Role of Trapping in Detection of a Small Bog Turtle (Glyptemys muhlenbergii) Population
ABSTRACT
We used data collected from a small but increasing population of bog turtles to evaluate the surveying methods for detecting such a small population and the level of trapping effort required to reliably detect it. Trapping with unbaited traps detected this small population more reliably and with less investigator-time investment than did widely used visual and manual search techniques. This population was most easily detected in May and June, but we still needed more than 9000 trap-hours of trapping effort to reach a 95% likelihood of capturing one of the turtles present. Based on our results, we suggest a standard of 20 traps per hectare for 20 days (a 20–20 rule) as an effort level that has a very high likelihood of detecting such small but potentially viable populations.
The bog turtle, Glyptemys muhlenbergii, formerly assigned to the genus Clemmys (Holman and Fritz 2001), is declining throughout its range as a result of habitat destruction and alteration. Bog turtles are naturally vulnerable to loss of small wetlands because they occur in discrete and isolated populations (Ernst 2001). The turtle is now protected as threatened or endangered in all 12 US states of occurrence. In 1997, the northern population was federally listed as threatened and the southern population was listed as threatened by similarity of appearance (USFWS 1997a, b).
Surveys are being conducted throughout the range of the bog turtle to assess potential sites for various reasons. Natural areas inventories are being conducted by state agencies to assess the quality of habitat and the presence of rare species. Nongovernmental entities, e.g., Project Bog Turtle, conduct surveys to identify new sites and to determine the conservation status of the turtles throughout the south. Developers also initiate surveys in sites scheduled for projects to satisfy the regulatory obligation for mitigation, to establish the need for buffer areas, or justify project modifications. Since the bog turtle was federally listed in 1997, there has been an increased demand for surveys to meet regulatory obligations.
When potential bog turtle sites are slated for development, surveyors are called upon to make a determination as to the presence or absence of these animals. In the experience of ABS, these requests come at all times of year and are often purported to be urgent. The searching-probing method, also known as the visual search method, is the most common method of detecting bog turtles (Smith 1994; Herman 1994). This method is described as a phase 2 survey in the Guidelines for Bog Turtle Surveys in the Bog Turtle Northern Population Recovery Plan (USFWS 2001). Researchers follow turtle tracks, look for basking turtles, and seek concealed animals by groping in the mud with their hands or probing with wooden sticks. Phase 2 surveys are conducted between 15 April and 15 June, and compliance with the protocol requires intensive investment of search time under defined weather conditions. Trapping is an additional method of surveying for bog turtles. Fahey (1992, 1998) first used this method in the south in the late 1980s. Most recent researchers have used trapping to some degree, usually in sites that look promising or in historic sites that have not yielded turtles by using the phase 2 survey method. Sites under permit review that are surveyed without finding evidence of bog turtles may then be destroyed.
Demographic data remain largely unavailable for most sites, but indications are that most populations are small (Table 1). In addition, bog turtles are difficult to detect because they are small, rare, cryptically colored, and behaviorally inconspicuous. It is probable that some sites that do contain turtles, especially those with small populations, are being overlooked. The question then arises: how and when should potential sites be examined to minimize the chances of overlooking a potentially viable population of bog turtles?
Under one commonly used evaluation system (Klemens 1993; USFWS 2001), sites occupied by as few as 6–10 turtles are regarded as “potentially viable” or “fair” if other habitat factors are favorable. To avoid the loss of such small populations through poorly timed or insufficiently intensive search, it is important to investigate the likelihood of detecting such small populations with commonly used search techniques. We drew upon data collected in the course of a habitat-use study of a small population of bog turtles of not less than 7 animals to explore this issue. By using data collected from 1993 to 2001, we compared the effectiveness of 2 common search techniques (searching-probing vs. trapping), determined the best season for such survey work, compared the detection success at this site with that for a larger population, and estimated the level of trapping effort required to reliably detect this small group.
Methods
The Sites
S3 is a complex of wetlands on a privately owned farm in the Piedmont region of North Carolina, approximately 450 m (1500 ft) above sea level. There are 3 very small and distinct wetland patches contained within a diameter of about 0.5 km and separated by a gravel-and-cobble dominated stream. Together they amount to 1.025 ha of wetlands suitable for bog turtles. Survey work at the site began in May of 1993. Based on extensive survey effort, we are confident that the population size, including both adults and hatchlings recruited during the study, was 7 turtles in 1993 and 13 in 2001. According to the habitat and population factors presented in Klemens (1993), the S3 site ranks as “possibly viable” among 3 ranks: “viable”, “possibly viable”, or “nonviable”. These categories were later revised to “good”, “fair”, and “poor” (USFWS 2001).
G2 is also on a privately owned farm located in the Piedmont in North Carolina. Although configured somewhat differently, the size of the wet area suitable for bog turtles is similar to that of S3. A study conducted during an overlapping time period (Green 1994, 1995, 1996) resulted in 48 turtles being captured at the site (Herman 1999), and the actual population is undoubtedly larger. G2 ranks as “viable” or “good” on the Klemens assessment scale.
Capture Methods
Turtles were captured by using visual searches, probing, and handmade unbaited traps. Traps were shaded, placed in shallow water, and checked every 24–48 hours. Captured bog turtles were permanently marked and released at the point of capture. Traps were used in S3 for slightly more than 175,000 trap-hours over the 8-year period (1 trap-hour = 1 trap in site for 1 hour). In 1993 and 1994, traps were distributed across the 3 patches of S3. After 1995, traps were placed in areas where turtles were expected, and areas that had been nonproductive in the past were avoided. Visual search times were estimated and recorded at the end of each visit. Time invested by the workers tending traps was also estimated.
Data Analysis
For comparison of the phase 2 techniques with trapping, a turtle yield value was computed for each calendar month of effort at the site as the number of turtles captured per hour of investigator time. The calendar year was divided into 2-month categories to achieve a statistically useful number of yield values for each time category. Two-way analysis of variance (ANOVA) was used to examine the effect of time period and capture technique.
For a comparison of trapping success between sites, we aggregated trap hours on a monthly basis:
Monthly trap-hours = number of traps present × hours of trap effort for month
We compared effectiveness of trapping turtles in S3 (N = 7–15) and G1 (N = 48+). Because of inter-year variation in turtle activity from weather conditions, only periods of simultaneous study were compared. Trapping occurred on both sites simultaneously during 4 months (June 1993, May 1994, June 1994, and June 1995). A Mann-Whitney U-test was used to compare trapping success, expressed as turtles/trap-hour, across the 2 sites.
To evaluate the intensity of trapping effort needed to detect the small S3 population, we computed the cumulative trap-hours that led up to each successful trapping event. We began either from the initial placement of traps in the site for the first capture or from the previous trap capture in the site. Sometimes two or more turtles were captured on the same day in the same or different traps. In these cases, all were assigned cumulative trap-hours from the most recent capture, rather than treating the second and subsequent turtles as requiring zero trap-hours for their capture. We transformed these data logarithmically to achieve a normal distribution and used the transformed data to calculate a cumulative trap-hour effort that would encompass 95% of this distribution. We performed this analysis on the whole data set and on a subset of captures that occurred within May and June during the 1993–1995 time frame in case improved trap placement late in the study influenced the time required.
We also derived an alternative estimator of the trapping effort necessary to detect the S3 population through an inverse prediction procedure. By using May and June data only, we modeled the number of turtles captured during a calendar month as a function of the trap-hours of trapping effort in the site during that calendar month. We then obtained a 95% confidence interval for the trap-hours needed to capture a single turtle. Because capture of one animal would suffice to confirm presence at the site, the upper bound of the 95% confidence interval represents the trapping effort required for this probability of detecting the population.
The R system (R Development Core Team 2005) was used for all statistical operations described.
Results
Technique and Season at S3
Trapping was much more effective than visual searching for detecting bog turtles at S3 (Fig. 1). Locating turtles by phase 2 methods at the S3 site required an average of 27.4 search hours per turtle captured for the years 1993 to 1997 and 2000. No turtles were captured this way in 1998 or 1999. Effects of both bimonthly category (p = 0.006) and capture technique (p < 0.001) were highly significant, and the trapping proved to be a more effective investment of the investigator's time. May and June were the most rewarding times to search for turtles by any method at this site.
Citation: Chelonian Conservation and Biology 7, 1; 10.2744/CCB-0683.1
Locating bog turtles remained difficult at S3 during May and June compared with G2 (Fig. 2). Trapping effort per turtle capture at S3 site for the period of overlap averaged 2600 trap-hours per turtle trapped as opposed to an average of 176 trap-hours at the G2 site. Average trapping success during the 4 months of concurrent effort was more than an order of magnitude lower at S3 (3.05 × 10−4 turtles/trap-hour) than at G2 (6.38 × 10−3 turtles/trap-hour). Despite the small sample size, the intersite difference was statistically significant (p = 0.021).
Citation: Chelonian Conservation and Biology 7, 1; 10.2744/CCB-0683.1
Detection of the Small Population at S3
The number of turtles trapped in a calendar month at S3 depended on the trapping time invested, as well as the month of the year. Turtles were not captured during every month of active work at the site. During 7 of the 23 months in which trapping was conducted, no turtles were caught; in 3 of those months > 1000 trap-hours were invested without success. Analysis of the cumulative trap presence times leading up to each S3 capture indicated that the average (log-transformed) capture occurred at 1820 trap-hours overall. In the subset of May and June effort early in the study, the corresponding value was 1850 trap-hours. The upper bound of the range into which a new capture was 95% likely to fall throughout the study was 9286 trap-hours across the entire study, or 10,079 trap-hours for the May-June subset of the data from early in the study.
The inverse prediction approach applied to May and June data yielded a much larger estimate, 18,645 trap-hours (Fig. 3). Two points that represented May and June of 1999, a drought year, had a very large effect on this analysis. Eliminating these 2 points from the inverse prediction yielded an investment of 10,550 trap-hours for a 95% likelihood of detection: a value in reasonable accord with those derived from the cumulative trap-presence calculations.
Citation: Chelonian Conservation and Biology 7, 1; 10.2744/CCB-0683.1
Discussion
Search Method and Season
Our findings agree with earlier studies (Lovich et al. 1992; Ernst et al. 1994; Carter 1997; USFWS 2001) that the optimal time for finding turtles is in the spring. In all seasons, trapping yielded more turtles per hour of investigator time spent in the site than did searching-probing. S3 was a difficult site in which to find bog turtles by these phase 2 methods, compared with other locations where such data have been collected. Rates reported in the literature for visual search success range from 0.27 to 0.6 turtles per investigator hour, with seasonal variation in success (Table 2). The highest hand yield for any month at the S3 site was approximately 0.1 captures per investigator hour, a lower rate than that found by either Lovich et al. (1992) or Smith (1994). We are not aware of any published data on captures per hour of investigator effort for the trapping method.
Detectability: 20–20 Rule
Turtles in small populations present a special difficulty in detectability. Results of this study suggests that small populations of turtles in potential sites may easily go undetected when using only visual searches. According to our cumulative trap-hour analysis, a minimum of roughly 9000 trap-hours per ha of trapping effort in May and June is needed to have 95% confidence of detecting a population of bog turtles of the size reported in this study (N ≊10). The considerably larger value obtained from our inverse prediction approach was heavily influenced by 2 data points (Fig. 3), which represented May and June of 1999, a very dry year. Omission of these points may be justified if we assume that no responsible surveyor would attempt to evaluate an unknown potential site in a year as dry as 1999 was in the North Carolina Piedmont. The approximately 10,500 trap-hour value obtained by eliminating these outliers appears reasonable because the cumulative trap-hour analysis could not include unsuccessful efforts. Because the effort levels derived from these 2 methods are roughly equivalent to setting 20 traps for 20 days in May and June, we call this our “20–20 rule”. The influence of 1999 on the data suggests that detecting the S3 population during a drought would require much more than a 20–20 effort.
Detection failures represent a serious problem for bog turtle conservation because such a large proportion of the known sites are occupied by relatively few animals. Herman and Tryon (1997) determined that only 23% of the sites in the southern range of the turtle could be considered “viable”. Viability was defined in that study as a population of 30 or more individuals with sufficient core habitat and evidence of reproduction or recruitment. When using a similar suite of characteristics as outlined in Klemens (1993), USFWS (2001) combined occurrences into population assessment sites (PAS) for evaluation and concluded that only 38% of ranked sites (104 of 275) in the northern range of the turtle could be considered “good”. By using the same PAS method of aggregating sites to assess the populations in the south, Herman (2003) concluded that 50% ranked as “good”.
By any reckoning, the sites that can be considered “good” or “viable” deserve protection. Because the majority of the known bog turtle sites rank below this level, identification and protection will also be needed for some of the “possibly viable” or “fair” sites. Caughley and Gunn (1996) provide convincing evidence that adequate conservation plans require the preservation of as many individuals as possible within the greatest possible area of high-quality, protected habitat. A wide variety of criteria have been used to evaluate bog turtle sites (Collins 1990; Klemens 1993; Herman and Tryon 1997; USFWS 1997b, 2001), and, by most such assessments, the S3 site would be ranked as “possibly viable” or “fair”. Such sites may warrant consideration for conservation or site enhancement, especially if they are part of a meta-population. The population increase at S3 over the course of this study (Somers 2000) suggests that conservation efforts for populations of this size may be worthwhile.
Any question of the appropriate conservation effort for a marginally viable site is irrelevant if a turtle population that uses the site goes undetected. Detection failures at S3 present convincing evidence that sites with small populations can be and most likely are being overlooked when using only phase 2 survey techniques. Such oversights can result in irreversible habitat degradation if sites are under permit review, scheduled for alteration or project development. It will take further assessment that involves known and suspected bog turtle sites in both the northern and southern ranges, and which have the widest possible array of site characteristics, to determine if the 20–20 rule outlined above will be fully adequate to identify potentially viable populations that would otherwise have gone undetected.
Current guidelines (USFWS 2001, 2006) do not require phase 3 surveys (trapping); however, the revised guidelines (USFWS 2006) allow for the recommendation of additional surveys (phase 2 and/or phase 3 surveys) if phase 2 surveys are negative and habitat is of sufficient quality and quantity. Results of our study suggests that traps should be used to assess marginal sites within the turtle's range that show hydrology and vegetation characteristics capable of supporting bog turtles if searches fail to produce bog turtles or signs of bog turtles. Searching-probing surveys that fail to confirm the presence of bog turtles in a site slated for development should be considered inconclusive and trapping in May and June should be required.
Monthly yield of turtles per investigator hour of effort at S3 is shown as a function of season and search method. Solid bars represent trapping results; whereas, hatched bars show results of phase 2 survey techniques. Error bar = SEM.
Monthly turtle yields in turtles per trap-hour are compared for 4 months of overlapping trapping effort at a “fair” (S3, solid) and a “good” (G2, hatched) site. Error bar = SEM.
The number of bog turtle captures during May and June trapping attempts are plotted as a function of trapping time in thousand trap-hours invested during that calendar month. The regression line is forced through the origin.
