g glue-on radio transmitter with a winged mesh attachment following protocols outlined by Estes-Zumpf and Rachlow (2007). Effort was made to recapture and reaffix transmitters to juveniles at regular intervals.
Cottontails over 800 g were affixed with a 24-g radio transmitter with a zip-tie collar attachment (Advanced Telemetry Systems, Isanti, Minnesota). Transmitters were equipped with 8-hr mortality switches. Following an adjustment period of 72 hr, cottontails were monitored weekly until a mortality event occurred, until the signal was lost, or at the termination of the study in April 2017. Any cottontails that died within the first 72 hr postcollaring were censored from analyses. To obtain fine-scale habitat data, we tracked cotton- We obtained estimates of leaf-off season canopy closure, shrub height, and stem density of shrub species every 50 m in a grid across used portions of study sites as determined from telemetry data.
Canopy closure during the leaf-off season was estimated for the four cardinal directions at each sampling location with a densiometer held at a height of 1 m (Cheeseman et al., 2018). These estimates were averaged by sampling location to obtain the final canopy closure estimate. The number of woody stems less than 10 cm DBH and originating within a 10-× 1-m plot was tallied by species and totaled to obtain total stem density of the plot. Stems were classified as native or invasive, and palatable native species were identified using Pringle (1960). Vegetation data were geolocated and raster-

| Survival
Cottontails can be aged using hind foot length until they reach adult size and sexual maturity at around 4 months of age (Chapman, 1975;Chapman et al., 1980), after which point age of mature cottontails cannot be assessed noninvasively (Bothma et al., 1972). We therefore analyzed survival data at two age classes, postnestling juveniles and adults. Juveniles were available for capture after leaving their nest (i.e., around 14 days) and considered juveniles until their hind foot length was >80 mm and their weight was >800 g (Bothma et al., 1972;Chapman et al., 1980). As it was not consistently possible to recapture individuals to reassess weight and hind foot length, all remaining juveniles were considered adults 4 months after the end of the recorded breeding season (i.e., 1 December). Juvenile and adult survival were assessed separately using logistic exposure models, which allow for ragged entry, removal of individuals, and time-varying covariates, and accommodate uneven intervals between observations (Shaffer, 2004). We used a sequential modeling approach and first developed three candidate sets of models with alternative environmental (seven adult and six juvenile models), health (14 adult and five juvenile models), and resource and landscape (23 adult and 19 juvenile models) parametrizations thought to influence survival of juvenile and adult cottontails ( Table 1, Table A1; Arnold, 2010). We developed a combined candidate set of 13 adult and seven juvenile models including combinations of variables from the environmental, heath, and resource and landscape factor models with a ΔAIC c value ≤ 2.0 (Table A1). We excluded correlated (r > 0.7) and nested variables (i.e., Japanese barberry stem counts are nested within total stems) from consideration in the same model.
We considered the environmental model set to control for variability in survival due to weather, which has been hypothesized to impact survival of cottontails at northern latitudes (Bond et al., 2001;Weidman & Litvaitis, 2011). As we were interested in examining the difference between New England and eastern cottontails, "species" was included as a factor in all but the null models. Season has been found consistently an important predictor of lagomorph survival at northern latitudes (Keith & Bloomer, 1993;Trent & Rongstad, 1974;Weidman & Litvaitis, 2011). Therefore, we included season in all models of adult survival. As parturition occurs spring through late summer and most juveniles have matured to adults by the leaf-off season, we did not include season in models for juvenile survival.
Each model set was evaluated using an information theoretic approach and Akaike information criterion corrected for small sample size (AIC c ; Burnham & Anderson, 2002). The final combined model set included combinations of variables within all supported weather, health, and landscape models and a null model. We performed model selection of the final candidate set using AIC c , and models with a ΔAIC c ≤ 2.0 and with a corresponding reduction in deviance were considered to have support (Arnold, 2010). All analyses were conducted in Program R 3.5.3 (R Core Team, 2019). We used model pre-

| Capture per unit effort
Low recapture rates and dense habitat features prevented estimation of density using mark-recapture or distance survey methods so we estimated capture per unit effort (CPUE) as an indicator of local abundance (Barbour & Litvaitis, 1993). Our metric of CPUE was the number of new captures per trapping session. We modeled CPUE as a function of cottontail species and shrubland type, and included an index of competition for New England cottontails (i.e., interaction between New England cottontails, shrubland type, and the number of eastern cottontails known in the patch). Patches within sites were manually digitized, and shrubland type was identified as early, mid, late, or persistent based on canopy closure, shrub height, and dominant shrub communities inferred from aerial imagery, ground truthing, and vegetation surveys. We fit the model using a mixedeffect zero-inflated Poisson regression in the "glmmTMB" package in Program R, with site as the random effect because we had data from multiple sessions from most sites. No New England cottontails were captured in early successional patches where eastern cottontails were present, so the interaction between the early successional level of the shrubland-type factor and eastern cottontail abundance was not included in the model. The number of trap nights and area TA B L E 1 Covariates considered in three candidate logistic exposure model sets for assessing the effects of weather, individual health, and resource and landscape factors on survival of juvenile and adult New England cottontails (Sylvilagus transitionalis) and eastern cottontails (Sylvilagus floridanus) in New York, 2013-2017

| Juvenile survival
We monitored survival of 27 juvenile New England cottontails and 25 juvenile eastern cottontails. From these individuals, we observed 4 New England and 4 eastern cottontail mortalities from 345 and 315 observations, respectively. Species, body condition, tick burden, and palatable stem densities had support from both top combined models of juvenile survival (Table 2). However, the stem density effect and the main effect for body condition lacked power as a predictors in our dataset ( Table 3). Survival of juvenile New England cottontails was higher than eastern cottontails (Table 3, predicted seasonal survival in Table A2). Tick burden had no effect on survival of juvenile rabbits in good body condition ( (Table 2).

| Adult survival
We monitored survival of 82 adult New England cottontails and 76 adult eastern cottontails. From these individuals, we observed 59 mortalities from 3,829 observations of New England cottontails and 69 mortalities from 4,753 observations of eastern cottontails.
Survival of both New England and eastern cottontails was influenced by weather, health, and landscape features, but did not appear to vary by levels of interspecific competition, which was not supported in any top model (Table 2).
Based on overlap of prediction intervals (PI), annual probability of survival for individuals in good condition and with covariates held at their seasonal means was not different between New England cottontails (x̄ = 0.32, 95% PI = 0.19-0.44) and eastern cottontails (x̄ = 0.19, 95% PI = 0.10-0.31; predicted seasonal survival in Table A2). Although all canopy closure terms in the regression model had low predictive power (Table 3) Survival was negatively correlated with movement and snow depth and did not differ between species (  Figure 4b).
Predicted cottontail survival in hypothetical shrublands varied by patch type and species. Survival of New England cottontails was lower in invasive-dominated early successional shrublands and patches with full tree canopy but low shrub understory than other patch types and only nonzero for eastern cottontails within early and mid-successional and persistent shrublands ( Figure 5). Although no other comparisons were significantly different, the results suggest that in general, New England cottontail survival was lower in early successional shrubland than late successional and persistent shrublands ( Figure 5).  (Table 4). Eastern cottontail CPUE was higher in early and mid-successional shrublands than late successional and persistent shrublands ( Figure 6). Given the mean number of eastern cottontails known alive, New England cottontail CPUE was greater in mid-successional shrubland than any other shrubland type ( Figure 6). However, in all shrubland types except persistent shrubland the CPUE of New England cottontails decreased with increasing eastern cottontails, such that with one eastern cottontail present, there was no difference between CPUE of mid-successional and persistent shrublands (Figure 7). In fact, CPUE for New England cottontails in early successional shrubland was only nonzero in patches where there were no known eastern cottontails (Figure 7). Only a single New England cottontail was captured during a session where two eastern cottontails were known present in a patch (mid-successional shrubland).

| D ISCUSS I ON
For New England cottontails and many other imperiled species that are data-limited, conservation has often proceeded based on the standard of "best available science" (Doremus, 2004). In such cases, it is common for knowledge from one part of a species' range to be applied to conservation planning in a less-studied location, or ecological theory may be leveraged to make broader inferences from correlative studies (Murphy & Weiland, 2014 is a driver of ecological traps for several understory nesting birds (Battin, 2004). It is possible that New England cottontails are using similarly inappropriate cues to evaluate habitat quality in this novel context, resulting in maladaptive selection for these areas. However, invasive exotic shrubs, such as Japanese barberry, also support higher tick abundances than areas of low or native shrub cover (Elias et al., 2006;Lubelczyk et al., 2004;Williams et al., 2009) and have been linked to increased tick burdens on New England cottontails when compared to native shrub-dominated habitats (Mello, 2018).
High tick burdens are also associated with lower fecundity (Keith & Cary, 1990), greater morbidity (Scott et al., 2014), and greater F I G U R E 3 Predicted seasonal survival and 95% prediction intervals for adult cottontails in the leaf-off (November-April) and leaf-on (May-October) seasons by (a) (Jones et al., 2018) in wildlife populations, and have been implicated in population declines of species including moose (DeIgiudice et al., 1997), several African ungulates (Lightfoot & Norval, 1981), and cottontails (Smith & Cheatum, 1944 Smith and Litvaitis (2000) found eastern cottontails better exploited resources and avoided predation within low cover areas than New England cottontails. As a result, eastern cottontails likely have an advantage in low cover patches, which could explain the species-specific trends in survival in early successional shrublands. Regardless, early successional shrublands appear to provide high-quality habitat to eastern cottontails where they are able to exclude New England cottontails (Cheeseman et al., 2018;Probert & Litvaitis, 1996). We do not know the exact mechanisms for increased survival of New England cottontails with shrubland age. One hypothesis is density-dependent predation, which has been well documented in leporids (Krebs et al., 1995). Overall cottontail density was highest in early and mid-successional shrublands, likely a result of abundant forage and cover in early and mid-successional shrublands. However, higher density in early and mid-successional shrublands could result in higher rates of density-dependent predation than in late successional and persistent shrublands. Further, New England cottontails appear less well adapted to avoiding predation than eastern cottontails and incur higher survival costs in less structurally complex habitat (Smith & Litvaitis, 1999. As a result, New England cottontails may have greater predation risk in co-occupied patches than eastern cottontails. High dispersal costs and reduced movement capacity limit the ability for species to distribute among habitats in response to density (Stamps et al., 2005). Movement distance had a high survival cost for both species. Studies examining dispersal of New England cottontails have suggested a high rate of exploratory movement, but low rates of successful dispersal, possibly due to scarcity of suitable habitat within dispersal distance (Cheeseman, 2017). These conclusions are supported by levels of fragmentation and movement observed in population genetic studies (Bauer, 2018;Fenderson et al., 2014). With high cost of movement and high search cost, New England cottontails may face high rates of diminishing fitness as population densities increase and they are unable to freely disperse across the landscape. While the impacts of limited dispersal at the metapopulation scale are widely recognized, high-risk exploratory and dispersal movements could also serve to reduce population viability at the patch level by reducing survival below sustainable levels. Reducing the distance between suitable F I G U R E 5 Survival of New England cottontail (Sylvilagus transitionalis) and eastern cottontail (Sylvilagus floridanus) in hypothetical shrublands. Predicted annual survival and 95% prediction intervals shown for hypothetical shrublands based on typical habitat metrics in New York, 2013-2017. Shrubland types include early invasive (EI, 0% tree canopy, 50 stems/10 m 2 Japanese barberry, 0 stems/10 m 2 native palatable shrubs), early native (EN, 0% tree canopy, 0 stems/10 m 2 Japanese barberry, 20 stems/10 m 2 native palatable shrubs), late invasive (LI, 100% tree canopy, 50 stems/10 m 2 Japanese barberry, 0 stems/10 m 2 native palatable shrubs), late low shrub cover (LLC, 0% tree canopy, 0 stems/10 m 2 Japanese barberry, 0 stems/10 m 2 native palatable shrubs), and persistent (PER, 100% tree canopy, 0 stems/10 m 2 Japanese barberry, 20 stems/10 m 2 native palatable shrubs) patches through placement of newly created habitat or the creation of stepping-stone patches or corridors may improve survival of dispersing individuals at both patch and metapopulation scales.
In summary, we caution that by focusing management and restoration efforts on apparently attractive, but low-quality habitat, the conservation strategy for New England cottontails may be fostering patches that promote eastern cottontails but act as ecological traps for New England cottontail. Early successional shrublands have been a target of New England cottontail conservation over the past decade, at considerable expense to regional wildlife management agencies and federal partners. Management actions should consider the trade-offs between survival, density, shrubland, age, and presence of eastern cottontails when planning habitat restoration for the New England cottontail. In general, early successional shrublands are unlikely to provide high-quality habitat, particularly where We further note that management to reduce invasive shrubs within shrublands without providing alternative cover will result in the destruction of habitat, or risk creation of sink habitat for New England cottontails.
Conservation of wildlife species and their habitat must necessarily be undertaken with incomplete knowledge of system dynamics. However, in doing so, we risk enacting management strategies that are ineffective or even detrimental to management goals.
Uncertainty is often high when ecosystems are anthropogenically disturbed, and management must account for multiple interacting variables. In the case of the New England cottontail, fitness, density, and habitat selection are decoupled in highly impacted environments, leading to inaccurate inferences of habitat quality from previous habitat selection studies or when extrapolated from studies in less impacted areas. We demonstrate how complex interactions can alter habitat quality and result in wide-scale implementation of potentially maladaptive management strategies. It is imperative that population responses to conservation based on such assumptions be monitored and evaluated regularly, and that ideally, management in the face of uncertainty proceed adaptively (Murphy & Weiland, 2016).

ACK N OWLED G M ENTS
The

CO N FLI C T O F I NTE R E S T
The authors declare no conflicts of interest.

AUTH O R CO NTR I B UTI O N S
Amanda E. Cheeseman: Conceptualization (lead); data curation (lead); formal analysis (lead); funding acquisition (equal); investigation (equal); methodology (lead); project administration (  d ΔAIC c = difference in the AIC c between a given model and the top model. e w i = AIC c weights, or the probability that of the models tested the given model fits the data best.

DATA AVA I L A B I L I T Y S TAT E M E N T
f Interaction did not include all main effects.

TA B L E A 2
Leaf-off (November -April) and leaf-on (May-October) predicted probability of survival for New England cottontails (Sylvilagus transitionalis) and eastern cottontails (Sylvilagus floridanus) when covariates are held at their means