Winter cave and mine surveys have been the primary method to monitor status of bat populations but they are not equally effective across regions or species. Many species of bats that roost in rock outcrops during the nonhibernation period are difficult for researchers to monitor with existing methods. Although some wildlife biologists have speculated visual surveys could be used to monitor populations of bats roosting on talus slopes, they did not know the efficacy of the method. We used standardized plot-based visual surveys to quantify presence and abundance of eastern small-footed bats Myotis leibii on talus slopes in Virginia, and studied sources of variation and error. Detection probability for talus surveys was relatively high but varied based on search effort and site characteristics. Both abundance and detection probability varied more among plots within sites than among sites or years. In trials with radio-tagged bats to study the causes of false negatives, 18% of bats roosted where surveyors could not see them, and 18% of bats were visible but overlooked due to human error. Less experienced surveyors counted slightly fewer bats than the principal investigator, we suspect because the principal investigator simply found the bats first. There also was a slight learning curve among less experienced observers. Visual surveys have strong potential to aid in the study of bats that roost on talus slopes. Talus surveys, unlike longer-established methods to monitor bat populations, provide ways to assess error. We recommend using talus surveys to monitor other rock-roosting bat species with poorly understood populations, such as many species in western North America.
Wildlife biologists have used visual surveys of caves and mines in winter extensively to monitor bat populations in eastern North America, where karst geology is relatively abundant. These winter surveys were the primary method they used to document bat mortality from the fungal disease white-nose syndrome (WNS) during its emergence and initial spread (Gargas et al. 2009; Turner et al. 2011; Ingersoll et al. 2013). Declines in overwintering bat populations suggested WNS devastated multiple species of bats in eastern North America, especially little brown bats Myotis lucifugus, northern long-eared bats Myotis septentrionalis, and tri-colored bats Perimyotis subflavus (Ingersoll et al. 2013; Frick et al. 2017).
If WNS continues spreading to western North America where karst geology is less widespread and where bats generally do not overwinter in large numbers in caves and mines, winter surveys alone will likely be insufficient for understanding population trends (O'Shea et al. 2003; Loeb et al. 2015). Even in eastern North America, effectiveness of winter surveys depends on the biology of the species in question. Lasiurine bats do not hibernate in caves or mines (Hein et al. 2005; Mormann and Robbins 2007), and other species that do hibernate in caves or mines occur in relatively low numbers, such as the eastern small-footed bat Myotis leibii (Best and Jennings 1997).
Surveys using mist nets and acoustic surveys of bat activity during the nonhibernation period are the main alternatives for detecting changes in abundance of bats, independent of their overwintering habits. These techniques also have widely recognized limitations. Results of mist-net surveys can be negatively influenced by environmental conditions, net placement, and sampling frequency (Larsen et al. 2007; Geluso and Geluso 2012; Marques et al. 2013), and capture probabilities vary among bat species (Pedersen et al. 2005; Weller and Lee 2007). Mist netting for rare species often requires substantial sampling effort to discern trends in abundance, which must be balanced against the potential for intensive repeated sampling to cause bats to become net-shy (Weller and Lee 2007; Francl et al. 2012; Marques et al. 2013). Acoustic monitoring is less expensive than mist-net surveys, but changes in activity levels do not necessarily mean changes in abundance; calls of some species are particularly difficult to detect, and species-level identification can be unreliable enough that many researchers instead classify calls by phonic groups (e.g., Myotis; Braun de Torrez et al. 2017; Rydell et al. 2017; Russo et al. 2018).
The eastern small-footed bat is an example of a species for which existing monitoring methods have been problematic. Wildlife biologists consider the species rare across most of its range, and its distribution and population status are poorly resolved (Best and Jennings 1997). In winter, surveyors typically count eastern small-footed bats in much lower numbers than other cave-hibernating bat species in eastern North America (Best and Jennings 1997; Trombulak et al. 2001; Turner et al. 2011). This may be because of their tendency to hibernate in cracks and crevices that surveyors overlook during winter surveys, and because the species overwinters in nonkarst features such as rock ledges (Krutzsch 1966; Moosman et al. 2017). Surveyors rarely capture eastern small-footed bats in mist net surveys and the bats are relatively difficult to detect and identify acoustically. As a result, and despite heightened concern after the emergence of WNS (USFWS 2013), status of the eastern small-footed bat is still somewhat in question. Two studies using mist nets observed substantial declines (64–84%) in capture rates in West Virginia and New Hampshire relatively soon after the emergence of WNS (Francl et al. 2012; Moosman et al. 2013); later studies did not detect significant changes, including capture rates from a mist-net study in North Carolina and Tennessee, and counts from winter cave and mine surveys (Frick et al. 2017; O'Keefe et al. 2019). It is unclear whether the different trends reported by past studies of eastern small-footed bat abundance reflect meaningful biological patterns or methodological differences. A host of western bat species are subject to similar monitoring challenges (O'Shea et al. 2003).
Eastern small-footed bats aggregate at exposed rock outcrops during the nonhibernation period, including cliffs and boulder fields known as talus slopes (Saugey et al. 1993; Johnson and Gates 2008; Johnson et al. 2011). This behavior is remarkably common—around 40% of bat species in North America roost in exposed rock outcrops in at least part of their range (Wilson and Ruff 1999). Much like caves and mines, rock outcrops are discrete habitat features that may make bats accessible to researchers, and could offer opportunities to monitor bat populations that are difficult to assess otherwise. Visual searches of talus slopes, boulder-strewn bluffs, and rock ledges have proven useful for finding roosting eastern small-footed bats and big brown bats Eptesicus fuscus (Roble 2004; Whitby et al. 2013; Huth et al. 2015; Moosman et al. 2017). Moosman et al. (2015) proposed using repeated randomized visual surveys for monitoring population trends on talus slopes, but effectiveness of visual surveys for quantifying abundance of eastern small-footed bats, or any other bats for that matter, is unknown. Wildlife biologists need to understand important aspects of visual surveys, including the method's ability to provide biologically meaningful estimates, the degree to which visual counts underestimate abundance, and its consistency across observers. Therefore, we expanded on the pilot study of Moosman et al. (2015) by conducting randomized plot-based visual surveys on six talus slopes in Virginia from 2013 to 2019. Our goals were to 1) assess performance of visual surveys from the standpoint of two common modeling approaches (models of bat abundance and those focused on detection probability), 2) obtain empirical data on rates of false negatives and their causes, and 3) assess potential for observer bias.
We conducted surveys between 26 May and 31 July 2013–2019, at nonforested talus slopes in western Virginia that ranged in area from < 0.01 ha to 3.25 ha, in Augusta (n = 3), Rockbridge (n = 2), and Rockingham (n = 1) counties. Sites were in the Blue Ridge, and Valley and Ridge, physiographic provinces of the Central Appalachian Mountains, on relatively large tracts of public lands dominated by mixed-deciduous forest. During the initial 3 y of the project, work focused on two talus sites in the James River Face (Figure S1, Supplemental Material) and Saint Mary's (Figure S2, Supplemental Material) wilderness areas, both in Rockbridge County, and one talus site at Sherando Lake Recreation Area (Figure S3, Supplemental Material), in Augusta County (hereafter, Marbleyard, St. Mary, and Sherando, respectively). We added two more sites to the project in 2016, including talus slopes near Humpback Rocks (Figure S4, Supplemental Material) and the Slacks Overlook (Figure S5, Supplemental Material), both in Augusta County (hereafter, Humpback and Slacks, respectively). We added Blackrock Summit (Figure S6, Supplemental Material), in Rockingham County, to the project in 2019. Size of rocks varied substantial between sites (Figure 1).
We measured abundance of eastern small-footed bats in circular plots of various sizes, the placement of which we predetermined using ArcMap GIS software (ESRI, Redlands, CA). To select plots we outlined the perimeter of each talus slope with the polygon tool in ArcMap on digital ortho-images, and created a pool of nonoverlapping randomly distributed points in the polygon to serve as the centers of plots. Observers navigated to within 3–5 m of the center of each plot using a handheld GPS device (Triton, Magellan, Corp., San Dimas, CA), then used an aerial image to judge final placement relative to landmarks such as particularly large boulders or prominent evergreen trees. We used a measuring tape to determine radius of the plot, by affixing one end to a backpack at the center of the plot and using colored chalk at the other end to mark the perimeter. Size of plots varied over time and among sites, out of necessity. In the first year of the study we used plots with a 5-m radius, but increased to plots with 7–10-m radius afterwards to minimize the number of plots with no bats and reduce zero-inflation in our statistical models. However, we needed smaller plots on particularly small talus slopes to prevent them from extending into the forest where eastern small-footed bats generally do not roost. We also increased size of plots seasonally to reduce the zero-inflation that we presumed would increase when adult females clustered into maternity colonies. In Virginia, eastern small-footed bats of both sexes tend to roost alone until parturition in early to mid-June, when females begin to cluster into colonies (Moosman et al. 2015). We considered 14 June as the beginning of the colonial roosting period but adjusted this date if there was evidence colonies had begun forming earlier during a given year. Such evidence included the observation of more than two adult female bats roosting together in the same crevice, or of lactating adults or pups. When possible, we used plots with a 7-m radius during the period when bats roosted solitarily and 10-m-radius plots during the colonial roosting period.
Teams of two to three people conducted surveys; they attempted to search all crevices in the plot using 400-lumen LED flashlights. We considered any crevice formed within or between rocks that was accessible from the surface as a potential roost site. However, eastern small-footed bats usually roosted in crevices with openings about 1 cm wide, or sometimes in those with wider openings if part of the internal recesses narrowed to ≤ 1 cm wide (see Figures S1–S6, Supplemental Materials). Note that we defined crevice width as the distance between the two planar surfaces forming the crevice—that is, the top and bottom surfaces for horizontal crevices or the two lateral surfaces for vertical ones. We attempted to reduce potential for observer bias by giving new surveyors an onsite tutorial about the roosting habits of eastern small-footed bats before their first survey, and by pairing newly trained surveyors with more experienced surveyors. We paused searches to allow the more experienced surveyor to show any bat they found to the less experienced surveyor, to speed development of their search image.
Each surveyor started searching in a different portion of the plot and used uniquely colored chalk to mark crevices they inspected. Surveyors initially focused their efforts on portions of the plot that still had unexamined crevices. After the team had searched the entire plot, each person examined their partner's portion of the plot to check for overlooked crevices. Searching ceased once both surveyors had examined the entire plot and could no longer find overlooked crevices. Surveyors used handheld analog counters to record the number of crevices searched, and they recorded the number of bats found on a datasheet. We did not capture bats because it would have substantially increased the time it took to complete surveys, and because we wished to minimize disturbance. We included crevices that overlapped the edge of a plot in counts, but included bats only if they were inside the plot. We limited surveys to one or two plots per day so surveyors did not become fatigued. The number of plots surveyed at each site varied based on the size of each talus slope. At the three smallest sites (Humpback, Slacks, and St. Mary) we surveyed an average of two to three plots (or 10–34% of each talus slope) annually. We spent more total effort annually (average of four to nine plots) at the remaining sites, but because they were substantially larger this amounted to only 6–7% of each talus slope.
We gathered empirical data on how frequently surveyors overlooked bats during surveys (false negatives) at the largest site (Marbleyard) in 2017 and 2018, using 11 trials with a total of nine different radio-tagged bats and eight different surveyors. Each trial involved a single radio-tagged bat, with seven bats used once and two bats used twice. We captured solitary bats from their roosts using two 30-cm lengths of solid-core grounding wire to coax them to the surface, with one wire to block the bat from retreating deeper into the crevice and the other to prod it to the entrance so we could capture it by hand. We glued radio transmitters (0.20-g Pip41, Lotek Wireless, Newmarket, ON; or 0.23-g LB-2N, Holohil Systems, Ltd., Carp, ON) that were 3.8–5.1% of body weight to the medial scapular region of bats, using medical adhesive (Torbot Bonding Cement, Torbot Group, Inc., Cranston, RI). Each bat was radio-tracked to its roost 1 or 2 d after its initial capture by a solitary observer, while one or two surveyors waited out of sight below the base of the talus slope (∼ 100 m away). Surveyors were “naïve” in that they did not know whether the plots they searched contained a radio-tagged bat. Backgrounds of surveyors ranged from volunteers with no prior experience, to relatively experienced student researchers, and the principal investigator (PI). Upon locating the bat, the solitary observer marked two adjacent plots: one containing the radio-tagged bat and one in a random direction to serve as a dummy plot. Surveyors chose which plot to search first and surveyed both using the same methods as in plots used to monitor abundance. The solitary observer was present to record outcome of each trial, but did not interact with the surveyors to reduce the chances of providing cues that might influence surveyor behavior. We recorded whether surveyors successfully documented presence of each radio-tagged bat, and used those data to understand sources of error. For trials that resulted in false negatives, we recorded whether bats were visible but simply overlooked (observer error), or if bats roosted in a recess of the crevice that made them difficult to see (methodological error). We did not include trials used to study false negatives in the dataset to assess patterns in abundance or detection probability. The Animal Subjects Committee at the Virginia Military Institute approved our methods of handling bats, and we conducted our study under wildlife collecting permits from the Virginia Department of Wildlife Resources, and a scientific research permit from the U.S. Park Service.
The sampling unit for statistical analyses was the individual survey (i.e., each instance a plot was searched). We modeled the number of bats counted per plot using mixed effects models (Pinheiro and Bates 2000). All models included a random intercept for site (i.e., talus slope) and for plot within site. In one model, we included year as a random intercept to allow for variation in bat counts among years. In another model, we added year as a linear predictor (i.e., fixed effect) to test for a trend in bat counts over time. We modeled roosting period (i.e., solitary vs. colonial) as a fixed effect, as we did number of crevices searched, which we included to account for differential search effort. We considered crevices searched as a more accurate index of search effort than plot radius, because bats only roost in crevices, and crevice abundance varied among plots due to habitat heterogeneity within and among sites. Thus, plot radius did not necessarily reflect the amount of roosting habitat contained within a given plot. Bat counts were overdispersed from a Poisson (dispersion = 1.96, P = 0.03) and we therefore modeled them with a negative binomial distribution. We also considered models that used a zero-inflated negative binomial for bat counts, but in all cases the zero-inflated model had a poorer fit to the data (i.e., higher Akaike's Information Criterion [AIC]). We fit models using the glmmTMB package (Brooks et al. 2017) in R version 3.5.2 (R Development Core Team 2018).
In some cases, researchers may be interested in monitoring presence or absence of bats at talus slopes rather than the actual counts of bats at these habitats. Talus surveys are amenable to such “occupancy” modeling (MacKenzie et al. 2002), in which the observation of animals reflects both an underlying state variable for true presence or absence and a detection process. Specifically, the probability of observing an animal at a site is equal to the probability that the animal is present (i.e., ”occupancy probability”) multiplied by the probability that the animal will be detected given that it is present (i.e., “detection probability”). The detection process may itself depend on covariates such as time of year, search effort, and characteristics of sites. Such modeling approaches are widely used in wildlife biology (MacKenzie et al. 2018) and a variety of tools exist for parameterizing biologically realistic models (MacKenzie et al. 2003; Kéry and Chandler 2012).
We used a static model without extinction or colonization (MacKenzie et al. 2002) to estimate detection probabilities for the three sites with the most surveys, using repeated surveys of each plot across years to build detection histories (Marbleyard, Sherando, and St. Mary). We did not have sufficient data to parameterize models in which covariates varied from year to year, though these models should be feasible with larger data sets. We knew that bats occupied these three sites in each year (i.e., surveyors found some bats), so we assumed occupancy probability to be one, and our analysis focused on the factors that influence detection probability in each individual survey. We considered three covariates that might affect the observation process: 1) the number of crevices searched within each surveyed plot, 2) roosting period of the survey (solitary vs. colonial), and 3) a variable for site reflecting differential abundance at those sites. We used AIC to compare all possible combinations of these three detection variables: an intercept-only model, a model for each individual detection variable, three two-variable models, and the full model with all three variables. We estimated detection probability for each site and survey based on model-averaged parameters (Burnham and Anderson 2002). Additionally, we used the mean detection probability for each site to estimate the number of surveys needed at that site for an overall detection probability of either 0.80 or 0.90. This was calculated as po = 1 − (1 − ps)s where po is the overall detection probability, ps is the detection probability for one survey, and s is the number of surveys (MacKenzie et al. 2006). We fit models assessing detection probability using the “occu” function in the R “unmarked” package (Fiske and Chandler 2011) and compared fits using the AICmodavg package (Mazerolle 2019).
Lastly, we modeled the observation process to examine the effects of surveyor experience on bat detection. In these models, the response variable was observation outcome (detection vs. nondetection), which was modeled using a binomial distribution. In one model, we compared detections by the PI to detections by student researchers to determine the effect of experience on the efficacy of surveys. In another model, we created a subset of detection data from student researchers only and examined the effects of surveyor experience (specifically, the number of prior surveys by that student) on the probability of detection. These models also included a fixed effect for crevices counted to account for search effort, and random intercepts for year and site. The model for observer experience also included a random effect for each individual surveyor in order to isolate the effects of experience from differences among individuals. We fit observer models using the glmer function in lme4, in the program R (Bates et al. 2015). We report linear model coefficients (β) ± standard error (SE) for fixed effects and variance (σ2) for random effects.
We counted a total of 141 eastern small-footed bats, in 95 solitary roosts and 11 roosts containing groups of 2–10 bats, out of a total of 119 surveys in 43 plots (Table S1, Supplemental Material). We also encountered big brown bats but in numbers too low to include in the analysis. Surveys of most plots contained one or more solitary eastern small-footed bats; others had relatively high bat counts because they contained maternity colonies (Figure 2). In the model of bat counts with year as a random effect, variance was higher among plots within sites (σ2 = 0.63) than among years (σ2 = 0.10) or among sites (σ2 = 0.10). Bat counts tended to increase with crevices searched, but this relationship was not strong (Figure 3; β = 0.007 ± 0.003 SE). Histograms of bat counts suggested there may be underlying differences in spatial distribution of bats between roosting periods, but we detected no clear effect of roosting period on bat counts (Figure 4; β = 0.13 ± 0.31 SE). When we modeled year as a linear predictor rather than a random effect, the trend in bat counts over times was negative (β = −0.07), depending on site, and the standard error for this estimate was large (SE = 0.07; Figure 5). Other results of the trend model were similar to the random effects model. Variance in counts was somewhat higher between plots within sites (σ2 = 0.41) than among sites (σ2 = 0.23), crevices searched was positively related to counts (β = 0.009 ± 0.002 SE), and there were no substantial differences in counts between roosting periods (β = 0.23 ± 0.28 SE).
Two models for detection probability clearly performed best at Marbleyard, St. Mary, and Sherando (Table 1): a model including number of crevices counted and differences among sites, and a model including these variables plus roosting period. Detection probability was positively related to the number of crevices searched (β = 0.027 ± 0.006 SE), and detection probability was higher at St. Mary compared to Marbleyard (β = 0.61 ± 0.75 SE) and lower at Sherando (β = −2.86 ± 0.84 SE). The effect of roosting period (which, based on AIC, had only limited support) corresponded to a higher detection probability during the solitary period (β = 0.73 ± 0.64 SE), with confidence intervals that overlapped zero.
Mean site-specific detection probability across all plots surveyed and years was 0.58 for Marbleyard, 0.75 for St. Mary, and 0.17 for Sherando. However, the detection probability for individual plot surveys varied from 0.99 for surveys of plots with many crevices searched at St. Mary to 0.07 for surveys of plots with fewer crevices searched at Sherando. Based on the average detection probability for each site, achieving an 80% detection probability in any given year at Marbleyard and St. Mary would require surveying only two plots, whereas surveyors at Sherando would need nine plots. To get a 90% detection probability in any given year, Marbleyard would require surveying three plots, St. Mary would require two plots, and Sherando would require 13 plots. In trials to obtain empirical data on false negatives at the Marbleyard site, surveyors detected 64% of radio-tagged bats (or 7 out of 11 trials). Out of the four trials that resulted in false negatives, two bats were visible but overlooked due to observer error, and two bats roosted in parts of the crevice that were not visible to observers (Table 2).
Bat counts conducted by the PI, who had the most extensive experience with eastern small-footed bat roosting habits, tended to be higher than those of student researchers (β = 0.96 ± 0.39 SE). However, this does not appear to have been from less search effort by students—mean number of crevices searched was similar for students (53.8 ± 3.0 SE) and the PI (52.9 ± 2.8 SE). In the model assessing effects of prior searches among students (i.e., excluding the PI; Table S2, Supplemental Material), there appeared to be small but variable effects of experience (β = 0.04 ± 0.02 SE).
Our results broadly suggest that visual surveys are useful for assessing populations of eastern small-footed bats on talus slopes, and the technique has significant potential to improve monitoring efforts for other species of rock-roosting bats. Detection probabilities for talus surveys were comparable or better than those reported for other vertebrates for which visual surveys are an important monitoring tool, including some amphibians, birds, and marine mammals (Conway and Simon 2003; Pollock et al. 2006; Iwai 2017). Modeled detection probabilities for talus surveys were also close to the empirical observations from radio-tagged bats at the Marbleyard site. Although our empirical estimate of false negatives had small sample size, together with modeled detection probabilities, they demonstrate that it is possible to assess reliability of talus surveys. This ability could make talus surveys especially valuable. Error associated with other methods of surveying bat populations (especially counts in winter) are largely unknown, and this is a widely recognized weakness that has yet to be addressed (O'Shea et al. 2003).
The tendency for talus surveys to produce relatively high variation among plots within sites, both in terms of detection probability and abundance, likely will dictate sample size in studies aimed at detecting differences between sites or years. We presume that both the within- and between-site variation in bat counts and detection probability was at least partly a reflection of habitat quality. Researchers have devoted only rudimentary study to factors influencing roost selection by eastern small-footed bats (Johnson and Gates 2008; Johnson et al. 2011), but wildlife biologists widely understand that thermoregulatory concerns drive roost selection in bats, including for rock-roosting species (Chruszcz and Barclay 2002; Neubaum et al. 2006). Size, shape, orientation, and solar exposure of rocks may influence the thermal traits of the crevices they create (Chruszcz and Barclay 2002; Lausen and Barclay 2003; Neubaum et al. 2006). All of our study sites received relatively high solar radiation because they lacked a tree canopy and were predominantly south facing, but solar exposure varied among and within sites from differences in topography and solar aspect. Rock size and shape also varied dramatically among and within sites. Rock size at Marbleyard and Blackrock Summit was several orders of magnitude larger than that at Sherando and Slacks, and there was further heterogeneity in rock size within each talus slope. Geophysical processes that create talus slopes tend to sort rocks longitudinally by size, with either the largest or smallest rocks farther downslope, depending in part on shape of the fragments (Statham 1972; Bones 1973; De Blasio and Sæter 2010). Such factors could explain why particular plots or sites had especially high bat densities, but researchers need to test this idea.
It is also worth considering that talus slopes tend to be part of larger networks of rock outcrops, as was the case for our study sites. Surrounding sites that we did not sample may have influenced abundance and occupancy at a given study site. Many questions remain about the degree to which bats travel between rock outcrops, and the importance of this connectivity to populations of rock-roosting bats. A number of authors have commented about this idea but there has been little direct work on the topic (Arnett and Hayes 2009; Anthony and Sanchez 2018). We did not attempt to investigate factors influencing abundance and habitat suitability because we had relatively few study sites, but talus surveys applied across larger spatial scales could be valuable in this regard.
In addition to influencing abundance of bats, rock size and shape may also affect surveyor efficiency. The sites with the lowest bat counts and detection probability (Slacks and Sherando) had much smaller rocks than our other sites. Small rocks created more crevices that observers had to inspect, but most were marginally suitable or too small for eastern small-footed bats to occupy. Smaller rocks also made the slope less stable, requiring us to choose our steps carefully to prevent rocks from shifting, whereas large rocks typically do not move when stepped on. On the other hand, searching for bats amongst especially large boulders was also challenging. Researchers planning to implement talus surveys should consider these points from a risk management perspective. They should perhaps exclude sites that carry excessive risk of injury to bats or surveyors from monitoring studies, depending on the goals of the research.
Models of both detection probability and abundance also demonstrated that surveys were prone to a moderate degree of interannual variation. This is relevant to monitoring efforts because it means that researchers will likely need longer time series to detect changes in population. Our exploratory assessment of population trends for eastern small-footed bats did not detect a strong overall pattern across sites, only potential emerging site-specific trends that warrant continued monitoring. These findings are not surprising given that we analyzed counts across only 7 y total, and fewer years for most sites. The reason for interannual variation in bat counts and detection probability are unclear, but we suspect frequent roost switching by eastern small-footed bats played a role. A small number of short-duration studies have suggested the species moves to new roosts within the same rocky slopes (e.g., < 50–60 m away), on a nearly daily basis (Johnson and Gates 2008; Johnson et al. 2011). Annual or seasonal changes in the spatial distribution of bats on talus slopes, relative to plot placement, could account for high interannual variation in count data and detection probability. Additionally, although we did not observe substantial differences in results between roosting periods, clustering by females into colonies did appear to contribute to higher variance in counts, and the behavior was linked to slightly lower detection probabilities. Interestingly, eastern small-footed bats also exhibit relatively high interannual variation in counts during winter, compared to co-occurring species of hibernating bats (Trombulak et al. 2001). Perhaps comparable movements between hibernacula affect winter counts of eastern small-footed bats.
Researchers who conduct visual surveys are frequently concerned about the potential for bias within and among observers (Thompson and Mapstone 1997; Potvin and Breton 2005). We detected only moderate differences between counts by the most experienced surveyor and student researchers, and the learning curve does not appear to be steep. This suggests that talus surveys can be generally effective in detecting bats even when observers do not have extensive prior experience with the technique. The fact that we used multiple surveyors for each plot probably promoted reliability in bat counts by reducing the number of overlooked crevices. Researchers consider this double-observer approach more reliable than surveys by single observers, and there have been calls to adopt it for bat surveys (Nichols et al. 2000; O'Shea et al. 2003). But we also suspect that, because each plot was surveyed by more than one person simultaneously, the slightly higher counts of the more experienced surveyor may reflect differential efficiency rather than true observer bias. That is, the most experienced surveyors may have found bats that their less experienced partners would have eventually found if they had been working alone. Trials with radio-tagged bats suggested that false negatives happened because bats sometimes hid in parts of crevices that researchers could not see, and because researchers sometimes overlooked bats due to carelessness or fatigue. We note that in trials with radio-tagged bats (which mostly did not use dual observers) inexperienced surveyors performed well. Both of the false negatives caused from true observer error were in plots searched by relatively experienced surveyors. Perhaps using a boroscope to search crevices that are difficult to inspect by naked eye would improve results in this regard.
Because of risks to bats and researchers, we recommend a few precautions when conducting talus surveys. First-time surveyors should either gain experience with the roosting habits of the focal species through radio-tracking, or receive onsite training from more experienced people to learn how to avoid stepping on any rock that forms a potential roost or is likely to move. For example, flat horizontal rocks make inviting steps but they tend to create crevices that are suitable as roosts, and stepping on them could kill bats. Surveyors should also consider wearing personal protective gear, when warranted, such as rock-climbing helmets and perhaps gaiters or chaps capable of protecting against bites from venomous snakes. Some of the talus sites we surveyed were habitat for eastern timber rattlesnakes Crotalus horridus and northern copperheads Agkistrodon contortix. Surveyors could easily avoid these dangers at our study sites but they deserve consideration.
Overall, the performance of talus surveys for assessing eastern small-footed bat populations in Virginia warrants expanding the technique to include other species of rock-roosting bats with poorly understood population status. Even in regions where researchers have monitored bat populations with longer-established methods, talus surveys would offer another line of evidence to assess local populations. If talus surveys prove effective with other species of rock-roosting bats, we suggest the technique be included in the suite of approaches advocated by the North American Bat Monitoring Program (Loeb et al. 2015).
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.
Figure S1. The Marbleyard study site, a 3.0-ha talus slope composed of relatively large sedimentary quartzite boulders, James River Face Wilderness Area, Rockbridge County, Virginia. Size, shape, and orientation of rocks were heterogeneously distributed across the contoured surface of the talus slope. Rock sizes ranged from bedrock slabs (A) and large boulders (B), mostly near the top of the slope, to < 1-m-wide blocks (C) downslope. This site had relatively high densities of crevices suitable for eastern small-footed bat Myotis leibii roosts.
Found at DOI: https://doi.org/10.3996/122019-NAF-103.S1 (467 KB PDF).
Figure S2. The Saint Mary's Wilderness Area study site, a 0.15-ha patch of talus with rocks ranging from 10 cm to > 2 m wide, Rockbridge County, Virginia. This site had relatively high densities of crevices suitable for eastern small-footed bat Myotis leibii roosts.
Found at DOI: https://doi.org/10.3996/122019-NAF-103.S2 (225 KB PDF).
Figure S3. The Sherando Lake Recreation Area study site, a 3.25-ha complex of talus slopes in Augusta County, Virginia. We surveyed a 1.18-ha subset of the site for eastern small-footed bats Myotis leibii, which had rocks ranging from ∼ 1-m-wide blocks (A) to smaller fragments ∼ 10–30 cm wide (B). We did not survey the more extensive area of smaller rocks (scree) lower on the slope (C). Yellow arrows indicate roosts. This site had lower densities of suitable crevices than other sites.
Found at DOI: https://doi.org/10.3996/122019-NAF-103.S3 (717 KB PDF).
Figure S4. The Humpback Rocks study site, a 0.15-ha talus slope in Augusta County, Virginia. Boulders at this site were flat and oblong, and created an unusually high density of horizontal crevices that were suitable as roosts for eastern small-footed bats Myotis leibii (A). Roosts (indicated by yellow arrows) sometimes were under small slate-like pieces of rock (B and C) that required extreme caution when choosing steps.
Found at DOI: https://doi.org/10.3996/122019-NAF-103.S4 (761 KB PDF).
Figure S5. The Slacks Overlook study site, a 0.2-ha thin strip of talus in Augusta County, Virginia. The site was dominated by small (∼ 10–20-cm-wide) rocks, which formed relatively few crevices suitable as roosts for eastern small-footed bats Myotis leibii.
Found at DOI: https://doi.org/10.3996/122019-NAF-103.S5 (234 KB PDF).
Figure S6. The Blackrock Summit study site, in Rockingham County, Virginia, is an extensive area of talus surrounding the summit. Rocks ranged from 30 cm to > 1 m wide and were largest near the top of the rock outcrop (A and B). We visually surveyed a subset of the talus slope surrounding the summit itself for eastern small-footed bats Myotis leibii, but excluded lower portions (C) with more lichen cover to avoid damaging lichens. Crevices were oriented in a variety of ways, including this diagonal ∼ 1-cm-wide crack in a large boulder that housed a maternity colony (D). The site had relatively high densities of suitable crevices.
Found at DOI: https://doi.org/10.3996/122019-NAF-103.S6 (709 KB PDF).
Table S1. Raw data from visual surveys for eastern small-footed bats Myotis leibii on talus slopes in Virginia, 2013–2019.
Found at DOI: https://doi.org/10.3996/122019-NAF-103.S7 (61 KB XLSX).
Table S2. The outcomes of visual surveys for eastern small-footed bats Myotis leibii on roosting on talus slopes in Virginia, 2013–2019, relative to experience level of surveyors. Surveyors included student researchers that varied in level of experience at the time of each survey vs. the principal investigator (PI).
Found at DOI: https://doi.org/10.3996/122019-NAF-103.S8 (56 KB XLSX).
Reference S1.Loeb SC, Rodhouse TJ, Ellison LE, Lausen CL, Reichard JD, Irvine KM, Ingersoll TE, Coleman JTH, Thogmartin WE, Sauer JR, Francis CM, Bayless ML, Stanley TR, Johnson DH. 2015. A plan of the North American Bat Monitoring Program (NABat). Asheville, North Carolina: U.S. Department of Agriculture, Forest Service, General Technical Report SRS-208.
We are grateful to the reviewers and editors that worked to help improve this paper from its initial version. Additionally, we thank D. Warner, R. Hendren, C. Ivey, M. Briggs, A. Barlow, S. Sell, K. Powers, S. Custer, M. Beckner, S. Bryan, W. Haslam, D. Pody, H. Thomas, J. Huth, W. Ford, E. Hill, S. Dickenson, D. Moosman, N. Kalen, M. Furr, D. Wright, and K. Simms for help with various aspects of the project, and the U.S. Forest Service and National Park Service for allowing us to conduct the study. This work was funded by the Virginia Department of Wildlife Resources (contract 2013-13772), through a Wildlife Restoration Grant from the U.S. Fish and Wildlife Service.
Any use of trade, product, website, or firm names is for descriptive purposes only and does not imply endorsement by the U.S. Government.
Citation: Moosman PR Jr, Marsh DM, Pody EK, Dannon MP, Reynolds RJ. 2020. Efficacy of visual surveys for monitoring populations of talus-roosting bats. Journal of Fish and Wildlife Management 11(2):597–608; e1944-687X. https://doi.org/10.3996/122019-JFWM-103
The findings and conclusions in this article are those of the author(s) and do not necessarily represent the views of the U.S. Fish and Wildlife Service.