Comparison of Two Detection Methods for a Declining Rodent, the Allegheny Woodrat, in Virginia

Effective management of rare or declining wildlife species depends on reliable and repeatable methods of accurately sampling such populations to estimate site occupancy. The effectiveness of monitoring methods varies by target species, location, and research interests (Smith et al. 1975; Garden et al. 2007; Thompson and Thompson 2007). Various sampling methods have been developed to monitor terrestrial small mammals (Garden et al. 2007; Thompson and Thompson 2007). Still, conventional small-mammal monitoring practices often involve invasive approaches such as livetrapping and handling (Smith et al. 1975; Williams and Braun 1983; McCay et al. 1998).

For small mammals, livetrapping methods can facilitate the estimation of population abundance and demographic vital rates (Nichols and Pollock 1983; Jones et al. 1996). However, livetrapping often is costly and labor-intensive, trap success varies by species and gear, and handling causes stress to the captured animals (Stanley and Royle 2005; Wiewel et al. 2007). Additionally, low detection probability or trap avoidance may lead to biased results (Diggins et al. 2016). Thus, if the primary research objective is to ascertain the probability of occurrence of a species in an area, survey methods such as track plates and remote-detection cameras (camera traps), which do not require physical capture and handling of animals, may provide effective alternatives or complements to livetrapping. The use of camera traps is common in large-mammal research and recently has gained popularity as an effective survey method for small mammals as camera technology and data storage have improved (De Bondi et al. 2010; Di Cerbo and Biancardi 2013; Diggins et al. 2016).

Allegheny woodrats Neotoma magister are emergent rock-obligate species distributed predominantly in the central Appalachian Mountains and Interior Low Plateau from New Jersey southward to Kentucky and Tennessee, with many populations concentrated in western Virginia and eastern West Virginia (Mengak 2002; Castleberry et al. 2006). Allegheny woodrats are listed as “Near Threatened” on the International Union for Conservation of Nature Red List (Linzey et al. 2008) and are likely extirpated from Connecticut and New York (Monty and Feldhamer 2002). This species is listed as “Endangered” in Maryland (MNHP 2021; NESCA 2021), New Jersey (NJESCA 1973; NJDFW 2012; Fowles 2018), and Ohio (BORCA 1978; ODNR 2020). Also, this species is considered “Threatened” in Pennsylvania (PWRCA 1986; PGC 2021), of “Special Concern” in North Carolina (NCGS 1987; NCWRC 2021), and “In Need of Management” in Tennessee (NETWSCA 1974; TWRA 2021). Although West Virginia currently has no formal state-level protected wildlife species legislation, Allegheny woodrats are described as “Vulnerable” in the West Virginia State Wildlife Action Plan (WVDNR 2015). In Virginia, the Allegheny woodrat is not provided state-level protection of “Threatened” or “Endangered” (VAC 2022); however, the species is designated as “Vulnerable” (S3; Campbell et al. 2010) and in need of moderate conservation (tier IV species) in Virginia's Wildlife Action Plan (VDGIF 2015). Alteration or loss of suitable habitat (e.g., undisturbed forests surrounding emergent rock outcrops and caves), isolation of populations from landscape-scale forest fragmentation, the near extinction of the American chestnut Castanea dentata to chestnut blight Cryphonectria parasitica, and exposure to raccoon roundworm Baylisascaris procyonis from increasing raccoon Procyon lotor populations likely have contributed to rangewide population declines (Balcom and Yahner 1996; LoGiudice 2003, 2006; Owens et al. 2004; Castleberry et al. 2006; Ford et al. 2006). Further, predation by hawks and owls (Fisher 1893; Powers et al. 2020a), fishers Pekania pennanti (McNeil et al. 2017), and timber rattlesnakes Crotalus horridus (G. Turner, Pennsylvania Game Commission, personal communication) may be components of population decline in even seemingly optimal habitat conditions. Habitat change and population declines across a large historic range have dictated the need for an effective and efficient monitoring approach for this species.

Historically, the presence or absence of Allegheny woodrats was ascertained using livetrapping methods (Mengak 2002; Ford et al. 2006). However, the central Appalachian Mountains include a large expanse of potentially suitable habitat, much of which is difficult to access, particularly with live traps. Therefore, the logistics of daily trapline visits often limit the size and scope of surveys (Mengak et al. 2008). From 2017 through 2020, we primarily trapped “legacy” woodrat sites (those with woodrats documented in the last century) in western Virginia to compare detection probability predicted using remote-detecting cameras (herein cameras) with that of the traditional method of livetrapping. During 2017, 2018, and 2020, we sought to determine whether cameras alone were sufficient to confirm site occupancy for future monitoring. Additionally, we used a subset of these sites in 2019 with confirmed occupancy to compare the efficacy of bait at cameras to increase the detection probability of Allegheny woodrats. We hypothesized that the detection probability of Allegheny woodrats would differ between trapping methods and predicted that cameras would produce a detection probability greater than that of live traps. Also, we predicted that detection probability would be greater when using baited vs. unbaited cameras. Further, we predicted that detection probability would increase with an increasing number of trap-nights for all trapping methods.

We surveyed for Allegheny woodrats from May through October 2017 (38 sites), 2018 (29 sites), 2019 (24 sites), and 2020 (11 sites) in the Blue Ridge and Ridge and Valley physiographic provinces in western Virginia. In 2017, 2018, and 2020, the number of sites varied by year because of site accessibility, funding, and personnel limitations. In 2020, fewer sites were surveyed than in 2017 or 2018 because of travel and personnel limitations related to the Covid-19 pandemic. Because we paired baited and unbaited cameras at sites in 2019, equipment limitations further reduced the number of sites we were able to survey in that year. Further, to focus solely on detection probability in 2019, we limited study sites to those in which Allegheny woodrats had been detected in 2017 and 2018. Our study sites included open-canopied or close-canopied erosion-resistant sandstone emergent rock outcrops, boulder fields, and limestone-solution caves (inside and at or around entrances; Figure 1) throughout the George Washington and Jefferson national forests, Virginia state landholdings, and numerous private lands currently or historically occupied by Allegheny woodrats. Distance between nearest neighboring study sites ranged from 180.7 to 21,231.0 m (mean = 5,189.0, SE = 115.6 m), which greatly exceeds the average nightly foraging distance and home range extent for Allegheny woodrats in the central Appalachians (Castleberry et al. 2001). Thus, we assumed within-season closure of study sites. Elevations ranged from approximately 200 to 1,200 m among study sites.

Overall, the Valley and Ridge and Blue Ridge provinces are generally characterized by long mountain ridges with moderate-to-steep side slopes. Above cleared valleys, our study areas were heavily forested (Homer et al. 2015). Dominant forest types were Appalachian oak (e.g., Quercus alba, Q. prinus, Q. rubra, Q. velutina) and mixed pine (e.g., Pinus pungens, P. rigida.) with smaller patches of other hardwood (e.g., Acer saccharum, Liriodendron tulipifera, Fagus grandifolia) and eastern hemlock Tsuga canadensis depending on aspect and elevation. Dense shrub layers of mountain laurel Kalmia latifolia and rosebay rhododendron Rhododendron maximum occurred throughout the study area (Kniowski and Ford 2018).

We paired 3–18 baited Tomahawk live traps model #201 (40.6 × 12.7 × 12.7 cm) and #202 (48.3 × 15.2 × 15.2 cm; Tomahawk Live-Trap Co., Tomahawk, WI) with an equal number of infrared remote-detecting cameras (Stealth Cam, model STC-G45NGX, Grand Prairie, TX; Bushnell Trophy Cam HD, model 119739, Overland Park, KS; or Reconyx HypeFire, model HC500, Holeman, WI) at discrete emergent rock formation or cave entrances separated from other emergent rock by > 100 m. Because topography and site accessibility varied at each site, the number of livetrap/camera pairs deployed was based on suitable, accessible habitat, with an upper limit per site (18) determined by the number of personnel assisting. Allegheny woodrats are strongly associated with emergent rock and movements typically follow nonrandom paths between ridgeline and side-slope topographic features such as drainages where colluvium occurs (Wood 2008). Thus, within study sites, we chose to place paired live traps and cameras (trapping sites) strategically to increase our probability of detecting Allegheny woodrats. Trapping sites were selected on the basis of two criteria: 1) the location of the trapping site must be characterized as having a cave entrance, a rocky crevice, or a rocky overhang present to place a live trap in or under, and 2) the trapping site must contain a tree or large rock to affix a camera to within 1–3 m of where the live trap would be placed. Cameras were fixed to a tree at ground level or set against a large rock facing directly at the trap. We separated the trapping sites by approximately 10 m within each study site (Figure 1). We programmed the cameras to take a rapid set of three photos at a minimum of 1-min intervals. In 2017, 2018, and 2020 our live traps were baited with half an apple following the protocol used by state agencies (Mengak et. al. 2008) and the methods of Castleberry et al. (2014). During these years, we deployed live traps and cameras for three consecutive nights at 31, 26, and 11 total sites, respectively. We revisited the live traps each morning to check for captured woodrats and replace missing bait but left cameras undisturbed until the conclusion of the survey period. Because of personnel limitations in 2020, only 11 sites were included in the survey. Additionally, live traps and cameras were only used for two consecutive nights when woodrats were detected during the first night. After live capture, we weighed, aged (adult, subadult, juvenile; Mengak 2002), sexed, and ear-tagged (National Band and Tag Co., Newport, KY; style 1005) the captured individuals using methods described by Castleberry et al. (2002; Figure 2). All camera and livetrapping methods were approved by the Radford University Institutional Animal Care and Use Committee (protocol no. FY17-07 and FY20-08) and Virginia Polytechnic Institute and State University Institutional Animal Care and Use Committee (protocol no. 17-070).

In 2019, we surveyed 24 sites known to be occupied in 2017 or 2018 using cameras only (Figure 2). We paired 3–12 unbaited cameras with cameras baited with 1 cup of commercially available livestock sweet feed (Multitext 12.5% multipurpose livestock feed, Augusta Cooperative Farm Bureau, Staunton, VA) over four consecutive nights (sample occasions). In 2019 we used sweet feed as bait instead of apples to reduce the likelihood of the entire bait source being removed during the night because the bait was not secured inside a Tomahawk trap as it was in 2017, 2018, and 2020. We did not replenish bait daily; however, we were able to discern that the bait was not completely diminished by the fourth night by examining photos taken on subsequent nights for presence of bait. For all years, we defined “sample occasion” as 24-h intervals starting from the day of camera deployment (i.e., first night, second night, third night, and fourth night). A “livetrap detection” occurred with the successful capture of a woodrat in at least one live trap at a site within a sample occasion and a “camera detection” occurred when at least one photo of a woodrat was captured at a site within a sample occasion.

We created sitewide detection histories of Allegheny woodrats for each trapping method each year (i.e., live trap vs. camera in 2017, 2018, and 2020; baited vs. unbaited in 2019). We estimated the probability of detection (p) and probability of site occupancy (ψ) using single-season models (MacKenzie et al. 2002) with package unmarked (Fiske and Chandler 2011) in Program R (R Core Team 2021; Text S1, Supplementary Material). We examined the effect of trapping method (live traps vs. camera), sample occasion (first, second, or third night), and cumulative trapping effort (cumulative number of trap-nights) on the detection probability of Allegheny woodrats. Because we selected survey sites with known or historic occurrences of Allegheny woodrats that were characterized by similar habitat features (emergent rock, rock overhangs, karst habitat), we held occupancy constant (Ψ(.)) in all models under the assumption that habitat features did not have an appreciable effect on site occupancy per se. We constructed 10 a priori models to examine additive and interactive combinations of the detection covariates (Table 1). We assessed model performance using Akaike's information criterion corrected for small sample sizes (AIC_c) and model weight (w_i; Burnham and Anderson 2002). We identified models with a difference in ΔAIC_c ≤ 2.0 and the highest w_i as the best-performing models (Burnham and Anderson 2002). Data from 2017, 2018, and 2020 were pooled to increase the sample size. To assess the effect of the use of bait on the detection probability of Allegheny woodrats, we created a set of eight a priori additive and interaction detection models (Table 2). Detection covariates included bait method (baited vs. unbaited), sample occasion (first, second, third, or fourth night), and cumulative trapping effort. Occupancy probability was held constant. We assessed the model fit of the best-performing models for both the trap- and bait-method analyses using the MacKenzie and Bailey (2004) goodness-of-fit test using 5,000 permutations on the best-performing trap- and bait-method models.

Our number of trap-nights varied among years with 595 camera trap-nights and 604 Tomahawk trap-nights in 2017, 633 camera trap-nights and 585 Tomahawk trap-nights in 2018, and 279 camera trap-nights and 279 Tomahawk trap-nights in 2020. In 2019, we sampled 1,204 baited camera trap-nights and 1,236 unbaited camera trap-nights. Overall, the number of detections of Allegheny woodrats varied among years (Figure 3). In 2017, we detected Allegheny woodrats at 16 of 31 sites including 11 sites where they were captured in traps and 16 where they were detected by cameras. In 2018, we detected woodrats at 13 sites including three sites where woodrats were captured in traps and 13 where they were detected by a camera. In 2019, we detected woodrats at 13 of 24 sites; woodrats were detected by baited cameras at 12 sites and unbaited cameras at 13 sites. In 2020, a total of 6 of the 11 sites had woodrat detections by both cameras and live traps at each of the six sites.

For the trap-method assessment, Ψ(.)p(detection method) was the best-performing model (Table 1; Data S1, Supplementary Material) with no lack of fit indicated (χ² = 11.46, df = 155, P = 0.18). The probability of detection (p) using cameras was 0.78 (SE = 0.04), whereas p using live traps was 0.51 (SE = 0.09). Occupancy probability was 0.42 (SE = 0.05). For the bait-method assessment, the Ψ(.)p(bait method + trap night) model outperformed all other models (Table 2; Data S1, Supplementary Material) with no lack of fit indicated (χ² = 11.95, df = 153, P = 0.74). Detection probability of Allegheny woodrats was greater using bait than without bait. For both bait methods, detection probability increased from the first night to the third night but did not increase after the third night (Figure 4). Detection probability using bait was 0.63 (SE = 0.06) on night 1, 0.78 (SE = 0.06) on night 2, 0.91 (SE = 0.03) on night 3, and 0.78 (SE = 0.06) on night 4. Without bait, p = 0.27 (SE = 0.07) on night 1, p = 0.43 (SE = 0.09) on night 2, p = 0.69 (SE = 0.09) on night 3, and p = 0.43 (SE = 0.09) on night 4. Occupancy probability was 0.38 (SE = 0.04).

In this study, we found that detection of Allegheny woodrats was influenced by detection method (live traps vs. camera traps) and whether bait was used with cameras. When using camera traps, the detection probability of Allegheny woodrats was almost twice that when using live traps, which is consistent with our predictions. By aiming the cameras directly at the live traps rather than placing traps near cameras (but not in the frame), we were able to observe that Allegheny woodrats often examined baited live traps but did not enter the traps. We suspect two potential explanations for these observations. First, “trap avoidance” can be attributed to many variables such as trap size, the appearance of the trap to the animal, residual scent, trap placement, previous exposure to capture and handling, and individual variation (Sealander and James 1958; Jolly and Dickson 1983). This species has been generally described as “trap happy,” whereby exposure to a bait source increases an individual's likelihood of being captured (Castleberry et al. 2002). Additionally, traps baited with apples have been successfully used in numerous studies of Allegheny woodrat with a capture–recapture focus (e.g., Johnson 2002; Chamblin et al. 2004; Mengak et al. 2008; Manjerovic et al. 2009; Castleberry et al. 2014). Nonetheless, at multiple sites in this study, we detected woodrats on camera during all three nights but did not capture woodrats in live traps that had remained open. This suggests that studies using only livetrapping methods underestimate the occurrence of Allegheny woodrats by recording false absences. A second potential explanation may be the premature or errant trigger of traps (e.g., disturbance of trap by a nontarget species or other cause), which prevents the capture and therefore detection of a target species investigating a closed trap. In contrast, camera traps may be less perceptible or intimidating to an animal investigating bait and remain “open” regardless of the number of animals that approach the bait, affording greater opportunity to detect the target species.

Our results also suggest that the use of baited cameras is more effective at detecting Allegheny woodrats than the use of unbaited cameras, thereby supporting our second prediction. These results are consistent with recent research in which the use of bait was found to increase the detectability of cryptic species including rodents (Paull et al. 2010; Rendall et al. 2021) and mesocarnivores (Eng and Jachowski 2019; Ferriera-Rodríquez and Pombal 2019; Mills et al. 2019) at camera traps. Although passive camera sampling is often used successfully for large species, Allegheny woodrats are a small, cryptic species and hence have less “apparency” in the environment (Castleberry et al. 2006). Because population densities of Allegheny woodrat are low, nonrandomly distributed, and movements are restricted to isolated and disjunct emergent rock habitats (Castleberry et al. 2001), the use of bait could increase the chance of detecting this species at an occupied site, thereby reducing the chance of a false negative (type II error; Stewart et al. 2019).

When using baited cameras, we found that the detection probability of Allegheny woodrats was greater on the first night of sampling than when using unbaited cameras. Detection probability increased through the third night for both bait methods; however, it did not increase on the fourth night. This suggests that three nights of survey effort are adequate to detect Allegheny woodrats at occupied sites. In contrast, the lack of additional gain in detection probability on the fourth night also may be explained by a dispersal or depredation event that may be falsely perceived as avoidance. Although no direct observations of dispersal or trap mortality were made during this study itself, we did observe one event of predation by a barred owl (Strix varia) during this time in other research efforts (Powers et al. 2020a). Additionally, the removal of bait by target or nontarget species may also reduce the likelihood of animals visiting the camera site. Thus, a longer-lasting bait source such as a solid bait block (Gooley and Schauber 2018) may reduce the risk of the complete consumption of bait and improve detection over longer survey periods.

Our results concur with similar studies of small-mammal detection methodology in montane habitats (Castleberry et al. 2014; Diggins et al. 2016; Smith and Weston 2017) and contribute to mounting research underlining the importance of imperfect detection of wildlife in occupancy studies (e.g., Kéry and Schmidt 2008; Otto and Roloff 2011; Brubaker et al. 2014). Baited camera surveys may provide a detection method that maximizes detection probability on the basis of the behavior and ecology of the target species. Further, the importance of addressing imperfect detection and maximizing the detection probability of Allegheny woodrats is emphasized by the low estimates of occupancy probability in our study. Despite sampling sites with known recent or historic detections of Allegheny woodrats, our number of captures was low and our best-performing trap-method model produced an occupancy estimate of approximately 0.42. Moreover, our bait-method analysis found occupancy probability to be approximately 0.38, though we sampled only sites where Allegheny woodrats were detected in 2017 and 2018. Both estimates were lower than that found by Mengak et al. (2008).

Our study provides a framework (remote camera surveys) for the development of a large-scale, long-term monitoring protocol of Allegheny woodrats, specifically with the goals of identifying changes in the distribution of the species and quantifying local extinction and colonization rates in a manner that does not require physical capture and handling of animals. Long-term monitoring and large databases are necessary for effective predictive modeling in small areas or across a large geographic or temporal scale (Busch and Trexler 2003). Thus, we suggest that the rangewide monitoring protocol of Allegheny woodrat optimally be set for 3 nights with baited camera traps at legacy and potentially occupied sites to monitor the persistence of contemporary populations. In addition to higher detectability, the use of cameras is more cost-effective than conventional livetrapping methods for small mammals (Welbourne and MacGregor 2015; Thomas et al. 2020). Further, cameras can generate other data or insights into Allegheny woodrat behavior (e.g., foraging times) and inter- and intraspecies interactions including competition and predation (Bridges and Noss 2011; Lazenby et al. 2015; Powers et al. 2020a).

For research goals in which demographic information such as abundance and vital rates are desired, we suggest that camera traps be used complementary to livetrap methods to ensure species presence before livetrapping and to improve livetrapping success. Additionally, the complementary use of cameras and live traps may lessen the amount of capture and handling of individuals and thus lower the chance of subsequent stress and capture myopathy. Whereas capture–recapture methods require repeat trapping and handling of individuals, mark–resight methods for estimating population density involve initial capture and marking of individuals and subsequent use of cameras for resighting marked individuals (Arnason et al. 1991; White and Shenk 2001; McClintock et al. 2009). Mark–resight methods also provide an alternative for demographic studies that is less invasive than mark–recapture, which requires repeated capture and handling (McClintock et al. 2009, Rich et al. 2014). Additionally, mark–resight methods may be a viable option when similarly less-invasive methods such as passive integrated transponder units (tags) may be cost-prohibitive. Because individual Allegheny woodrats do not have unique markings, artificial markings such as patterns dyed or shaved into the fur are required for individual identification. Accordingly, the initial trapping of Allegheny woodrats can be used to mark individuals and collect ancillary data (e.g., morphometrics, parasite presence). Subsequent camera observations can be used to estimate population density (Powers et al. 2020b).

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.

Text S1. Program R code used to conduct the trapping method and bait method single-season occupancy analyses for Allegheny woodrats Neotoma magister from May through October 2017, 2018, and 2020 in the George Washington and Jefferson national forests and surrounding public and private lands in Virginia.

Available: https://doi.org/10.3996/JFWM-21-037.S1 (49 KB RTF)

Data S1. Detection history and detection covariates for Allegheny woodrats Neotoma magister throughout the George Washington and Jefferson national forests and additional state and private lands in Virginia from May through October 2017, 2018, and 2020 and from April through June 2019.

Available: https://doi.org/10.3996/JFWM-21-037.S2 (44 KB XLSX)

Reference S1. Monty AM, Feldhamer GA. 2002. Conservation assessment for the Eastern woodrat (Neotoma floridana) and the Allegheny woodrat (Neotoma magister). Milwaukee, Wisconsin: U.S. Forest Service, Eastern Region. Available: https://doi.org/10.3996/JFWM-21-037.S3 (479 KB PDF) and https://www.fs.usda.gov/Internet/FSE_DOCUMENTS/fsm91_054316.pdf

Funding and equipment provided by the Virginia Department of Wildlife Resources through a Wildlife Restoration Grant from the U.S. Fish and Wildlife Service and Radford University. We thank many private landowners for providing access survey sites. We are grateful to J. Bentley, J. Brown, C. Bryan, J. Crotts, H. Custer, M. Dimas, E. Gladin, A. Leon, N. McDonald, B. Mullen, K. Nelson-Anderson, L. Platt, L. Van Meter, and C. Wozniak for field assistance. We also thank three anonymous reviewers and the Associate Editor who provided comments that improved an earlier version 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.