An Initial Habitat Suitability Analysis for the Red Wolf Across its Historical Range

Habitat suitability modeling is an approach for predicting the suitability of a location for a taxon based on its observed relationship with environmental conditions (Elith and Leathwick 2009). Conducting such habitat suitability analyses (HSA) for species of conservation concern are difficult when knowledge of the species derives from just a remnant of its historical range. Yet HSA are also necessary to underpin potential population restoration efforts. On the one hand, HSA based on habitat use by a relict population risk inferences that are spatially unsupported, especially if the landscape where the population persists does not reflect the diversity of landscapes being assessed. On the other hand, HSA are a primary and often necessary method for identifying potential landscapes that can accommodate reintroduced populations. The critical issues are the adequacy of the predictors used for model building (Araújo and Guisan 2006) and the recognition that model development methods and HSA functionality are customizable based on available data (Carvalho et al. 2012; Reza et al. 2013). Here we assess these issues in the context of the red wolf Canis rufus.

The red wolf is endemic to the southeastern, lower Midwestern, and mid-Atlantic regions of the United States. Little was known about the taxon until the 1960s, when the red wolf was federally listed as “threated with extinction” (USFWS 1967) under the Endangered Species Preservation Act of 1966 (ESPA 1966, as amended), and research efforts focused on determining the status of the red wolf and identifying individuals to place into a captive breeding program (Phillips et al. 2003). In 1973, the Endangered Species Act became law, and the red wolf was federally listed as endangered pursuant to the U.S. Endangered Species Act (ESA 1973, as amended). A captive breeding program was established from 14 individual canids deemed as red wolves, which were collected from the wilds of Louisiana and Texas between 1973 and 1980. Red wolves were then declared extinct in the wild in 1980. This captive breeding program formed the basis for two population restoration efforts: a nonessential experimental population (which is a determination under Section 10(j) of the Endangered Species Act that means that the specific population is not essential to the listed species' continued existence) on the Albemarle Peninsula of northeastern North Carolina in 1987, and a nonessential experimental population within the boundaries of Great Smoky Mountains National Park (GSMNP), Tennessee, in the early 1990s. The latter restoration effort ended in 1998 (USFWS 1998), meaning the only persisting wild population is the Eastern North Carolina Red Wolf Population (ENC RWP), which currently estimates 19–21 individuals (USFWS 2022).

Habitat use of the NC NEP has been well-studied. These red wolves select agricultural fields over other available habitat types (Hinton and Chamberlain 2010; Hinton 2014), perhaps because these areas offer high-quality foraging opportunities (Hinton and Chamberlain 2010) and provide cover. However, these habitats are heavily manipulated and subject to flooding for waterfowl management. As such, lands available to red wolves periodically expand and contract, and Chadwick et al. (2010) observed strong seasonality in habitat use. Red wolves used agriculture fields between July and October while they selected forested and “grass-brush areas” from November to May (Chadwick et al. 2010). Human population density is also important to red wolf habitat selection. Dellinger et al. (2013) suggest that red wolves select agricultural land and early successional fields as well as areas near secondary dirt roads and areas with lower human presence, such as forests and marshes.

Similarly, Karlin et al. (2016) assessed telemetry locations collected between 1999 and 2008 from 178 red wolves and noted that characteristics most predictive of presence were human population density, secondary road density, and agricultural area. They found that while red wolves tend to select agricultural landscapes on the Albemarle Peninsula, they are also able to persist within other landscapes. Indeed, Dellinger et al. (2011) documented a pack that successfully raised a litter of pups in a home range primarily consisting of upland forest surrounded by areas with higher human density. Thus, red wolves show habitat preferences but nonetheless successfully can use a diversity of habitat types.

In contrast, the relatively brief (1991–1998; USFWS 1998) restoration effort that occurred in GSMNP generated data on habitat use. These red wolves selected deciduous forests, pastures, and woody wetlands (Mauney 2005), which may have reflected habitat availability. Large open areas are also a predominant landscape feature in one of three sections within the park where reintroductions occurred, and home ranges established within this portion centered on these open areas. Red wolves used woody wetland areas for hunting and for resting, perhaps because human use in these areas of the park was low compared with other areas. Thus, similar to the restored ENC RWP, the Great Smoky Mountain population used a diversity of open and forested habitats.

Despite information from the two restoration efforts, it is important to note that the landscape occupied by these populations represents a small subset of the entirety of habitat features found across the historical range of the species. The dominant habitat features and the relative use of habitat types in the Albemarle Peninsula and GSMNP do not easily translate directly across the broader historical geographic range of the red wolf. Knowledge on how the red wolf used the entirety of its historical range is poorly known, although the species was once found in a wide array of habitats (Carley 1979), including habitats not found in the extant range. For instance, Russell and Shaw (1971) reported that red wolves in Texas used three broad landscape types: marshlands, prairie grasslands, and woodlands, with woodlands indicated as having the lowest relative density. Paradiso and Nowak (1972) noted that red wolves rested in open landscapes such as weedy fields, grass, or brush pastures. Red wolves may also have occupied bottomland forests and wetlands along rivers throughout their historical range (Paradiso and Nowak 1972; Riley and McBride 1972). The last strongholds of red wolves were marshes and coastal prairies within counties along the coastlines of southeastern Texas and southwestern Louisiana (Phillips et al. 2003; USFWS 2007), and introgressed red wolf alleles persist in canids inhabiting the region (Heppenheimer et al. 2018; Murphy et al. 2019; vonHoldt et al. 2021). Although care must be taken in generalizing habitat preferences from residual populations, these historical reports, like the findings from the two reintroduced populations, illustrate the ability of red wolves to occupy a wide diversity of habitats if not subject to human persecution (e.g., extensive mortality due to vehicle-impacts, shooting).

To date, five HSA have been conducted for the red wolf. Three focused on particular sites (Shaffer 2007; Jacobs 2009; Desmul 2013) and two focused on the broader historical range (van Manen et al. 2000; O'Neal 2018). Most prominently, in the late 1990s, the USFWS provided a list of 31 prospective release areas for which van Manen et al. (2000) created models to predict potential red wolf release success. Nonetheless, two decades later, red wolves remain limited to the ENC RWP, and no new population establishment efforts have occurred since GSMNP. In 2016, the USFWS announced that identifying new reintroduction locations was a priority for recovery efforts (USFWS 2016).

Our goal in this study was to use insights from previous studies to identify and rank suitable habitat and potential reintroduction sites across the southeastern United States. Presence points from the NC NEP population are not seamlessly translatable to the varied broader landscape that comprises the historical range of the species, and scant habitat use research occurred on the original red wolf population; therefore, our approach was to conduct a literature review of red wolf landscape use to identify primary habitat metrics. We then use the results of this literature review to build an HSA based on a variety of generalized biotic and abiotic factors and focused on determining the suitability of the current landscape that comprises the historical range of the species. Further, we examine federal landscapes that are presumably large enough to sustain a red wolf population, and rank these landscapes based on results of the HSA. Our approach differs from previous HSA in that we assess the entire historical range of the species rather than a subset of pre-identified localities and combine generalized baseline parameters to build models and gain initial insights. These baseline models can easily be modified or supplemented as new information becomes known.

For this HSA, the study area stretches across the 22 states (Figure 1) that comprise the majority of the hypothesized red wolf historical range. We examined theses, dissertations, government documents, and published articles for red wolf habitat use and preference (Supplemental References S1–S31). We identified these documents by examining references cited in the Red Wolf Recovery Plan (USFWS 1990), red wolf 5-y status reviews (USFWS 2007, 2018a), and the Red Wolf Species Status Assessment (USFWS 2018b), and we supplemented this through Google Scholar searches (keywords: “red wolf” and permutations of “habitat suitability” and “habitat use”). We evaluated both landscape and human-influence characteristics in the 22-state region to aid in identifying habitat suitability (Table 1). Characteristics chosen for inclusion in the analyses are widely used in carnivore HSA (Schadt et al. 2002; Martin et al. 2012; Reed et al. 2016) and provide a general understanding of suitable habitat availability based on landscape type and human disturbance. Data layers were sourced from the U.S. Census Bureau TIGER Database (Roads; U.S. Census Bureau 2020), U.S. Geological Survey Gap Analysis Protected Areas Database (Land Ownership Patterns, Agency Oversight; U.S. Geological Survey Gap Analysis Project 2020), U.S. Geological Survey (Human Population; Cities and Towns of the United States and Earth Resources Observation and Science [EROS] Data Center), and National Land Cover Database (NLCD, Landscape Type).

We uploaded the NLCD layer into ArcMap (v. 10; https://www.esri.com/en-us/arcgis/products/arcgis-desktop/overview), clipped it to the 22-state study area, and resampled it from a 30 × 30 m cell size to a 100 × 100 m cell size using the “NEAREST” technique. Using NEAREST minimizes changes to pixel values and is suitable for discrete data use, including land cover data. We reclassified the detailed NLCD layer into distinct, general categories: forest, grassland, shrubland, wetland, cropland, and pasture. We categorized human population by size as high (population ≥ 100,000 individuals), medium (population < 100,000 and ≥ 10,000 individuals), and low (population < 10,000 individuals). For interstate and highway data, we assumed that proximity to roads negatively correlates with risk of mortality. Therefore, we deemed cells located 0–1 km from a highway (Shaffer 2007) and 0–2 km from an interstate as less suitable. We also deemed as less suitable any cells located less than 2 km from a low population (Shaffer 2007), less than 5 km from a medium population, and less than 10 km from a large population. To account for this, we created buffers around each road type and human population type using Euclidean Distance.

We assigned values to each landscape, population, and road type characteristic based on a multicriteria decision-making analysis framework. When used with a geographic information system, this process combines geographical data and value judgements to construct hypothesized habitat valuations (Malczewski 2004). We used the Reclassify tool to assign a value to each of the above-listed characteristics. Values were based on information from the literature (Table 1), with higher values associated with the specific landcover, population, or road type characteristic indicating more favorable habitat (Malczewski 2004; Belongie 2008; Table 1). We categorized all characteristics on a scale of 9 (most beneficial to the red wolf) to 1 (least beneficial to the red wolf). For example, cropland is the habitat type that the red wolf has been documented to use extensively, and so it received a value of 9. Humans are the leading cause of mortality for red wolves, so cells near high populations received a value of 1. Although the use of shrubland habitat is not widely documented, we assumed that as habitat generalists, shrubland use would be favored because of the provision of cover and potential prey (e.g., leporids), and thus the habitat type received a value of 7. Pasture is a potentially important characteristic (e.g., the GSMNP red wolf population used the restored meadow landscape of Cades Cove area, which the NLCD data layer classifies as pasture), but we excluded it because there is no further account of red wolves using true pasturelands and thus it is unclear whether such habitats are suitable.

Using the Weighted Sum tool, we multiplied each characteristic by its assigned weight, and then summed all individual characteristics. For this analysis, the weight multiplier kept the characteristic positive or turned it negative. We identified landscape characteristics as beneficial to the red wolf because they provide shelter and prey, so their weight multiplier was 1, which kept the characteristics positive. We identified population and road characteristics as nonbeneficial to the red wolf because they are causes of mortality, so their weights multiplier was −1, which resulted in negative values. Each 100 × 100 m cell resulted in a number ranging from 9 to −8. We then organized numbers into four classes, or thresholds, as determined by natural breaks classification in ArcGIS. We labeled each threshold with a number: 4 (most suitable), 3 (suitable), 2 (intermediate), 1 (less suitable). We chose natural breaks because it creates classes that have lowest within-class variations and highest among-class variations (Jacobs 2009). Additionally, given the generalist nature of red wolves, we suspect that small adjustments in characteristic values would have little impact on the output of the natural breaks classification. We then clipped the analysis area to illustrate the extent of the historical range (USFWS 2022).

With the goal of identifying landscapes that would support a minimum of 10 wolf packs (thereby facilitating a self-sustaining population and reducing inbreeding concerns; Robinson et al. 2019) while minimizing conflicts that might occur on private lands, we conducted a second analysis contrasting the suitability of large, federally managed landscapes (primarily lands managed by a federal agency, such as the U.S. Forest Service). Although red wolves have variable home range sizes (Riley and McBride 1972; Phillips et al. 2003; Mauney 2005; Hinton et al. 2016), we selected an estimated mean home range size of 100 km² such that 10 nonoverlapping home ranges would require approximately 1,000 km². We therefore identified federally managed landscapes of greater than 1,000 km² and contrasted the percent of each management unit that was identified as suitable, as well as the absolute amount (in km²) of suitable habitat in each management unit. Additionally, to assess how our predicted suitability model reflected actual habitat use, we compared the ENC RWP with our model outcomes.

A review of the cell values across the historical range highlights that the majority of the landscape was ranked as a 7 on the suitability scale (43.7%) followed by rankings of −1, 8, and 6 at 19.2, 15.2, and 8.9%, respectively (Figure 1). When organized into thresholds, the majority of the landscape was classified as most suitable (66.8%), while the percentages of habitat designated as suitable, intermediate, and less suitable were 6.2, 20.4, and 6.6%, respectively (Figure 2; Data S1, Supplemental Material). Focal examination of the ENC RWP revealed that most of the landscape ranked as a 7 on the suitability scale (56.7%) followed by a ranking of 8 (23.9%), with all other numerical rankings falling at or below 6.0% (Figure 3). Results of the threshold analysis indicate most of the ENC RWP landscape was most suitable (84.8%; Figure 4), with only 1.9% of the area classified as less suitable (Data S2, Supplemental Material).

Among federally managed landscapes of greater than 1,000 km², all had greater than 50% most suitable habitat (Figure 5; Table S1, Supplemental Material), and 16 had greater than 80% of habitat classified as most suitable. Okefenokee National Wildlife Refuge in southern Georgia was identified as having the highest proportion of most suitable habitat (Figure 5). Seven additional sites each comprised greater than 85% most suitable habitat, including Apalachicola National Forest in the Florida panhandle, Ozark National Forest in southern Missouri, Talladega National Forest in east-central Alabama, DeSoto National Forest in southern Mississippi, Kisatchie National Forest in central Louisiana, and Monongahela National Forest in eastern West Virginia (Data S3, Supplemental Material).

Based on the amount of most suitable land available within each management unit, 18 landscapes were identified as containing over 1,000 km² of most suitable habitat (Figure 6). Four sites—George Washington and Jefferson National Forests in predominately eastern Virginia, Ouachita National Forest in northern Arkansas and Oklahoma, Mark Twain National Forest in southern Missouri, and Ozark National Forest in northern Arkansas—had notably large quantities (> 4,000 km²) of most suitable habitat. All remaining sites, except for Francis Marion National Forest and Shawnee National Forest, had greater than 1,000 km² of most suitable and suitable habitats.

Paradiso and Nowak (1972) suggested general landscape preference of the red wolf to be “warm, moist, and densely vegetated habitats, including virgin pine and lowland hardwood forests, coastal prairies, and marshes.” However, this description is not reflective of the landscape found throughout the entirety of their historical range, and indeed, at the time this characterization was made by Paradiso and Nowak, the species had already been lost from most of its historical range and remained only in isolated landscapes. Given its large historical geographic distribution, it is probable that red wolves “utilized a large suite of landscape types” (Kelly et al. 2004) and are likely landscape generalists. However, habitat preferences are nonetheless apparent. For instance, although the wolves in the ENC RWP use a variety of landscapes, they show a preference for open agricultural habitat (Chadwick et al. 2010; Hinton and Chamberlain 2010, 2016; Dellinger et al. 2013; Karlin et al. 2016).

Our HSA indicates that there remain multiple large landscapes of sufficient suitability for a red wolf restoration effort. Overall more than 50% of the historical range might remain viable for population restoration, albeit in the absence of information on political and local residential perceptions. This is unsurprising because large carnivore use of developed and heterogeneous landscapes is well-recognized (e.g., Riley et al. 2014; Vickers et al. 2015; Benson et al. 2016; Murray et al. 2016; Poessel et al. 2017; Mueller et al. 2018). Red wolves are no exception. The species is able to inhabit agricultural landscapes and use roads as movement corridors (Dellinger et al. 2013; Hinton et al. 2016; Karlin et al. 2016). Thus, as predictors of landscape use by a generalist species such as the red wolf, nuances in model construction and parameterization should have relatively little importance on the broader results. However, a caveat is that large carnivore persistence is dependent on human tolerance, and measures of tolerance were not incorporated into our modeling effort. For instance, despite the landscape inhabited by the ENC RWP ranking primarily as most suitable, human-caused mortality is the main reason the ENC RWP population is in decline. Regional variability in human tolerance for red wolves might ultimately result in different measures of the percent of and amount of habitats rated as “most suitable.”

van Manen et al. (2000) ranked 31 prospective release regions using parameters very different from those used in our analysis. However, similar to our analysis, those top-ranked prospective release regions contain large tracts of public land. All but one of our federally managed landscapes (Pinelands National Reserve) falls into one of the 31 prospective release areas of van Manen e al. The similarity of our findings to those of van Manen et al. underscore that sizable amounts of habitat persist where a red wolf restoration could be justified. van Manen et al. ranked nine sites as having high success potential (e.g., > 0.8 on a 0–1 scale), including regions in northwestern Alabama, eastern West Virginia, southern Mississippi, southern Missouri, Okefenokee ecosystem, and northwestern Arkansas. Talladega, Monongahela, DeSoto, Mark Twain, Ozark, and Ouachita national forests, along with Okefenokee National Wildlife Refuge, are federally managed landscapes that rank highly in our analysis of percent suitability or amount available and are located in or near those regions. Perhaps more importantly, Ozark, Ouachita, and Monongahela national forests rank high in both percent and amount of suitable lands. In contrast, Western North Carolina, coastal South Carolina, eastern Tennessee, and southern Ohio were given a score of less than 0.51 in the van Manen analysis. Pisgah, Francis Marion, Cherokee, and Wayne are National Forests located in each respective region and were ranked greater than 80% most suitable in percent suitability, and all but Francis Marion have more than1,000 acres (405 ha) of habitat classified as most suitable.

We treated federally managed land units as independent of one another. Importantly, however, there are landscapes within the study area that allow for natural red wolf expansion with relatively low movement barriers. For instance, federally managed lands in west-central and northern Arkansas, as well as southern Missouri are closely situated, as are lands in western Virginia and southern West Virginia. There are human populations and highways throughout these areas, but most populations are less than 100,000 and the major interstates and highways bisecting these lands are limited. Additionally, each of these forests within these areas rank highly in amounts of suitable lands available, and one federal agency (the Forest Service) has managerial oversight over each of these forests, despite being located across multiple states. As such, future analyses might examine the suitability of combined land-management areas.

An additional consideration is fragmentation of federal lands that are designated as a single National Forest, independent of road networks. Mark Twain National Forest and Wayne National Forest, for example, comprise multiple large patches. It is important to consider the ownership of the landscape surrounding these fragments and the distance between fragments. Landscapes overseen by nonfederal agencies or organizations, corporations or individuals often lack a unified conservation agenda, with associated negative effects for broad-scale ecosystem management (Dallimer and Strange 2015). Additionally, humans are the prime cause of mortality among red wolves. For the first 25 years of the reintroduction, humans were the cause of 50% of red wolf deaths (Rabon and Bartel 2013). Thus, to effectively manage restoration efforts, it is important to consider agency and private management of landscapes surrounding parcels of federally owned land along with the likelihood of partnerships and “buy-in.”

Our analyses constitute an initial, broad-scale assessment. This HSA model allows for customization of additional characteristics based on weight preference. For instance, potentially important factors such as prey density were not incorporated because such data are not uniform, or are lacking, across states. Similarly, information on coyote density across the southeastern United States is limited, and thus we excluded the importance of coyotes from our analyses, further assuming that all potential restoration sites would contain robust coyote populations and that adaptive management plans similar to those administered for the ENC RWP would be required independent of restoration site. Nonetheless, addressing prey density and coyote-associated risks, along with other important factors such as hunting pressures, management of pasture and agricultural lands, forest type and density, and the attitudes of the local community might be additional approaches to consider when refining site rankings and model outcomes.

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.

Data S1. Geospatial data set (raster and shapefile) of predicted habitat suitability (as of 2021) for the red wolf Canis rufus across its historical range.

Available: https://doi.org/10.3996/JFWM-21-003.S1 (28.196 MB ZIP)

Data S2. Geospatial data set (raster and shapefile) of predicted habitat suitability (as of 2021) for the red wolf Canis rufus across the eastern North Carolina red wolf population.

Available: https://doi.org/10.3996/JFWM-21-003.S1 (28.196 MB ZIP)

Data S3. Geospatial data set (raster and shapefile) of predicted habitat suitability (as of 2021) for the red wolf Canis rufus across federally managed landscapes greater than 1,000 km² in size across its historical range.

Available: https://doi.org/10.3996/JFWM-21-003.S1 (28.196 MB ZIP)

Table S1. Percent suitability and total amount of habitat ranked “suitable” available on federally managed landscapes over 1,000 km² in size, based on a habitat suitability analysis for the red wolf Canis rufus (2021). Sites are within the historical range of the red wolf and are ordered alphabetically by name of federal landscape.

Available: https://doi.org/10.3996/JFWM-21-003.S2 (17 KB DOCX)

Reference S1. Bartel RA, Rabon DR Jr. 2013. Re-introduction and recovery of the red wolf in the southeastern USA. IUCN Global Re-introduction Perspectives 2013:107–115.

Available: https://doi.org/10.3996/JFWM-21-003.S3 (171 KB PDF)

Reference S2. Carley CJ. 1979. Status summary: the red wolf (Canis rufus). Albuquerque, New Mexico: U.S. Fish and Wildlife Service. Endangered Species Report 7.

Available: https://doi.org/10.3996/JFWM-21-003.S4 (1.990 MB PDF) and https://ecos.fws.gov/ServCat/DownloadFile/49693?Reference=49261

Reference S3. Chadwick J, Fazio B, Karlin M. 2010. Effectiveness of GPS-based telemetry to determine temporal changes in habitat use and home-range sizes of red wolves. Southeastern Naturalist 9:303–316.

Available: https://doi.org/10.3996/JFWM-21-003.S5 (487 KB PDF)

Reference S4. Dellinger JA. 2011. Foraging and spatial ecology of red wolves (Canis rufus) in northeastern North Carolina. Master's thesis. Auburn, Alabama: Auburn University.

Available: https://doi.org/10.3996/JFWM-21-003.S6 (1.237 MB PDF) and http://etd.auburn.edu/handle/10415/2497 (May 2022)

Reference S5. Dellinger JA, Ortman BL, Steury TD, Bohling J, Waits LP. 2011. Food habits of red wolves during pup-rearing season. Southeastern Naturalist 10:731–740.

Available: https://doi.org/10.3996/JFWM-21-003.S7 (640 KB PDF)

Reference S6. Dellinger JA, Proctor C, Steury TD, Kelly MJ, Vaughan MR. 2013. Habitat selection of a large carnivore, the red wolf, in a human-altered landscape. Biological Conservation 157:324–330.

Available: https://doi.org/10.3996/JFWM-21-003.S8 (564 KB PDF)

Reference S7. Desmul L. 2013. Habitat connectivity and suitability for Canis rufus recovery. Master's thesis. Durham, North Carolina: Duke University.

Available: https://doi.org/10.3996/JFWM-21-003.S9 (4.343 MB PDF) and https://dukespace.lib.duke.edu/dspace/handle/10161/6791 (May 2022)

Reference S8. Henry VG. 1998. Notice of termination of the red wolf reintroduction project in the Great Smoky Mountains National Park. Federal Register 63:54152–54153.

Available: https://doi.org/10.3996/JFWM-21-003.S10 (27 KB PDF)

Reference S9. Hinton JW. 2014. Red wolf (Canis rufus) and coyote (Canis latrans) ecology and interactions in northeastern North Carolina. Doctoral dissertation. Athens: University of Georgia.

Available: https://doi.org/10.3996/JFWM-21-003.S11 (2.144 MB PDF) and https://getd.libs.uga.edu/pdfs/hinton_joseph_w_201405_phd.pdf (May 2022)

Reference S10. Hinton JW, Chamberlain MJ. 2010. Space and habitat use by a red wolf pack and their pups during pup-rearing. Journal of Wildlife Management 74:55–58.

Available: https://doi.org/10.3996/JFWM-21-003.S12 (934 KB PDF)

Reference S11. Hinton JW, Proctor C, Kelly MJ, van Manen FT, Vaughan MR, Chamberlain MJ. 2016. Space use and habitat selection by resident and transient red wolves (Canis rufus). PLoS ONE 11:e0167603.

Available: https://doi.org/10.3996/JFWM-21-003.S13 (1.373 MB PDF)

Reference S12. Jacobs T. 2009. Putting the wild back into the wilderness: GIS analysis of the Daniel Boone National Forest for potential red wolf reintroduction. Master's thesis. Cincinnati, Ohio: University of Cincinnati.

Available: https://doi.org/10.3996/JFWM-21-003.S14 (10.906 MB PDF) and https://etd.ohiolink.edu/apexprod/rws_etd/send_file/send?accession=ucin1248796842&disposition=inline

Reference S13. Karlin ML. 2011. The endangered red wolf (Canis rufus): spatial ecology of a critically imperiled species in a human-dominated landscape. Doctoral dissertation. Charlotte: The University of North Carolina at Charlotte.

Available: https://doi.org/10.3996/JFWM-21-003.S15 (3.145 MB PDF) and https://core.ac.uk/download/pdf/345080154.pdf

Reference S14. Karlin ML, Václavík T, Chadwick J, Meentemeyer R. 2016. Habitat use by adult red wolves, Canis rufus, in an agricultural landscape, North Carolina, USA. Mammal Study 41:87–95.

Available: https://doi.org/10.3996/JFWM-21-003.S16 (2,138 KB PDF)

Reference S15. Mauney HF. 2005. Using geographic information systems to examine red wolf home range and habitat use in the Great Smoky Mountains National Park. Master's thesis. Chattanooga: The University of Tennessee, Chattanooga.

Available: https://doi.org/10.3996/JFWM-21-003.S17 (22.319 MB PDF) and https://scholar.utc.edu/theses/730/ (May 2022)

Reference S16. O'Neal S. 2018. A comprehensive assessment of red wolf reintroduction sites. Master's project. Durham, North Carolina: Duke University.

Available: https://doi.org/10.3996/JFWM-21-003.S18 (4.755 MB PDF) and https://dukespace.lib.duke.edu/dspace/handle/10161/16521

Reference S17. Paradiso JL, Nowak RM. 1972. Canis rufus. Mammalian Species 22:1–4.

Available: https://doi.org/10.3996/JFWM-21-003.S19 (434 KB PDF)

Reference S18. Phillips MK, Henry VG, Kelly BT. 2003. Restoration of the red wolf. Pages 272–288 in Mech LD, Boitani L, editors. Wolves: behavior, ecology, and conservation. Chicago, Illinois: University of Chicago Press.

Available: https://doi.org/10.3996/JFWM-21-003.S20 (1.158 MB PDF)

Reference S19. Riley GA, McBride RT. 1972. A survey of the red wolf (Canis rufus). Scientific Wildlife Report No. 162. Washington, D.C.: U.S. Fish and Wildlife Service.

Available: https://doi.org/10.3996/JFWM-21-003.S21 (850 KB PDF) and https://ia802700.us.archive.org/23/items/specialscientifi162unit/specialscientifi162unit.pdf

Reference S20. Russell DN, Shaw JH. 1971. Distribution and relative density of the red wolf in Texas. Proceedings of the Annual Conference, Southeastern Association of Game and Fish Commissioners 25:131–137.

Available: https://doi.org/10.3996/JFWM-21-003.S22 (394 KB PDF)

Reference S21. Shaffer J. 2007. Analyzing a prospective red wolf (Canis rufus) reintroduction site for suitable habitat. Report. 32 pages.

Available: https://doi.org/10.3996/JFWM-21-003.S23 (1.985 MB PDF) and https://nywolf.org/wp-content/uploads/2019/01/Analyzing-a-Prospective-Red-Wolf-Canis-rufus-Reintroduction-Site-for-Suitable-Habitat.pdf

Reference S22.U.S. Census Bureau. 2020. 2020 TIGER/Line Shapefiles (machine readable data files).

Available: https://doi.org/10.3996/JFWM-21-003.S24 (38.040 MB ZIP) and https://www.census.gov/cgi-bin/geo/shapefiles/index.php?year=2020&layergroup=Roads

Reference S23.[USFWS] U.S. Fish and Wildlife Service. 1990. Red wolf recovery and species survival plan. Atlanta, Georgia: U.S. Fish and Wildlife Service.

Available: https://doi.org/10.3996/JFWM-21-003.S25 (2.488 MB PDF) and https://redwolves.com/newsite/wp-content/uploads/2016/01/3-USFWS-1989.pdf

Reference S24.[USFWS] U.S. Fish and Wildlife Service. 2007. Red wolf 5-year status review: summary and evaluation. Manteo, North Carolina: U.S. Fish and Wildlife Service.

Available: https://doi.org/10.3996/JFWM-21-003.S26 (619 KB PDF) and https://redwolves.com/newsite/wp-content/uploads/2016/01/5-USFWS-2007.pdf

Reference S25.[USFWS] U.S. Fish and Wildlife Service. 2016. Recommended decisions in response to red wolf recovery program evaluation. Memorandum. Atlanta, Georgia: U.S. Fish and Wildlife Service.

Available: https://doi.org/10.3996/JFWM-21-003.S27 (477 KB PDF) and https://www.fws.gov/sites/default/files/documents/news-attached-files/recommended-decisions-in-response-to-red-wolf-recovery-program-evaluation.pdf

Reference S26.[USFWS] U.S. Fish and Wildlife Service. 2018a. Red wolf 5-year status review: summary and evaluation. U.S. Fish and Wildlife Service.

Available: https://doi.org/10.3996/JFWM-21-003.S28 (208 KB PDF) and https://www.southernenvironment.org/uploads/words_docs/2018.04.24_Red_Wolf_Status_Review.pdf

Reference S27.[USFWS] U.S. Fish and Wildlife Service. 2018b. Red wolf species status assessment. U.S. Fish and Wildlife Service.

Available: https://doi.org/10.3996/JFWM-21-003.S29 (4.267 MB PDF) and https://nywolf.org/wp-content/uploads/2020/10/SSA_RedWolf_201804.pdf

Reference S28.[USFWS] U.S. Fish and Wildlife Service. 2022. Red wolf recovery program. U.S. Fish and Wildlife Service.

Available: https://doi.org/10.3996/JFWM-21-003.S30 (651 KB PDF) and https://www.fws.gov/project/red-wolf-recovery#current-status-of-wild-population-section

Reference S29.U.S. Geological Survey Gap Analysis Project (GAP). 2020. Protected areas database of the United States (PAD-US), 2.1: U.S. Geological Survey data release. https://doi.org/10.5066/P92QM3NT.

Available: https://doi.org/10.3996/JFWM-21-003.S31 (??? KB PDF) and https://www.sciencebase.gov/catalog/item/60259839d34eb12031138e1e (May 2022)

Reference S30. van Manen FT, Crawford BA, Clark JD. 2000. Predicting red wolf release success in the southeastern United States. Journal of Wildlife Management 64:895–902.

Available: https://doi.org/10.3996/JFWM-21-003.S32 (1.743 MB PDF)

Reference S31. Vaughan MR, Kelly MJ. 2011. Evaluating potential effects of widening US 64 on red wolves in Washington, Tyrrell, and Dare Counties, North Carolina. Report of Virginia Tech to North Carolina Department of Transportation, Raleigh, North Carolina.

Available: https://doi.org/10.3996/JFWM-21-003.S33 (12.594 MB PDF) and https://digital.ncdcr.gov/digital/collection/p249901coll22/id/486960

We thank the Point Defiance Zoo & Aquarium's Dr Holly Reed Conservation Fund and the Akron Zoo's Conservation Fund for financial support; K.E. DeMatteo, K.N. Key, and J. Lindstrom for their valuable insight; and the University of Missouri School of Natural Resources Writing Workshop group. We also thank the anonymous reviewers and Journal Editors for constructive comments that improved the quality of this manuscript.

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.