The US Endangered Species Act (ESA) of 1973 provides a framework for formal recognition of imperiled species and establishment of management guidelines to protect and recover populations. Terrestrial and freshwater species can become candidates for listing as endangered (E; in danger of extinction if not protected) or threatened (T; in danger of becoming endangered if not protected) by 1) direct action of the US Fish and Wildlife Service (USFWS) in the US Department of the Interior or 2) following petitioning by 1 or more individuals and organizations, a draft proposal to list the species followed by a period for public comment, and then a final ruling. While listing as E implies a more dire situation faced by the species, in most cases the level of federal protection and requirements for recovery planning under the ESA are equivalent for E and T species.

The emydid genus Graptemys consists of 14 recognized species of map turtles and sawbacks in eastern and central North America. The genus is distinguished by river drainage endemism (Lamb et al. 1994), the most extreme sexual size dimorphism known in the Tetrapoda (Gibbons and Lovich 1990; Stephens and Wiens 2009), and dietary diversity that is particularly strong among females, which vary greatly interspecifically in their trophic morphology (Lindeman 2000; Lindeman and Sharkey 2001). Five species make up a clade that is diagnosed by exceptional megacephaly in adult females and inhabit river drainages of the central Gulf Coast (Lamb et al. 1994; Lindeman 2013; Stephens and Wiens 2003; Praschag et al. 2017; Thomson et al. 2017). Large adult females of these 5 species subsist almost entirely on mollusks, while males eat primarily mollusks and aquatic insects (Sanderson 1974; Shealy 1976; Lindeman 2000, 2013, 2016; Selman and Lindeman 2015; J. Vučenović and P.V.L., unpubl. data, 2015–2018).

The species known today as the Pearl map turtle (Graptemys pearlensis) and the Pascagoula map turtle (Graptemys gibbonsi) spent almost a century and 4 decades, respectively, considered to be conspecific with other map turtles. The first 6 specimens of G. pearlensis were collected in the West Pearl River in 1891 and referenced in the description of Graptemys pulchra along with 2 additional syntype specimens collected in 1876 from the Mobile Bay river drainages (Baur 1893). A second species with megacephalic females, Graptemys barbouri, was described almost half a century later from the Apalachicola River drainage (Carr and Marchand 1942). Nothing new was published on G. pulchra sensu lato until Cagle (1952) and his students sampled in 2 river drainages, the Pearl and Escambia, and collected specimens that he placed under the name G. pulchra. Cagle's students collected the first 12 specimens of the species now known as G. gibbonsi in 1952 (Lindeman 2013), and these were also considered to be part of G. pulchra for the next 4 decades (e.g., Conant 1975; Mount 1975). Lovich and McCoy (1992) examined specimens of G. pulchra from all 4 river drainages. They designated 1 of the specimens from 1876 as lectotype and restricted the name G. pulchra to specimens from the Mobile Bay drainages, describing the Escambia Bay drainage specimens as Graptemys ernsti and the Pearl and Pascagoula river drainage specimens as G. gibbonsi; they also recognized some differences between specimens from the latter 2 rivers. Ennen et al. (2010) described the Pearl River drainage specimens as G. pearlensis, restricting G. gibbonsi to the Pascagoula River drainage specimens. Thomson et al. (2017) reported support for the monophyly of each of the 4 allopatrically distributed species formerly united as G. pulchra sensu lato.

As a consequence of the taxonomic situation, the conservation concern for the 4 species once lumped as G. pulchra has lagged behind that devoted to several congeners. The USFWS began review of 6 species of Graptemys in 1977 and added a seventh in 1982, but G. pulchra sensu lato was not among the species chosen for review (Lindeman 2013). Two of the species that were reviewed, the ringed sawback (Graptemys oculifera) of the Pearl River drainage and the yellow-blotched sawback (Graptemys flavimaculata) of the Pascagoula River drainage, were listed as T under the ESA 6 and 1 yrs prior, respectively, to the initial taxonomic splitting of G. pulchra in 1992 (USFWS 1986, 1991). In a subsequent series of basking surveys conducted in 1994 and 1995 throughout the geographic ranges of G. oculifera and G. flavimaculata, their predominance over G. gibbonsi sensu lato—which was outnumbered overall by the 2 sawbacks by > 5:1—generated the initial concern for the possible need for listing of G. gibbonsi sensu lato (Lindeman 1998, 1999). Historical data suggested more equitable ratios of sawbacks to megacephalic species in past decades, although a tendency for more extensive upstream and tributary ranges in the megacephalic species was also noted (Lindeman 1999). Similarly, subsequent publications have also reported imbalances in the ratio of sawbacks to megacephalic species for both river drainages (Selman and Qualls 2009; Selman and Jones 2017) and suggested the possible need for the megacephalic species' federal listing. The need to determine the importance of tributary populations to overall population viability has been noted in all of these studies, and, indeed, numerous new upstream and tributary records documenting previously unknown populations were reported for various species of Graptemys between 1995 and 2012, including G. pearlensis and G. gibbonsi (Lindeman 2013, 2014a).

In 2010, G. gibbonsi sensu lato was included in a petition, submitted by the Center for Biological Diversity and associated nonprofit organizations, to list 404 species associated with freshwater habitats in the southeastern United States (Alexander 2018). It was subsequently announced to be among 374 of the petitioned species that would be reviewed for possible listing under the ESA (USFWS 2011).

We report on our most recent (2015–2018) basking surveys and trapping aimed at determining the status of the megacephalic species G. pearlensis and G. gibbonsi throughout their respective ranges. We also combine these most recent data with data from earlier surveys in the 20th century (Selman and Qualls 2009; Selman and Jones 2017; P.V.L., unpubl. data, 2008–2013) to address the following objectives: 1) comparing the absolute and relative abundance of the 2 candidate species; 2) determining the status of the 2 species relative to their respective sympatric, federally listed sawback congeners, including how ratios of the sympatric pairs of congeners vary throughout the 2 river drainages; 3) determining as completely as possible the ranges of occurrence for the 2 candidate species with particular emphasis on critical examination of upstream range limits and occurrences in smaller tributaries of the river drainages; and 4) determining what percent of the estimated global population of each candidate species occurs in various main stem and tributary reaches. We then make recommendations regarding their status as candidates for federal listing.

METHODS

Description of Study Areas. — The Pearl River originates at the confluence of Tallahaga and Nanih Waiya creeks in northeastern Neshoba County, Mississippi. As it flows ∼ 675 river km (rkm; all drainage distances herein were determined by detailed tracings using the path function in Google Earth) toward the Gulf of Mexico, southwesterly at first and then more southerly, 8 major tributaries join it, and it splits into 2 main channels near its delta, the larger West Pearl and the smaller East Pearl, with additional connecting channels in the delta. Primary drainage modifications date to the 1950s and 1960s and are in the regions of Mississippi's largest city, Jackson, with 1) impoundment of the Pearl River and lower Pelahatchie Creek in Madison and Rankin counties forming the Ross Barnett Reservoir (13,500 ha) and 2) channel modification in the Pearl for flood control downstream in Hinds and Rankin counties (Selman 2020) and, in the lower drainage, with 3) construction of a navigation canal in the early 1950s and diversion of flow into the West Pearl. Prior to canal construction, the West Pearl had been a smaller channel than the East Pearl (Piller et al. 2004; Tipton et al. 2004). Upper portions of the Yockanookany and Tuscolameta Creek have also been channelized, with straight artificial channels installed alongside sinuous natural channels (Selman and Jones 2017).

The Pascagoula River is a shorter river (∼ 131 km) that originates in northern George County, Mississippi, at the confluence of 2 more extensive tributary systems: the Leaf River drainage to the west and the Chickasawhay River drainage to the east. The Leaf River originates in southern Scott County and flows ∼ 306 km in a southerly/ southeasterly route with several tributaries. The Chickasawhay River originates at the confluence of Okatibbee Creek and the Chunky River in northern Clarke County and flows ∼ 259 km in a southerly/southwesterly route, joined by a smaller number of tributaries. The Pascagoula drainage has the distinction of being the largest river system within the ranges of the species of the genus Graptemys that is relatively unaffected by flow regulation via large main stem dams or other forms of water regulation (Dynesius and Nilsson 1994). Its largest impoundment is Okatibbee Reservoir (1420 ha) on upper Okatibbee Creek in northern Lauderdale County, Mississippi.

Sandy forested riverbanks and sandy river substrates dominate throughout both drainages, with a common feature being flat inner-bend sandbars opposite steep outer-bend cut banks, particularly along larger stream reaches. Deadwood used by Graptemys species for basking and underwater grazing is generally abundant in both drainages (Lindeman 1999). In-stream rock formations are prominent features of some portions of the Strong River (Pearl River drainage) and the Chunky River (Pascagoula River drainage). Current speed varies greatly with discharge, with frequent rainy periods that can temporarily raise water levels by several meters.

Point Counts. — Basking and surface-active turtles were counted from bridges, boat ramps, and other access points on sunny, warm days between early May and early July 2015–2018. Turtles were generally sighted and identified with 18× Canon image-stabilizing binoculars, but a spotting scope that magnified up to 60× was also used in some cases. Because point counts are often low, many of the sites were visited multiple times to increase sample size and increase the chances of detecting resident species. Voucher photographs were taken at several localities, in particular to document upstream extensions of the known ranges of Graptemys species.

Boat Surveys. — Basking and surface-active turtles were counted from either a jon boat with an outboard motor or a canoe with a trolling motor, depending on stream width and depth, between 2015 and 2018. For these surveys, numbers of turtles recorded were divided by length of the survey, traced in Google Earth along the center of the stream, to generate basking densities (turtles/ rkm). Turtles that could not be identified to species, either because they jumped into the water at the approach of the boat or were partially obscured by deadwood or other turtles, were also noted. Boat surveys generated much larger sample sizes than point counts; thus, most reaches sampled by boat were not revisited, with the exception of mark-resight efforts made on a stretch of the Pearl River near Jackson (Selman 2020) and at 6 sites on the Pascagoula drainage. For reaches where surveys were repeated, an average basking density was calculated, with averaging weighted by rkm if survey length was variable.

We corrected basking densities for unidentified turtles by assuming that they occurred at the same frequency as identified turtles in the survey, that is, that there was no interspecific difference in the probability of not being able to identify a turtle due to it either reentering the water or being partially hidden from view. Basking densities for each Graptemys species and for pooled non-Graptemys species that are reported herein are corrected for unidentified turtles.

We designated the Pearl and West Pearl rivers and the Leaf, Chickasawhay, and Pascagoula rivers as main stem habitats of their respective drainages based on relative channel width and depth. Tributaries were divided into large and small categories, with the large tributaries being the Yockanookany, Strong, Bogue Chitto, and East Pearl rivers for the Pearl River drainage and Tallahala, Buckatunna, Black, Red, and Big Black creeks for the Pascagoula River drainage.

Trapping. — At select sites during 2015–2018, we captured turtles using basking traps made from crawfish wire (Selman et al. 2012) or modified hoop nets (Lindeman 2014b), fyke nets (Vogt 1980), and opportunistic capture by hand or dip net. We recorded sex of turtles based on relative tail size and measured midline plastron length (PL) of each Graptemys specimen captured to the nearest 1 mm, determined body mass with spring scales, and notched marginal scutes in combinations of 2 or 3 that were unique within each species, following Cagle (1939). We categorized G. pearlensis and G. gibbonsi as hatchlings, older unsexed juveniles, adult males, juvenile females, or adult females based on whether individuals had growth annuli and enlarged tails as well as minimum recorded sizes for mature males (60 m PL) and mature females (180 mm PL). Turtles were released at their sites of capture.

Survey and Trapping Data. — We examined data from point counts in terms of numbers seen per point count, while we examined data from boat surveys in terms of densities (turtles/rkm). We computed catch per unit effort (CPUE) for each species as the number of turtles captured in each trap type per trap day (TTD; 1 day of checking basking traps) or per overnight set period for fyke nets (TTN); no CPUE was calculated for hand or dip net captures due to their irregular and opportunistic use. We also examined 2 indices of relative abundance for point counts, boat surveys, and trapping: 1) ratio of Graptemys seen or captured (expressed as sawbacks per sympatric megacephalic species) and 2) percent of all turtles seen or captured for each Graptemys species and combined non-Graptemys species.

We combined data from point counts, basking density surveys, and trapping from 2015 to 2018 with earlier data sets collected between 2006 and 2014 (Selman and Qualls 2009, all 3 types, Pascagoula River drainage; Selman and Jones 2017, basking density surveys and trapping, Pearl River drainage; P.V.L., unpubl. data, 2008–2013, point counts, both river drainages). Two of the earlier surveys on the lower Pascagoula's Escatawpa tributary and 1 on its Bluff Creek tributary were conducted upstream of the known range of G. gibbonsi and are not included here. In addition, during 6 Pascagoula drainage surveys within the range of G. gibbonsi in 2006–2007, unidentified turtles were not recorded (and in 3 of these, only Graptemys were counted); these surveys were withheld from calculations and summaries of corrected basking densities but were used for calculation and summation of relative abundance. At 5 sites on the Pearl River in 2008–2014, data on sex or size were recorded for only G. oculifera (Jones 2017; Selman and Jones 2017), so those trapping data could not be used in summarizing of population structure of G. pearlensis.

Statistical Analyses. — We analyzed mean numbers of each Graptemys species recorded per point count and basking density in boat surveys in 2 × 3 analyses of variance (ANOVAs) with 2 time periods (2006–2014 vs. 2015–2018) and the 3 categories of river reaches (main stems and large and small tributaries). Due to large numbers of low values for point counts and basking density surveys, some of which were zeroes, each data point x was first transformed as ln(x + 1). Tukey's test was used post hoc to identify which reach categories differed in turtles/point count or basking density. We compared sympatric map turtle and sawback distributions across the 3 reach categories in totals recorded across point counts and across basking density surveys. Finally, drainage-wide population structures of G. pearlensis and G. gibbonsi were compared in a χ² test for independence with hatchling, older unsexed juvenile, adult male, juvenile female, and adult female categories.

Drainage-Wide Population Estimates. — To generate global population estimates for G. pearlensis and G. gibbonsi, we first traced the full extent of all occupied river reaches in the Pearl and Pascagoula river drainages using the path function in Google Earth. Because data at individual sites were not adequate for rigorous population estimation, we used a coarse-filter approach to estimate population density along river reaches. We first used combined basking survey data collected over series of sites on river main stems and tributaries to divide extended reaches from each drainage into 4 categories (high density, medium density, low density, and very low density). Based primarily on basking density survey data and secondarily on point count averages, we assigned a characteristic basking density to each of the 4 categories and assumed it represented the average density over the entire set of reaches so categorized. We then multiplied reach length in rkm by characteristic basking survey density and corrected for unobserved turtles by multiplying by 6.67, making an assumption based on previous mark-resight studies of G. gibbonsi, G. flavimaculata, and G. oculifera in which ∼ 85% of paint-marked turtles were not detected in basking counts (Selman and Qualls 2009, 2011; Brown et al. 2016; Selman and Jones 2017; P.V.L., unpubl. data, 2016; see also Selman, in press). We then summed these estimates of total turtles for sets of reaches defining various portions of the drainage and for the total drainage. We used these population estimates to estimate percentage of the global population found in tributaries vs. main stem reaches for each species.

RESULTS

Point Counts. — From 2015 to 2018, we conducted 233 total point counts at 125 different sites on the Pearl River drainage. Of these, 174 counts at 82 sites were determined to be within the range of G. pearlensis, including 23 sites on the main stem Pearl and West Pearl rivers and 59 sites on 8 tributary stream networks (Lobutcha Creek, Tuscolameta Creek, Yockanookany River, Pelahatchie Creek, Strong River, Pushepatapa Creek, Hobolochitto Creek, and Bogue Chitto River) and 2 connected channels (East Pearl River and Pearl River Navigation Canal; Table 1). On the Pascagoula River drainage, we conducted 344 point counts at 210 sites. Of these, 248 counts at 136 sites were determined to be within the range of G. gibbonsi, including 27 sites on the main stem Pascagoula, Leaf, and Chickasawhay rivers and 109 sites on 15 tributary stream networks (West Tallahala Creek, Oakohay Creek, Bowie River, Tallahala Creek, Bogue Homa, Thompsons Creek, Gaines Creek, Atkinson Creek, Chunky River, Okatibbee Creek, Souinlovey Creek, Shubuta Creek, Buckatunna Creek, Black Creek, and Red Creek). To these point counts, we added 125 earlier counts for the Pearl River drainage (2008–2013) and 151 earlier counts for the Pascagoula River drainage (2006–2013), with most of the earlier counts taken at sites that were included in the more recent data (Table 1).

Table 1. Point count results for sites within the known ranges of megacephalic map turtle species, broken down according to most recent data (2015–2018) and all data since 2006 (adding Selman and Qualls 2009 and P.V.L., unpubl. data, 2008–2013) as well as by main stem reaches, larger tributaries, and smaller tributaries, with total and average numbers and proportions of each Graptemys species and combined other species of turtles as well as the ratio of sawback to megacephalic Graptemys species. The Pearl River drainage is inhabited by G. pearlensis and G. oculifera, while the Pascagoula River drainage is inhabited by G. gibbonsi and G. flavimaculata.

View Fullscreen

Generally, abundance of each Graptemys species declined with the size of the river reach surveyed. All 4 species averaged higher numbers seen per point count on main stem reaches than on tributary reaches in the most recent surveys, and all species but G. flavimaculata were more abundant in counts made on large tributaries than in counts made on small tributaries. In the larger data set, the main stem–tributary disparity was considerably greater for the 2 sawbacks (∼ 10 times greater on main stem reaches for G. oculifera and 60 times greater for G. flavimaculata) than for the 2 megacephalic map turtles (approximately even for G. pearlensis and 4 times greater for G. gibbonsi). All other species pooled together likewise averaged higher at main stem sites than at tributary sites for the Pascagoula River drainage but averaged higher at tributary sites than main stem sites for the Pearl River drainage.

In ANOVA comparisons, time period (2006–2013 vs. 2015–2018) was not a significant term for numbers of either Graptemys species seen per point count in the Pearl River drainage (G. pearlensis: F_1,295 = 0.21, p = 0.65; G. oculifera: F_1,295 = 1.82, p = 0.18), but both species differed in number seen among reach categories (G. pearlensis: F_2,295 = 3.44, p = 0.033; G. oculifera: F_2,295 = 66.3, p < 0.0001). In Tukey's post hoc tests, G. pearlensis counts were significantly higher on large tributaries than on small tributaries, with main stem reaches intermediate and not different from the other 2 categories. Densities of G. oculifera differed in all comparisons, with greatest counts on main stem reaches and lowest counts on small tributaries.

For the Pascagoula drainage, both Graptemys species were recorded in higher numbers in earlier surveys than in more recent surveys (G. gibbonsi: F_1,395 = 5.24, p = 0.023; G. flavimaculata: F_1,395 = 5.44, p = 0.020), and both had significant differences among reach categories (G. gibbonsi: F_2,395 = 21.77, p < 0.0001; G. flavimaculata: F_2,395 = 68.97, p < 0.0001). In Tukey's post hoc tests, G. gibbonsi densities were significantly greater on main stem reaches than on either category of tributary, while large tributaries had nonsignificantly greater densities than small tributaries. For G. flavimaculata, densities were significantly greater on main stem reaches than on either category of tributary, while small tributaries had nonsignificantly greater densities than large tributaries.

Density Surveys by Boat. — Between 2015 and 2018, we conducted 40 jon boat or canoe surveys on river reaches totaling 187.8 rkm on the Pearl River drainage and 31 surveys on river reaches totaling 135.3 rkm on the Pascagoula River drainage (Table 2). We could not identify 9% of all turtles observed during Pearl surveys and 7% of all turtles observed during Pascagoula surveys, with unidentified turtles occurring at ∼ 3 times greater frequencies on tributary reaches than on main stem reaches.

Table 2. Weighted averages of basking density (corrected for unidentified turtles; see text) and relative abundance of each Graptemys species and combined other turtle species from jon boat and canoe surveys. Data are broken down according to most recent surveys (2015–2018) and all surveys since 2006 (including data from Selman and Qualls 2009, Selman and Lindeman 2015, and Selman and Jones 2017) as well as by main stem reaches, larger tributaries, and smaller tributaries. The Pearl River drainage is inhabited by G. pearlensis and G. oculifera, while the Pascagoula River drainage is inhabited by G. gibbonsi and G. flavimaculata. No surveys were conducted on tributaries of the Pearl River drainage prior to 2015.

View Fullscreen

For G. pearlensis, basking density in 2015–2018 was 2.2 times higher on main stem reaches as on tributary reaches and 2.1 times higher on large tributaries than on small tributaries. For G. gibbonsi, basking density was 1.6 times higher on main stem reaches than on tributary reaches and 1.2 times higher on large tributaries than on small tributaries. In both river drainages, the main stem–large tributary–small tributary disparity was greater for the sympatric sawback species. For G. oculifera, we observed 9.0 times greater density on main stem Pearl and West Pearl reaches than on tributaries and 1.9 times higher density on small vs. large tributaries. For G. flavimaculata, we observed 12.0 times greater density of on main stem Pascagoula, Leaf, and Chickasawhay reaches than on tributaries and 2.8 times higher density on large vs. small tributaries.

Additional surveys conducted earlier include 1 2.7-rkm reach in the Pearl River surveyed in 2006 (Selman and Jones 2017) and 20 in-range reaches totaling 218.5 rkm in the Pascagoula River drainage surveyed in 2006–2008 (Selman and Qualls 2009). Combining data from the earlier surveys for the Pascagoula River drainage, density of G. gibbonsi rose by 26% on main stem reaches and declined by 5% on tributary reaches; that is, inclusion of the earlier surveys increased the disparity in density of G. gibbonsi between main stem and tributary reaches from a 1.6-fold difference to a 2.2-fold difference.

In ANOVA comparisons, all 4 Graptemys species exhibited significant differences in corrected basking density among main stem reaches, large tributaries, and small tributaries (G. pearlensis: F_2,38 = 5.84, p = 0.006; G. oculifera: F_2,38 = 10.45, p = 0.0002; G. gibbonsi: F_2,38 = 3.71, p = 0.038; G. flavimaculata: F_2,38 = 8.74, p = 0.0008). Graptemys pearlensis had highest densities on main stem reaches and lowest densities on small tributaries, with only the 2 extreme categories different in a Tukey test. Tukey's test identified no pairs of categories different for G. gibbonsi, although the trend was similar to that for G. pearlensis, with larger reaches averaging highest basking densities. For both sawbacks, main stem reaches had significantly higher densities than both large and small tributaries, which did not differ significantly from one another. In the Pascagoula River drainage, earlier surveys (2006–2008) had significantly higher densities than later surveys (2015–2018) for both Graptemys species (G. gibbonsi: F_1,38 = 3.71, p = 0.038; G. flavimaculata: F_1,38 = 8.74, p = 0.0008).

Trapping Results. — Total trapping effort over all years was 1868 trap days for basking traps and 50 trap nights for fyke nets in the Pearl River drainage vs. 2920 trap days for basking traps and 79 trap nights for fyke nets in the Pascagoula River drainage (Supplemental Material 1; https://doi.org/10.2744/CCB-1414.1.s1). We captured 2808 total turtles of 14 native species: 183 G. pearlensis, 390 G. gibbonsi, 841 G. oculifera, and 1169 G. flavimaculata, plus 225 turtles of 10 other species (2 other emydids, 4 kinosternids, 2 trionychids, and 2 chelydrids). We also captured 1 specimen each of 2 nonnative species of Graptemys (Graptemys pseudgeographica and Graptemys sabinensis) in the Pearl River near Jackson. Totals include 224 recaptures recorded at sites where we marked all turtles, accounting for 10% of total captures at those sites.

More Graptemys were caught in basking traps than in fyke nets, both in absolute terms (2154 vs. 25) and in terms of CPUE (0.45 TTD vs. 0.19 TTN; numbers for all 4 native Graptemys species from all years in both cases; breakdown by species in Table 3). Fyke nets were particularly ineffective in sampling Graptemys in the Pearl River drainage (0.04 TTN vs. 0.30 in the Pascagoula River drainage). In contrast, basking traps caught Graptemys at a slightly higher CPUE in the Pearl River drainage (0.54 TTD vs. 0.41 in the Pascagoula River drainage).

Table 3. Summary of catch per unit effort (CPUE) and relative abundance for Graptemys species in the Pearl and Pascagoula river drainages based on numbers of turtles captured in basking traps and fyke nets (see Supplemental 1). The unit effort is a trap day, defined as a basking trap checked multiple times throughout a day, or a trap night, defined as a fyke net left in place overnight, with trap days and trap nights weighted equally for combined CPUE. Also given is the percent of each species in the total catch of turtles of all species and the ratio of each sawback species to its sympatric megacephalic map turtle species.

View Fullscreen

Relative Abundance in Basking Counts and Trapping. — In point counts and basking density surveys, the sawback:megacephalic ratio was consistently more favorable to G. gibbonsi (i.e., lower, with fewer sawbacks per megacephalic) than to G. pearlensis, except in small tributaries in the data for point counts from all years (Tables 1 and 2). There was a preponderance of the megacephalic species G. gibbonsi in point counts and basking density surveys on tributary reaches of the Pascagoula system, while there was a preponderance of the sawback species G. oculifera in tributary reaches of the Pearl system in basking density surveys but not in point counts. Relative abundance among all species of turtles observed in basking counts were largely similar by species and category of river reach between point counts and basking density surveys (Tables 1 and 2; Fig. 1), with the biggest discrepancy by type of survey being for G. oculifera on small tributaries, where they were much more relatively abundant in boat surveys than in point counts (55% vs. 4%). For both megacephalic Graptemys species, relative abundance among all turtle species (all years) varied little among main stem, large tributary reaches, and small tributary reaches (range, 11%–25% for G. pearlensis, range 21–38% for G. gibbonsi). In contrast, sawback relative abundance was generally much higher on main stem reaches (67% for point counts and 80% for boat surveys for G. oculifera, 46% and 61% for G. flavimaculata) than on tributary reaches (10% and 35% for G. oculifera, 5% and 12% for G. flavimaculata), as pooled other species were high in relative abundance in the tributaries of both river systems (range, 45%–67%) compared with main stem reaches (range, 8–33%). For 2015–2018 data from both drainages, the differences in distribution of sightings between megacephalic map turtle species and sympatric sawbacks among the 3 categories of river reach were highly significant, in both point counts (Pearl: χ²₂ = 98.2, p < 0.0001; Pascagoula: χ²₂ = 184.1, p < 0.0001) and boat surveys (Pearl: χ²₂ = 196.6, p < 0.0001; Pascagoula: χ²₂ = 177.0, p < 0.0001).

View Figure

• Download as Powerpoint
• Download Figure

Similar trends in relative abundance occurred in trapping data (Table 3). Sawback captures outnumbered megacephalic map turtle captures on main stem reaches of both river systems, almost twice as strongly so in the Pearl River as in the Pascagoula, Leaf, and Chickasawhay rivers; that is, the main stem Graptemys ratios were more favorable toward G. gibbonsi than toward G. pearlensis. In tributary reaches, megacephalic map turtles dominated in both river systems by lesser margins. Other turtle species became more prevalent in trapping results on tributaries, as pairs of sympatric Graptemys species made up 95% of the total turtle catch in main stem reaches of each river system, compared to just 71% in Pearl tributaries and 62% in Pascagoula tributaries.

We captured the 2 megacephalic Graptemys at lower CPUE rates than the 2 sawback species overall, and the megacephalics also contrasted with the sawbacks in having slightly higher CPUE on tributary reaches than on main stem reaches compared to much higher CPUE on main stem reaches for the sawbacks (Table 3). Graptemys pearlensis had slightly lower CPUE than G. gibbonsi.

Population structure based on 451 total captures (Table 4) were similar for the 2 species of megacephalic map turtles (χ²₄ = 5.39, p = 0.25). Sex ratios of captured turtles larger than minimum male size were 55♂:31♀ (or 1.8♂:1♀) for G. pearlensis and 141♂:124♀ (1.1♂:1♀) for G. gibbonsi, with the former but not the latter departing significantly from unity (χ²₁ = 6.70, p = 0.0010 and χ²₁ = 1.09, p = 0.30, respectively; Table 4). Operational sex ratios (i.e., discounting immature females) were significantly skewed toward males for G. pearlensis (55♂:12♀, or 4.6♂:1♀; χ²₁ = 28, p < 0.0001) and for G. gibbonsi (141♂:48♀, or 2.9♂:1♀; χ²₁ = 46, p < 0.0001). Unsexed juveniles, including hatchlings in their first full growing seasons and 1-yr-olds, were commonly captured at many sites (Table 4), accounting for 18% of all G. pearlensis and 24% of all G. gibbonsi, the latter of which had especially high frequencies of hatchlings, mostly via hand or dip net captures, at 2 sites on the upper Chickasawhay and Chunky rivers (58% and 63% of all turtles captured at those sites, respectively).

Table 4. Population structure as indicated by trapping results for Graptemys pearlensis and Graptemys gibbonsi, 2005–2018. Unsexed juveniles were smaller than the minimum size of adult males and in at least their second season of growth, while juvenile females were larger than minimum male size but lacked the male secondary sexual characteristic of an enlarged tail and were < 180 mm in plastron length. Recaptures are not included.

View Fullscreen

Total Range, with New Tributary Records and Upstream Range Extensions. — The total range we found to be occupied by G. pearlensis was 1279.6 rkm, with 632.6 rkm (49%) in the main stem Pearl and West Pearl rivers and the remaining 647.0 rkm (51%) in various tributaries. Of these totals, ∼ 33.6 rkm of the middle Pearl River's original channel is submerged under the Ross Barnett Reservoir, which has 61.5 km of shoreline. The total range we found to be occupied by G. gibbonsi was 1733.7 rkm, with main stem reaches of the Pascagoula, Leaf, and Chickasawhay rivers totaling 623.1 rkm (36%) and various smaller tributaries totaling 1110.6 rkm (64%). Of these totals, ∼ 10.1 rkm of upper Okatibbee Creek is submerged under Okatibbee Reservoir, which has 26.7 km of shoreline.

Between 2015 and 2018, G. pearlensis was detected for the first time in 5 tributaries, and other new records document upstream range extensions within 4 additional tributaries (Table 5). Assuming a continuous range within these stream reaches, the total newly discovered range for G. pearlensis sums to 188.3 rkm. Graptemys gibbonsi was detected for the first time in 3 tributaries, and other new records document range extensions within 4 additional tributaries (Table 5). Its newly discovered range sums to 133.3 rkm. New tributaries and upstream range extensions were also recorded for the sympatric sawback species (4 and 2, respectively, totaling 102.5 rkm for G. oculifera, and 1 and 3, respectively, totaling 55.5 rkm for G. flavimaculata). These records include a total of 8 new county records for Graptemys species (3 vouchering previous sight records and 5 establishing previously unknown county occurrences; Lindeman 2017a, 2017b, 2017c).

Table 5. Range extensions measured in river km (rkm) recorded for Graptemys species in 2 river drainages during most recent fieldwork (2015–2019) in newly documented tributaries and upstream of records in tributaries already known to be occupied. Known ranges as of 2014 are based on records published prior to the start of the present study (Lindeman 2013, 2014a, 2014b). Voucher photos deposited in the University of Florida digital collection are listed by catalog number where available.

View Fullscreen

We surveyed basking turtles in point counts on several additional tributaries that had no previous records of the 2 megacephalic map turtle species as well as upstream of the known ranges on many occupied streams (Supplemental Material 2; https://doi.org/10.2744/CCB-1414.1.s2). No G. pearlensis were detected in 16 tributaries (24 sites, 32 total counts), and we also conducted 42 surveys at 21 total sites upstream of their known range in 8 streams known to be occupied farther downstream. No G. gibbonsi were detected in 20 tributaries (38 sites, 52 total counts), and we also conducted 82 surveys at 49 total sites upstream of their known range in 18 streams known to be occupied farther downstream. Other turtle species were often recorded during these out-of-range point counts (n = 228 turtles, in 208 total counts).

Global Population Estimates. — Based primarily on basking density surveys made by boat and secondarily on results of point counts, we divided each river drainage into river reaches typified by high, moderate, low, and very low density of their resident megacephalic map turtle species (Fig. 2). In the Pearl River drainage, we assigned these categories typical basking densities of 10, 3, 1, and 0.7 G. pearlensis/rkm, respectively, and these categories comprised 3%, 68%, 16%, and 14%, respectively, of the total range of G. pearlensis. In the Pascagoula River drainage, we assigned these categories typical basking densities of 5, 4, 2.5, and 0.5 G. gibbonsi/rkm, respectively, comprising 33%, 6%, 38%, and 23%, respectively, of the total range of G. gibbonsi. On the assumption that we saw on average 15% of the population during basking surveys, we estimated global populations to be 21,841 G. pearlensis and 34,081 G. gibbonsi. Two main factors caused the 56% difference in global population estimates: 1) G. gibbonsi occurred at greater densities in main stem reaches, and 2) the Pascagoula River system has more occupied tributaries than the Pearl River system, particularly in terms of smaller tributaries.

View Figure

• Download as Powerpoint
• Download Figure

Table 6. Percent of occupied range and percent of total estimated population for Graptemys pearlensis in the Pearl River drainage and Graptemys gibbonsi in the Pascagoula River drainage. Tributaries are arranged from north to south according to where they join main stem rivers, with indented entries for tributaries of tributaries and impoundments (Fig. 2).

View Fullscreen

Comparisons of various drainage segments by the proportion of the total range length they constitute and the percent of the estimated global population that inhabits each segment are presented in Table 6. For G. pearlensis, 61% of the estimated global population occurs in main stem reaches, and 34% occurs in 4 large tributaries, with half of the latter total (3749 turtles) in the Bogue Chitto River, leaving 6% spread among 11 smaller tributaries, of which only Lobutcha Creek exceeded n = 300. For G. gibbonsi, 57% of the estimated global population occurs in main stem reaches, and 25% occurs in 5 large tributaries, primarily in Tallahala, Buckatunna, and Black creeks (each n > 2500), with 18% spread among 19 smaller tributaries, in particular Okatibbee and Souinlovey creeks, the Bogue Homa, and the Chunky River (each n > 750).

DISCUSSION

The general picture of basking turtle abundance for the Pearl and Pascagoula drainages is a gradient from larger channels at downstream localities that are dominated by their respective federally listed sawback species, with declining Graptemys abundance farther upstream and into tributaries, to the point where first sawbacks and later the megacephalic species disappear, but other species rise in relative abundance and continue to be seen farther upstream. While all 4 Graptemys species in these river systems decline in abundance in smaller river channels, the gradient is far steeper for the 2 sawbacks. The 2 focal species of megacephalic Graptemys show considerably less variation in their relative abundance than the sawbacks and combined other species, making up a more or less constant proportion of the basking turtle fauna throughout their respective ranges. Both absolute and relative abundance of G. gibbonsi in the Pascagoula River drainage are more robust than are those of G. pearlensis in the Pearl River drainage, although both species show evidence of long-term historical declines in abundance.

Main Stem vs. Tributary Abundance. — Generally, the 4 resident species of Graptemys were most abundant in basking surveys and in CPUE in larger main stem river reaches and least abundant in small tributaries, but the gradient of basking density or CPUE was considerably steeper for the 2 sawbacks. In main stem river reaches of both drainages, sawbacks achieved high abundance (G. oculifera basking densities: mean 12.4/rkm, maximum 88.1/rkm; G. flavimaculata basking densities: mean 25.1 turtles/rkm, maximum 99.6 turtles/rkm). Average and maximum basking densities of sympatric megacephalic species were considerably lower than the sawback levels (G. pearlensis: mean 3.8/rkm, maximum 16.1/rkm; G. gibbonsi: mean 4.3 turtles/rkm, maximum 17.2/rkm).

The above interspecific differences mirror previous reports of basking density in Graptemys. Basking surveys of narrow-headed Graptemys species commonly report basking densities averaging over 20 turtles/rkm and ranging to over 100/rkm (reviewed in Ilgen et al. 2014). Megacephalic species of Graptemys, in contrast, have been reported to occur at site-specific basking densities that average > 10/rkm in only 2 rivers and exceed 15/rkm for single sites only rarely (Table 7). Narrow-headed Graptemys feed on sponges and other invertebrates and algae that they graze from submerged deadwood (Shively and Jackson 1985; Lindeman 2016; Selman and Lindeman 2018; P.V.L., W.S., and R.L.J., unpubl.). These communities are sunlight dependent (Shively and Jackson 1985); thus, higher densities in larger river channels may reflect prey abundance. High densities of narrow-headed species are reported to occur in large, sluggish, low-gradient channels near the coast as well as at upstream confluences with reservoirs (Fehrenbach et al. 2016; Jones 2017). Megacephalic species rely primarily on benthic mussels (Sanderson 1974; Shealy 1976; Lindeman 2016; J. Vučenović and P.V.L., unpubl. data, 2015–2018), and no reports suggest that well-insolated or large, sluggish channels near the coast or upstream ends of reservoirs show any particular pattern regarding the abundance of these turtles.

Table 7. Basking densities reported for megacephalic species of Graptemys from rivers to compare to a similar table for narrow-headed Graptemys species in Ilgen et al. (2014). Indented drainages are tributaries of the rivers listed above them. No corrections are made for unidentified turtles observed. NR = not reported.

View Fullscreen

Tributary populations are not as dense as main stem populations, yet some tributaries contain large populations of G. pearlensis and G. gibbonsi, including many that have only recently been discovered. For example, presence of G. pearlensis in the Yockanookany was first established in 1994, and extensive upstream range extensions have now been reported in the Yockanookany, Strong, and Bogue Chitto rivers (Lindeman 1998, 2010, 2013, 2014a, 2017c); these 3 tributaries together are inhabited by nearly one-third of our estimated global population (Table 6), with nearly half occurring in stretches not known to be part of the species' range prior to the present century. Likewise, G. gibbonsi occurs in relatively large numbers in upper Tallahala and Tallahoma creeks, Souinlovey and Buckatunna creeks, and the Bogue Homa, which collectively house over one-fifth of our estimated global population; all but the lower sections of Tallahala Creek were unreported populations until the first decade of this century (Selman and Qualls 2009; Lindeman 2017b).

No major channel alterations or impoundments have occurred on the Pascagoula or the 2 major tributaries that form the Pascagoula. In contrast, 33.6 rkm of the middle Pearl River is inundated by the Ross Barnett Reservoir, and downstream channel alterations have diverted considerable flow from the East Pearl into the West Pearl and the Pearl River Navigation Canal (Piller et al. 2004; Tipton et al. 2004). We found the reservoir and all 3 downstream distributaries to be very-low-density or low-density reaches, suggesting that substantial loss of population in the lower reaches of the Pearl River drainage has occurred historically due to river engineering. Dickerson and Reine (1996) also reported much lower basking density of G. pearlensis in the West Pearl than in the upper Pearl (Table 7).

Relative Abundance. — The proportion of all Graptemys that were 1 or the other of the megacephalic species varied widely throughout both drainages, mainly due to strong variation in the basking density of their sympatric congeners, with low relative abundance of megacephalics on main stems that became higher on tributaries where sawback abundance declined precipitously, particularly on smaller tributaries, where the sawbacks were often absent. The most atypical part of this pattern was the relative abundance of G. oculifera on small tributaries in boat surveys (55% relative abundance compared to 23% on large tributaries), which may have been influenced by the fact that many of the small tributaries on the Pearl drainage were surveyed by boat at their downstream junctures with the Pearl and West Pearl main stems (Tuscolameta Creek, Pearl River Navigation Canal, Holmes Bayou, and Porter River) and may have supported populations that were continuous with those main stems. The proportion of all turtles that were megacephalic Graptemys also varied greatly due to congener trends and increased relative abundance of other species on small tributaries far upstream. Other species of turtles occurred well above the upstream range limits of the 2 megacephalic species but were far less important components of the basking turtle fauna on main stem reaches.

Population Structure of G. pearlensis and G. gibbonsi. — Both megacephalic Graptemys species showed evidence of considerable recruitment in trapping results, with unsexed juveniles accounting for 18% of G. pearlensis captures and 24% of G. gibbonsi captures and juvenile females being captured more frequently than adult females in both species. The overall sex ratio was somewhat male biased for G. pearlensis and almost exactly 1:1 for G. gibbonsi. Operational (i.e., adult) sex ratios were strongly male biased in both species, which is to be expected due to the exceptional sexual dimorphism in maturation age and thus size. Male Graptemys typically mature in 2–3 yrs (Lindeman 2013), while females mature much later; maturity may occur around 9 yrs of age in G. pearlensis and 11 yrs of age in G. gibbonsi (Vogt et al. 2019a, 2019b).

Without information on possible trapping biases, it is unclear what population structure indicates about population viability. For example, hand and dip net captures were almost exclusively juvenile turtles, particularly hatchlings, but hatchlings were captured in basking traps and fyke nets much less commonly. With the strong reliance on basking traps in the present study, any sexual differences or ontogenetic changes in basking proclivity would bias population structure. A comparison of trapping rates of different life stages with separate estimates of population size by life stage would be instructive. Estimating survival rates of adult males and adult females (e.g., Jones 2017) and a life table analysis would ultimately lead to a rigorous analysis of population viability.

Range Extensions and Their Significance to Conservation. — If range is measured in terms of the length of occupied stream reaches, our visual surveys increased the known ranges of the focal species by 15% for G. pearlensis (i.e., 188.3 of 1279.6 total rkm are range extensions) and by 8% G. gibbonsi (133.3 of 1733.7 total rkm). The positive nature of these findings is tempered, however, when the range extensions are viewed in light of our range-wide population estimates. Newly reported stream reaches account for just 9% of the global estimated population of G. pearlensis and 4% of the global estimated population of G. gibbonsi because most of the range extensions were in areas designated as low or very low density. In addition, without telemetry or wide-ranging mark-recapture sampling, it remains unknown how dependent turtles observed in the smallest tributaries may be on making seasonal forays into connecting larger tributaries and main stem river reaches for feeding, nesting, or refuge from drought and low water levels in small tributaries.

The USFWS recently decided not to list Barbour's map turtle (G. barbouri), another species with megacephalic females that had been part of the same listing petition as G. pearlensis and G. gibbonsi (Yarbrough 2017). Part of the rationale given for not listing G. barbouri was that the species had not disappeared from any portions of its known range, and in fact, during field surveys in the past decade and a half, the species was found in 2 new drainages and 3 tributary streams it had not been previously known to occupy in the Apalachicola River drainage (Hepler et al. 2015), with new findings extending the range of the species by 6% (Yarbrough 2017). As in our study, however, the newly reported tributaries and drainages had low populations of G. barbouri compared to main stem reaches of the Chattahoochee, Flint, and Apalachicola river system. We propose that the stated rationale likely overestimates the significance of the recent extensions of known range for G. barbouri and caution against making a similar conclusion in the case of G. pearlensis and G. gibbonsi.

Comparison of Abundance Between G. pearlensis and G. gibbonsi. — Graptemys gibbonsi was the more abundant of the 2 focal megacephalic Graptemys species in several metrics (Fig. 3): overall point counts (44% more G. gibbonsi than G. pearlensis for all counts, 333% more in counts on main stem reaches), overall basking density surveys (3% more G. gibbonsi for all counts, 13% more in main stem counts only), trapping results (22% more G. gibbonsi per trap day/night overall, 38% more on main stem reaches), total range length (35% more occupied rkm for G. gibbonsi), global population estimates (56% more G. gibbonsi than G. pearlensis), and more equitable ratios with its sympatric sawback species (1 G. gibbonsi per 1.3 G. flavimaculata vs. 1 G. pearlensis per 1.5 G. oculifera for point counts, 1 per 1.9 vs. 1 per 5.7 for surveys by boat, and 1 per 3.0 vs. 1 per 4.6 for trapping results). We believe it is clear that populations of G. gibbonsi are more robust overall than populations of G. pearlensis. A previous comparison of sawback:megacephalic species ratios in 1994–1995 (Lindeman 1998, 1999) did not consider the difference between the drainages because the 2 megacephalic species were at that time considered a single species under the name G. gibbonsi (Ennen et al. 2010). In replicated point counts conducted primarily on main stem sites (31 of 41 total sites, or 76%), the 2 sawbacks were 5.1 as abundant as the 2 megacephalics. However, in that study the sawback:megacephalic ratio was 9.6:1 for G. oculifera and G. pearlensis vs. 2.4:1 for G. flavimaculata and G. gibbonsi. Similarly, other reports, based on data included in the present study, also found more equitable ratios for the Pascagoula species pair (Selman and Qualls 2009) than for the Pearl species pair (Selman and Jones 2017).

View Figure

• Download as Powerpoint
• Download Figure

The 2 river drainages are occupied by their respective megacephalic map turtle species to nearly identical extents along their main stem reaches (632.6 rkm for G. pearlensis vs. 623.1 rkm for G. gibbonsi), but G. gibbonsi occurs at greater densities along main stem reaches, as evidenced in point counts, basking density counts, and trapping results (Fig. 3). Habitat degradation is more extensive along the Pearl River than along the Leaf, Chickasawhay, and Pascagoula rivers. The Ross Barnett Reservoir has greatly reduced habitat suitability of 5% of the main stem Pearl River, and we also found low densities of G. pearlensis in the West and East Pearl rivers, whose flow was greatly altered by construction of the Pearl River Navigation Canal, which also has very low densities. Near Jackson, river channelization has also impacted G. pearlensis habitat negatively (Selman 2020), and channelization has also occurred along Tuscolameta Creek and the upper Yockanookany River. In the Pascagoula River drainage, only 1 small reservoir has been constructed within the range of G. gibbonsi on upper Okatibbee Creek, and no channelization has been undertaken.

Riparian land preservation is also more extensive in the Pascagoula River drainage. Upstream portions of the Pascagoula River drainage flow through Bienville and DeSoto national forests, and extensive habitat protection occurs downstream in 2 Nature Conservancy preserves near the confluence of the Leaf and Chickasawhay rivers (Lindeman 2013) and the Pascagoula and Ward Bayou wildlife management areas (WMA), which protect riparian habitat for almost the entire length of the Pascagoula River. The upper reaches of Okatibbee Creek in Okatibbee WMA and a short portion of lower Red Creek in Red Creek WMA are also protected. The most extensive habitat preservation on the Pearl River is the Bogue Chitto National Wildlife Refuge along the upper West and East Pearl and lower Bogue Chitto rivers, which is contiguous with the Pearl River WMA, which protects the area between the West and East Pearl rivers downstream to the Gulf of Mexico. Additional protection of G. pearlensis habitat occurs in the uppermost Pearl River in Nanih Waiya WMA, the Pearl River as it flows into Ross Barnett Reservoir in Pearl River WMA, and the upper Strong River in Bienville National Forest.

The human populations of the 2 river drainages provide a direct index of the degree of river degradation in each: according to 2010 US census data, the 16 Mississippi counties inhabited by G. gibbonsi have a combined human population of 608,072, while the 17 Mississippi counties and 2 Louisiana parishes inhabited by G. pearlensis have a combined population of 1,168,261 (92% higher). Much of the difference is due to the Jackson metropolitan area, as Hinds, Madison, and Rankin counties account for 41% of the total human population for counties and parishes in the Pearl River drainage. Human populations may also relate inversely to habitat preservation, as less populated areas provide better opportunities for habitat acquisition.

Given the greater concentration of estimated global population of G. gibbonsi in tributaries (43% vs. 39% for G. pearlensis), in a river system that is more tributary rich (Fig. 2), populations of G. gibbonsi should also prove to be more buffered against catastrophic declines due to anthropogenic degradation of habitat, as many such threats would not extend upstream into tributary channels. The point regarding insulation against threats in tributaries is important to the conservation status of these 2 species because the relatively tributary-poor Pearl River drainage is afflicted with considerably more plans for projects that would degrade habitat for these 2 species. Plans for new reservoirs on the Pearl River both upstream and downstream of Jackson have been or are being considered (Lindeman 2013; Selman 2020).

Although they are ecologically similar species, the reasons given for listing G. oculifera and G. flavimaculata differ (USFWS 1986, 1991) in ways that are relevant to comparing the status of G. pearlensis and G. gibbonsi and making listing determinations. The listing of G. oculifera was predicated primarily on possible future threats to the species via further river engineering (USFWS 1986), and such projects, including the most recent “One Lake” proposal (Selman 2020), would be at least as deleterious for G. pearlensis, perhaps more so. The listing of G. flavimaculata, in contrast, was a reaction to perceived declines in populations that were considered to have already occurred (USFWS 1991). In the Pearl River drainage, G. oculifera outnumbers G. pearlensis to a much greater degree than G. flavimaculata outnumbers G. gibbonsi, a situation that has held steady for more than a quarter century (Lindeman 1998, 1999; Selman and Qualls 2009; Selman and Jones 2017); thus, the declines G. gibbonsi has suffered appear to be less severe.

Long-Term Historical Changes in Abundance of G. pearlensis and G. gibbonsi. — Basking survey data for sympatric species pairs of Graptemys in the Pearl and Pascagoula river drainages have been collected beginning only in the 1980s, when the USFWS began conducting surveys (Lindeman 2013). Trapping and specimen data date back to the late 1800s for the Pearl River and to 1952 (except for a single specimen of G. flavimaculata collected in 1930) for the Pascagoula River (Lindeman 2013) and may be used to give some idea of relative abundance in past decades. Overall, the data suggest that more equitable ratios of the 2 species occurred in southerly main stem reaches of each river drainage several decades ago. For his 2 species descriptions, Baur (1890, 1893) acquired 15 G. oculifera and 6 G. pearlensis collected in the West Pearl River between 1888 and 1891, a sawback:megacephalic ratio of 2.5:1. Cagle (1953) reported a ratio of 0.5:1 for a collection of 153 Graptemys specimens in the West Pearl River, and Tinkle (1958) reported the same ratio for 87 Graptemys specimens collected at 3 main stem Pearl and West Pearl sites. Cliburn (1971) collected at 7 sites throughout the Pearl River drainage and reported a 0.6:1 ratio for 114 Graptemys specimens. In 1978–1979, the ratio was 0.97:1 for 421 Graptemys captured in the middle Pearl River in Copiah and Simpson counties, Mississippi (McCoy et al. 2020). At a site on the middle Pascagoula, initial collection by the Tulane Field Crew in 1952 was strongly skewed toward G. flavimaculata, with a ratio of 11.6:1 for 202 specimens (Lindeman 2013), although Tinkle (1958) gave a much more equitable ratio of 1.4:1 for 36 Graptemys collected at a site on the lower Chickasawhay. The ratio for all 261 known specimens that were collected prior to 1960 is 6.9:1. Cliburn (1971) collected at 16 sites throughout the Pascagoula River drainage and reported a 0.97:1 ratio for 213 Graptemys specimens. In 1978–1979, the ratio was 0.6:1 for 141 Graptemys captured in the lower Chickasawhay River in Greene County, Mississippi (McCoy et al. 2020).

We found that both Graptemys species in the Pascagoula River drainage were less abundant in our most recent (2015–2018) point counts and basking density counts than in earlier data (2006–2014), while no difference by time period was found for either species in the Pearl River drainage. Due to differences in survey technique, however, it is difficult to know how much if any of these significant differences should be attributed to actual declines. Earlier surveys in the Pascagoula drainage depended more heavily on the use of spotting scopes with higher magnification, while the more recent surveys have been done primarily using binoculars. Differences in turtle detection among observers or differences in flight initiation by basking turtles viewed using different survey techniques may also contribute to differences. Using the same trapping protocol 5 times over 25 yrs, Selman and Jones (2017) found that G. pearlensis declined in abundance at 3 of 5 sampling sites on the Pearl River.

Federal Action Under the ESA. — Given their relatively low density throughout almost all of their range, historical decline in relative abundance, and existing and proposed threats to their habitat, we believe that G. pearlensis should be federally listed as T. Populations of G. gibbonsi are more robust than those of G. pearlensis, less threatened by existing habitat degradation, and more insulated from potential future threats by more extensive tributary populations. Listing G. gibbonsi as T would recognize historical declines in its abundance, and recovery efforts could aim to stop further declines and rebuild populations. An alternative would be to list G. gibbonsi as T-SA due to the similarity of G. pearlensis and G. gibbonsi. A T-SA listing would protect the species from exploitation through the pet trade while focusing federal and state recovery efforts on the more dire status of G. pearlensis. On the other hand, a T-SA listing would exempt G. gibbonsi from the requirement of interagency consultation for potentially harmful activities funded or authorized by federal or state agencies as would be required with a T listing.

Given their low relative abundance compared to the 2 sympatric sawback species that have already long been listed as T, the possibility of a listing as E might also be considered. Given that both species have ranges extending farther upstream into tributary streams than the sawbacks and the fact that neither species has been extirpated from any segment of its range, however, we believe a listing as E is probably not warranted at this time. Reports of the possible imminent extinction of G. pearlensis (Alexander 2018) are almost certainly overstated, but the need for protection from further habitat degradation and from exploitation for the pet trade is clear.

Recommended Recovery Efforts. — The primary need for recovery planning is a better understanding of anthropogenic factors that limit populations of megacephalic Graptemys species and that may contribute to their decline. The highest priority of this effort should concern the prey base of G. pearlensis and G. gibbonsi—primarily invasive Asian clams (Corbicula spp.) and native mussels but with aquatic insect larvae also important prey for males and juveniles (J. Vučenović and P.V.L., unpubl. data, 2015–2018)—and how these taxa responds to impoundment, channel dredging, and sedimentation. Nesting by G. pearlensis and G. gibbonsi has not been studied in detail, and habitat requirements for sandbars or other nesting habitats may also be important factors in population viability. Finally, deadwood abundance and the preservation of its source in riparian forests are important determinants of riverine turtle abundance (Lindeman 1999; Sterrett et al. 2010, 2015). More detailed studies are needed comparing map turtle abundance and ecology in deadwood-poor river reaches with little riparian forest source vs. deadwood-rich areas located in areas with well-managed riparian forests.

In a comparison of numbers of published articles, G. pearlensis and G. gibbonsi were among the most sparingly studied of North American turtle species (Lovich and Ennen 2013). Much of what we do know about these 2 species is from data collected during recovery efforts focused on federally listed sympatric congeners (Selman and Qualls 2009; Selman and Lindeman 2015; Selman and Jones 2017); additional focused study of the basic ecology and life history of each megacephalic species is much needed.

Basking density and trapping surveys are valuable methods for monitoring Graptemys populations, particularly if the same methodology is used at the same sites (e.g., Jones 2017; Selman and Jones 2017). Given the diversity of techniques used to quantify basking density of Graptemys in the Pearl and Pascagoula river systems as well as in other river systems, there is a need to standardize basking survey methodology for future monitoring efforts along with designation of sites that will be monitored iteratively in the future. For the Pascagoula River drainage, we were unable to definitively determine whether data trends during the first 2 decades of the 21st century are indicative of real declines in abundance or simply artifacts of variation in survey methodology, with earlier basking density surveys having been more dependent on higher-magnification equipment (spotting scopes) employed from greater cover (bridges, riverbanks, and sandbars). Strong drainage-wide gradients in the number of basking turtles and their relative abundance also complicate efforts to compare sites over time if they are not from the same river reaches. We suggest that the USFWS and its cooperators establish standard survey methodologies for basking density surveys and/or trapping surveys on at least 6 stretches of the Pearl and West Pearl as well as 2 each of the Yockanookany, Strong, and Bogue Chitto rivers. Similarly, we suggest designation of at least 2 stretches each on the Pascagoula, Leaf, and Chickasawhay rivers and 1 each on Tallahala, Okatibbee, and Buckatunna creeks, the Chunky River, and the Bogue Homa.

Endangered species listing prohibits take for the pet trade, which would benefit populations of G. pearlensis and G. gibbonsi. Enforcement of the take prohibition, via detection of illegal collecting in the wild and illegal sale combined with prosecution of offenders, would be of paramount importance. Take is likely to be greatest near public boat ramps that facilitate access (Selman and Jones 2017); thus, spot checks by state or federal wildlife biologists for evidence of illegal trapping and educational outreach via boat ramp signage and pamphlets that encourage the public to report suspicious collecting activities they observe should be incorporated into recovery planning. Training of officers to recognize turtle species of the 2 river systems would be an important component of these efforts. A comparison of abundance and population structure in reaches near public access and areas insulated from public access would also be of interest. In addition, online trade in the 2 species should be monitored in order to be able to interdict poaching operations.

Finally, active habitat management will be necessary for rebuilding populations of G. pearlensis and G. gibbonsi. If mussel prey abundance is determined to be a limiting factor as suspected, mussel habitat restoration efforts (Haag and Williams 2014) and reduction of sedimentation will be an important part of recovery efforts. Sandbar nesting habitats may require control of nonnative vegetation and contouring to improve access to female turtles and control of nest predators, such as raccoons (Procyon lotor) and fish crows (Corvus ossifragus; Jones 2006). Experimentation with placement of deadwood in river reaches with less riparian cover may also be fruitful (Lindeman 2013).