Understanding species distributions and relative abundance patterns is paramount in facilitating proper conservation and management. Many species, however, are rare or cryptic, or they frequent hard-to-survey areas, including aquatic environments. It can be particularly challenging to delineate a species' distribution, which can span a large geographic area, and simultaneously estimate abundance, which requires labor-intensive estimates of detection probabilities, although both are necessary when assessing the status of any species.

Barbour's map turtle (Graptemys barbouri; Carr and Marchand 1942) is endemic to rivers and creeks in a portion of 3 states of the southeastern United States; the species is state-listed as threatened in both Florida and Georgia and is of “high conservation concern” in Alabama. The US Fish and Wildlife Service recently assessed its conservation status in response to a petition for federal listing (Center for Biological Diversity 2010, 2012) that cited population declines due primarily to overexploitation, habitat loss, and degradation. Yet information is limited on the distribution, relative abundance, and population trends of G. barbouri across its range, which could be used to evaluate whether protections are warranted.

Barbour's map turtle has been described as a denizen of fast-moving sections of limestone-bottomed streams and rivers with ample basking sites, consisting of snags and fallen trees (Ernst and Lovich 2009; Sterrett et al. 2015). In Florida, it can also occupy sandy bottoms of alluvial rivers with low visibility (Choctawhatchee, lower Apalachicola), impoundments (Dead Lakes on the Chipola River), and blackwater rivers (Ochlockonee), although densities appear to be low in some of these areas. Historically, this species was known only from the Apalachicola–Chattahoochee–Flint river basin in southeastern Alabama, southwestern Georgia, and the Florida Panhandle (Carr and Marchand 1942; Cagle 1952; Sanderson 1992). Between 1996 and 2003, the known range expanded to include the Choctawhatchee River drainage in Florida and Alabama (Godwin 2002; Enge and Wallace 2008; Godwin et al. 2014) and the Ochlockonee River in Florida (Enge et al. 1996). We focused on these 4 core Florida populations (Apalachicola, Chipola, Choctawhatchee, and Ochlockonee) in our surveys. These 4 rivers comprise the entire known range of this turtle in Florida, although a female is known to have oviposited fertile eggs at the headwaters of the Wacissa River (Aucilla River drainage) in Jefferson County (Jackson 2003).

Although previous survey efforts have focused on G. barbouri in Florida, those surveys covered smaller sections of the range than our study. These previous surveys were done in 1985–1986 along 59 km of the Chipola River (Moler 1986), in 1990 in the lower Apalachicola River south of Brickyard Cutoff (Ruhl 1991), in 1992 in the upper (below Jim Woodruff Dam) and middle Apalachicola River (near Wewahitchka and lower Chipola Cutoff; Stewart 1992), in 1999–2000 in the Choctawhatchee and Ochlockonee rivers (Enge and Wallace 2008), and in 1999–2014 in the Apalachicola, Chipola, and Brothers rivers within the Apalachicola River Wildlife and Environmental Area (Ricketts 2014).

The objectives of the present study were to determine the current distribution and relative abundance of Barbour's map turtle throughout its range in Florida by surveying the Choctawhatchee, Apalachicola, and Ochlockonee rivers and their major tributaries using 1) single-pass basking surveys and 2) multiple-pass basking surveys and N-mixture modeling. Single-pass surveys are relatively rapid and allowed us to cover nearly the entire known geographic range to assess the distribution. Multiple-pass surveys are more time-consuming because they require repeated surveys of smaller areas, but they provide information on detection for more accurate estimation of abundance.

METHODS

To assess the distribution of Barbour's map turtle, we conducted single-pass basking surveys (Moler 1986; Lindeman 1999; Enge and Wallace 2008) on the Choctawhatchee, Apalachicola, Chipola (a prominent tributary of the Apalachicola), and Ochlockonee rivers from the Florida state line downstream to pretidal areas where the main-stem river channel began to braid (Fig. 1). For the Ochlockonee River, however, we ended standardized surveys just south of Whitehead Lake after covering > 25 river kilometers (rkm) with no G. barbouri observations. All surveys were performed during 2014–2015 in spring (April–June) or fall (October), between 0925 and 1900 hrs Eastern Standard Time (typically between 1000 and 1500 hrs). Each section was surveyed via kayak by a 2-person team traveling downstream in parallel (1 kayak per shoreline); observers used binoculars (×8 power) to scan ahead for turtles and basking sites. On the Ochlockonee and Choctawhatchee rivers, we surveyed the same sites as those surveyed in 1999–2000 by Enge and Wallace (2008). On the lower Apalachicola River and the Chipola River below the Dead Lakes, because the river is wider and the distance greater between boat ramps, we used a 4-person team using 2 motorboats traveling upstream, with 1 pilot and 1 observer per shoreline. For the motorboat surveys, a constant speed (8–9.5 km/hr) and distance from the bank (20 m) were maintained in an attempt to mimic the kayak surveys. All observations of G. barbouri (as well as all other reptiles) were recorded, along with observed number, sex/age class, time of day, and location (latitude and longitude). Graptemys barbouri are sexually dimorphic and nonjuveniles were sexed (male/female) and aged (adult/subadult/juvenile) based on a combination of size and sexual characteristics, or recorded as “unknown” if a determination could not be made. We classified subadults as individuals noticeably smaller than the average adult size but in the early stages of developing secondary sex characteristics. Subadult females were distinguished from similarly sized adult males by their wider heads, blunted carapacial spines, and small tails. Individuals classified as “subadult unknown” were large enough to have male secondary sex characteristics, but the sex could not be ascertained by the observers before the turtle escaped view. Although single-pass surveys focused on documenting the distribution of G. barbouri, we also report observed counts and densities (turtles/rkm). We note that these counts are the minimum number of turtles present during our surveys and should be interpreted as relative indices of abundance, as they do not account for detectability.

View Figure

• Download as Powerpoint
• Download Figure

To complement our observed counts from the single-pass surveys, we implemented multiple-pass basking surveys that included repeated observations of shorter sections of the rivers. These surveys were done independent of the single-pass surveys and allowed us to estimate abundance by incorporating detectability (probability of detecting an individual at a given site at a given time). We established 11 multiple-pass survey sections, five 5-km sections on the Ochlockonee and six 2-km sections on the Apalachicola; sections were separated by 1-km buffer sections (Figs. 2 and 3). When selecting multiple-pass survey sites, we chose the rivers from our single-pass distribution surveys with the lowest (Ochlockonee) and the highest (Apalachicola) observed turtle densities. Each section was surveyed 5 times over 5 d within a short period for each river (< 10 days). Abundance estimates and detection rates were calculated for these multiple-pass surveys using N-mixture models (Royle 2004) and the “unmarked” package in R (Fiske and Chandler 2011). We assumed a Poisson distribution for the counts and a binomial distribution for detection. We used Akaike's Information Criterion for small samples sizes (AICc) to compare models. We expected that detection varied from day to day (due to such factors as different weather conditions or increased boat traffic), so we considered models for each river with either constant detection or detection that varied by day. In the field, we tended to find turtles in sinuous parts of the rivers, so we quantified the sinuosity of each section (using a Python toolbox in ArcGIS) and considered sinuosity as a single covariate influencing our counts for all models. We also considered constant-intercept abundance models for each river. For the Ochlockonee River, we noticed in the field that our counts tended to vary by section (> 80% of our observations were in the first 2 sections), so we additionally considered a categorical single covariate that distinguished the upper portion of the river (sections 1 and 2) from the lower portion of the river (sections 3, 4, and 5). From the top N-mixture model for each river section (lowest AICc), we obtained estimates of abundance and detection rates.

View Figure

• Download as Powerpoint
• Download Figure

View Figure

• Download as Powerpoint
• Download Figure

To approximate G. barbouri abundance across the entire Florida range, we used the highest and lowest detection rates calculated by the N-mixture models to adjust our single-pass survey counts (approximate abundance = count/detection probability). These estimates provide a range of abundance values for our single-pass distribution counts, but assume that detection for all single-pass surveys was within the range of detection values calculated during multiple-pass surveys of shorter river sections.

RESULTS

During the single-pass survey, 502 rkm were surveyed and we observed 5917 Barbour's map turtles (Fig. 4; Table 1). The known range of G. barbouri was extended 17.3 km downstream on the Choctawhatchee and we documented the first records from Walton County (vouchered via digital photos, UF177222 and UF177223; Hill and Mays 2016). We extended the range 7.5 km upstream and 51.4 km downstream on the Ochlockonee and documented the first records from Wakulla County (UF177224 and UF177225, Mays and Hill 2016). The Apalachicola River produced the most G. barbouri observations, with 3779 individuals observed (21.8 turtles/rkm), followed by the Choctawhatchee, with 1245 (10.6 turtles/rkm), the Chipola, with 826 (7.1 turtles/rkm), and the Ochlockonee, with 67 (0.7 turtles/rkm; Fig. 4).

View Figure

• Download as Powerpoint
• Download Figure

Table 1. Sex and age structure of observed Graptemys barbouri, total counts, and densities (turtle counts/river kilometer [rkm]) from single-pass basking surveys in the Choctawhatchee, Chipola, Apalachicola, and Ochlockonee rivers, Florida, 2014–2015. Also shown are approximate abundance estimates for each river, calculated using our range-of-detection values estimated from multiple-pass surveys of smaller sections of rivers (Apalachicola, Ochlockonee) and N-mixture modeling (21%–51% detection) to adjust our raw counts (count/detection = approximate abundance).

View Fullscreen

Site-specific survey data are presented in Table 2. The proportion of map turtle age classes (adult, subadult, juvenile) and sexes observed differed among the rivers (Table 1; Fig. 5), but overall, adults were the most abundant (45%), followed by subadults (25%) and juveniles (14%). The remaining 16% of turtles were of unknown age or sex class and consisted almost entirely of subadults or adult males that could not be seen clearly enough to differentiate. Overall, 8629 reptile observations representing 23 species were tallied during our surveys; most of these were turtles (96%, n = 8260). Barbour's map turtle was the most commonly observed species, comprising 72% of all turtles and 69% of all reptiles observed (Table 3).

Table 2. River sections surveyed and number of Graptemys barbouri observed per each single-pass basking survey on the Choctawhatchee, Chipola, Apalachicola, and Ochlockonee rivers, Florida, 2014–2015.

View Fullscreen

View Figure

• Download as Powerpoint
• Download Figure

Table 3. Reptile observations tallied during Graptemys barbouri distribution surveys on the Choctawhatchee, Chipola, Apalachicola, and Ochlockonee rivers, Florida, 2014–2015.

View Fullscreen

The multiple-pass survey abundance estimate for the six 2-km sections (12 km total) along the Apalachicola River was 2079 map turtles (or 173.3 turtles/rkm), with an estimated 21%–32% daily detection rate (Tables 4 and 5). For the Ochlockonee, the multiple-pass survey abundance estimate for the five 5-km sections (25 km total) was 292 map turtles (or 11.7 turtles/rkm), with an estimated 32%–51% daily detection rate (Tables 4 and 5). The best supported model (lowest AICc) for the Apalachicola included the abundance covariate for sinuosity and the detectability covariate for day; the best supported model for the Ochlockonee included abundance covariates for river section (categorical) and sinuosity and the detectability covariate for day. We obtained abundance and detectability estimates based on these top N-mixture models (Tables 4–6).

Table 4. Modeled abundance estimates by survey section using multiple-pass basking surveys and N-mixture modeling for Graptemys barbouri in the Apalachicola and Ochlockonee rivers, Florida, 2015. Data are presented as mean ± SD and range (in parentheses).

View Fullscreen

Table 5. Modeled detection estimates per day using multiple-pass basking surveys and N-mixture modeling for Graptemys barbouri in the Apalachicola and Ochlockonee rivers, Florida, 2015. Data are presented as mean ± SD and range (in parentheses).

View Fullscreen

Table 6. Daily observed and estimated abundance counts for Graptemys barbouri during multiple-pass basking surveys on the Ochlockonee and Apalachicola rivers, Florida, 2015. No data are presented for Ochlockonee River sections 3–5 on 14 May because rain hampered the survey at those sections. Observed no./rkm was calculated using the high count from the 5 dates. Modeled estimates were calculated from our N-mixture models.

View Fullscreen

Adjusting the single-pass survey counts based on the range of detectabilities we found in our multiple-pass surveys (21%–51%) gives approximate abundance estimates of 7410–17,995 individuals (43–104 turtles/rkm) for our surveys on the Apalachicola River, 2441–5929 (21–50 turtles/rkm) for the Choctawhatchee, 1620–3933 (14–34 turtles/rkm) for the Chipola, 131–319 (1–3 turtles/rkm) for the Ochlockonee, and an overall approximate abundance estimate across all surveys of 11,601–28,176 (23–56 turtles/rkm). We emphasize that these estimates assume that detection for all single-pass surveys was within the range of detection values calculated during multiple-pass surveys of shorter river sections.

Both the single-pass distributional surveys and the multiple-pass abundance surveys resulted in high counts and estimates of G. barbouri. The 3 densest sections for G. barbouri during our distributional surveys were AR-3 on the Apalachicola (926 G. barbouri observed in 18 km; 51.5 turtles/rkm), AR-1 on the Apalachicola (457 G. barbouri observed in 10 km; 45.6 turtles/rkm), and Cp-9 on the Chipola (374 G. barbouri observed in 9 km; 41.1 turtles/rkm). For the multiple-pass surveys, a single 2-km section (section 3) on the Apalachicola produced an observed single-day count of 159 map turtles (79.5 turtles/rkm) and an estimated abundance of 419 map turtles (209.5 turtles/rkm). Another 2-km section of the same river (section 5) produced an observed single-day count of 170 map turtles (85.0 turtles/rkm) and an estimated abundance of 316 map turtles (158.0 turtles/ rkm).

DISCUSSION

We surveyed the entire known range of Barbour's map turtles in Florida and observed 5917 G. barbouri during single-pass surveys across 502 rkm of the Choctawhatchee, Chipola, Apalachicola, and Ochlockonee rivers. Adjusting these counts based on the range of detectabilities from our multiple-pass surveys (21%–51%) gives an overall approximate abundance estimate across all surveys of 11,601–28,176 (23–56 turtles/rkm). We emphasize that these are estimates and assume detection for all single-pass surveys was within the range of detection values calculated during multiple-pass surveys of shorter river sections. We recommend interpreting our observed 5917 G. barbouri as a minimum count, while acknowledging that the overall Florida population could be as high as 11,601–28,176 map turtles if our range of detection estimates reflect all rivers surveyed. For our multiple-pass surveys, we purposefully selected the 2 rivers that had the lowest and highest observed counts from our single-pass surveys in an effort to obtain a range of detection values that were representative of all rivers addressed in the present study. Future surveys could improve on these methods by conducting multiple-pass surveys for an increased number of sections and rivers and exploring additional covariates that may influence detection.

We also recommend several potential improvements to our survey methodologies. First, we noticed multiple disturbances that may have impacted our counts of basking turtles, specifically boat traffic (both boater behavior and wake overwash) that often caused basking turtles to drop into the water. We tried to minimize these effects by conducting surveys on weekdays when there was less traffic, but future work could improve on our efforts by quantifying these events and their impacts on the basking behavior of river turtles. Second, we attempted to make both our kayak and motorboat efforts similar for all of our surveys by always using 2 teams (1 observer per shoreline) and maintaining a steady speed and course. However, we did not account for potential differences in counts among watercraft types and recommend that future studies consider these potential differences. Third, only visible turtles during the time of survey can be counted using basking surveys and it is unknown what percentage of the population is available at a given time. Additional study to determine basking rates and availability would provide a better understanding of observed counts and help improve future abundance and detectability estimates.

The numbers of Barbour's map turtles observed in our study represent the highest densities reported for the species (as high as 51.5 turtles/rkm from observed counts and 209.5 turtles/rkm from estimated abundances) and are among the highest documented for megacephalic Graptemys (Lindeman 2013; Ilgen et al. 2014; Lindeman et al. 2020). Moler (1986) averaged 2.6 G. barbouri/rkm along a 59-km section of the upper Chipola River, with a reported high of 13.7 turtles/rkm for a section from Spring Creek to County Road 280 (Magnolia Road). Ruhl (1991) observed only 0.8 G. barbouri/rkm along the lower Apalachicola River but experienced high water levels that likely impeded detection. Stewart's surveys on the Apalachicola River for the US Fish and Wildlife Service (1992) tallied 5–8 G. barbouri/rkm along 4 separate stretches of the upper and middle sections. Enge and Wallace (2008) reported 2.0 G. barbouri/rkm along a 145.2-km stretch of the Choctawhatchee River, with the highest densities, 5–7 turtles/rkm, seen in the river's upper reaches north of Caryville. The same study documented only 4 G. barbouri along 162-km stretch of the Ochlockonee River.

Surveys within the Apalachicola River Wildlife and Environmental Area averaged 25.6, 6.1, and 6.0 map turtles/rkm along the Chipola, Apalachicola, and Brothers rivers, respectively; the highest single survey was from October 2012 on the Chipola, with 1010 turtles observed along the 27.5-km stretch for an average of 36.7 map turtles/rkm (Ricketts 2014). Although they used a different methodology from basking surveys, Chaney and Smith (1950) hand-captured 397 G. barbouri in the Chipola River over a 3-night period that Lindeman (2013) calculated as 62 turtles/rkm for the stretch they worked.

In Georgia, Hepler et al. (2015) used basking surveys to target G. barbouri along the Chattahoochee and Flint rivers and major tributaries. The highest density reported was 9.7 map turtles/rkm along the Flint River where it was the most common turtle observed, with 3323 individuals. However, on the Chattahoochee River, only 731 G. barbouri were observed (2.4 map turtles/rkm), second behind Pseudemys concinna with 983 observations.

Because we used the same survey sections on the Choctawhatchee and Ochlockonee rivers as did Enge and Wallace (2008), we can compare counts to make inferences about changes. In the present study, numbers of G. barbouri observed were 4.2 times greater on the Choctawhatchee and 16.5 times greater on the Ochlockonee than in the earlier study. It cannot be determined whether the larger numbers of turtles observed during the present study resulted from differences in methodologies or were an actual increase in population, although such a large increase in population may be unlikely given the relatively short time (14 yrs) between the studies. While the 2-observer method used in the present study likely contributed to higher counts than previously reported, colonization and proliferation of G. barbouri, especially in the Ochlockonee River, does appear to be a recent phenomenon supported by an absence of reports pre-1996. In general, spatial trends in relative abundance were similar to those observed by Enge and Wallace (2008); for the Choctawhatchee, we observed the highest densities of map turtles in the upper stretches, with decreasing numbers as we moved farther downstream. The same trends as reported by Enge and Wallace (2008) were also seen on the Ochlockonee River, with most map turtles detected directly upstream and downstream of Lake Talquin.

While our study greatly expanded the known range and observed number of G. barbouri in both the Choctawhatchee and Ochlocknee rivers, the factors that influence population trends in this species remain unknown, with previous accounts suggesting that a lack of food may be limiting distributions (Enge and Wallace 2008). Corbicula spp. (introduced clams from Asia) are a potential food source for other megacephalic map turtles (Selman and Lindeman 2015; but see Ewert et al. 2006) and were first detected in the Ochlockonee River during the 1960s (J. Williams, pers. comm.). Anecdotally (we did not record shell abundances), Corbicula now appears to be abundant in the Ochlockonee River, and we tended to observe large aggregations of map turtles adjacent to large numbers of Corbicula shells on the river bank. It is possible that these introduced clams may be providing food and shaping the distribution and abundances of Barbour's map turtles in Florida, although this hypothesis remains to be tested.

Our distributional surveys covered all rivers in the reported Florida range of G. barbouri, with the exception of the single observation from the upper Wacissa River (Jackson 2003). As the present study was under way, a second adult female G. barbouri was observed and photographed on the lower part of the Wacissa River (R. Means, pers. comm.); thus, future attention should be expanded to investigate whether a population is present in that river.

This present effort is the most comprehensive distributional survey for G. barbouri in Florida to date and demonstrates that the species is one of the more commonly observed turtle species in the Choctawhatchee, Chipola, Apalachicola, and Ochlockonee rivers. Based on our recent surveys, and in comparison to those done in 1999–2000 (Enge and Wallace 2008), Barbour's map turtle in the Choctawhatchee and Ochlockonee rivers appear to be expanding its range and may be increasing in abundance. Our estimated densities (turtles/rkm) from the multiple-pass surveys, in addition to the high observed counts during our single-pass surveys, suggest that Barbour's map turtle is currently secure in Florida. However, because G. barbouri is endemic to only 3 watersheds, continued monitoring of this species is important, as underlying resources and threats can change rapidly in a river system. In addition, state regulations prohibiting take and maintaining protections for riparian areas and water quality should continue. We recommend that future attempts to evaluate changes in G. barbouri population size incorporate standardized, repeatable methodologies that account for and improve on detectability, such as multiple-pass surveys and N-mixture modeling (described here) or capture–mark–resight surveys using basking traps (described by Selman and Lindeman 2015).