Colymbosaurus sp.

Colymbosaurus sp.

Figs. 5–7 View Fig View Fig View Fig , 9 View Fig .

Material.— NHMD 189834 , a partial skeleton consisting of 22 non-consecutive vertebrae from the cervical, dorsal, and caudal regions, hundreds of rib fragments, partial girdle elements, all four propodials, and several distal limb elements from Reptile Creek , approximately 3 km away from Eureka Weather station , Ellesmere Island , Nunavut, Canada, uppermost Berriasian to lowermost Valanginian .

Description.— Preservation: NHMD 189834 preserves all four propodials along with several limb elements, partial girdle elements, 22 vertebrae and numerous partial ribs. The elements are preserved as three-dimensional, mostly undistorted bones, and many of them show well-preserved surface details and facets. However, all of the bones are weathered, and many are incomplete or preserved in many fragments. The propodials have been partly reconstructed using plaster, but not to an extent that makes descriptions misleading.

Ontogeny: Where preserved, the neural arches are fused to the vertebral centra, indicating that the individual was osteologically mature ( Araújo and Smith 2023), and therefore an “adult” as per the criteria of ( Brown 1981). In addition, the relatively coarse, finished bone outer layer, as well as the large size of the skeletal elements and the well-defined distal facets on all propodials are indicative of adulthood.

Body size: The humeri of NHMD 189834 are 390 mm (left) and 400 mm (right) long, and the left and right femora are 350 mm and 360 mm long, respectively. All of these bones are slightly shorter than the respective elements in the three specimens of Colymbosaurus svalbardensis from the Slottsmøya Member ( Knutsen et al. 2012c; Roberts et al. 2017), and the femora are slightly shorter than in Colymbosaurus megadeirus (CAMSMJ.29596, etc.) from the UK ( Benson and Bowdler 2014), but fit within the large and small size classes of Russian colymbosaurines ( Arkhangelsky et al. 2020). In the specimen of C. megadeirus , the neck length was estimated to be 2 m and the trunk 1.7 m ( Benson and Bowdler 2014), so, based on femoral length, the Ellesmere Island specimen would have been somewhat shorter than that.

Axial skeleton: In total, 22 non-consecutive vertebrae are preserved ( Fig. 5 View Fig ). 21 vertebrae are complete or almost complete centra, and some have partial neural arches preserved intact. One vertebra is represented by only a small fragment. Sachs et al. (2013) is used for referring to the regions of the axial skeleton. As the specimen is neither complete nor articulated, the total number of vertebrae is unknown for the different body regions. Three cervical centra are preserved Fig. 5A View Fig ), and since this was a long-necked species, this means the vast majority are missing. Two pectoral vertebrae are preserved ( Fig. 5B, C View Fig ). Two dorsal vertebrae are preserved with an intact transverse process on one side, but these processes have broken off the other vertebrae from the same region. Five centra were too weathered or incomplete to be assigned to any specific vertebral region. In addition to the centra, several isolated transverse process portions are preserved. There are also five large disarticulated neural arches, but which region of the vertebral column they belong to is unknown. Centrum measurements of the preserved vertebrae are provided in Table 1.

Cervical vertebrae: Three cervical vertebrae are preserved in NHMD 189834, all missing the neural arch and neural spine ( Fig. 5A View Fig ). As they were disarticulated, it is not known for certain which ones are from the more anterior position in the neck. One cervical centrum is complete (centrum C). It is shorter anteroposteriorly than tall dorsoventrally, as in Colymbosaurus megadeirus , which is slightly shorter than tall ( Benson and Bowdler 2014), Muraenosaurus sp. specimen MUHNCAL.20176 ( Otero et al. 2020), and Tatanectes laramiensis (O’Keefe and Street 2009; USNM 536970, LLD personal observations). However, the latter is significantly shorter anteroposteriorly than NHMD 189834. This contrasts with Spitrasaurus wensaasi and Pantosaurus striatus (USNM 5782, LLD personal observations), where the cervical vertebrae are longer than tall, and S. larseni and Abyssosaurus nataliae , which have cervical centra that are equally long and tall ( Berezin 2011; Knutsen et al. 2012d). However, in Ophthalmothule cryostea , which preserves a complete neck, the relative measurements of the vertebrae vary between different parts of the neck ( Roberts et al. 2020), which might also be the case for NHMD 189834. The cervical centrum is also significantly wider mediolaterally than tall dorsoventrally, as in Colymbosaurus megadeirus ( Benson and Bowdler 2014) , Muraenosaurus leedsi ( Andrews 1910) , Abyssosaurus nataliae ( Berezin 2011) , Spitrasaurus larseni ( Knutsen et al. 2012d) and Pantosaurus striatus (USNM 5782 and 536963, LLD personal observations; Wilhelm and O’Keefe 2010). The vertebral length index is 0.74, similar to the posterior vertebrae of C. megadeirus ( Benson and Bowdler 2014) . For C. svalbardensis , cervical vertebrae are not known ( Roberts et al. 2017).

On the dorsal surface, the neural canal is equally wide mediolaterally for its entire length, as also figured for C. megadeirus ( Benson and Bowdler 2014) , but in contrast to Spitrasaurus wensaasi , where it is hour-glass shaped ( Knutsen et al. 2012d). Two anteroposteriorly elongated foramina are situated on the floor of the neural canal, as in Spitrasaurus wensaasi ( Knutsen et al. 2012d) and Ophthalmothule cryostea ( Roberts et al. 2020) .

Two of the cervical centra are platycoelous (centrum A and B), with one side flat and the other only very slightly concave, whereas the other centrum (centrum C) is amphicoelous with two slightly concave surfaces. The latter morphology is the same as in the cervical centra of C. megadeirus , Spitrasaurus wensaasi , S. larseni , and Abyssosaurus nataliae ( Berezin 2011; Knutsen et al. 2012d; Benson and Bowdler 2014), in contrast to Cryptoclidus eurymerus , where all centra are amphicoelous ( Brown 1981). A groove around the perimeter of the articular faces, as described in Colymbosaurus megadeirus ( Benson and Bowdler 2014) , is absent. The cervical centrum is spool shaped, with expanded articular ends compared to the main body of the centrum.

The lateral surface of the centrum lacks a longitudinal ridge, as it does in Colymbosaurus megadeirus ( Benson and Bowdler 2014) , Spitrasaurus wensaasi and S. larseni ( Knutsen et al. 2012d) . However, as only three cervical centra are preserved it is possible that a lateral ridge was present in other centra. The cervical rib facet is single headed and situated on the ventralmost portion of the lateral surface of the centrum ( Fig. 5A View Fig 4). This is similar to the situation in Muraenosaurus leedsi , Colymbosaurus megadeirus , Pantosaurus striatus (USNM 5782, LLD personal observations) and Spitrasaurus wensaasi ( Andrews 1910; Knutsen et al. 2012d; Benson and Bowdler 2014). The facet is anteroposteriorly longer than dorsoventrally tall, as in Spitrasaurus wensaasi ( Knutsen et al. 2012d) . The facet does not contact the anterior or posterior margin of the centrum. The facet in centrum C is fused to the cervical rib.

On the ventral surface, two nutritive foramina are positioned posteriorly. These are anteroposteriorly oriented ovals in centrum A, but circular in centrum C ( Fig. 5A View Fig 3). They are widely spaced, several centimetres apart, in contrast to Colymbosaurus megadeirus ( Benson and Bowdler 2014) and Spitrasaurus wensaasi , where the foramina are positioned close to each other ( Knutsen et al. 2012d), even more so than in Tatanectes laramiensis (O’Keefe and Street 2009). The ventral surface is almost flat, with a slight bulge between the foramina. Spitrasaurus wensaasi has a keel running along the entire ventral surface ( Knutsen et al. 2012d), whereas Tatanectes laramiensis has no ventral keel (O’Keefe and Street 2009).

Pectoral vertebrae: Two vertebrae are interpreted as pectoral vertebrae ( Fig. 5B, C View Fig ), because their rib facets are dorsoventrally tall and contact the base of the neural arch. Alternatively, they might represent sacral vertebrae, which have a rather similar morphology. Their centra are mediolaterally wider than dorsoventrally tall as in Colymbosaurus megadeirus and Ophthalmothule cryostea ( Benson and Bowdler 2014; Roberts et al. 2020), but not to such an extent as in the cervicals. The articular surface is only slightly concave. As in the cervical vertebrae, the neural canal is evenly mediolaterally wide. On the left lateral side of one vertebra (centrum D) is an anterodorsally oriented ridge that connects the rib facet and the neural arch base. This morphology is also found in the pectoral and sacral vertebrae of Colymbosaurus megadeirus and Ophthalmothule cryostea Benson and Bowdler 2014 ; Roberts et al. 2020). The paired foramina on the ventral surface are more widely spaced than in the cervical centra, so they are also visible in lateral view. The distribution of foramina is also asymmetrical, with an additional foramen on the right side of centrum D. The rib facets are incomplete but confirm that the centrum had single-headed ribs.

Dorsal vertebrae: The ten dorsal vertebrae are determined based on the more or less circular outline of the articular surfaces of the centra, the lack of facets for ribs on the centrum body and, where preserved, a neural arch that bears dorsolaterally directed transverse processes for articulation with the dorsal ribs ( Sachs et al. 2013). Except for one centrum (centrum L, anteroposterior length:dorsoventral height = 1.01), all are slightly dorsoventrally taller than anteroposteriorly long, as in Colymbosaurus megadeirus ( Benson and Bowdler 2014) . Most of the dorsal centra are also slightly dorsoventrally taller than mediolaterally wide, whereas some are slightly wider than tall, as in C. svalbardensis ( Roberts et al. 2017) . The dorsal centra are platycoelous with a completely flat anterior articular surface, and an almost flat but slightly concave posterior articular surface. They are strongly spool shaped, with articular ends mediolaterally wider than the main body of the centrum. In addition, some centra have a small ridge or lip around the articular margins where the centrum becomes mediolaterally wider still. A few possess a small knob in the centre of the articular face. Vertebrae ARROW, G, and R probably belong to the anteriormost portion of the dorsal series. In these, the floor of the neural canal widens at both ends, in contrast to most others, where it is evenly wide throughout. In one centrum (centrum M), the floor of the neural canal narrows at one end.

Three centra preserve one or both of their transverse processes (centra H, I, and L); one of these processes is almost complete (centrum I). The neural canal in the dorsal vertebrae is significantly taller than wide and is dorsoventrally ovate in cross-section. In centrum I, the most complete dorsal vertebra, it is 30 mm tall, twice the maximum mediolateral width of the canal ( Fig 5G View Fig ). This character has been suggested as an autapomorphy of C. svalbardensis Roberts et al. 2017 ), differentiating it from Kimmeridge Clay Formation cryptoclidids. However, although Benson and Bowdler (2014) figure a plesiosauroid vertebra from the Kimmeridge Clay (OXFUM J.55491) with a circular dorsal neural canal to show the possible anatomy in C. megadeirus , the dorsal canal is not known for certain in C. megadeirus .

In C. megadeirus , the neural canal in the posterior cervical vertebrae is dorsoventrally tall, and one dorsal vertebra also appears to show a dorsoventrally tall neural canal ( Benson and Bowdler 2014; Fig 6F View Fig ); however, this vertebra is distorted, and the tall canal could be an artefact of crushing. The ventral portion of the neural canal is wider than the top, exactly as seen in C. svalbardensis ( Knutsen et al. 2012c) . In Ophthalmothule cryostea , the neural canal is oval in anterior view ( Roberts et al. 2020).

A cross-section through the neural arch pedicles is visible in centrum R. These are mediolaterally widest about 1/3 from the anterior margin where they form a triangular buttress below the transverse process. They taper anteriorly to a sharp anterior edge, and posteriorly to a flat posterior edge.

The transverse processes are long, gracile, and inclined dorsolaterally and slightly posteriorly, as in C. svalbardensis ( Knutsen et al. 2012c; Benson and Bowdler 2014; Roberts et al. 2017). The distal extremity is expanded dorsoventrally to form an anterodorsally oriented facet that faces slightly posterolaterally. The rib facet is oval, as in C. svalbardensis and Tatanectes laramiensis (USNM 536974, LLD personal observations; Knutsen et al. 2012c). The bases of the transverse processes are oriented anteroposteriorly, and cover the entire length of the centrum except for a short anterior portion. Where it can be evaluated, a transverse subdiapophyseal fossa, such as the one shown in Hampe (2013), is absent. The arrangement of foramina is asymmetrical in some of the dorsal centra and at least two centra (centra G and M) have up to six small foramina on the base of the neural canal. On the ventralmost portion of the lateral side the dorsal centra have up to five anteroposteriorly elongated foramina, a condition also observed in anterior dorsal vertebrae in Tatanectes laramiensis (USNM 536974, LLD personal observations). For example, in centrum K, there are five clear foramina on the left side of the centrum, but only one foramen on the right side.

In lateral view, several of the dorsal vertebrae possess one articular margin that extends more ventrally compared to the other articular margin. In the holotype of C. svalbardensis ( PMO A27745 ), a similar condition, termed “anterior lip on the ventral margin”, was observed in one dorsal vertebra and interpreted as a taphonomical artefact ( Knutsen et al. 2012c). However, as it is observed in several dorsal centra of the studied specimen we here interpret it as a genuine morphological feature in some of the dorsal vertebrae in this species .

Caudal vertebrae: Two caudal vertebrae (O and P) are preserved, but are missing the neural arch and spine. The caudal rib facets and chevron facets contribute to an almost pentagonal outline in articular view. In Abyssosaurus nataliae the shape is hexagonal ( Berezin 2011). The caudal vertebrae in NHMD 189834 are mediolaterally wider than dorsoventrally tall, as in C. svalbardensis and C. megadeirus ( Benson and Bowdler 2014; Roberts et al. 2017), and anteroposteriorly shorter than dorsoventrally tall, but not as tall as in C. megadeirus ( Benson and Bowdler 2014) .

The caudal vertebrae bear processes for the caudal ribs positioned approximately at mid height on the lateral surface of the centrum, ventral to and disconnected from the neural arch, as in most cryptoclidids ( Benson and Bowdler 2014). The rib facets are also positioned midway anteroposteriorly. The rib facets are subtriangular ovals with a flat to concave anterior surface. This feature is known to vary along the caudal series in C. svalbardensis ( Knutsen et al. 2012c) .

On the ventrolateral corners are small chevron facets that are triangular in ventral and lateral views. The posterior facets are significantly larger than the anterior facets. Triangular chevron facets are also found in C. svalbardensis but differ from C. megadeirus , which has oval chevron facets ( Knutsen et al. 2012c; Benson and Bowdler 2014; Roberts et al. 2017). The chevron facets are significantly smaller than the caudal rib facets. The ventral surface of the centrum is almost flat and bears four elongated foramina arranged in two pairs, one in front of the other.

The neural arch pedicels appear to have been fused to the centrum but have broken off at their base. They are mediolaterally narrow for their entire length, and the floor of the neural canal is equally mediolaterally wide for its entire length. The pedicels extend anteriorly to the anterior edge of the centrum but terminate posteriorly well before the posterior edge of the centrum.

Ribs and gastralia: Approximately 100 rib fragments and gastralia are preserved, but none are complete. The longest pieces are 150 mm. It is possible that some pieces may match up. Some elements preserve the proximal ends of dorsal ribs including the articulation facets. The facets are oval with concave articular faces, and the main part of the ribs are flattened. Fragments of other ribs are larger and more robust, and most likely belong in the dorsal section. Their cross-sections vary greatly in outline, from oval to almost circular, and from nearly figure-of-eight-shaped to thickened T-shaped. As discussed for C. svalbardensis Roberts et al. 2017 ), cross sections vary along the length of the ribs. One complete and one incomplete sacral rib are preserved ( Fig. 5D View Fig ). These elements are slenderer than the dorsal ribs. The proximal articular facet is trilobate in proximal view, anteroposteriorly widest ventrally. The single facet is divided into distinct upper and lower parts, with the slightly smaller upper part inclined about 40° relative to the larger ventral part. The distal end of the rib is dorsoventrally expanded into a rounded bulb, but it is not expanded anteroposteriorly. Although both ends of the rib are dorsoventrally expanded, the proximal end of the sacral rib is dorsoventrally much thicker than the distal end.

Girdle elements: One incomplete element is large, flat and 1–2 cm thick, but partly covered with matrix ( Fig. 6A View Fig ). It was found attached to one of the humeri, and although this was probably not its real position in life, this proximity suggests this is part of the pectoral girdle, probably the coracoid. The element has a partially preserved outline including a long gently concave margin, which might represent the lateral margin of the coracoid. A tapering, slightly squared-off end could represent the posterolateral cornua of the coracoid.

One incomplete fan shaped element is interpreted as a partial scapula ( Fig 6B View Fig ). The posterior process with the glenoid contribution is preserved, whereas the anterior outline is incomplete, so that the full size and outline of the bone is unknown. The process is thicker and more robust than the rest of the element, as in Spitrasaurus wensaasi ( Knutsen et al. 2012d) . The glenoid contribution is complete and has an almost circular cross-section. The process with the glenoid contribution is offset ventrally from the plate of the scapula, likely medially and ventrally.

A separate, small fragment has a tapering shape that resembles the posterolateral cornua of the other coracoid, or the lateral part of a clavicle ( Fig. 6C View Fig ). For example, the shape of this element matches the posterolateral cornua of C. svalbardensis figured by Roberts et al. (2017: fig 8A, B) and MM LL.5513-8 figured by Brown (1981: fig. 43e). It could also be a partial ilium, resembling the proximal end of this element in C. svalbardensis ( Knutsen et al. 2012c; Roberts et al. 2017). However, it does not appear broken in the narrowest portion, and if so, it is far shorter and more massive than the ilia in C. svalbardensis and Tatanectes laramiensis (USNM 536974, LLD personal observations; O’Keefe et al. 2009).

Two unidentified, massive partial elements probably belong to the coracoid or pubis, but they are too incomplete to be certain. The specimen also preserves four relatively large fragmentary bones and many smaller fragments that likely are parts of the girdles but are too incomplete to be assigned with certainty. As material is preserved from both the fore and hindlimbs, the partial girdle elements might belong to either the pectoral or pelvic girdle.

Limbs: The overall shapes and proportions of fore- and hindlimbs are relatively similar to each other in plesiosaurs, which is one of the reasons for the continued debate on swimming style. Four propodials are preserved in this specimen. The humeri and femora exhibit a significant difference in size and shape from each other, so they may become helpful to differentiate isolated forelimbs and hindlimbs in other specimens where this is ambiguous. Two propodials are proximodistally longer, more robust, anteroposteriorly wider, have more clearly defined distal facets, and one more distal facet than the other two. These two limb propodials also have a more posteriorly displaced prominence above the proximal head (epipophysis). In plesiosaurs, this prominence in the humerus (tuberosity) is typically displaced posteriorly relative to the epipophysis, whereas the prominence in the femur (trochanter) is situated directly above the epipophysis/capitulum ( Andrews 1910; Brown 1981; Knutsen et al. 2012d; Roberts et al. 2017). The humerus and femur in this specimen can therefore be confidently differentiated on this basis.

Many long-necked plesiosaurs have longer forelimbs than hindlimbs ( Krahl 2021), and within cryptoclidids, the humeri are more robust than femora ( Arkhangelsky et al. 2020).

Humerus: The larger and more robust pair of propodials is interpreted as the humeri ( Fig. 7A, B View Fig ) based on the posteriorly positioned tuberosity, as in e.g., C. svalbardensis and Spitrasaurus wensaasi ( Knutsen et al. 2012d; Roberts et al. 2017). The humerus of Tatanectes laramiensis appears less robust than the one in NHMD 189834, and has a smaller proximal head (O’Keefe and Street 2009; USNM 536976, LLD personal observations). The left humerus is almost complete, with a small part of the preaxial margin of the shaft reconstructed in plaster, and some damage to the distal facet for the radius. The tuberosity is posteriorly positioned ( Fig. 7A View Fig ). In proximal view, the tuberosity (dorsal) has a slightly smaller surface area than the capitulum (ventral), and the tuberosity and capitulum are partly separated by a constriction, the isthmus, which is more pronounced anteriorly than posteriorly. The right humerus is significantly more reconstructed in plaster but still the majority is intact. It is broken in the distal portion, missing parts of the anteriormost distal facet.

The Ellesmere Island specimen (NHMD 189834) is similar to C. svalbardensis (PMO 222.663) in having a mediolaterally wider humeral proximal head compared to the femora ( Roberts et al. 2017). Because the tuberosity and the capitulum are not situated directly above each other, the proximal heads of the humeri are wider than the shafts in dorsal and ventral view. The Ellesmere Island specimen is also like C. svalbardensis (PMO 222.663) because the shafts of the humeri are more robust than those of the femora Fig. 7 View Fig ). The shaft is at its minimum anteroposterior width directly distal to the proximal head, then increases in anteroposterior width very gradually, until a large increase for the expanded distal portion in the last third of the anteroposterior length. This contrasts with Spitrasaurus larseni , which widens anteroposteriorly throughout the shaft ( Knutsen et al. 2012d). The distal portion widens more posteriorly, and only slightly towards the anterior direction. This is similar to C. svalbardensis , Tricleidus sp. , and Tatanectes laramiensis ( Brown 1981; O’Keefe and Street 2009; Knutsen et al. 2012c; Roberts et al. 2017), but differs from Spitrasaurus wensaasi and S. larseni which have evenly expanded distal ends anteriorly and posteriorly ( Knutsen et al. 2012d).

The distal margin of the humerus has four distinct facets in dorsal view, as in Tatanectes laramiensis (O’Keefe and Street 2009) and “ Plesiosaurus ” manselii ( Hulke, 1870). Colymbosaurus svalbardensis ( Roberts et al. 2017) and Pantosaurus striatus ( Wilhelm and O’Keefe 2010) have three facets, whereas in C. megadeirus the number of distal facets is unknown ( Benson and Bowdler 2014; Fig. 8 View Fig ). The osteologically immature holotype of Spitrasaurus larseni has no facets, but it has three elements in the epipodial row ( Knutsen et al. 2012d), which suggests that three humerus facets would have been present in osteologically mature individuals. Based on comparison to C. svalbardensis PMO 222.663 ( Roberts et al. 2017), the anterior facet is interpreted as for the radius, the second facet is interpreted as the ulnar facet, and the third facet for the postaxial ossicle. The fourth facet probably articulated with a second postaxial ossicle, as seen in a complete articulated forelimb of a colymbosaurine (tentatively identified as Colymbosaurus sp. ) from the Kimmeridge Clay of Wiltshire, UK (MJML K2334) ( Fig. 8 View Fig ). The ulna facet is probably the widest one, but this is not fully certain as the anterior facet (radius) is not complete in either humerus. The ulna facet is also longest in Colymbosaurus svalbardensis ( Roberts et al. 2017) . In Pantosaurus striatus the radius facet is much longer and broader than the ulna facet ( Wilhelm and O’Keefe 2010), in contrast to NHMD 189834, where the facets show the opposite relationship. The facet for the ulna is slightly concave, whereas the facets for the radius and the posterior accessory ossicle are flat.

On the distal surface of the right humerus ( Fig. 7B View Fig 2 View Fig ), a slight ridge bisects the epipodial facets. This ridge is less defined than in C. megadeirus in Benson and Bowdler (2014). In the left humerus, the presence of the ridge is uncertain.

Femora: Both femora are preserved ( Fig. 7C, D View Fig ). The femur is shorter and narrower than the humerus and has slenderer proximal ends in dorsal view due to the tuberosity being situated directly dorsal to the capitulum. In dorsal and ventral view, the shaft is narrower than the proximal head. In contrast, a proximal head and shaft of equal anteroposterior width is a unique trait of C. svalbardensis compared to all other taxa in the family Cryptoclididae ( Roberts et al. 2017) . The shaft widens anteroposteriorly, more prominently posteriorly, in the distalmost quarter of the element. This is similar to C. svalbardensis and C. megadeirus but differs from Spitrasaurus wensaasi ( Knutsen et al. 2012d; Benson and Bowdler 2014; Roberts et al. 2017; Figs. 7 View Fig , 8 View Fig ).

The distal margin of the femur bears only three facets, which differentiates it from the four-faceted humerus. Colymbosaurus svalbardensis also has three facets on the femur ( Knutsen et al. 2012c; Roberts et al. 2017; Fig. 7C View Fig 1 View Fig , C 4, D). The femoral facets are also less pronounced than the facets on the humeri. The fibula facet, situated in the middle, is the anteroposteriorly longest of the three. The posteriormost facet for the postaxial element is the shortest. The fibula facet is concave, whereas the tibia facet is flat, and the posterior accessory ossicle facet is slightly convex.

On the distal surface of one femur ( Fig. 7C View Fig 2 View Fig ), a shallow ridge bisects the epipodial facets, whereas in the other femur, this feature is less clear.

Limb elements: In total 20 limb elements are preserved in addition to the propodials. Twelve of these are flattened elements that are interpreted as epipodials and mesopodials, and eight hourglass-shaped elements, interpreted as metapodials or phalanges.

Six elements can be confidently reconstructed along sharp and tightly fitting articulations ( Fig 9A View Fig ). These comprise an epipodial row and the proximal mesopodial row. The reconstructed epipodial row forms a combination of proximal facets that most closely matches the distal facets of the humeri in size and orientation ( Fig. 9A View Fig ). On this basis the reconstructed elements are interpreted as belonging to a forelimb. The epipodial row comprises at least three subequally sized elements: the radius, ulna, and a postaxial accessory ossicle that articulates with the ulna on the postaxial margin of the limb.

The radius is almost complete but missing part of the postaxial distal margin. It has a long straight facet for the humerus, a shorter facet for the ulna, and a postaxial facet that contacts two elements in the proximal mesopodial row (a short contact with the intermedium and a long contact with the radiale). The preserved part of the preaxial margin is convex. The overall shape and size of the radius are more similar to the proportions of the tibia of Colymbosaurus megadeirus (the ulna is not known in that species) than to those in the radius of C. svalbardensis . The facet on the radius for the ulna is relatively much shorter in C. svalbardensis ( Roberts et al. 2017) .

The ulna is of a pentagonal shape in dorsal view, with distinct facets for the humerus, radius, intermedium, ulnare, and postaxial accessory ossicle. This is unlike in C. svalbardensis (PMO 222.663), which has a diamond shape ulna with strongly reduced facets for the radius and postaxial accessory ossicle ( Roberts et al. 2017). The preaxial facet on the ulna for the radius is slightly shorter than the postaxial facet for the postaxial accessory ossicle. The facets for the intermedium and ulnare are subequal. A similar condition in the hindlimb—“fibula … with facets for the fibulare and astragalus subequal in length”—was regarded as a synapomorphy of Colymbosaurus by Roberts et al (2017: 3). An additional isolated element can be identified as the second ulna because its shape and size match the equivalent element in the more complete forelimb ( Fig. 9B View Fig ). Both ulnae possess a distinct prominence in the middle of the distal margin between the facets for the intermedium and ulnare, similar to the fibula in the hindlimb of Colymbosaurus megadeirus Benson and Bowdler 2014 ).

A further three limb-bones fit well along the distal margin of the epipodial row and articulate with each other, to form a proximal mesopodial row. These are therefore identified (preaxially to postaxially) as the radiale, intermedium, and ulnare. A smaller trapezoid-shaped element with a wedge-like transverse cross section tapering preaxially or postaxially) also articulates well with this row ( Fig. 9A View Fig ), but fits equally well in the position of the second postaxial accessory ossicle, in articulation with the first postaxial accessory ossicle and the humerus (i.e., the fourth bone in the epipodial row). So, this bone cannot be placed with confidence. It could also potentially be from elsewhere in the forelimb, as many small subrectangular accessory ossicles are present along the preaxial margin of a complete articulated forelimb of a colymbosaurine (tentatively identified as Colymbosaurus sp. ) from the Kimmeridge Clay of Wiltshire, UK (MJML K2334), one in the proximal mesopodial row, one in the distal mesopodial row, and one in the metapodial row (ASS personal observations). Alternatively, this ossicle could also be from the hindlimb.

Four additional mesopodials are preserved but are difficult to identify ( Fig. 9C–F View Fig ). None of these bones seems to articulate well with the reconstructed proximal mesopodial row of the forelimb, so they are probably not from the distal mesopodial row of the forelimb, and could be mesopodials from the hindlimb instead. The fore- and hindlimbs in plesiosaurs often show the same general features ( Caldwell 1997), so comparison of these isolated bones with identified bones in the forelimb may help determine their position in the hindlimb. One of the mesopodials is a hexagonal wedge shape in dorsal view ( Fig. 9E View Fig ). It is similar, but not identical, in shape and size to the radiale. So, it could be from the equivalent position in the hindlimb, i.e., one of the tibiale. A second ossicle ( Fig. 9D View Fig ) is also hexagonal, but slightly elongate with two longer facets parallel to each other. It is similar, but not identical, in shape and size to the intermedium, so it could be the astragalus, the bone in the equivalent position in the hindlimb. The third element ( Fig. 9F View Fig ) is also hexagonal, with two longer facets subparallel to each other. It could also be an astragalus, but it is different to the one figured in Fig. 9D View Fig , so only one of them (or none of them) can be the astragalus. The fourth element ( Fig. 9C View Fig ) is the largest mesopodial in dorsal view. It is a disc-like hexagonal element with all facets subequal in length. It is also dorsoventrally shallower than the other mesopodials. It does not match any of the mesopodials in the preaxial row of the forelimb, but it is also larger in dorsal view, so it is unlikely to be a distal mesopodial. It is probably from the proximal mesopodial row of the hindlimb, may be the fibulare.

Eight phalanges are also preserved (four are shown in Fig. 9G–J View Fig ), although the largest of these may be a metapodial. They are all expanded at their distal and proximal ends, and constricted in the middle, to form an hour-glass shape in dorsal or ventral view. Some are more proximodistally elongate than other, and one is almost as wide pre-postaxially as it is long proximodistally. Some of the elements also have an additional small facet on the proximal (?) surface, similar to the notches described in the tarsals of C. megadeirus ( Benson and Bowdler 2014) .