Abstract: The Jurassic was an important period in the evolution of Testudinata and encompasses the origin of many clades, and this is especially true of Jurassic turtles from Western Europe. A new genus and species of Late Jurassic turtle, Hispaniachelys prebetica gen. et sp. nov. from the upper Oxfordian of the Prebetic (Southern Spain), is described on the basis of postcranial material. The specimen is the only known tetrapod from the Mesozoic of the Prebetic and the oldest turtle from southern Europe. A mosaic of characters indicates this is a new genus: it displays basal features including dorsal epiplastral processes/reduced cleithra, no medial contact of the extragulars and a long first thoracic rib, alongside derived characters including an absence of mesoplastra and the vertebral 3/4 sulcus crossing neural 5. The phylogenetic position of the new taxon is hard to resolve, and it might be either a paracryptodire or a basal testudine, but it is distinct from Plesiochelys. A complex taphonomic history is shown by a range of overlying grazing traces and bioerosion on the carapace. The carapace was subsequently overturned and buried ventrally up, terminating grazing activity, and was then bored by sponges before final burial. Scanning electron microscopy reveals phosphatic microspheroids associated with bacterial decay in the vascular cavities of the cancellous bone, suggesting the carapace may have acted as a closed microenvironment in which decay-derived authigenic minerals formed.
The earliest known turtle is Odontochelys semitestacea from the Late Triassic of Guizhou, China, which is 220 myr old (Li et al. 2008), some 10 myr older than Proganochelys quenstedti from Germany, previously regarded as the basal sister group to all other turtles (Gaffney and Meylan 1988; Gaffney 1990; Joyce 2007). Modern turtles are divided into two clades, Pleurodira and Cryptodira, based on the manner in which they withdraw their head and neck into the shell: pleurodires retract the head by making a horizontal ‘S’ bend in the neck, whereas cryptodires make a vertical bend. These neck-folding modes are hard to determine in early fossil turtles, and the assignment of some Jurassic taxa has been debated. For example, Kayentachelys aprix from the Early Jurassic (190 Ma) was previously thought to be the earliest cryptodire (Gaffney et al. 1987), but more recent studies (Joyce 2007) have suggested it is a more basal turtle and that Pancryptodira, the clade including modern cryptodires and their stem taxa, did not diverge until the Middle Jurassic (Danilov and Parham 2006, 2008; Sterli 2008).
Complete turtle specimens are rare in the Jurassic (Joyce 2000), being confined usually to fragmentary remains of either the skull or shell in isolation. Much previous work on turtles has therefore focused on identifying phylogenetic relationships through cranial (e.g. Gaffney 1975a, 1984) or shell characters (Joyce 2003). The skull displays wide disparity between species and contains many diagnostic characters (Gaffney 1979), but it is often not preserved, or only in isolation from carapace and plastron elements. Joyce (2007) provided a comprehensive character matrix of both cranial and postcranial characters, helpful here in that our specimen consists primarily of dermal elements of the shell, with some isolated vertebrae.
Late Jurassic turtles are distributed widely across Europe, with specimens from Britain, Germany, France and Switzerland (Lapparent de Broin 2001). The majority of these have been assigned to the families Plesiochelyidae, Eurysternidae and Thalassemydidae, but the distinctiveness of these families has been questioned in recent phylogenetic analyses (Joyce 2007). Late Jurassic turtles from the Iberian microcontinent include those from the Lourinhã Formation (Kimmedigian–Tithonian) of western Portugal (Antunes et al. 1988) and trackways from the Late Jurassic Vega, Tereñes, and Lastres formations (Kimmeridgian) of Asturias, Spain (Avanzini et al. 2005). Turtles are poorly represented in the Oxfordian (Bardet 1995), with only an incomplete Plesiochelys specimen from the Oxfordian of Germany (Kuhn 1949). Many fossil turtles are known from the Oxford Clay of England, but this is largely Callovian in age (Martill and Hudson 1991). Outside Europe the Oxfordian turtle Caribemys oxfordiensis has been found in Cuba (De la Fuente and Iturralde-Vinent 2001). Considering their paucity in the Oxfordian, the discovery of a new turtle of this age from Spain is significant.
The aim of this paper is to describe the new specimen in detail, to identify it and determine its relationships, to present information on its bone histology and taphonomy and to discuss its palaeoecology.
Geographical and geological setting
The specimen was found by Matías Reolid (MR) in bed 62 (upper Oxfordian) of the Riogazas-Chorro (RGCH) section of the Lorente Formation of the External Prebetic, the northernmost part of the Betic Cordillera in southeast Spain (Text-fig. 1). Bed 62 is within a 22-m sequence of well-stratified limestone and marl rhythmites dominated by limestones, which make up the Middle Oxfordian to Lower Kimmeridgian age Lorente Formation (Olóriz et al. 2002, 2003). The ammonite biostratigraphy indicates that the turtle occurs in the Bimammatum Biozone (Olóriz et al. 1999). The sediments were deposited in a mid-shelf environment on a carbonate epicontinental shelf system. The marl-limestone rhythmites represent fluctuations in carbonate productivity and delivery of siliciclastic material derived from the proximal Iberian palaeomargin, which is defined by the extent of Iberian Massif tabular sediments immediately to the north (Olóriz et al. 2003, 2006; Reolid et al. 2010). The RGCH section has been heavily studied for its rich invertebrate fauna (Olóriz et al. 2002, 2003, 2006; Reolid 2003, 2008; Reolid et al. 2008), but vertebrates have not so far been reported.
For scanning electron microscopy (SEM), fragments of carapace and indeterminate fossil bones were mounted and coated with gold and carbon. Secondary electron images were made to study internal ultrastructure and crystal morphology, and back-scattered electron (BSE) imaging and energy-dispersive X-ray (EDX) analyses to obtain textural and chemical data. These analyses were performed using the FEI Quanta 400 at the Centro Andaluz de Medio Ambiente (CEAMA, Granada).
Derivation of name. Genus after the Latin for Spain and Greek for turtle, species after the Prebetic of Andalucía (Spain), where the specimen was found.
Holotype. RGCHSP-62-52 (Text-fig. 2), housed in the Museo de Paleontología of the Universidad de Granada, Spain.
Type stratum and locality. Bed 62 of the Lorente Formation; Bimammatum Ammonite Biozone, Oxfordian, Late Jurassic; Sierra de Cazorla: Riogazas-Chorro-Sponges section, Prebetic, Betic Cordillera, western Andalucía, east of Jaén, southeast Spain.
Diagnosis. Moderately large turtle with a thick fully ossified carapace; eight hexagonal neurals; eight costals; trapezoidal nuchal; a dorsal epiplastral process (cleithra); diamond shaped entoplastron that does not separate the anterior portions of the epiplastra; single cervical scute covering half of the nuchal; vertebral 3–4 sulcus crossing neural 5; anteroposteriorly narrow pectoral scutes; a partially sinusoidal pectoral/humeral sulcus; anal scute confined to xiphiplastron.
Hispaniachelys prebetica gen. et sp. nov. is a moderately large turtle with a thick carapace, similar in its oval outline and thickness to the specimen described as Plesiochelys solodurensis by Bräm (1965). H.prebetica differs from P. solodurensis and all other Plesiochelys in its possession of a dorsal epiplastral process. The presence of this basal character means the new turtle can therefore be excluded from the ‘Plesiochelyidae’. It has previously been demonstrated that the dorsal epiplastral process may be a remnant of the cleithrum (Joyce et al. 2006). H. prebetica shares many characters with Eileanchelys waldmani from the Bathonian (Anquetin et al. 2008), such as a dorsal epiplastral process (reduced cleithra), a trapezoidal nuchal, an osseous connection of the plastron to the carapace, an entoplastron that does not divide the epiplastra anteriorly and an anal scute that does not extend to the hypoplastron. H. prebetica differs from E. waldmani, however, in lacking mesoplastra. Additionally, the sulcus between vertebrals III and IV crosses neural 6 in E. waldmani and another stem turtle, Condorchelys antiqua (Sterli 2008), rather than neural 5, as in H. prebetica. The basal taxon Kayentachelys aprix also shows features present in H. prebetica, including dorsal epiplastral processes/cleithra (Joyce et al. 2006) and a single, wide but narrow cervical. H. prebetica differs from K. aprix and other taxa such as Glyptops plicatulus because it lacks mesoplastra.
General preservation. The specimen comprises solely postcranial elements, including the majority of the carapace and plastron (Text-fig. 2), some disarticulated cervical and thoracic vertebrae, and some of the internal bones of the shell such as the dorsal vertebrae, and the scapula. Elements of the skeleton that protrude from the carapace such as the skull, tail and limbs are missing, as is commonly the case in fossil turtles (Meyer 1991). The posterior elements of the carapace are also disarticulated or missing, including much of the pygal region. The carapace itself is preserved in three dimensions and therefore displays its original curvature with no signs of flattening. The right-lateral peripherals are disarticulated from the carapace, although some are preserved in isolation. The plastron collapsed into the carapace, as the specimen was preserved with the ventral surface facing upwards, and consequently, the plastron is broken up but still relatively well preserved. A number of isolated fragments of the carapace are also preserved.
The carapace (Text-figs 2, 3A) is 430 mm long in the midline, 340 mm wide at right angles through peripheral V, and 175 mm deep at the dorsoventral axis, the highest point. The carapace is significantly domed at the anterior end, but becomes progressively flattened in the posterior half. The preserved elements of the carapace include the nuchal, 12 of the 16 costals, three of which are partially preserved, neurals 1, 3, 4, 5, 6, 7 and 8, and part of the first suprapygal. Sixteen of the 22 peripherals are preserved. The carapacial plates have a relatively even and continuous thickness of c. 9 mm. The outline of the carapace is oval, as in Plesiochelys solodurensis (Rütimeyer 1873; Bräm 1965). Carapace elements lack sculpture and keels or other external ornament, as is the case in most other turtles (Joyce 2007). The smooth carapace separates the specimen from Platychelys oberndorferi (Wagner 1853; Bräm 1965), which exhibits angular sculpturing of the carapace.
Nuchal. The nuchal bone is poorly preserved as its dorsal surface is heavily weathered and pitted, and the right anterolateral margin is damaged by bioerosion, but the outline is clear. The nuchal is subrectangular, with a partially convex anterior margin, shallow trapezoidal posterior margin and linear lateral edges, similar to that seen in the Plesiochelyidae and Eurysternidae (Joyce 2003). The posterior margin is serrated and interlocks with the first costals and first neural. The nuchal is domed and does not become depressed at the anterior end, in contrast to Xinjiangchelys latimarginalis and Solnhofia parsonsi. The carapacial morphology of Eileanchelys waldmani is domed (Anquetin et al. 2008), though all reported specimens are variably fragmented and compressed. A trapezoidal nuchal is also present in Eileanchelys waldmani (Anquetin et al. 2008). The first thoracic vertebra is preserved on the underside of the nuchal.
Neurals. The specimen possesses eight neural plates, the first of which is partially preserved but can be seen to be subrectangular, with two small ‘V’-shaped notches on the anterior margin that interdigitate with the posterior margin of the nuchal. The second neural is not preserved, and only its length can be inferred through the border with the costal. Neurals III–VII are of a distinctive hexagonal shape with short anterior sides, similar to those of Solnhofia parsonsi (Joyce 2000) and Plesiochelys solodurensis (Bräm 1965), with a relatively narrow anterolateral edge that enlarges increasingly in more posterior neurals (Text-fig. 3A). Neurals II–VII also decrease in length towards the posterior. In contrast to P. solodurensis, in which the longest neurals occur in the centre (Oertel 1924; Bräm 1965), neural IV in this specimen is shorter than neurals III and V. Neural VIII is the smallest in total area, being wider than it is long and asymmetrical in form. The shape of the eighth neural is variable in some species of turtles, and this may therefore be a developmental feature (Peng and Brinkman 1993). In Plesiochelys and ‘Craspedochelys’picteti (Rütimeyer 1873; Bräm 1965; considered a junior synonym of Plesiochelys by Gaffney 1975b), the position of the vertebral IV/V sulcus is variable (Bräm 1965). The dorsal row of neurals has no gaps, which prevents medial contact of the costals, as in most basal eucryptodires (Joyce 2007). The neurals differ from those of Tropidemys langi (Rütimeyer, 1873; Bräm 1965) because they lack the distinct ridge that runs the length of the neurals. The neurals of Eileanchelys waldmani are badly preserved in comparison to Hispaniachelys prebetica, and it is not possible to discern whether it possessed 8 or 9 neurals (Anquetin et al. 2008).
Peripherals. There are 11 pairs of peripherals, an unequivocal synapomorphy of Testudines and some basal clades (Joyce 2007). The anteriormost peripherals are well preserved, with obvious sulci portraying the arrangement of overlapping marginal scutes, each of which overlaps two peripherals. The peripherals are all sutured to the costals (Text-figs 2, 3A). The first peripheral is elongate and contacts a large portion of the first costal. This is the condition (Peng and Brinkman 1993) in Plesiochelys, Solnhofia parsonsi and Kayentachelys aprix, whereas more derived taxa such as Xinjiangchelys latimarginalis have a shortened first peripheral. The second peripheral is shorter than the first and contacts both the first and second costals. The second peripheral also has a serrated protrusion at its dorso-posterior edge that intersects between the first and second costals, forming a ‘V’-shaped triangular division between the peripheral margins of the two costals. Peripheral III is proportionally longer than in P. solodurensis, with a serrated contact to peripheral IV, and contacting only costal II. Peripheral IV does not contact the third costal as in P. solodurensis, only bordering costal II, which is also the case in Kayentachelys aprix. Peripheral V is only sutured to costal III. Peripheral VI contacts both costals III and IV. Peripheral VII is shorter and is fused to both costals IV and V. Peripheral VIII has a long contact with costal V and a relatively short contact with costal VI. The suture between peripherals VIII and IX is partly sinuous. Peripheral X is different in that it intersects deeply between costals VII and VIII, giving it a concave lateral margin with costal VIII. The margin of peripheral X is also elongate, meaning that the suture between peripherals X and XI is curved. Peripheral XI contacts costal VIII and the first suprapygal, and the posterior suture to peripheral XI is deeply serrated. Peripheral XI would contact the second suprapygal and the pygal in life, but these are not preserved. Several disarticulated peripherals (Text-fig. 4) are bridge peripherals from the right-hand side and show a triangular pit with a rugose surface on their internal side (Text-fig. 4D, G), which presumably accommodated the osseous connection between carapace and plastron. Similar pits were also described in Xinjiangchelys chowi (Matzke et al. 2005), though these were interpreted as sockets for pegs that protrude from the plastron into the peripherals, which is a more derived feature (Tong et al. 2002). A small tunnel structure extends from the pits on the internal surface of the bridge peripherals (Text-fig. 4E–H), which likely housed a blood vessel in life. The internal morphology of the bridge peripherals is often not described in turtle fossils, as it is generally not visible. All preserved peripherals are longer than they are wide.
Costals. The presence of eight costal bones can be inferred, though the eighth is only partially preserved. The anterior costals are deeply arched, whereas posterior costals are flattened, giving the carapace a lower region towards the rear. Costal I is similar to that of Plesiochelys solodurensis, although the sutures to the nuchal and peripherals I and II are less linear with more protrusions. Costal II has a shorter contact with the neural than the peripherals, matching its laterally flared shape (Text-fig. 3A). Costals III–VII are perpendicular to the longitudinal axis and possess relatively parallel borders. Costal VIII is narrower, giving it a wedge-shaped appearance, and it has a rounded lateral margin with peripheral X rather than the triangular peripheral-costal contacts of the other costals. There are no carapacial fontanelles, as in Plesiochelys, but differing from other Late Jurassic turtles such as Eurysternum wagleri (Zittel 1877), Idiochelys fitzingeri (Meyer 1839; Rütimeyer 1873), Solnhofia parsonsi (Joyce 2000) and Thalassemys hugii (Bräm 1965) which all have carapacial fontanelles. Where visible in ventral view, the costals have pronounced rib heads (Text-fig. 5), which shallow rapidly towards the centre of the costal.
Sulci are well developed across the entire carapace in RGCHSP-62-52, although bioerosion has removed or obscured their configuration in some places, such as the right side (Plate 1F).
Cervical scute. RGCHSP-62-52 shows only one cervical scute, a character that excludes the specimen from the Plesiochelyidae (Bräm 1965; Lapparent de Broin et al. 1996; Joyce 2003, 2007), all of which possess three cervical scutes. The cervical scute is close to rectangular in shape, differing from the trapezoidal cervical scute of Solnhofia parsonsi (Joyce 2000), and from that of Idiochelys fitzingeri (Meyer 1839; Rütimeyer 1873), which is wider than the entire nuchal.
Vertebral scutes. The vertebral scutes are hexagonal in form, as revealed by the prominent sulci (Text-fig. 3A), similar in shape but narrower than those of P. solodurensis, being more akin to the hexagonal vertebrals of Eurysternum wagleri (Zittel 1877), Palaeomedusa testa (Meyer 1860; Joyce 2003) and Idiochelys fitzingeri (Meyer 1839; Rütimeyer 1873). Tropidemys langi (Rütimeyer 1873; Bräm 1965) also possesses narrow vertebral scutes, but these are somewhat more rectangular. Consequently, the four pleural scutes on either side are wide. Nonetheless, the vertebral scutes are wider than long, the primitive condition (Peng and Brinkman 1993). The anterior and posterior margins of the vertebral and pleural scutes run parallel to those of the costal bones. The vertebral III–IV sulcus is on neural V, differing from both Eileanchelys waldmani and Condorchelys antiqua where it is on neural VI. Anterolaterally, the margins of the first vertebral scute contact the cervical scute and the first and second marginals (Text-fig. 3A).
Pleural scutes. The pleural scutes (Text-fig. 3A) are proportionally wider than in Plesiochelys solodurensis, more akin to the extension over a large area of the costals as seen in Tropidemys langi (Rütimeyer 1873; Bräm 1965). All pleurals have a margin with two vertebrals and three marginal scutes, aside from the first pleural which contacts marginals II–V (as well as a small supramarginal scute) and has no border with the first marginal.
Marginal scutes. There are 12 pairs of marginal scutes (Text-fig. 3A) that do not encroach onto the surface of the costals, as in Plesiochelys and Tropidemys langi. A small supernumerary submarginal scute is visible on the left side between marginals III and IV and the first pleural. This could be an anomaly particular to this specimen, as it cannot be checked on the right side. Zangerl (1957) reported that such abnormalities are common in extant turtles, and a similar submarginal is reported from Solnhofia parsonsi (Joyce 2007).
The individual elements of the plastron are well preserved, but they have collapsed after fossilization. The original form (Text-figs 2B, 3B) can be discerned, however, as the margins of each element of the plastron and the buttresses are visible. The plastron is highly ossified, as in Plesiochelys (Bräm 1965; Joyce 2000), lacking any bridge fontanelles. The anterior plastral lobe differs from that of P. solodurensis in being more elongate. The bridge is longer than the anterior or posterior plastral lobes (Text-fig. 3B), as in Plesiochelys. The absence of mesoplastra is a synapomorphy of the Eucryptodira (Gaffney and Meylan 1988). Musk ducts in the plastron, known from Siamochelys peninsularis, are unknown in H.prebetica. A central plastral fontanelle is present, although its apparent size is enhanced by damage. The central plastral fontanelle is exhibited by specimens described as P. ‘etalloni’ (Oertel 1924; Bräm 1965), but not by P. solodurensis, which exhibits a fully ossified plastron. Plastral fontanelles may be a juvenile feature (Joyce 2007), although the features and proportions of this specimen suggest it is an adult.
Epiplastron. The anterior portions of the epiplastra meet each other at the midline, and the entoplastron and hyoplastron posteriorly (Text-figs 3B, 6A, B). The epiplastra also have shallow notches at their anterior edges centred upon the sulci of the gular scutes (Text-fig. 3B), but these are not anterior plastral tuberosities that give the anterior plastral rim a sculptured margin in the basal Proganochelys quenstedti and Proterochersis robusta. A dorsal epiplastral process (Text-fig. 6B) extends from the middle of the epiplastron and is concave on the anterior surface. A dorsal epiplastral process is a primitive feature seen in a larger form than that described here in Proganochelys quenstedti (Baur 1887; Gaffney 1990), Proterochersis robusta (Fraas 1913) and Palaeochersis talampayensisRougier et al. 1995, and in a reduced form equivalent to that seen in our specimen in Eileanchelys waldmani (Anquetin et al. 2008), Kayentachelys aprix (Gaffney et al. 1987; Joyce et al. 2006), Xinjiangchelys latimarginalis (Peng and Brinkman 1993), Meiolania platyceps (Owen 1886; Gaffnet 1996), Mongolochelys efremovi (Khosatzky 1997) and Glyptops plicatulus (Cope, 1877). Joyce et al. (2006) demonstrate that the dorsal epiplastral processes actually represent the cleithra, as was suspected in earlier work (Jaekel 1915), but then reinterpreted as dorsal outgrowths emanating from the epiplastra (Gaffney 1990).
Entoplastron. One lateral half of the entoplastron is preserved, and its overall shape can be discerned from this and the opposing gap in the hyoplastron. The entoplastron is smaller than the epiplastron and forms a small diamond shape with slightly curved anterior margins, being sutured to both epiplastra at the anterior margin and to the hyoplastra at the posterior margins (Text-figs 3B, 6A). The anterior extension of the entoplastron forms an arrow shape between the epiplastra similar to that described in Eileanchelys waldmani (Anquetin et al. 2008).
Hyoplastron. The anterior elements can be identified as the hyoplastron in a slightly displaced position because of the position of the pectoral/humeral sulcus, which runs anterior to the axillary notch (femur hole) and the attachment of the buttress to the peripherals. The midline edge of the hyoplastron is lined by a short row of processes that interdigitate with the opposite element. The axillary buttresses of the hyoplastron (Text-fig. 6C) are strongly developed and contact the internal surface of peripherals II and III and extend anteriorly and dorsally to contact the visceral surface of the first and second costals (Text-figs 3B, 6C). This differs from Plesiochelys solodurensis and the specimen described as P.‘etalloni’ by Bräm (1965), in which the axillary buttress only reaches peripheral III. The anterior extension of the buttresses to peripheral II is seen in Xinjiangchelys latimarginalis (Peng and Brinkman 1993) and Solnhofia parsonsi (Joyce 2000). The hyoplastron is sutured to peripherals II, III, IV and V. The internal surface of isolated bridge peripherals can be seen in Text-figure 4, displaying holes that accommodated the connection from the margins of the hyoplastron.
Hypoplastron. The abdominal/femoral sulcus runs into the inguinal notch. The inguinal buttresses attaching the hypoplastron to the carapace run from the very anterior of peripheral V, along peripherals VI, VII and VIII (Text-figs 3B, 6G, F). The inguinal buttress also contacts costal VI; Kayentachelys aprix and Plesiochelys classically possess large buttresses (Gaffney et al. 1987; Gaffney and Meylan 1988), which provide the shell with support by joining the plastron to the peripherals (Joyce 2007).
Xiphiplastron. The xiphiplastra (Text-figs 2, 3B, 6H) are similar in outline to those of the specimen described as ‘Plesiochelys hannoverana’ by Oertel (1924), with the suture to the hypoplastra sloping anterolaterally, rather than horizontally as in P. solodurensis. The suture to the hypoplastra is heavily serrated, possessing a rounded interdigitation at the midline and a sub-triangular interdigitation at the anterolateral edge.
Plastral scutes. There is a single row of inframarginal scutes (Text-fig. 3B) laterally narrower than those of Plesiochelys solodurensis. The full sequence of inframarginal scutes is unclear because some of the hypoplastron is obscured (Text-fig. 2). The first inframarginal overlaps peripherals III and IV and the hyoplastron, contacting marginals III and IV, the pectoral scute and the second inframarginal at the posterior margin. Inframarginal II contacts the pectoral scute and marginal V and is approximately square. Inframarginal III is an elongate version of the preceding one, contacting the pectoral, marginals V and VI and the abdominal scute at the anterior margin. Inframarginal IV is an elongate rectangle that is laterally thin, covering peripherals V, VI and VIII, but it is uncertain whether there is another inframarginal between this and the abdominal scute, as the area is partially obscured.
The gular scutes (Text-fig. 3B) cover the anterior portion of the epiplastra and together form a ‘V’ shape that meets posteriorly at the margin between the epiplastra and entoplastron. The extragulars lay either side of the gulars with their posterior margins perpendicular to the midline plastral sulcus. There is no medial contact of the extragulars, as in Kayentachelys aprix. Both the gulars and extragulars are confined to the epiplastra, with no extension onto the entoplastron.
The humeral scutes collectively cover the entoplastron, the posterior part of the epiplastron and the anterior plastral lobe of the hyoplastron (Text-fig. 3B). The humerals are elongate anteroposteriorly in comparison to those of Plesiochelys.
The pectorals are short and broad. The pectoral/abdominal sulcus is linear and runs parallel to the hyoplastron/hypoplastron suture (Text-fig. 3B).
The abdominal scutes are similar to those of Solnhofia parsonsi and P. solodurensis, but the abdominal/femoral sulcus bisects the plastral midline further behind (Text-fig. 3B). The abdominals appear to extend onto peripherals VI and VII, but there may have been an additional inframarginal scute.
The femorals are positioned further back than those of Solnhofia parsonsi and P. solodurensis, but in other respects are identical. The posterior midline femoral sulcus is partly sinusoidal and terminates at the anterior margin of the xiphiplastra.
The anal scutes only overlay the xiphiplastra and do not extend onto the hypoplastron, as in Eileanchelys waldmani (Anquetin et al. 2008) and Plesiochelys solodurensis.
Vertebrae. There is an isolated partly damaged cervical centrum (Text-fig. 7A–E, K–O), and a partly preserved cervical (Text-fig. 7V–A1). The isolated cervical centrum, possibly the fifth, shows a strongly developed ventral keel and a single transverse process. It is short and wide, with stout pre- and postzygapophyses. The centrum is amphicoelous and ventrally keeled, as in Proganochelys, Xinjiangchelys and Plesiochelys, and has a single short transverse process over the middle of the centrum. Cervical ribs are not preserved. The neural canal is large and circular, while the neural arch is relatively short and extends almost the length of the centrum but can only be observed in lateral view as matrix obscures anterior and posterior views. The diapophysis and parapophysis are separate, a primitive cryptodiran feature (Gaffney and Meylan 1988; Peng and Brinkman 1993). The first thoracic vertebra (Text-figs 7F–J, P–T; 8E, F) is partially preserved, the first thoracic rib being very long. The second thoracic vertebra (Text-fig. 8A–D) has large parapophyses, and a dorsoventrally elongate neural arch. Where visible, the sequential dorsal vertebrae are similar in appearance and bordered by strongly developed rib heads. The centrum of the second thoracic is partly covered by matrix, which obscures some of the surface detail. The third and fourth dorsal vertebrae are also partly visible in dorsal view where the carapace is damaged.
Ribs. The first thoracic rib (Text-fig. 8) extends almost the length of the first costal, being very elongate. The second thoracic rib almost reaches the axillary buttress.
Scapula. There is a partially preserved right scapula (Text-fig. 4M), which shows a damaged portion of a somewhat flattened acromial process.
Manus. Some isolated fragments of bone are preserved which are possibly part of a phalange and a carpal, but they are too damaged to be conclusive.
At one time, the Spanish specimen might have been assigned to the family ‘Plesiochelyidae’, together with many other apparently aquatic Late Jurassic forms, and so placed unequivocally within Eucryptodira, as one of several basal outgroups to Cryptodira. However, ‘Plesiochelyidae’ turns out to be an ill-defined ragbag of species in need of revision, it has proved hard to retrieve in phylogenetic analyses, and it is regarded as paraphyletic by Joyce (2007). Further, the classic assumption that cryptodires and pleurodires are distinct clades among turtles that both originated in the Early Jurassic has not been confirmed by more recent phylogenetic analyses (e.g. Joyce 2007; Anquetin et al. 2008), and many Jurassic taxa have proved hard to resolve in the cladogram.
In view of these difficulties, we performed a cladistic analysis to determine the phylogenetic position of the Spanish taxon. We used the cladistic data matrix from Joyce (2007, pp. 80–95), with the addition of the three taxa noted by Anquetin et al. (2008), namely Eileanchelys waldmani, Heckerochelys romani and Condorchelys antiqua, as well as the three additional characters they added. We excluded the three ‘rogue’ taxa identified by Joyce (2007) and Anquetin et al. (2008), namely Portlandemys mcdowelli, Sandownia harrisi and Mongolemys elegans. We coded Hispaniachelys according to those 139 characters (Table 1) and ran a cladistic analysis, in PAUP (Swofford 2002) using a heuristic search and a tree-bisection algorithm with 5,000 replicates.
Table 1. Character codings for Hispaniachelys prebetica gen et sp. nov.
Codings according to the data matrix (136 characters) of Joyce (2007), as modified by Anquetin et al. (2008) by the addition of three further characters (139 characters).
The phylogenetic analysis (Text-fig. 9) gave mixed results, confirming the difficulty of achieving strong resolution in parts of the tree, as reported earlier by Joyce (2007) and Anquetin et al. (2008). The Adams and 50 per cent majority rule consensus trees (Text-fig. 9A, B) differentiated the major clades in Joyce’s (2007, p. 62)‘preferred’ tree, but with less resolution. Pleurodira, Paracryptodira, Eucryptodira and Cryptodira are distinguished, but not always with the exact composition noted by Joyce (2007). For example, in the Adams and majority rule consensus trees (Text-fig. 9A, B), the Pleurodira encompasses Platychelys, Caribemys and Notoemys, whereas these are close outgroups of Pleurodira in Joyce’s (2007) preferred solution, although his consensus tree shows similar issues. The bootstrapped strict consensus tree (Text-fig. 9C), however, identified the components of Pleurodira correctly. On the other hand, Paracryptodira is retrieved in the Adams consensus tree (Text-fig. 9A), but not in the 50 per cent majority or strict trees (Text-fig. 9B, C) in which the two paracryptodire subclades, the Pleurosternidae and Baenidae, are separated as parts of an unresolved polytomy. The major clades Eucryptodira and Cryptodira are identified in the first two consensus trees (Text-fig. 9A, B), but they are lost when only bootstraps over 50 per cent are retained (Text-fig. 9C).
In the light of the incomplete resolution of several major clades, as noted earlier by Joyce (2007) and Anquetin et al. (2008), it is hard to be entirely confident about the placement of a new taxon. According to the three consensus trees, Hispaniachelys prebetica appears at first to occupy very different phylogenetic positions: in the Adams consensus tree (Text-fig. 9A), it is part of a basal tritomy among Testudines, neither definitively a pleurodire/paracryptodire, nor a eucryptodire; in the 50 per cent majority rule consensus tree (Text-fig. 9B), it is a paracryptodire, associated with the baenid clade; and in the strict consensus tree (Text-fig. 9C), it is one of a large number of unresolved testudines that are not assigned to any major subclade. One observation is that, although the new Spanish turtle might well have been named as a species of Plesiochelys, this genus falls definitively at the base of Eucryptodira according to the first two consensus trees (Text-fig. 9A, B), and thus, it is distant phylogenetically from Hispaniachelys.
In terms of character evolution, Hispaniachelys could be coded for 46 of the 139 characters (33 per cent coding), and of these coded characters, only 15 (11 per cent) are in a derived state. These are sufficient to determine that Hispaniachelys is a testudine. Comparison with Joyce’s (2007, fig. 18) preferred phylogeny confirms that Hispaniachelys belongs within all the most inclusive clades in succession. It is a member of Testudinata (Joyce’s node 1) based on the carapace, plastron and other classic turtle apomorphies, but cannot be assessed at the Palaeochersis node (node 2) because diagnostic apomorphies are all cranial. It is an unequivocal member of the Proterochersis + Testudines clade (node 3) based on derived states of characters 71 (reduction in number of supramarginals), 72 (five or more vertebrals) and 80 (absence of posterolateral entoplastral process), and of the Kayentachelys + Testudines clade (node 4) based on derived states of characters 65 (11 pairs of peripherals), 71 (supramarginals absent), 94 (anterior plastral tuberosities absent) and 120 (cleithra reduced and with no osseous contact with carapace). Hispaniachelys is not a member of the Mongolochelys-Meiolania clade (node 5), but does belong to the Mongolochelys + Testudines clade (node 6), based on the derived state of character 78 (medial contact of epiplastra), to the Kallokibotion + Testudines clade (node 7), based on derived states of characters 84 (axillary buttresses contact peripherals and first costal) and 86 (axillary buttresses contact peripherals and costal V), and to Testudines (node 8), based on derived states of characters 62 (articulation of nuchal with eighth cervical vertebra absent) and 79 (reduced posterior entoplastral process). Each of these nodes is associated with other apomorphies, but those could not be coded on our material, and there are no contradictory codings in Hispaniachelys.
Within Testudines, the situation is less clear, largely because the codable characters themselves have equivocal evolutionary histories, according to Joyce’s (2007) phylogenetic analysis, and we lack evidence in the material of Hispaniachelys to assess the unequivocally distributed cranial and postcranial characters. For example, characters 73 (narrow vertebrals) and 74 (vertebral II–III sulcus shifts from neural VI to V) provide some evidence for assignment to Pleurodira or Pancryptodira, but cannot indicate a particular clade assignment. Character 85 (mesoplastra absent) is an apomorphy of Pancryptodira (but also Cryptodira and Chelidae), and 113 (first dorsal rib intermediate in length) is an apomorphy of several pleurodire and cryptodire clades. Hispaniachelys does not appear to be a member of the Platychelys clade (node 9) because it lacks derived states of characters 87, 91 and 100, even though it equivocally shows characters 76 and 85. Most apomorphies of Pleurodira (node 12) are cranial, and Hispaniachelys lacks the derived state of character 68, but shows 73, 74, 76 (?) and 120, but these can also apply to other nodes, as noted. Assignment to certain key clades cannot be assessed because relevant characters cannot be coded: Paracryptodira (node 15), Eucryptodira (node 18) and the Santanachelys clade (node 20). Hispaniachelys could belong to the Solnhofia clade (node 19), based on possession of the derived state of characters 84 and 86, but it lacks 75, and the Xinjiangchelys clade (node 21) based on the possession of the derived state of characters 73 and 113, but it lacks 99, but not to the Hangaiemys clade (node 22) because it lacks the derived states of characters 83 and 92, or to the Cryptodira (node 25) because the cleithra are reduced, but not absent (character 120(2)).
More material of the vertebrae, limbs and especially of the skull of Hispaniachelys would assist in placing this taxon phylogenetically. However, as noted by previous authors (Joyce 2007; Anquetin et al. 2008), much more work has yet to be carried out on all Jurassic turtles, and further unequivocal apomorphies have yet to be discovered, before the phylogeny becomes clearer.
Taphonomy and palaeoecology
Preservation and bone histology
The specimen is largely intact, with most of the elements of the carapace in original position. When found, the specimen was overturned. The degree of fragmentation is low and appears largely to affect the distal parts of some bones, which are fractured and not rounded. Other components display some fragmentation and disarticulation but are not significantly displaced (e.g. carapace plates). The plastron has collapsed into the cavity of the carapace. The external surface of the carapace displays traces of grazing composed of radial grooves, isolated circular borings (<5 mm) and small interlaced boring fields resembling Entobia, which are created by endolithic sponges (Farinati and Zavala 2002). The plastron and other bones do not have trace marks on their surface.
Thin sections of the bone (Plate 1A–D) show details from a fragment of the carapace. The cortical bone layers (Plate 1D) are separated into an inner and outer layer, which surround the porous cancellous region (Plate 1B). The porous cancellous bone can also be seen under the SEM (Plate 2F). The external cortical region comprises densely compacted regular layers (Plate 1C). The collagen fibril bundles and thin vascular structures of the external cortical layer are well preserved and can be made out in thin section and under the SEM (Plate 2A, B, E). Towards the periphery of the external cortex, the collagen fibril bundles become more densely packed and increasingly interwoven. Growth layers can be seen as distinct parallel lines particularly in the outer half of the external cortex and are wavy and less distinct in the internal half (Plate 1A). The number of growth lines cannot be determined because of weathering and pitting of the outer surface, regrettable because these have been used in modern turtles to determine individual age (Wilson and Tracy 2003). Vascular structures are confined exclusively to the inner half of the cortical layers and their abundance obscures any layering, although they are less abundant than in the porous cancellous region. Large primary vascular canals are scattered throughout the inner half of the cortical layers. The internal cortical layer is like a reversed external layer in appearance, the only significant difference being the absence of oxidation or weathering on the outer surface. In the cancellous region, vascular cavities average 0.1 to 0.2 mm in diameter and constitute c. 75 per cent of the total area in cross section. The vascular spaces become increasingly flattened and elongated with increasing proximity to the cortical bone.
An SEM EDX spectrum analysis (not illustrated) showed that the remains are preserved as calcium phosphate (francolite). Aligned francolite crystal bundles are visible within the bone (Plate 2E), perhaps pseudomorphs of the original collagen fibrils of the bone, revealing a linear structure to the cortical region. Further, filaments (<5 μm in diameter; Plate 2A, B) and globular structures (10–50 μm in diameter; Plate 2C, D) composed of authigenic francolite are also present in the cancellous region. Some surfaces of the bones are densely covered by microspheres (<1 μm; Plate 2B, C; Schmitz and Ernst 1994).
The specimen was probably overturned before burial, as shown by its original orientation when collected and the plastron fragments filling the carapacial bowl. This presumably explains the absence of borings and traces on the plastron, which would have been buried and so sheltered from borers and grazers. The collapse of the plastron perhaps resulted from early burial beneath loose sediment, causing it to cave in to the internal cavities. This unusual burial position is also recorded in the Solothurn turtle limestone (Late Jurassic, Kimmeridgian) of the Jura Mountains, Switzerland (Meyer 1991; Billon-Bruyat 2005).
The carapace is relatively intact, and the fragments are not widely dispersed, although distal elements including the limbs and head are missing, perhaps from scavenging of more accessible parts. A convex-down position of a turtle carapace prevents or slows down disarticulation, as shown in burial experiments using hawksbill turtles (Meyer 1991); in these studies, a carapace buried convex-up was completely disarticulated within 10 days, whereas a carapace buried convex-down was still almost intact after 15 days. Previous taphonomic studies showed that the skull, neck, tail and limbs of turtles become disarticulated early, but also that the carapace usually becomes disarticulated before the plastron (Bertini et al. 2006).
There are some fractures in the distal portions of the carapace, which suggests the action of larger scavengers or that sediment compaction under loading caused the fractures. These fractures are unlikely to have resulted from reworking, as sedimentological data (Olóriz et al. 2003; Reolid 2008) indicate low-energy conditions: the microfacies is a fine-grained wackestone with peloids, and there are no sedimentary structures indicating turbulence. Associated macroinvertebrates are intact and unfragmented, including ammonoid, bivalve and brachiopod shells, which are articulated. The distal portions of the bones also lack signs of abrasion. The sedimentary setting was a softground at mid-shelf and a depth of c. 60–80 m (Olóriz et al. 2006). The action of scavengers is cited as a further possible reason for carapace-down burial of turtles in both modern and fossil death assemblages, in studies of turtle taphonomy by Corsini et al. (2006) and Corsini and Chamberlain (2009).
The carapace went through a series of colonization events after death (Text-fig. 10). Bioerosion indicates initial colonization of the exposed carapace primarily by grazers, Gnatichnus (A. Santos and E. Mayoral, pers. comm. 2009), before the overturning of the carapace (Corsini and Chamberlai 2009). Meyer (1991) noted that the costal plates of modern hawksbill turtle skeletons are often covered in grazing traces formed by sea urchins as they remove algae and other organic material encrusting the carapace, seen also in carapaces from Solothurn. In our specimen, borings along the right-lateral edge of the costals on the right side of the carapace where the marginals are disarticulated are interpreted as clionaid sponge encrustation marks, possibly Entobia based on the interlaced morphology of the bored chambers. Other types of sponges (mainly Hexactinellida) are found in this section as isolated clumps and as sponge-microbialite buildups (Reolid 2010), so it is possible that the carapace acted as a suitable firm ground for them, amidst the surrounding loosely consolidated sediment. The rest of the carapace is covered by a network of interlaced Gnatichnus borings produced by the grazing activity of regular echinoids while the carapace was resting in the sediment with the colonization of clionaid sponges and disarticulated section exposed at the surface (Plate 1E, F, G). The possibility of colonization by some trace-makers during postmortem transport and floating, however, should not be discounted. No epibionts were found encrusting the specimen, even though barnacles and algae are common encrusters in recent turtles. Indeed, many barnacle species are found exclusively attached to turtle shells today, although these would be removed when the turtle died and the dermal scutes to which they were attached rotted away (Pfaller et al. 2006, 2008).
The filaments, globular structures and microspheres observed under the SEM (Plate 2C) are related to microbial (fungal or bacterial) activity associated with decay of the organic matter. Similar microspheroids were recorded by Elorza et al. (1999) in Upper Cretaceous turtle bone fragments from the Laño Quarry in the Basque-Cantabrian region of northern Iberia. Schmitz and Ernst (1994) proposed that such microspheroids, which also occur in the Eocene Messel Oilshale Pit in Germany, are by-products of bacterial action, perhaps associated with the decay of organic tissues in the cancellous layer of the bone. These microspheroids and crystal bundles suggest that the carapace may have acted as an enclosed microenvironment favourable to the development of decay-induced authigenic mineralization, and therefore facilitating its own preservation (Briggs 2003).
Hispaniachelys prebetica was preserved in a mid-shelf environment of 60–80 m water depth (Olóriz et al. 2006). Based upon this depositional setting, H. prebetica could be considered the oldest marine crown turtle, but it is possible for freshwater turtles to be washed out into marine basins from river deltas (Joyce and Gauthier 2004). An aquatic rather than terrestrial habit is interpreted for H. prebetica on the basis of its bone histology, based on its abundant circular primary vascular canals in the inner cortical layers, a feature characteristic of extant aquatic turtles (Scheyer and Sander 2007). The overall structure of the bone is more vascularized than that typical of terrestrial species (Scheyer and Sander 2007).
Hispaniochelys prebetica exhibits a rather domed anterior portion of the shell, which might be interpreted to suggest the individual was female. This suggestion follows Oertel (1924) and Bräm (1965), who argued that female P. solodurensis had an arched or domed anterior portion of the carapace. This gender assignment is presumably based on the need for females to house larger shoulder muscles and girdles to excavate nests on the shore. Specimens of Plesiochelys described by Bräm (1965) are distorted to various degrees, so the reality of this supposedly gender-specific character is dubious (Gaffney 1975b), and nothing can be said about the gender of our specimen.
The thick, heavily ossified carapace of H. prebetica would have been useful in a coastal marine habitat where manoeuvrability and protection are more important than weight reduction and energy conservation. Ocean wandering turtles today reduce drag by having a shallow streamlined shell. However, they also typically have a dorsoventrally deep-set anterior carapace, as in highly aquatic freshwater turtles (Depecker et al. 2006), but more massive and typically dorsoventrally longer. This creates an anterior dome that is intermediately arched between that of highly aquatic freshwater turtles and the extremely arched carapaces of terrestrial species.
Acknowledgements. We thank Simon Braddy (University of Bristol) for providing advice and literature on trace fossils, and personal comments by Ana Santos and Eduardo Mayoral (Universidad de Huelva) about grazing traces on the turtle carapace. We are also enormously grateful to Walter Joyce, Juliana Sterli and Jean-Paul Billon-Bruyat for their advice on turtle taxonomy and morphology, informally and formally through the review process. This study is part of a research project by BJS, carried out in partial fulfilment of the MSc in Palaeobiology at the University of Bristol.