A Nearly Complete Ornithocheirid Pterosaur from the Aptian (Early Cretaceous) Crato Formation of NE Brazil

A partial ornithocheirid, representing a rare example of a pterosaurian body fossil from the Nova Olinda Member of the Crato Formation, NE Brazil, is described from the collections of the State Museum of Natural History, Karlsruhe. While similar in preservation and taphonomy to Arthurdactylus conandoylei, it is distinguished by slight differences in biometric ratios, but the absence of a skull prevents closer identification. Mostly complete body fossils belonging to ornithocheiroid pterosaurs appear to be relatively more abundant in the younger Romualdo Member of the Santana Formation, making the described specimen one of only two well documented ornithocheiroids known from the Nova Olinda Lagerstätte.


Introduction
The Araripe Basin of NE Brazil contains two Early Creta− ceous Lagerstätten that are world renowned for their excep− tional preservation of insects and vertebrate fossils (Unwin 1988;Unwin and Martill 2007). Pterosaurs from the Aptian/ Albian−aged Crato Formation and Albian/?Cenomanian− aged Santana Formation (Martill 2007) confirm that taxa at− tributed to both the Ornithocheiroidea (sensu Unwin 2003) and Azhdarchoidea inhabited the area of this inland lagoon, a setting in which marine influences became more dominant towards the end of the Albian after the formation of the Santana sea (Kellner and Tomida 2000).
A new specimen from the Nova Olinda Member in the collections of the State Museum of Natural History, Karls− ruhe (SMNK PAL 3854), is described here, representing the rare occurrence of a largely complete ornithocheirid ptero− saur from this locality. The specimen is taxonomically inde− terminate, missing the skull, cranially located elements of the cervical column and the second to fourth phalanges of the wing−fingers, but is otherwise in a fairly good state of preser− vation. As is typical of fossils from the Crato Lagerstätte, the bones are crushed and few three−dimensional details can be observed. A greater degree of damage is observed along the caudally located thoracic vertebrae and the pelvic girdle. The skeleton has collapsed upon itself following contact with the lagoon floor, exposing the majority of bones in their dorsal or dorsolateral aspects (Fig. 1 Description.-The cervical column is represented by four vertebrae, identified as cervicals 5-8 (C5-8) (Figs. 2, 3). The morphology of these vertebrae is typical for ornithocheiroid taxa (Table 1), in that the cervical vertebrae are 2-2.7 times as long as they are wide, they have a wide neural canal and widely diverging pre− and postzygapophyses, and the pre− zygapophyses are located lateral to the postzygapophyses (Bennett 2001). The neural spines of C5-6 are broken and no comment can be made on their relative height. Caudal to the 6 th cervical vertebrae the remaining cervicals have been dis− placed from their natural position and are now visible in their craniolateral (C7) and cranial (C8) aspects. The neural spine of the 7 th cervical is tall and thin with respect to the vertebral body while the most caudally located cervical vertebrae (i.e., C8-9) preserve large robust ribs that remained in situ, sug− gesting that these had fused to the transverse processes. Al− though the 9 th cervical itself is not visible, being overlain by the 8 th , its presence is confirmed by a single large rib situated caudal to that of the 8 th cervical (Fig. 3C). The capitulum and tuberculum are widely spaced, by approximately 16 mm, and the shaft is narrow, decreasing rapidly to 6 mm by the mid corpus. The rib of the 9 th cervical is complete and terminates in a robust, slightly convex surface after a length of 89 mm. The centra of the thoracic vertebrae are missing, buried and/or badly damaged such that the description of these ele− ments is restricted to the neural spines, three of which are partially fused and form a supraneural plate. Thin suture lines separating the individual neural spines are visible (Fig. 2). The neural spine of the first visible thoracic vertebra lies sep− arate from the supraneural plate, although it is uncertain whether this is due to damage or displacement of the skele− ton, or whether the neural spine simply did not form part of this structure. A large oval depression with a raised rim occu−  pies the lateral flank of the third visible thoracic vertebra, forming the articulation for the medial articular surface of the scapula. Caudal to the notarial vertebrae the vertebral col− umn is kinked and the more caudal thoracic vertebrae are badly damaged, indistinct and partly overlain by the left fe− mur. These are also seen in lateral aspect and they have a maximum height of 19 mm from the base of the centrum to the top of the neural spine.
Nothing can be said about the sacral vertebrae, which are obscured from view by the overlying pelvic girdle. At least six gastralia have separated from the main body of the fossil and were displaced to overlie the distal portion of the right 104 ACTA PALAEONTOLOGICA POLONICA 57 (1), 2012 Fig. 3. New ornithocheirid pterosaur specimen (SMNK PAL 3854) from the Nova Olinda Formation, Crato Basin, Brazil. A. 5 th cervical vertebrae. B. 6 th and 7 th cervical vertebrae. C. 8 th and 9 th cervical ribs. Photographs (A 1 , B 1 , C 1 ) and explanatory drawings (A 2 , B 2 , C 2 ). Black shading highlights gaps/foram− ina with the bone and the collapsed neurocentral canal.
ulna following the disintegration of the dorsal column. These gastralia are preserved as three opposing sets and are associ− ated with a limited amount of mineralised tissue. Eleven cau− dal vertebrae lie loose from the axial column and are scat− tered between the notarial vertebrae and the right scapulo− coracoid. These are identical in appearance to those of C. piscator (Kellner and Tomida 2000), being long and narrow, typically between 2 and 4 times as long as they are wide. Where visible, the neural spines are positioned cranially and protrude past the cranial margin of the vertebrae, indicating that these bones belong to the middle portion of the tail. The terminal−most caudal vertebrae remain in articulation and are mediolaterally compressed to a greater degree than the other caudals. These are conical in form and do not expand at their articular surfaces (Fig. 2). The right scapulocoracoid is preserved in cranial view. No suture line between the two elements is visible, indicating that the structure was fully fused. The glenoid body measures 23 mm across its widest part and the scapula is shorter than the coracoid, the condition typical of ornithocheiroids (Frey et al. 2003c;Kellner 2003;Unwin 2003). The scapula is off− set against the glenoid body at an angle of 121°and the dorsal rim of the articular surface is visible as a flat, slightly convex surface that would have braced against the sub−oval facet of the notarium. The coracoid is about half as thick as the scap− ula in the mid−shaft region and forms an angle of 60°against the scapula. The proximal portion of the coracoid, where it would have articulated with the sternum is obscured by over− lying sediment. The vertical distance between the dorsal sur− face of the scapula articular surface and the sternocoracoid articulation is estimated to have been no greater than 100 mm.
The individual bones of the forearm are mostly preserved in either near articulation or bent beyond bone lock. Both hu− meri are visible in dorsal aspect and preserve a short, warped, sub−triangular deltopectoral crest that is approximately one third of the total humeral length. The caudal tuberosity is short, and no pneumatic foramina are observed where it con− verges with the shaft of the humerus, although the compacta of these surfaces is slightly damaged. On the distal portion of the left humerus, along the craniodorsal margin of the shaft, a pronounced flange, 15% the humeral length, may be the in− sertion for the flexor muscles of the carpus (Bennett 2003). Towards the distal margin of the humerus, the humeral shaft expands to about twice its width at the elbow joint, preserv− ing an almost straight dorsal margin perpendicular to the long axis of the shaft. The epiphyseal gap is partially open.
The left ulna lies at an angle of 72°to the humerus while the right makes an angle of 53°. Both have become slightly disarticulated from their natural positions. Both ulnae are about six times longer than they are wide and preserve no ob− vious muscle scars. While the right radius lies in situ along the right ulna the left radius has separated from the ulna and preserves a mid shaft width of 10-13 mm.
The carpals have fused to form two distinct syncarpal blocks but these have disarticulated from one another and are badly damaged so that no anatomical details can be observed. The fourth metacarpals have been displaced by approxi− mately 180°and lie sub−parallel to the ulnae; the distal roller joints are located near the proximal ends of the ulnae. Both wing metacarpals are also preserved in ventral view, with the ventral part of the roller joint measuring 16 mm across at its widest point, and they retain a natural articulation with the first phalanges of the wing finger. At least two additional metacarpals, presumably belonging to the left wing, are also preserved. Both are long and narrow and at least one would have made contact with the distal syncarpal during life.
The digits of the left manus are well preserved, and al− though the individual phalanges have disarticulated from each other they have not been displaced any great distance. The distal bones retain a contact with their respective un− guals. The phalanges are about half as wide as the proximal margins of the unguals and slightly curved, with expanded proximal and distal margins at the articular facets to accom− modate the neighbouring elements. The unguals are slightly curved with longitudinally concave ventral faces, forming a sulcus that extends almost the entire length of the bone. The phalanges of the right manus are mainly obscured and only two partial phalanges and one ungual are visible.
Only the first phalanges of the wing fingers are preserved and lie in natural articulation with the fourth metacarpals, making angles of 89°and 68°against the long axis of the left and right metacarpals respectively. The extensor tendon pro− cess has fused to the proximal face of the first phalanx and a large pneumatic foramen is present on the caudoventral sur− face of the right phalanx, adjacent to the proximal articula− tion. At its distal terminus the cranial margin of the first wing finger phalanx gradually merges with the gently convex dis− tal articulation. Caudally this articulation facet forms a sharp caudally directed process that is formed by the steeply con− vex caudodistal margin of the shaft. This creates an ex− panded surface for the second wing phalanx that would have directed the distal phalanges caudally.
The individual elements of the pelvis have fully fused to form two complete pelvic plates, although these are in a poor condition. The compacta is damaged and the bone itself is crushed over several underlying elements. The left ischio− pubic plate is observed in lateral view while the right is seen in caudolateral aspect, the acetabulum dominating the lateral faces of each and the obturator foramen being fully enclosed within the pubis. The cranial and caudal margins of the left pubis are shallowly concave, being almost symmetrical about the long axis of the bone. The bone is narrowest about its mid point before expanding towards its ventral margin, however, the ventral margin of the pubis is broken and its original shape cannot be determined. While the long axis of the pubis is directed ventrally, that of the ischium is directed caudoventrally at an angle of 46°against the pubis. As with the pubis, it is narrowest about its mid point and expands to− wards its ventral margin. Approximately half the caudal por− tion of the left ischium has been broken and lost from the http://dx.doi.org/10.4202/app.2010.0079 specimen while the right has been distorted and is directed through the bedding plane of the slab. The preacetabular pro− cess of the ilium is long and thin but its cranial portion is ei− ther overlain by the ulna (left) or broken (right). The post− acetabular process of the right pelvic plate partially overlies that of the left, which in turn has been crushed over the re− mains of the sacral vertebrae. The postacetabular process ex− pands caudodorsally for approximately one quarter of its to− tal length due to the curvature of the dorsal margin of the bone. Caudal to this expansion, the dorsal and ventral mar− gins are directed caudoventrally, creating a caudoventrally directed process more or less uniform in width.
The femora lie adjacent to the pelvis and preserve a long, narrow femoral collum femoris, approximately 17 mm long, and a moderately bowed, laterally curving shaft. The femoral head is offset against the shaft at an angle of 20°. Although both tibias have disarticulated from their respective femurs, they have not been displaced much. The tibia is long and slender, becoming increasingly narrow towards its distal ar− ticulation. Neither fibula is preserved. The distal articulation of the left tibia is broken and missing but the right is complete and terminates as a flat surface, 7.5 mm in width, indicating that the calcaneum and astragalus had not fused with this bone to create a tibiotarsus (Fig. 4). Two tarsals, the astra− galus and a distal tarsal, lie level with the distal margin of the tibia but off to one side. The astragalus is weakly crescentic in shape and is observed in medial aspect. A large foramen pierces the craniomedial portion of the bone. Although its caudomedial face is damaged, the preserved remains suggest that the medial surface was convex, projecting medially as described by Kellner (2004). The dorsal margin of the astra− galus is concave to form the articulation with the distal sur− face of the tibia, while the ventral margin is strongly convex. The distal tarsal is as large as the astragalus itself, with a gen− erally smooth convex surface and lies in contact with three of the four metatarsals.
The right metatarsals are very long and thin, the longest reaching a length of 58 mm with a mid−shaft diameter of only 1 mm. The length of these bones ranges from 54-58 mm, but at least one of the first four metatarsals is noticeably shorter with a length of only 45 mm. The fifth metatarsal is clearly distinct from the others, and is about one fourth to one fifth the length of metatarsals I-IV and has a sub−triangular out− line with slightly convex medial and lateral margins.
The pedal phalanges are delicate and show only slight curvature. Their proximal and distal margins are slightly ex− panded at the interphalangeal articulations and in all cases the unguals maintain a natural articulation with the penulti− mate phalanx. The unguals themselves are equal in width to the preceding phalanges and are only slightly curved.

Discussion
Largely complete ornithocheiroid pterosaurs appear to be much less common in the Nova Olinda Lagerstätte than in the geologically younger Romulado Member of the Santana For− mation, NE Brazil. A thorough review of these was provided by Unwin and Martill (2007), who noted that specimens from the Crato Formation attributed to the Ornithocheiroidea are limited to SMNK PAL 1132 (Arthurdactylus conandoylei, Frey and Martill 1994), a headless skeleton preserving the ma− jority of the postcranial elements; the proximal part of a wing belonging to an indeterminate ornithocheiroid (SNMK PAL 3842); an isolated tooth (Sayão and Kellner 2000); the com− plete cranium of SMNK PAL 3828 (Ludodactylus sibbicki, Frey et al. 2003a); and the rostral fragments of cf. Brasileo− dactylus (Sayão and Kellner 2000). Unwin and Martill (2007) include two further specimens, MPSC R−739 and MPSC R−779 (Nuvens et al. 2002) in the Ornithocheiroidea. A com− parison of the present specimen with these other named taxa from the same locality, however, is problematic because no 106 ACTA PALAEONTOLOGICA POLONICA 57 (1), 2012 postcranial elements are known for L. sibbicki, while the taxo− nomic validity of Brasileodactylus is uncertain (Unwin and Martill 2007) and diagnostic features are restricted to the skull (Veldmeijer et al. 2009). It is fortunate that pterosaurs generally display isometric growth, in which individual bones grow at the same relative rate through ontogeny, and this allows the calculation of bio− metric ratios to distinguish taxa. Omitting the pteranodontids (studied by Bennett 2001) and nyctosaurids, the wing meta− carpal and femur of the Ornithocheiridae + Istiodactylidae show significantly negative allometric relationships, so that the ratios of lengths of these elements to humerus length in− crease at larger sizes (see Appendix 1). Limb bone propor− tions place SMNK PAL 3854 within the Ornithocheiroidea (Table 2) where it is distinguished from pteranodontid and nyctosaurid pterosaurs by the relative shortness of the wing metacarpal, and from the Istiodactylidae by a set of ratios that lie outwith the observed range of values known for these taxa. This suggests that the current specimen may be tenta− tively assigned to the Ornithocheiridae. Despite the similari− ties in preservation and taphonomy between SMNK PAL 3854 and SMNK PAL 1132, the described specimen is dis− tinguished from A. conandoylei through biometric ratios, where its humerus is relatively longer, forming ratios of 0.63 and 0.41 with the ulna and first wing−finger phalanx respec− tively (0.74 and 0.52 for SMNK PAL 1132). A more distinct ratio is observed in the hind limb, where the femur is substan− tially shorter, based on a large difference in the forelimb/ hindlimb ratio (1.48 for SMNK PAL 3854 versus 1.81 for SMNK PAL 1132). A possible relationship is suggested with a larger specimen attributed to Santanadactylus sp. (SMNK PAL 1250) by an almost identical suite of bone ratios (note that this latter fossil is in a private collection and has not been described or diagnosed; Frey and Martill 1994).
SMNK PAL 3854 is estimated to have had a wing span of 3.4 m, based on the observations of Veldmeijer (2003), who noted that in Santanadactylus pricei (Wellnhofer 1991) the ratio between the total length of the wing finger and the hu− merus plus radius/ulna is 2.7. A similar ratio is also recorded for A. conandoylei, suggesting that this relationship is useful across a range of ornithocheirid taxa. The length of the torso, measured from the first thoracic vertebra to the caudal mar− gin of the ilium, is 140 mm, namely 4% of the total wingspan.
Ontogenetic age.-Despite the relatively advanced state of suture closure in parts of SMNK PAL 3854 shows no fusion between the astragalus and the tibia, and so is not yet at maxi− mum age. The relative timings of suture closure in pterosaurs have been briefly reviewed by Bennett (1993) and Kellner and Tomida (2000), and the former tentatively proposed that for Pteranodon the atlas−axis complex, scapulocoracoid, second− ary centres of ossification in the humerus, the cranial−most no− tarial vertebrae and ribs appear to precede all others. The su− ture between the extensor tendon process and the first wing finger phalanx appears to close shortly before skeletal matu− rity is reached. An examination of the suture states in SMNK PAL 3854 indicates that fusion of the tibia and proximal tarsals, the caudal thoracic ribs and their respective vertebrae, complete suturing of the notarium, and fusion of the humeral epiphyses to the humerus itself should all be considered indi− cators of late ontogeny and are preceded by the closure of the suture between the extensor tendon process and the first wing−finger phalanx and elements of the pelvic plate. Such ob− servations do not contradict the proposals of Bennett (1993) and Kellner and Tomida (2000) but rather provide an example of interspecific variation in developmental timings that likely existed even between closely related taxa.
Life style.-The estimated length of the wing spar combined with the short length of the torso and hind limbs indicates that SMNK PAL 3854 developed the high aspect−ratio configura− tion noted for other ornithocheiroids and likely shared their interpreted lifestyle, being a relatively fast, open water glider, using dynamic or thermal soaring (e.g., Chatterjee and Templin 2004). The view that such taxa spent most of their life on the wing is further supported by the structure and relative size of the pes with respect to the wings; these bones in SMNK PAL 3854 are exceedingly thin compared with the pedal elements of azhdarchoid pterosaurs with which they shared the Cretaceous Crato lagoon. The comparison of a similar−sized azhdarchoid, SMNK PAL 3900, effectively il− lustrates this difference where the mid−shaft diameter of the phalanges ranges between 2-2.5 mm, with the unguals being http://dx.doi.org/10.4202/app.2010.0079 Table 2. Selected long bone ratios of ornithocheiroid taxa. Abbreviations: fe, femur; FL, fore limb (humerus + ulna + metacarpal IV); HL, hind limb (femur + tibia); hu, humerus; mc IV, wing metacarpal; wph 1, first wing finger phalanx; ti, tibia; ul, ulna. 17 mm long (Fig. 4A 2 , B). Similar values are also noted for the medium−sized azhdarchoids SMNK PAL 3830 and 6409, and the former demonstrates that sharp keratinous sheaths doubled the total length of the pedal claws. In contrast, the phalanges of SMNK PAL 3854 are only 1 mm in diameter with a length of 7 mm. The delicate nature of the pes in SMNK PAL 3854 therefore suggests that it was predomi− nantly a soaring animal that spent little time on the ground. While one reviewer noted that blunted claws would be more in keeping with a terrestrially active lifestyle, it is difficult to explain the long, sharp keratinous sheaths of azhdarchoids without invoking a role in terrestrial support, as pterosaurs did not utilise their pedal claws for either the manipulation of prey or arboreal perching. Nonetheless, the lack of wear on the ungual sheaths of SMNK PAL 3830, a morphologically mature individual, perhaps indicates that these animals spent a significant portion of their life on the wing, while still being more terrestrially competent than ornithocheirids such as that described here.
Taphonomy.-The resting position and taphonomy of SMNK PAL 3854 strongly resembles SMNK PAL 1132, where the skeleton has collapsed upon itself following the destruction of the ventral body wall and the subsequent re− lease of air from the pneumatic system. In both individuals, the lack of a preferred orientation indicates that bottom cur− rents were absent, while the overlapping bones indicate that the carcass sank left wing first, and the hind limbs were the last elements to settle. In contrast to SMNK PAL 1132, the neck did not detach at its base but between the 4 th and 5 th cervicals (Frey and Martill 1994). The sternum is also miss− ing from SMNK PAL 3854, the reason perhaps being the rapid bacterial degradation of the large pectoral muscle mass. While the pectoral muscles insert at the humerus with mas− sive tendons, their origin on the sternum is fleshy and thus subject to more rapid decay and earlier detachment of the sternum, where the sternocostal and sternocoracoideal artic− ulations must have been weak. The carcasses of ornitho− cheiroid pterosaurs likely followed the pattern of decay ob− served in birds and mammals (Oliver and Graham 1994;Da− vis and Briggs 1998), in which the head detaches first from the body, followed by the sternum. Long−term floating at the water surface is typical for highly pneumatised tetrapods (Schäfer 1962(Schäfer , 1972, buoyed up by their air sacs. The state of decomposition when the specimen settled on the lagoon floor was probably relatively advanced, based on the degree of disarticulation. While this indicates that the interarticular ligaments had largely lost their ability to restrict the mobility of the bones, the carcass nonetheless must have been largely intact when it settled, as is shown by the preservation of easily displaced elements, such as manual and pedal phalanges. It then seems likely that the larger elements missing from the slab, such as the head and neck and the distal wing phalanges, were likely lost dur− ing sedimentary transport rather than during collection.

Conclusions
A new specimen from the Crato Formation of NE Brazil is an ornithocheirid pterosaur, distinguishable from Arthurdac− tylus conandoylei, the only other nearly complete postcranial skeleton from the Crato locality, by biometrics of the long bones. Almost identical bone ratios indicate possible rela− tionship with a specimen of ?Santanadactylus sp. (Frey and Martill 1994), but the lack of a formal description of the latter prevents a taxonomically useful comparison.
The inferred high aspect ratio of the wing, combined with a delicate pes, supports the established hypothesis that ornitho− cheiroids spent the majority of their life in the air and that, while fully capable of terrestrial locomotion, spent little time on the ground when compared to other pterosaurs.
The animal was ontogenetically relatively mature when it died, based on the state of its sutures, which include fusion of the extensor tendon process to the first wing−finger phalanx, the scapula to the coracoid, the proximal and distal syncar− pals of the wrist, the ischiopubic plate, and a partial notarium. The animal was not fully mature, however, because a tibio− tarsus had not yet formed, and suture lines separate the epiphyses from the humerus. These elements can therefore be regarded as late−forming structures and may be of use in diagnosing the relative ontogenetic maturity of pterosaurs in future studies.

Appendix 1
Long bone measurements of ornithocheiroid specimens used for calculating ratios and isometric/allometric relationships in this paper, where "R" denotes a value reconstructed from personal observations or the source literature. At the 95% CI with the humerus acting as the independent variable, the relationships of selected bone elements are defined as: ulna = 0.325*x 0.936 +/− 0.119 (R 2 = 0.96); mc IV = 0.493*x 0.789 +/− 0.082 (R 2 = 0.97); femur = 0.444*x 0.798 +/− 0.183 (R 2 = 0.93); and wph 1 = 0.16*x 1.071 +/− 0.13 (R 2 = 0.97). With respect to the humerus, the wing metacarpal and femur show a significantly negative deviation from isometry. Specimens attributed to pteranodontids and nyctosaurid pterosaurs were omitted for several reasons: the elongation of the wing metacarpal in relation to the humerus is unusual for ornithocheiroids; certain bone relationships e.g. ulna/mc IV appear to be different between Pteranodon/Nyctosaurus and the remainder of the Ornithocheiroidea; the large number of specimens known for Pteranodon would unduly bias the regression relationships towards this single genus rather than the broader trend across the entire division. The observed taxonomic level of regression relationships is thusly important and the state at one level need not be directly applicable to others.