An Extinct Map Turtle Graptemys (Testudines, Emydidae) from the Late Pleistocene of Florida

An Extinct Map Turtle Graptemys (Testudines, Emydidae) from the Late Pleistocene of Florida Author(s) :Dana J. Ehret and Jason R. Bourque Source: Journal of Vertebrate Paleontology, 31(3):575-587. 2011. Published By: The Society of Vertebrate Paleontology URL: http://www.bioone.org/doi/full/10.1080/02724634.2011.562267 BioOne (www.bioone.org) is a nonprofit, online aggregation of core research in the biological, ecological, and environmental sciences. BioOne provides a sustainable online platform for over 170 journals and books published by nonprofit societies, associations, museums, institutions, and presses. Your use of this PDF, the BioOne Web site, and all posted and associated content indicates your acceptance of BioOne s Terms of Use, available at www.bioone.org/page/terms_of_use. Usage of BioOne content is strictly limited to personal, educational, and non-commercial use. Commercial inquiries or rights and permissions requests should be directed to the individual publisher as copyright holder. BioOne sees sustainable scholarly publishing as an inherently collaborative enterprise connecting authors, nonprofit publishers, academic institutions, research libraries, and research funders in the common goal of maximizing access to critical research.

Journal of Vertebrate Paleontology 31(3):575 587, May 2011 2011 by the Society of Vertebrate Paleontology ARTICLE AN EXTINCT MAP TURTLE GRAPTEMYS (TESTUDINES, EMYDIDAE) FROM THE LATE PLEISTOCENE OF FLORIDA DANA J. EHRET * and JASON R. BOURQUE Florida Museum of Natural History, University of Florida, Gainesville, Florida 32611, U.S.A., dehret@flmnh.ufl.edu; jbourque@flmnh.ufl.edu ABSTRACT Graptemys kerneri, n. sp., from the Suwannee River drainage of north-central Florida, represents the most southeastern occurrence of the genus. This species is morphologically and geographically most similar to the extant Barbour s map turtle, Graptemys barbouri. G. kerneri exhibits sexual dimorphism similar to extant G. barbouri, G. ernsti, G. pulchra, and G. gibbonsi, with females being megacephalic and attaining a much larger size than males. The new species possesses a very wide skull and mandible, making it the most blunt-headed member of its clade. Specimens described here include a nearly complete skull, eight mandibles, two epiplastra, 34 neural bones, and an assortment of other shell fragments. Previously reported fossil material from Florida was collected in the 1960s along the Santa Fe River and referred to both the Pliocene and Pleistocene. Rare earth element (REE) analysis of this material is reinterpreted here as being Rancholabrean in age. INTRODUCTION Map turtles are highly aquatic emydid turtles comprising 15 described species, which occur primarily within Gulf of Mexico river drainages in the United States (Turtle Taxonomy Working Group, 2009). The fossil record for Graptemys is poor, consisting mainly of postcranial shell fragments recovered from states within its extant range including Texas, Tennessee, Alabama, Michigan, and Florida (Wilson and Zug, 1966; Jackson, 1975; Guilday et al., 1978; Holman et al., 1990). Jackson (1975) was the first to report on fossil remains of Graptemys from the Santa Fe River, Columbia County, Florida. Recent discoveries from multiple Pleistocene localities along the Suwannee (Dixie, Columbia, Lafayette, and Suwannee counties), Santa Fe (Columbia County), Waccasassa (Levy County), and Aucilla (Taylor County) rivers have yielded new Graptemys material that, along with the previously reported specimens, are attributed to a new species. This species, representing the eastern-most occurrence of the genus known, exhibits sexual dimorphism and may have attained a slightly larger size than extant Graptemys species (Jackson, 1975). BACKGROUND The evolutionary history of Graptemys remains enigmatic due to the paucity of fossils prior to the Pleistocene. Eocene (Chadronian) shell material from South Dakota previously referred to as both Chrysemys inornata and Graptemys inornata Loomis, 1904, was recombined as Pseudograptemys inornata by Hutchison (1996) in order to distinguish it from both of the extant genera. However, assignment to the Emydidae was noted as suspect, as it is based on geography and a small number of shell characteristics, some of which are also present in the Geoemydidae. Definitive placement within the emydids was withheld until additional material (i.e., skulls) could be collected (Hutchison, 1996). Other authors have proposed a Pseudemys or Trachemys sister group for Graptemys species and its purported sister taxon Malaclemys (Bickham et al., 1996; Lamb and Osentoski, 1997; Spinks et al., * Corresponding author 575 2009; Wiens et al., 2010). However, no other fossils or transitional forms have been reported to support this hypothesis (McKown, 1972; Dobie, 1981). The absence of any such forms and incongruence between mitochondrial (mtdna) and nuclear (nudna) analyses of the emydids has led to some debate over the origination and divergence of both genera (McKown, 1972; Wood, 1977; Dobie, 1981; Lamb and Osentoski, 1997; Spinks et al., 2009, Wiens et al., 2010). Although these molecular inconsistencies are not uncommon, this does illustrate the need for more complete information using numerous different sources, i.e., both mitochondrial and nuclear DNAs (Wiens et al., 2010). Using mtdna forgraptemys and Malaclemys, Lamband Osentoski (1997) estimated a divergence time of 7 11 million years ago (late Miocene). Meanwhile, the lack of fossil evidence is most likely related to collecting bias and the habitat requirements for both Graptemys (river endemism) and Malaclemys (estuarine habitats). Graptemys fossils were first reported from Florida by Jackson (1975). They consisted of a small sample from the Santa Fe River in Columbia County, Florida, collected from riverine deposits at two adjacent sites, Santa Fe I and Santa Fe II, respectively. A nuchal (UF 10572) and hyoplastron (UF 19246) were recovered from Santa Fe I (a temporally mixed fauna), whereas the mandible (UF 19161) and neural bone (UF/TRO 100) were from Santa Fe II (late Pleistocene). Santa Fe I contains fossils that are both early Pleistocene ( 2.2 Ma) and late Pleistocene ( 15 ka) in age that are difficult to distinguish based on general appearance (Jackson, 1975; MacFadden et al., 2007). However, Jackson referred to the nuchal as being Pliocene in age based on coloration and preservation, whereas the other fossils were attributed to the late Pleistocene (late Rancholabrean). We believe that this material, combined with recently discovered specimens from the Aucilla, Waccasassa, and Suwannee rivers, represents a new Graptemys species, Graptemyskerneri,n. sp., distinct from, but closely related to, Graptemys barbouri.the original description of the Santa Fe material identified the specimens as both Graptemys barbouri and/or G. barbouri-like ; however, Jackson (1975) did recognize that the material might also represent a new species. Fossil occurrences in Dixie, Columbia, Lafayette, Levy, Suwannee, and Taylor counties are outside the current geographic distribution of Graptemys barbouri and Graptemys ernsti Lovich and McCoy, 1992 (the two extant

576 JOURNAL OF VERTEBRATE PALEONTOLOGY, VOL. 31, NO. 3, 2011 FIGURE 1. Southeastern United States showing the localities for Graptemys kerneri and the extant ranges for the megacephalic clade Graptemys species: 1, Waccasassa R.; 2, Suwannee R.; 3, Santa Fe R.; 4, Aucilla R.; 5, Apalachicola R.; 6, Chattahoochee R.; 7, Flint R.; 8,YellowR.;9,Escambia R.; 10, Conecuh R.; 11, Alabama R.; 12, Tombigee R.; 13, Coosa R.; 14, Pascagoula R.; and 15, Pearl R. endemic Florida species), and extend the range for the genus into north-central Florida (Ewert et al., 2006). This material exhibits characteristics consistent with highly sexually dimorphic Graptemys species within the megacephalic clade such as G. gibbonsi Lovich and McCoy, 1992, G. barbouri, G. ernsti, andg. pulchra Baur, 1893 (Lindeman, 2008). Based on the size of G. barbouri, where females attain carapace lengths up to 33 cm whereas males can grow to 13 cm (Ernst et al., 1994), all of our specimens represent females. GEOLOGY AND AGE Graptemys specimens reported here are from numerous river localities in north-central Florida. Materials were recovered from the Aucilla River, Taylor County; the Santa Fe River, Gilchrist and Columbia counties; the Waccasassa River, Levy County; and the Suwannee River, Dixie, Columbia, Lafayette, Levy, and Suwannee counties, all of which drain into the Gulf of Mexico along the west coast of Florida (Fig. 1). All specimens were surface collected along river bottoms, from in situ sediments. Although the exact ages of the specimens are uncertain, the dominant sediments in these rivers are Pleistocene in age. Based on the previous work by MacFadden et al. (2007), the Santa Fe I specimen presented us with an opportunity to use rare earth element (REE) analysis to assess the age of the specimen with more refined accuracy. Exposed basement rocks in the region where the Graptemys fossils were collected include the Eocene Ocala and the Oligocene Suwannee limestones (Vernon, 1951; Hulbert, 2001; Kendrick, 2006). These marine limestones were created during

EHRET AND BOURQUE EXTINCT GRAPTEMYS FROM FLORIDA 577 periods when higher sea levels inundated a majority of peninsular Florida. Fossils found in these limestones include echinoids, foraminifera, mollusks, bryozoans, ostracodes, sharks, and other chondrichthyans (Hulbert, 2001; Kendrick, 2006). These limestone formations form the river beds in most of the drainages in the panhandle and western peninsular Florida, and are the preferred habitat for G. barbouri in the Apalachicola River (Ernst et al., 1994). Aucilla River The Aucilla River is well known for the late Pleistocene early Holocene Page-Ladson site excavated by the Florida Museum of Natural History for over 20 years (Webb, 2006). As discussed above, the river bottom is composed mainly of Suwannee limestone. This formation is unconformably overlain by undifferentiated marine sands that vary in thickness from 1 to 15 m across the region, with some of the larger sand deposits representing barrier islands, dunes, and bars associated with late Pleistocene sealevel high stands (Kendrick, 2006). Additionally, lenses of weakly phosphatic, clayey sands are scattered throughout the river in Taylor and Jefferson counties, which represent reworked middle Miocene sediments and possibly some Plio-Pleistocene paleosinkhole fill (Rupert, 1996; Kendrick, 2006). Kendrick (2006) noted that many of these sediments in basin fills and sinkholes represent complex in situ records of Pleistocene flora and fauna. Carbon dates of specimens collected at the Page-Ladson site resulted in ages spanning the latest Pleistocene and early Holocene (Webb and Dunbar, 2006). Santa Fe River The Santa Fe is a spring-fed river that is one of the main tributaries of the Suwannee River (Fig. 1) (Vernon, 1951; Jackson, 1975; Bryan et al., 2008). The river bed cuts into the Ocala Limestone and is lined with alluvium that contains both early and late Pleistocene vertebrate faunas (MacFadden et al., 2007). Of the three sites where Graptemys fossils were collected, two contain late Pleistocene faunas (Santa Fe II and Santa Fe VII), and one contains a temporally mixed fauna (Santa Fe I). Fossils from Santa Fe I have been found to be both early Pleistocene (ca. 2.2 Ma) and late Pleistocene (ca. 15 ka) in age and include terrestrial taxa such as Titanis walleri, Bison sp., Mammuthus columbi, Megalonyx jeffersonii, Holmesina floridanus, andnannipus peninsulatus (Jackson, 1975; Hulbert, 2001; MacFadden et al., 2007). MacFadden et al. (2007) recently used REE analysis to separate the two faunas, which we have applied to one of the Graptemys specimens and will discuss below. Suwannee River The Suwannee River has its headwaters in the Okefenokee Swamp of southern Georgia and flows southwest before emptying into the Gulf of Mexico in Suwannee County, Florida (Vernon, 1951; Puri et al., 1967). It is entrenched in a shallow solution valley in Florida and traverses the Eocene Ocala Limestone, the Oligocene Suwannee Limestone, and the Miocene Hawthorn Group (Schmidt, 1997). The Suwannee likely existed as a smaller stream during the Pliocene and became a larger river with the capture of waters from an ancestral river originating in Georgia sometime in the early Pleistocene (Schmidt, 1997). Graptemys materials have been recovered from numerous localities in the southern Suwannee River, from Lafayette County south to the mouth of the river along the Gulf of Mexico. The age of the sediments in the Suwannee River has not been examined extensively or correlated with other localities (R. Portell, pers. comm., January, 2010). Other fossil materials collected from the Dixie County locality in the FLMNH include Holmesina septentrionalis and Megalonyx jeffersonii, which would indicate a late Pleistocene age. Materials collected from the Lafayette County locality that are cataloged in the collections at FLMNH include Holmesina septentrionalis, Bison sp., Tapirus veroensis, and Mammut americanum, which also indicates a late Pleistocene age. Waccasassa River The Waccasassa River in Levy County is another karst river that flows into the Gulf of Mexico, to the south of the Suwannee. The river bed flows over the Avon Park Limestone (middle Eocene) and contains fossils that are considered to be both late Hemphillian and late Pleistocene in age, likely eroding out from the surficial sands and clays along the banks (Vernon, 1951; Domning et al., 1982; Bryan et al., 2008). Vernon (1951) originally proposed that the Waccasassa may have once been much larger, enveloping the Santa Fe River in its northern reaches. However, this theory has been discredited by Puri et al. (1967) and Schmidt (1997). Recent invertebrate and vertebrate fossils collected in situ from areas along the Waccasassa are also believed to be from the Avon Park Formation (middle Eocene) (R. Portell, pers. comm., January, 2010). RARE EARTH ELEMENTS In recent years, the diagenetic uptake of rare earth elements (REEs) during bone fossilization has been instrumental in separating time-averaged vertebrate faunas and faunal sequences (Trueman and Benton, 1997; Metzger et al., 2004; MacFadden et al., 2007). Living vertebrate tissues contain very low levels of REEs (ranging in ppb), but they are rapidly taken up and concentrated during the early stages of bone diagenesis (MacFadden et al., 2007). REE values in bones from different sedimentary environments have been shown to vary temporally as well as spatially (Trueman, 1996; Trueman and Benton, 1997). The concentrations of REEs in fossil bones are indicative of the geochemical conditions and redox potential of the pore waters present at the time of diagenetic recrystallization (Henderson et al., 1983; Mac- Fadden et al., 2007). As local pore-water filters through the bone, uptake of REEs begins within a few years and may continue for over 10,000 years (Patrick et al., 2001; Trueman et al., 2004; Mac- Fadden et al., 2007). Furthermore, once a bone is altered in the early stage of diagenesis, it is not prone to additional exchange or uptake of elements (Trueman and Tuross, 2002; MacFadden et al., 2007). Therefore, bones that have been reworked temporally or spatially with other fossils can be distinguished by their REE signatures (Trueman et al., 2006). MATERIALS AND METHODS The assignment of the Santa Fe nuchal (UF 10572) to the Blancan ( 2.2 Ma) by Jackson (1975) was based solely on taphonomic considerations. He noted that Blancan fossils from Santa Fe I were typically glossy black, whereas Rancholabrean specimens were coarse and brown. Recent work by MacFadden et al. (2007) on this taphonomically mixed fauna utilized rare earth elements (REEs) in an attempt to more accurately separate these faunas. In order to determine the age of this specimen, the nuchal was sampled for REE concentrations following the methods described in MacFadden et al. (2007). Institutional Abbreviations UF, Florida Museum of Natural History, Gainesville, Florida, UF/TRO, specimens formerly in the collection of the Timberlane Research Organization, Lake Wales, Florida, now housed at the Florida Museum of Natural History. Anatomical Abbreviations In reference to bones: C, costal (e.g., C1 = first costal); N, neural;p, peripheral. In reference to scutes: M, marginal; PL, pleural; V, vertebral.

578 JOURNAL OF VERTEBRATE PALEONTOLOGY, VOL. 31, NO. 3, 2011 SYSTEMATIC PALEONTOLOGY Order TESTUDINES Batsch, 1788 Suborder CRYPTODIRA Cope, 1868 Family EMYDIDAE Bell, 1825 Subfamily DEIROCHELYINAE Agassiz, 1857 Genus GRAPTEMYS Agassiz, 1857 GRAPTEMYS KERNERI, sp. nov. Holotype UF 239000, nearly complete skull (Figs. 2, 3). Etymology Species name is in honor of Andreas Kerner, in recognition of numerous donations that he has contributed over the years to the Florida Museum of Natural History. This includes the majority of the specimens presented in this paper, particularly the holotype, UF 239000. Type Locality and Age Suwannee River (DI005), which forms the boundary between Gilchrist and Dixie counties, Florida, 29 36 57 N, 82 58 11 W (Fig. 1). The specimen was recovered from unconsolidated sediments from the river bed. These sediments are presumed to be Rancholabrean, late Pleistocene in age ( 15 ka). Referred Materials UF 23900, partial skull; UF 10572, partial nuchal bone; UF 246210, partial nuchal bone; UF 249151, nuchal bone; UF 19161, partial mandible; UF 235104, partial mandible; UF 238540, partial mandible; UF 238541, partial mandible; UF 238542, partial mandible; UF 238543, partial mandible; UF 238544, partial mandible; UF 255356, partial mandible; UF 239751, right epiplastron; UF 254852, left epiplastron; UF 254853, 1st neural bone; UF 254854, 1st neural; UF 254855, 1st neural; UF 254868; UF 239752, 2nd neural; UF 245005, 2nd neural; UF 245006, 2nd neural; UF 246200, 2nd neural; UF 246201, 2nd neural; UF 254856, 2nd neural; UF 254861, 2nd neural; UF 239753, 3rd neural; UF 244991, 3rd neural; UF 245007, 3rd neural; UF/TRO 100, 3rd neural; UF 246202, 3rd neural; UF 246203, 3rd neural; UF 246206, 3rd neural; UF 254857, 3rd neural; UF 254858, 3rd neural; UF 254859, 3rd neural; UF 254860, 3rd neural; UF 254865, 3rd neural; UF 254869, 3rd neural; UF 245008, 4th neural; UF 246207, 4th neural; UF 246208, 4th neural; UF 254862, 4th neural; UF 254864, 4th neural; UF 254866, 4th neural; UF 239754, 5th neural; UF 254863, 6th neural; UF 246209, 7th neural; UF 254867, 7th neural; UF 239759, left 9th peripheral; UF 239758 right 8th peripheral; UF 246211, left hyoplastron; UF 254851, left hyoplastron; UF 239755, entoplastron; UF 239757, left hypoplastron; UF 249153, right hypoplastron; UF 244994, left xiphiplastron. Distribution Materials were recovered from north-central Florida, from the Big Bend region in the eastern panhandle, east and south into peninsular Florida (Fig. 1). Although most of the fossil materials described here were found in the Santa Fe, Suwannee, and Waccasassa rivers (Columbia, Dixie, Gilchrist, Lafayette, and Levy counties) of peninsular Florida, two mandibles were collected from the Aucilla River in Taylor County. These records are eastern and southern range extensions for the genus in Florida. Temporal ranges for extant species within the megacephalic clade include G. barbouri,which is found in the Apalachicola, Chattahoochee, and Chipola rivers in Florida, the Chattahoochee River in Alabama, and the Chattahoochee and Flint rivers in Georgia; G. ernsti, which is found in the Escambia River in Florida, the Conecuh (which is the same river as the Escambia), Sepulga, Pea, and Yellow rivers in Alabama; G. pulchra, which is found in the Tombigee, Tensaw, Black Warrior, and Coosa rivers in Alabama, and the Coosa and Tallapoosa rivers in Georgia; and G. gibbonsi, which is found in the Pascagoula and Pearl rivers of Alabama and the Pearl River in Louisiana (Ernst et al., 1994). Diagnosis (Female) skull and mandible shorter in length relative to overall width, in contrast with G. barbouri, inwhichthe skull is longer (measured from the nasal canal at the maxilla/ premaxillary suture to the occipital condyle); external surface of maxilla deep dorsoventrally; enlarged protuberance of the processus trochlearis oticum on the parietal bone that overhangs the foramen for the trigeminal nerve, in contrast with G. barbouri, which only has a slight processus trochlearis oticum on the parietal that continues onto the prootic bone; canal for the stapedial artery deep, trench-like, high on the parietal bone, unlike G. barbouri, which has a shallow and low canal on the parietal; anterior edge of the otic capsule (comprised of the prootic and quadrate) relatively straight transversely, in contrast to G. barbouri, where this edge is angled posteriorly (when viewed dorsally); reduced inferior process of the parietal associated with an expanded passage between the orbital and inferior temporal fossae, compared with G. barbouri, where the inferior process of the parietal bulges along the parietal/palatine sulcus greatly reducing passage between the orbital and inferior temporal fossae (Figs. 2 4); mandible wide, U-shaped, and relatively flat with highly expanded triturating surfaces, differing from G. barbouri in which the mandible is dorsoventrally deeper and more V-shaped; symphyseal sulcus for triturating pads shallow, in contrast to that of G. barbouri, which is deep (more trench-like) and longer anteroposteriorly; broad thickened bulge at symphysis that runs the entire length of the symphysis, unlike G. barbouri, which has a relatively narrow symphyseal bulge; coronoid processes relatively low and rounded, in contrast to G. barbouri where the coronoid processes are taller and directed posteriorly (Figs. 4, 5). DESCRIPTION Skull The holotype skull (UF 239000) was collected from the Suwannee River, Dixie County, Florida (Figs. 2, 3). It represents an adult female, which we speculate exhibits megacephaly similar to that in other species in the megacephalic clade of map turtles that have evolved to facilitate a largely molluscivorous diet. It exhibits ontogenetic distortion (including outward flaring of the maxillae and parietals) characteristic of older individuals of Graptemys barbouri examined in the FLMNH collection (Cagle, 1952). Although clearly representing an older individual based on its size, it is smaller than the largest G. barbouri skulls measured at the FLMNH. The length of the skull (measured from the nasal canal at the maxilla/premaxillary suture to the occipital condyle) is 60.9 mm. The width of the skull (measured at the anterior-most point of the suture between the quadrate and quadratojugal when held in lateral view) is 60.2 mm. Compared to measured specimens of G. barbouri; this specimen is shorter in length relative to overall width, giving it a blunt-faced or snub-nosed appearance. The skull is relatively complete, shows wear consistent with water tumbling, and is missing the premaxillae, the distal portion of the sagittal crest, and portions of the supraoccipitals, postorbitals, and parietals (Figs. 2, 3). The postorbital bar is completely missing on the left side of the skull and is represented by a thin portion of the quadratojugal on the right side, which preserves the suture with the jugal. Although incomplete, we believe that the postorbital bar was at least as thick as that of G. barbouri based on breaks along the quadratojugal, jugal, and postorbital bones. The prefrontals are short and wider anteriorly at the suture with the maxilla than what is seen in G. barbouri. The fissure ethmoidalis is narrow and keyhole-shaped. The frontals contribute to the orbital rim and are shorter than those of G. barbouri,drawing the frontoparietal suture closer to the orbits. There is no medial inflection of the inferior processes of the frontal. The parietals are anteriorly flared, a feature only seen in older, sexually mature females of G. barbouri. This combination of characters results in the shortening of the nasal region, and a flared back of the skull producing a snub-nosed or short-faced appearance. A similar skull morphotype is also seen in some bothremydid and baenid turtles (Gaffney et al., 2006; Lyson and Joyce, 2009).

EHRET AND BOURQUE EXTINCT GRAPTEMYS FROM FLORIDA 579 FIGURE 2. The skull (holotype) of Graptemys kerneri (UF 239000). A, dorsal view; B, ventral view; C, side view (right); D, anterior view. The medial process of the jugal is well developed and contacts both the pterygoid and parietal. The jugal also contacts the palatine. The foramen orbitonasale is large, being more than 1/6 of the orbit length (Joyce and Bell, 2004). The quadratojugal is large and contacts the jugal and maxilla. There is no contact between the quadratojugal or the pterygoid and the articular facet of the quadrate. The incisura columella auris, the notch within the quadrate that holds the stapes and eustachian tube, is open, a feature commonly seen in emydids (Gaffney, 1972; Joyce and Bell, 2004). The protuberance on the processus trochlearis oticum of the parietals is enlarged and overhangs the foramen for the trigeminal nerve in dorsal view. This protuberance tapers at the sulcus where the parietal and prootic bones meet. In most G. barbouri, G. ernsti, G. pulchra, and G. gibbonsi specimens examined, this protuberance is not as enlarged and is formed by both the parietal and prootic bones. The epipterygoid does not participate in the trigeminal foramen. The canal for the stapedial artery is trench-like and very high on the parietal as compared to other Graptemys species. The passage between the orbital and inferior temporal fossae is enlarged, as a result of a flattened inferior process of the parietal. In G. barbouri, this process is enlarged along the parietal/palatine suture, greatly reducing this passageway. There is no contact between the inferior process of the parietal and the maxilla. The stapediotemporal foramen is positioned close to the sutural junction of the supraoccipital, prootic, and opisthotic bones and located more posteriorly than in G. barbouri. The anterior margin of the otic capsule, formed by the anterior edges of the prootic and quadrate bones in Graptemys, is deep and straight in transverse view. In other species of the megacephalic clade, the processus trochlearis oticum is shallower and faces anterolaterally. These features outline the anterior portion of the fossa temporalis superior, which contains the main mass of the M. adductor mandibulae externus, which is used for hard biting when the jaw is nearly closed (Gaffney, 1972; Jones et al., 2009). The maxilla is deep and robust with a strongly curved labial ridge. There is no medial contact of the maxillae along the anterior margin of the jaw. However, there is contact between the maxilla and vomer. The triturating surfaces are broad and flat (i.e., lacking lingual ridges or serrations) like other species of the megacephalic Graptemys clade, which is most likely related to a molluscivorous diet. There is participation of the palatine in the triturating surface of the upper jaw. The articular facets of the quadrates are larger and broader than those seen in G. barbouri. The basisphenoid of UF 239000 is diamond-shaped, in contrast to that of other species, which tend to be heart or spade-shaped. Mandibles Eight partial mandibles represented mainly by isolated dentaries and one more complete specimen (Fig. 5) were recovered from the Aucilla River, Taylor County (UF 235104, UF 255356),

580 JOURNAL OF VERTEBRATE PALEONTOLOGY, VOL. 31, NO. 3, 2011 FIGURE 3. Line drawings of the skull of Graptemys kerneri (UF 239000). Abbreviations: bo, basioccipital; bs, basisphenoid; ex, exoccipital; fr, frontal; ju, jugal; mx, maxilla; op, opisthotic; pa, parietal; pal, palatine; pf, prefrontal; po, postorbital; pr, prootic; pt, pterygoid; qu, quadrate; so, supraoccipital; vo, vomer.a, dorsal view; B, ventral view; C, side view (right); D, anterior view. Santa Fe River, Gilchrist and Columbia counties (UF 19161, UF 238540), and Suwannee River, Dixie, Columbia, and Suwannee counties (UF 238541, UF 238542, UF 238543, UF 238544). All are broader and flatter than those of G. barbouri, which exhibits the most expanded mandible among extant species of Graptemys. This massive mandibular structure, characteristic of the sexually dimorphic females of the megacephalic clade Graptemys species, is considered an adaptation for molluscivory (Carr, 1952; Sanderson, 1974; Jackson, 1975; Lee et al., 1975; Ewert et al., 2006). The labial ridge in all specimens is thickened and relatively flat compared to those of other representatives of the megacephalic clade of Graptemys. At thesymphysis, the anterior tip is also flatter than and not as upturned as that of G. barbouri or G. ernsti. There is a distinctive broad, low ridge that runs anteroposteriorly along the symphysis of the triturating surface. Directly along the symphysis and bisecting this ridge is a shallow trench, which gets deeper posteriorly. The triturating surface does not extend posteriorly behind the symphysis of the dentary. In dorsal view, the dentary bones meet at a wide angle, giving the mandibles a Ushaped appearance (Jackson, 1975). This is in contrast to G. barbouri and other species of Graptemys where the dentaries meet at a more acute angle, giving the mandibles a more V-shaped appearance. UF 255356 represents the most complete specimen, which, in addition to the dentaries, also preserves the angular, surangular, articular, and prearticular bones (Fig. 5H J). The suture between the surangular and dentary is simple, without interdigitation. The foramen dentofaciale majus is small. As noted in the partially preserved specimens, the specimen is wider and more U-shaped when compared with other species of Graptemys (Figs. 4, 5). The preserved coronoid process is low and rounded, unlike G. barbouri, which has a process that is much taller and angled posteriorly, as noted by Joyce and Bell (2004). The area articularis mandibularis is also preserved, although it is moderately worn. Based on its size, these facets appear to be more mediolaterally compressed when compared with G. barbouri.

EHRET AND BOURQUE EXTINCT GRAPTEMYS FROM FLORIDA 581 FIGURE 4. Extant Graptemys skulls and mandibles (in dorsal view) and mandibles (in ventral view) within the megacephalic clade. A, G. barbouri skull, UF 5000; B, G. barbouri mandible, UF 5000; C, G. ernsti skull, UF 34989; D, G. ernsti mandible, UF 34989; E, G. pulchra skull, UF 85226; F, G. gibbonsi skull, AUMP 2683. Carapace Neurals Neurals described here include unassociated N1, N2, N3, N4, N5, N6, and N7 (Fig. 6). Sculpture of these bones varies from low, shallow striations that run anterodorsally along the surface, to small, irregular pits, to completely smooth. Four N1s have been recovered from the Suwannee River, Lafayette County (UF 254853, UF 254854, UF 254855, UF 254868). The bones are oval in shape, and are longer than they are wide. Lengths range from 30.9 to 38.7 mm and widths from 22.8 to 27.3 mm. All specimens exhibit a broad, low dorsal keel that is present on the anterior half of the neural bone, ending at the sulcus for the posterior edge of V1. Seven N2s have been recovered from the Suwannee River, Dixie County (UF 239752, UF 245005, UF 245006), Lafayette County (UF 254856, UF 254861), and the Waccasassa River, Levy County (UF 246200, UF 246201). The bones are hexagonal in shape with a strongly concave anterior suture, which is typical of emydids (Auffenberg, 1974), whereas the posterior suture of the N3 is convex. Most neurals are broader than long, with the exception of UF 245006, which is slightly longer than it is broad. Widths range from 22.7 to 31.4 mm, whereas lengths range from 20.3 to 32.1 mm. All specimens are relatively thick and exhibit a broad, low dorsal keel that extends onto N3 to form a blunt knob, which is characteristic of the genus Graptemys and some Malaclemys terrapin. Thirteen N3s have been recovered from the Suwannee River, Dixie County (UF 239753, UF 245007, UF 246203), Lafayette County (UF 245857, UF 245858, UF 254859, UF 254860, UF 254869, UF 254865), the Waccasassa River, Levy County (UF 244991, UF 246202, Fig. 6C; UF 246206, Fig. 6E), and the Santa Fe River, Columbia County (UF/TRO 100, Fig. 6B). They are hexagonal in shape, with a strong anterior concavity along the suture. The posterior suture with N4 is convex. All of the complete specimens are slightly longer than they are broad, with lengths ranging from 22.7 to 36.8 mm and widths from 20.4 to 31.4 mm. Jackson (1975) noted that UF/TRO 100 exhibited more tapering from the anterior to the posterior borders than seen in G. barbouri. This is also true of other N3 described in this study. Specimens are very thick, and exhibit an anterior dorsal keel that ends in a blunt knob midway along the midline of the bone. This knob is associated with the sulcus for the posterior edge of V2 that runs perpendicular to the length of the bone and follows the contour of the keel. A series of three knobs down the midline of the carapace is characteristic for Graptemys species (commonly referred to as sawbacks ) and some Malaclemys terrapin. Theshapeof this knob varies, in some specimens being more pointed than in others. This feature is likely related to age. Many of these neurals are larger than some of the largest specimens of Graptemys barbouri in the FLMNH collections. Six N4s are known from the Suwannee River, Dixie, Lafayette, and Levy counties (UF 245008, Fig. 6D; UF 246207, UF 246208, UF 254862, UF 254864, UF 254866). They are hexagonal in shape, and also exhibit an anterior concave suture. They are very similar in appearance to N2, exhibiting a posterior convex suture, which is associated with another wide midline dorsal keel that is somewhat taller than that seen on N2. This keel becomes much larger posteriorly, and projects beyond the posterior edge of the bone. This is similar to what is seen in G. barbouri, G. ernsti, G. pulchra, andg. gibbonsi. The keel extends onto N5, and ends in another knob. Unlike N2, N4 is almost equilateral. Overall lengths range from 22.3 to 33.6 mm and widths from 21.9 to 33.3 mm.

582 JOURNAL OF VERTEBRATE PALEONTOLOGY, VOL. 31, NO. 3, 2011 FIGURE 5. Mandibles of Graptemys kerneri, Graptemys barbouri, and Graptemys ernsti for comparison in dorsal view. A, G. kerneri, UF 238543; B, G. kerneri, UF 235104; C, G. kerneri, UF 19161; D, G. kerneri, UF 238544; E, G. kerneri, UF 238541; F, G. kerneri, UF 238542; G, G. kerneri, UF 238540; H, G. kerneri, UF 255356; I, G. kerneri, UF 255356 (in 3/4 view); J, G. kerneri, UF 255356 (in lateral view); K, G. barbouri, UF 5000; L, G. barbouri, UF 73797; M, G. ernsti, UF 34989.

EHRET AND BOURQUE EXTINCT GRAPTEMYS FROM FLORIDA 583 FIGURE 6. Graptemys kerneri bones of the carapace and plastron in dorsal view except H, J and K in ventral view. A, nuchal, Santa Fe R., UF 10572; B, 3rd neural, Santa Fe R., UF/TRO 100; C, 3rd neural, Waccasassa R., UF 246202; D, 4th neural, Suwannee R., UF 245008; E, 3rd neural, Waccasassa R., UF 246206; F, 5th neural, Suwannee R., UF 239754; G, H, epiplastron, Suwannee R., UF 254852 (dorsal and ventral views, respectively); I,left9th peripheral, Suwannee R. UF 239759; J, epiplastron, Suwannee R. UF 239751; K, hyoplastron, Suwannee R., UF 246211. One N5 is preserved from the Suwannee River, Dixie County (UF 239754, Fig. 6F). It is hexagonal in shape, with an anteriorly concave and a posteriorly convex sutural surface. Its low, broad midline keel is lower than on N3. The keel ends in a low, blunt knob three-quarters of the way down the bone that is much smaller than the knob present on N3. This knob ends at the sulcus for the posterior edge of V3. N5 is much broader (29.6 mm) than long (27.6 mm), in contrast to N3, which is longer than broad. One N6 has been found from the Suwannee River, Lafayette County (254863). The neural is hexagonal in shape, with anterior and posterior sutures being relatively straight. There is a low, broad keel running down the midline of the neural, which is more apparent on the anterior half of the specimen. It is much broader (31.6 mm) than it is long (26.2 mm). Two N7s have been recovered from the Suwannee River, Lafayette (UF 254867) and Levy (UF 246209) counties. This neural is hexagonal in shape and as broad as it is long. It is much smaller than the other neurals reported for the species due to its position near the posterior of the carapace. It exhibits a low, broad keel, which is developed along the entire midline of the

584 JOURNAL OF VERTEBRATE PALEONTOLOGY, VOL. 31, NO. 3, 2011 bone. The keel presumably continues onto N8 and ends in what would be a third, small knob as seen in other Graptemys species (Ernst et al., 1994). However, no N8 of G. kerneri has been recovered to date. Peripherals Two peripherals have been recovered from the Suwannee River, Dixie County (UF 239758, UF 239759, Fig. 6I). UF 239758 represents an 8th right peripheral from an adult, female individual based on its large size. This bone is widest distally, where it measures 28.3 mm. The distal edge is serrated, although both serrations are chipped. These serrations are separated by a perpendicular sulcus on both the dorsal and ventral faces of the bone that represents the contacts of M8 M9. On the dorsal side, this sulcus continues up the bone until it meets the outside edge of the pleuromarginal sulcus at a right angle near the costoperipheral suture. On the ventral side, this sulcus ends at the edge of the marginal sulci. There is a depression near the posterior edge of the costoperipheral suture for the insertion of the rib end of C6. The assignment of these bones to G. kerneri is related to the overall thinness of the bone, relative to its size. Other emydid species that exhibit serrated peripherals (i.e., Pseudemys and Trachemys) tend to have bones that are much thicker proportionally. UF 239759 represents a left P9 from a very large female based on its size. The specimen has a maximum length of 54.3 mm and a maximum distal width of 38.5 mm. The bone is serrated along the distal edge, with two broad serrations that face towards the posterior of the shell. Much like the other peripheral, these serrations are separated by a sulcus forming the edge of M9 and M10. On the dorsal side, this sulcus meets the pleuromarginal sulcus at a right angle near the costoperipheral suture. On the ventral surface, this sulcus meets the proximal edges of the M9 and M10 sulci approximately two-thirds of the way up the bone. The costoperipheral suture contacts C6 and C7, and contains a deep depression for the insertion of the C7 rib end. Nuchals Three nuchals referred to G. kerneri have been recovered from the Santa Fe River, Gilchrist County (UF 10572) and the Suwannee River, Dixie (UF 249151) and Levy (UF 246210, Fig. 6A) counties. UF 10572 was collected in 1963 and was originally described by Jackson (1975). This bone is similar to G. barbouri, having a broad cervical scale, sulci for M1 and M2, and V1 scutes. Although there is variation in the extant Graptemys species, the cervical scale overlaps much deeper on the ventral side of the bone than other taxa. Jackson (1975) noted the lack of a distinct keel along the midline of the nuchal. Although this is a variable feature in Graptemys, it should also be noted that the posterior portion of this bone is broken and does not preserve the neural suture. Two other nuchal specimens have been recovered (UF 249151 and UF 246210) from the Suwannee River. UF 249151 is nearly complete, although it is moderately water-worn. Scute sulci patterns are similar to those described and seen in UF 10572. This specimen and the posterior portion of another highly water-worn nuchal (UF 246210) both preserve the N1 suture and partial sutures with the left and right C1s. This neural suture is concave, whereas the other partially preserved sutures are straight. Both specimens are very thin and highly arched along the midline of the bone running posteriorly similar to specimens of G. barbouri. Costals Two partial costals from the Suwannee River, Dixie County (UF 239795, UF 239796), have been identified as cf. Graptemys kerneri. Both specimens are referred to Graptemys based on the relative thinness of the specimens in relation to their size. UF 239795 represents the anterior portion of a right C1. The dorsal surface contains the V1 and V2 sulci as well as that of PL1. The ventral surface of the bone preserves a partial, broken rib head. UF 239796 represents a right C2. The anterior half of the bone is preserved, including the V2 and PL2 PL3 sulci on the dorsal surface. Sculpture is limited to three irregular pits and shallow annular grooves that run around the edges of the bone, perpendicular to the sutures. The ventral surface preserves the partial head of the C2 rib. Plastron Epiplastron One right and one left epiplastron were recovered from the Suwannee River, Dixie (UF 239751, Fig. 6J) and Lafayette (UF 254852, Fig. 6G, H) counties. The right epiplastron is broader than long with a flat anterior edge, indicating that the forelobe of the plastron would have been relatively squared off anteriorly, a feature commonly seen in other macrocephalic taxa such as Platysternon, Sternotherus minor, and some Graptemys species. The highly flattened anterior edge of the plastron is not shared with any other freshwater emydid turtles. Modern Graptemys species of the megacephalic clade have a small knob or protuberance on the corners of the anterior edge of the epiplastron. Although this feature is not exhibited in UF 239751, it is present in UF 254852 (Fig. 6G, H). The hyoplastral suture is straight for most of its length, but is somewhat concave in the center in both specimens. The gular-humeral sulcus is arcuate, and runs anteriorly from the entoplastral suture to the outer edge of the epiplastron, just posterior to the anterior corner of the bone. In UF 254852 there is a strong concave dip posteriorly in the sulcus near the anterior protuberance. The gular and humeral scutes overlap very slightly onto the visceral face of the bone. Entoplastron One entoplastron was recovered from the Suwannee River, Dixie County (UF 239755). The bone is much broader (62.5 mm) than it is long (45.2 mm). It is also much thinner and broader than other emydid entoplastra. The gularhumeral sulcus is present and forks on the anterior half of the surface. The visceral face is water worn; however, the scars for the ligaments attaching to the acromion process appear to be solely contained on the entoplastron. Hyoplastron Two left hyoplastra were recovered from the Suwannee River, Levy and Lafayette counties (UF 246211, Fig. 6K, UF 254851). UF 246211 is nearly complete, preserving most of the sutures and a partial axillary buttress. The bone is thickest along the interhyoplastral suture and thins out both anteriorly and posteriorly. The humeral-pectoral and pectoral-abdominal sulci run perpendicular to the interhyoplastral suture and curve sharply posteriorly as they reach the outer margin. UF 254851 is also nearly complete and almost identical to UF 246211, only slightly larger in size. The axillary buttress of UF 254851 is better preserved than in UF 246211. It is also broken; however, it does show a U-shaped margin along the anterior opening of the shell. Hypoplastron Two hypoplastra, one left and one right, referred to Graptemys were recovered from the Suwannee River, Dixie County (UF 239757, UF 249153). UF 239757 is very thin and preserves the hyo-hypoplastral, interhypoplastral, and the hypo-xiphiplastral sutures. It also preserves only a small portion of the inguinal buttress. The bone preserves the abdominalfemoral sulcus, which runs from the posterior edge of the inguinal buttress anteriorly across the bone towards the hypoxiphiplastral suture. UF 249153 is significantly larger and thicker than the other specimen. The abdominal-femoral sulcus is present and is similar in shape to that of UF 239757. A portion of the inguinal buttress is also preserved, and creates a U-shaped opening for the posterior opening of shell. This buttress projects posteriorly and flares outward; however, it is incomplete and the overall size and shape cannot be determined. Xiphiplastron One left xiphiplastron was found in the Waccasassa River, Levy County (UF 244994). This specimen is from an adult female and is extremely thin for its size. The hypoplastral and interxiphiplastral sutures are both nearly straight. The posterior lobe is broken, so it is not possible to confirm the presence of a small caudal notch present in other species of

EHRET AND BOURQUE EXTINCT GRAPTEMYS FROM FLORIDA 585 the genus. The femoral-anal sulcus runs diagonally across the bone, starting at the anterior corner of the interxiphiplastral suture and ending on the distal margin approximately midway through the xiphiplastron. Viscerally, the femoral scute margin is much wider than that of G. barbouri or other Graptemys species. DISCUSSION AND CONCLUSIONS The origin and evolution of Graptemys still remains poorly understood (Wood, 1977; Dobie, 1981; Lamb and Osentoski, 1997; Wiens et al., 2010). This problem is compounded by a lack of fossil material and conflicting molecular analyses. Based on scant fossil evidence found throughout the eastern half of the United States, we know that many species were established by the late Pleistocene (Wilson and Zug, 1966; Jackson, 1975; Guilday et al., 1978; Holman et al., 1990). However, no fossil materials older than the late Pleistocene have been definitively identified. Female members of the megacephalic subclade of Graptemys exhibit a shift in skull proportions from the west to east along the Gulf coast (Fig. 1). This trend begins with G. gibbonsi, which exhibits a rather elongate skull, in particular the nasal regions. Skulls become progressively wider and shorter to the east, with G. barbouri being the most blunt-faced of the extant species (Cagle, 1952; Figs. 2 4). Skull proportions of G. kerneri accord with this trend by exhibiting the most blunt-faced and being the most easterly occurring species in the group. Interestingly, however, G. kerneri exhibits signs of the same ontogenetic distortion observed in the other species (such as the flaring of the maxilla, jugal, and quadratojugal and the upward inflection of the parietal at the frontoparietal suture) but at a smaller overall skull size than observed in G. barbouri (Cagle 1952; Figs. 2 4). This indicates that G. kerneri may have stopped growing, in terms of length, at a smaller size than G. barbouri. Extreme sexual dimorphism and molluscivory in female Graptemys are characteristics of all species within the megacephalic clade (Ernst et al., 1994; Lindeman, 2000; Lindeman and Sharkey, 2001; Lindeman, 2008; Stephens and Wiens, 2009). Males of the same species are significantly smaller and do not show adaptations for strict molluscivory, although they may take small gastropods in their diet (Sanderson, 1974; Lee et al., 1975). This clade of megacephalic species is also supported by molecular genetic analysis based on a character set composed of mtdna sequences of cytochrome b and control region areas of the mitochondrial genome (Lamb et al., 1994; Lamb and Osentoski, 1997; Wiens et al., 2010). Based on molecular analyses, characters exhibited in the new fossil materials, and geographical proximity, G. barbouri is the likely sister species of G. kerneri. The assignment of the nuchal (UF 10572) from the Santa Fe River to the Blancan (Jackson, 1975) is refuted here by REE analysis. Comparisons of the average REE N patterns for the Blancan versus Rancholabrean samples of MacFadden et al. (2007) with that of UF 10572 show that the G. kerneri sample clearly resembles the latter (Fig. 7). This trend demonstrates that the Graptemys nuchal underwent diagenesis and accumulated rare earth elements at approximately the same time interval as the Rancholabrean samples. A late Pleistocene age for the Santa Fe, Aucilla, and northern Suwannee River materials and the potential for a middle Pleistocene age of the southern Suwannee and Waccasassa River specimens can also tell us something about the distribution of G. kerneri (Fig. 1). Graptemys are rarely found traveling on land, which has led to a characteristic river endemism with regards to several of the extant species (McKown, 1972; Jackson, 1975; Ernst et al., 1994; Lindeman, 2008). Therefore, the distribution of species through geologic time is most likely related to shifts in river drainage patterns. Jackson (1975) proposed two possible scenarios for the spread of Graptemys into the Suwannee River FIGURE 7. Rare earth element (REE N normalized to post-archean Australian Shale; McLennan, 1989) data for the temporally mixed Santa Fe River fauna, comparing averaged values for Blancan ( 2.2 Ma) and Rancholabrean ( 15 ka) fossils compared to the G. kerneri nuchal, UF 10572 (after MacFadden et al., 2007). drainage: (1) the confluence or relative closeness of the northern reaches of the Flint River (a major tributary of the Apalachicola River) with the Withlacoochee and Alapaha rivers (tributaries of the Suwannee River) in Southern Georgia during the Pleistocene; (2) the confluence of river drainages (including the Apalachicola and Suwannee rivers) during a lower sea level event. Either of these two scenarios could have allowed the potential for G. barbouri, G. kerneri, or the common ancestor to invade the Suwannee River drainage without traversing land. The first hypothesis requires the confluence or at least a weak connection between the Flint, Withlacoochee, and/or Alapaha rivers in Georgia during the Pleistocene. However, analysis of the Okefenokee Swamp and the northern Suwannee River in Georgia suggests that this connection is a relatively young event ( 6700 BP) (Schmidt, 1997; Wright et al., 2005). Although Schmidt (1997) alludes to an ancestral river north of the Okefenokee that later connected with the Suwannee, the timing of this event is not congruent with the fossil record of G. kerneri in the Santa Fe River at 15 ka. Therefore, we reject this hypothesis as a potential distributional pathway. A second hypothesis proposes that confluence of river drainages off of the coast of present day Florida during a period of lower sea level stand could have produced a pathway for the spreadof Graptemys into the Suwannee River drainage (Jackson, 1975). This hypothesis seems more feasible given the geologic history of the Gulf coast of Florida, the ages, and distribution of the fossils. Although it is difficult to trace Florida s shoreline and the meander of its rivers out into Gulf of Mexico during these lower stands, such as the Wisconsinan stage ( 20,000 BP), sea level exceeded 100 m below that of today (Webb, 1990). During one of these lower sea level stands, it is possible that two or more of the river drainages along the Gulf coast would have met or nearly so, allowing Graptemys to enter new river drainages. Additionally, late Pleistocene and early Holocene records of the Suwannee River delta show that the mouth has meandered both northward and southward in recent past (Wright et al., 2005). Therefore, G. barbouri, G. kerneri, or a related species could have entered the Suwannee River, and moved up into the Santa Fe, Aucilla, and Waccasassa rivers during a glacial period and then been isolated from the Apalachicola population as the sea level rose during an interglacial period.