A prevalence of Arthropterygius (Ichthyosauria: Ophthalmosauridae) in the Late Jurassic—earliest Cretaceous of the Boreal Realm

Phylogenetic analysis

For the phylogenetic analysis, we used recent matrix focused on ophthalmosaurids, presented by Zverkov & Efimov (2019). One unit, Keilhauia nui, was removed, as the specimen it is based on is considered undiagnostic in the present contribution, see ‘Discussion’. Two other units, Arthropterygius volgensis and A. chrisorum PMO 222.669 were added to the dataset. The scores for species of Arthropterygius were extended and partially changed based on new data (see Supplemental Materials for details). Six new characters related to the morphology of the supratemporal, parietal, quadrate, coracoid, and humerus were added to the dataset (for details see Table S10). The new characters were coded from the literature for taxa that we have not personally examined (Table S11; Gilmore, 1905; Broili, 1907; Andrews, 1910; Fraas, 1913; Sollas, 1916; Romer, 1968; McGowan, 1972, 1973a; Johnson, 1979; Kirton, 1983; Wade, 1984, 1990; Godefroit, 1993; Fernández, 1994, 1997, 1999, 2007a; Bardet & Fernández, 2000; Maisch & Matzke, 2000; McGowan & Motani, 2003; Kear, 2005; Motani, 2005; Maxwell & Caldwell, 2006; Druckenmiller & Maxwell, 2010; Kolb & Sander, 2009; Zammit, Norris & Kear, 2010; Fischer et al., 2011, 2012, 2014a, 2014b; Maxwell, Fernández & Schoch, 2012; Fernández & Talevi, 2014; Marek et al., 2015; Paparella et al., 2017). The analysis was performed using TNT 1.5 (Goloboff & Catalano, 2016), applying traditional search with 10,000 replicates and tree bisection and reconnection with 100 trees saved per replication. The RAM allocation was extended to 1,024 MB and the memory to 10,000 trees. Decay indices (Bremer support, “suboptimal” = 5) and resampling methods to estimate the robustness of nodes (standard bootstrapping and jackknifing, 1,000 iterations) were also computed in TNT 1.5.

In order to eliminate problematic “wildcard” taxa, we used a posteriori approach of Pol & Escapa (2009), that is, directly implemented in TNT 1.5. The two taxa (Athabascasaurus bitumineus Druckenmiller & Maxwell, 2010 and Platypterygius platydactulus Broili, 1907) were identified as unstable and pruned from the second analysis. The pruned dataset was analysed using the exact same procedures as was used for the full dataset.

Principal component analysis

To compare humeri of ophthalmosaurids we gathered a series of metrics and ratios that collectively summarize morphology of the humerus (Tables S6 and S7). The metrics are: proximodistal length of the humerus, anteroposterior width of humeral proximal and distal ends, thickness of humeral proximal end; dorsoventral width of humeral distal end; anteroposterior width at midshaft, anteroposterior and dorsoventral width of the distal facets, and the angle between the ulnar and radial facets (for details see Fig. S1). Based on the metrics the following ratios were calculated (Table S7):

1. Humeral proximal expansion: anteroposterior width of humeral proximal end divided by the humeral proximodistal length.

2. Humeral distal expansion: anteroposterior width of humeral distal end divided by the humeral proximodistal length.

3. Humeral stoutness: humeral minimal anteroposterior width at diaphysis divided by the humeral proximodistal length.

4. Humeral proximodistal proportionality: anteroposterior width of humeral proximal end divided by the same measurement of its distal end. The character based on this ratio is used in current phylogenetic analyses and distinguish ophthalmosaurids, which commonly have nearly equal proximal and distal humeral ends or proximal end slightly wider than the distal end see, for example, Fischer et al. (2011: Character 32).

5. Isometry of the humeral proximal end (or “anteroposterior elongation” of the humeral proximal end): anteroposterior width of humeral proximal end divided by the thickness of humeral proximal end (see Fig. S1). This ratio has extremely high value in “Grendelius” zhuravlevi (2.587) for which strongly compressed humeral proximal end is considered as autapomorphic (Zverkov, Arkhangelsky & Stenshin, 2015); the standard values for ophthalmosaurids are 1.8–1.5; for taxa with “isometric” humeral proximal end this value could be close to one (e.g., Undorosaurus nessovi and Platypterygius platydactylus see Table S7).

6. Humeral distal compression: anteroposterior width of humeral distal end relative to the maximal dorsoventral width of humeral distal end.

7. Relative anteroposterior width of facet for preaxial accessory epipodial element and radial facet.

8. Relative anteroposterior width of ulnar and radial facets. As well as for ratio 4, there is a character based on similar ratios in current phylogenetic analyses, see, e.g., Motani (1999: Character 52) and Moon (2019: Character 209). However, the referred character use “relative size” of ulnar and radial facets, which is not always clear as ulnar facet could be longer than radial facet but the same time, less wide dorsoventrally (as in most specimens of Arthropterygius). In this regard, it is better to consider separately relative anteroposterior width of ulnar and radial facets and relative dorsoventral width of ulnar and radial facets.

9. Relative dorsoventral width of ulnar and radial facets.

The dataset is resolved at the specimen level with left and right humeri considered separately in order to reveal the existing humeral asymmetry within an individual and to assess its possible effects on the results. Data (see Tables S6 and S7) were collected based on personal observations of NGZ and completed by measurements and in rare cases analysis of pictures of the following references: (Broili, 1907; Nace, 1939; Kuhn, 1946; Wade, 1984; Delair, 1986; McGowan, 1972; Arkhangelsky, 1998; Kolb & Sander, 2009; Maxwell, 2010; Maxwell & Kear, 2010; Moon & Kirton, 2016). Only humeri with all documented ratios were considered, in rare cases, we completed our dataset by approximate ratios estimated based on oblique views (the case of B. extremus and P. platydactylus) or proportionally translated from other conspecific individuals (the case of P. americanus). The final dataset consisted of 55 humeri belonging to 45 individuals and 10 variables (Table S8). The ratios and angle between the ulnar and radial facets (in rad) were used as variables for the PCA. Data were scaled to equal variance by subtracting the mean value for each variable and then dividing each variable by the standard deviation. We then created a distance matrix with these data (Table S8). The dataset was analysed in PAST v. 3.20 (Hammer, Harper & Ryan, 2001).

Remarks on the synonymy of Arthropterygius, Palvennia, Janusaurus, and Keilhauia

Based on the incomplete type specimen of A. chrisorum solely, the autapomorphic features of Arthropterygius are: basisphenoid with foramen for the internal carotid arteries opening posteriorly; basioccipital facet of the basisphenoid facing posterodorsally and occupying a half of the element in dorsal view; basioccipital with extracondylar area wide in lateral view and practically unseen in posterior view; shifted anteriorly stapedial and opisthotic facets of the basioccipital; presence of “ulnar torsion,” with ulnar facet not as dorsoventrally wide as the radial facet, forming a distal skew of the humeral ventral surface (Maxwell, 2010; Zverkov et al., 2015; Nikolay G. Zverkov, personal observations, 2015–2016). All these features could be observed in the type specimens of genera that are here synonymized with Arthropterygius, except for cases where an element is unknown or obscured from observation: basisphenoid is mostly hidden in the holotype of Janusaurus lundi; humerus is incomplete in the holotype of Palvennia hoybergeti; basicranial elements are not preserved in the holotype of Keilhauia nui, and both the basioccipital and humerus are absent in the holotype of Ichthyosaurus volgensis.

The epipodial elements angular in outline for articulation with humerus in the holotype of Arthropterygius chrisorum were also considered as an autapomorphy of Arthropterygius (Maxwell, 2010). However, this condition was later reported for another ophthalmosaurid (Zverkov et al., 2015) and, in this regard, should be further considered with caution.

From our observations on the holotype of Keilhauia nui (PMO 222.655), we are unable to confirm any of the evidence proposed by Delsett et al. (2017) as supporting the maturity of the specimen. The specimen is too fragmentary and poorly preserved, thus we consider it referable to Arthropterygius only in open nomenclature. For details on this decision see Discussion.

In previous phylogenetic analyses, the position and relations of Arthropterygius, Janusaurus, Palvennia, and Keilhauia varied sufficiently and these taxa never formed a clade (Roberts et al., 2014; Delsett et al., 2017; Fischer et al., 2016; Maxwell et al., 2016; Motani et al., 2017; Paparella et al., 2017; Moon, 2019). In some phylogenetic hypotheses, Arthropterygius, Janusaurus, and Palvennia were recovered as sister taxa to each other, but also to a clearly distinct ophthalmosaurid “Cryopterygius” (Roberts et al., 2014; Maxwell et al., 2016; Delsett et al., 2016). These results could have been considered as an argument for the validity of all these genera, however, the recent results of Zverkov & Efimov (2019) recovered Arthropterygius, Janusaurus, Palvennia, and Keilhauia in a clade distantly related to “Cryopterygius.” This clade, called the “Arthropterygius clade” in Zverkov & Efimov (2019), has relatively good support and its potential genus rank was announced in that paper. The phylogenetic results of the present contribution (see below) helps to further develop this idea. In this research, Arthropterygius clade is supported by nine autapomorphies: posterior position of the foramen for internal carotid arteries; dorsally facing basioccipital facet of the basisphenoid; raised opisthotic facet of the basioccipital; anteriorly shifted stapedial and opisthotic facets of the basioccipital; gracile stapedial shaft; weak quadrate condyle; robust clavicles; ulnar facet/radial facet ratio less than 0.83; pronounced angle between the articulated coracoids. For the taxa within the Arthropterygius clade, no unambiguous autapomorphies are found so that their generic independency could hardly be justified.

• Arthropterygius chrisorum (Russell, 1994)

• (Figs. 2–8)

• v.1910? Ophthalmosaurus thyreospondylus Owen; Bogolubov: 474

• *1994 Ophthalmosaurus chrisorum Russell: 198, fig. 3

• 2010 Arthropterygius chrisorum (Russell, 1994); Maxwell: 404, figs. 2–5

• v.2018 Palvennia hoybergeti Druckenmiller et al., 2012; Delsett, Druckenmiller, Roberts, Hurum: 8, figs. 5–13 [pars.]

Holotype: CMN 40608, fragmentary skeleton of a large mature individual (for details see Maxwell, 2010).

Referred specimens: PMO 222.669, SGM 1573, CCMGE 3–16/13328, CCMGE 17–44/13328. See Table 1

Emended diagnosis: A moderately large (four to five m) ichthyosaur, diagnosed relative to other species of Arthropterygius by the following unique characters: quadrate with strongly ventrally shifted articular boss, V-shaped in posteromedial view; absence of pronounced angular protrusion of the quadrate (known for PMO 222.669, CCMGE 3–16/13328, CCMGE 17–44/13328); basisphenoid trapezoidal in outline with maximum mediolateral width in its anterior part; posterior foramen for the internal carotid arteries not visible in ventral view in adults, separated from the ventral surface by a thin shelf (known for CMN 40608, PMO 222.669, CCMGE 3–16/13328, CCMGE 17–44/13328); dorsoventrally high opisthotic with extremely reduced and robust paraoccipital process (hitherto found only in PMO 222.669); blunt termination of the lateral extremities of the interclavicle (this complete element is found only in SGM 1573); strongly anteroposteriorly elongated proximal end of the humerus with reduced deltopectoral crest shifted to its anterior edge (humeri are known for all the referred specimens); extremely pronounced ventral skew between the ulnar and radial facets of the humerus; facet for the anterior accessory epipodial element of the humerus as wide as, and equal in size to the radial facet.

Occurrence: Upper Jurassic, Deer Bay Formation (likely its Volgian part) of Melville Island, Northwest Territories, Canada (type locality); Middle Volgian Promza Formation (Dorsoplanites panderi Ammonite Biozone) of Ulyanovsk Region, Russia; upper part of the Hofer Formation (uppermost Volgian to lowermost Ryazanian, Berriassian) of Franz-Josef Land, Russian Extreme North; Slottsmøya Member of the Agardhfjellet Formation (Middle Volgian part of the section) of Svalbard, Norway.

Appendicular skeleton

Scapula. The left scapula is completely preserved in CCMGE 17–44/13328 (Figs. 7J–7M). The element is robust: its proximodistal length is shorter than the coracoid anteroposterior length. It is similar to that of Ophthalmosaurus icenicus in general morphology (Seeley, 1874; Andrews, 1910; Moon & Kirton, 2016). The scapular shaft is mediolaterally flattened and elongated-oval in cross-section. The glenoid contribution is well developed and equal in length to the coracoid facet. The acromial process is massive and well-prominent; it curves ventrolaterally, forming a nearly right angle with the lateral surface of the scapula (Fig. 7N).

Coracoid. The coracoid of CCMGE 3–16/13328 is slightly longer anteroposteriorly than wide mediolaterally (Fig. 7P). It is similar to that of Ophthalmosaurus icenicus and Undorosaurus gorodischensis (Andrews, 1910; Moon & Kirton, 2016; Zverkov & Efimov, 2019), but differs in relative size, being anteroposteriorly longer than scapular proximodistal length. The medial symphysis is lenticular in outline; it occupies anterior two-thirds of the medial surface. The anteromedial process is prominent, laterally limited by an extensive anterior notch (anterior notch is relatively smaller in CCMGE 3–16/13328 than in the holotype, most likely as a reason of immaturity). The posterior portion of the coracoid is strongly compressed and convex posteriorly (Fig. 7P). The most interesting trait is that articulated coracoids form a pronounced angle of 100° (Fig. 7O); this condition is unique for Arthropterigius. The scapular facet and glenoid contribution are offset by an angle of c. 140°. Their surfaces are slightly convex and tuberous. The glenoid contribution surface is parallel to the medial symphysis of the coracoid, thus coracoid mediolateral length is constant, unlike caudally constricting coracoids of Sveltonectes (Fischer et al., 2011), Nannopterygius (Hulke, 1871; Kirton, 1983), and “Paraophthalmosaurus” (Arkhangelsky, 1997; Efimov, 1999a) and caudally expanding coracoids of Undorosaurus (Efimov, 1999b).

Clavicle. The clavicle of CCMGE 3–16/13328 (Figs. 7X–7Z) is a large and robust element. It is very similar to that of A. lundi, being dorsoventrally high and anteroposteriorly thick, compared to other known ophthalmosaurids. On its medial surface, there is a rugose circular facet for articulation with the acromial process of the scapula (Fig. 7Y). This facet is pronounced, but not as well developed as in A. lundi (see below).

Interclavicle. The interclavicle of SGM 1573 is a large and slender T-shaped element. The anterior transverse bar of the interclavicle is straight, with a high dorsally rising wall; its lateral extremities extend far laterally, and their ends are rounded (Figs. 7CC and 7DD). There is no ventral knob observed in Undorosaurus gorodischensis and Grendelius alekseevi (Zverkov, Arkhangelsky & Stenshin, 2015; Zverkov & Efimov, 2019). The posterior median stem is slender and bears a shallow trough along its dorsal surface. There is a prominent bulge in the middle of the ventral surface of the stem (Figs. 7CC and 7DD). In PMO 222.669 a displaced portion of the clavicle was erroneously interpreted as a wide interclavicle posterior median stem (Delsett et al., 2018). In fact, the interclavicle of PMO 222.669 is heavily distorted and broken into several disarticulated pieces due to a collapsing of pectoral girdle during the taphonomic process, but judging from the preserved fragments, its posterior median stem was quite slender.

Humerus. Humeri are known for all the referred specimens. The humerus is a large and robust bone with wide and dorsoventrally compressed midshaft. The humeral “torsion” (angle between the long axes of the proximal and distal ends of the humerus) is c. 70°. The dorsal process is prominent and plate-like, extending up to the half of the humeral midshaft (Figs. 7C, 7F and 7S). The deltopectoral crest is poorly developed and shifted to the anterior border of the humerus (Figs. 7A, 7E, 7G, 7I, 7T and 7W). The proximal end is semi-rectangular in outline, being anteroposteriorly longer than dorsoventrally thick (Figs. 7E, 7I and 7W). There are three distal concave facets for the preaxial accessory element, radius, and ulna. The facet for the preaxial accessory element is large and semicircular in outline; it occupies nearly equal space as the radial facet. The radial facet is irregularly pentagonal in outline; its ventral edge is angular, forming in posterior half an abrupt skew to the ulnar facet (Figs. 7D, 7H and 7V). A ratio of the dorsoventral width of the radial facet to ulnar facet is 0.7–0.8 (see Table S7). Among ophthalmosaurids, the pronounced constriction between the ulnar and radial facet with a ventral skew is also known for “Macropterygius” and Sisteronia, which at the same time lack a well-developed facet for an anterior accessory epipodial element characteristic of Arthropterygius (Fischer et al., 2014b; Moon & Kirton, 2018).

Epipodial elements. The articular surfaces of the epipodial elements are convex for a peg-and-socket articulation with the concave distal humeral facets; however, this condition varies even in mature specimens from extremely deep in CMN 40608 to more shallow in SGM 1573. The anterior accessory epipodial element is circular in dorsal view; its anterior edge lacks perichondral ossification as in Ophthalmosaurus icenicus (Andrews, 1910; Moon & Kirton, 2016). This element rapidly tapers anteriorly. The radius is pentagonal in dorsal and ventral views (Figs. 6A and 6F). The ulna is the largest element in the epipodial row, its dorsal and ventral cortical parts are roughly hexagonal in outline. The element gradually constricts in dorsoventral width posteriorly. A perichondral ossification of the posterior edge of the ulna is absent (Fig. 6A). The intermedium wedges between the radius and ulna, but not reach the humerus, however, a distance between the humerus and intermedium varies from relatively short in CCMGE 3–16/13328 and CMN 40608 to relatively long in CCMGE 17–44/13328. Distally intermedium bears two slightly demarcated facets for distal carpals three and four, indicating a “latipinnate” forefin architecture. Maxwell described the distal margin of the intermedium of CMN 40608 as “gently curved” but indicated that the distal edge of the intermedium forms a surface for the articulation of a single distal carpal’ (Maxwell, 2010: 410), considering this uncertainty and the new data on other referred specimens (CCMGE 3–16/13328, CCMGE 17–44/13328, PMO 222.669), it is more likely to interpret the presence of the two poorly demarcated facets for distal carpals three and four in the holotype (CMN 40608) rather than a single convex facet.

Distal limb elements. All the mesopodial and autopodial elements in CCMGE 17–44/13328 and PMO 222.669 are strongly dorsoventrally thickened, circular in outline and loosely packed, indicating a large amount of cartilage in the forefin, which is most similar to the condition observed in Ophthalmosaurus icenicus (Andrews, 1910; Moon & Kirton, 2016). One of the elements in CCMGE 17–44/13328 has a semicircular outline in dorsal view and bears a perichondral ossification along one of its edges, this probably represents a pisiform (Fig. 6A). The pisiform of exact same morphology is present in the left limb of PMO 222.669 (Nikolay G. Zverkov, 2018, personal observation).

Pelvic girdle. The only central portion of the ischiopubis has been collected for CCMGE 17–44/13328, which complicates the description of the element. The ischiopubis is plate-like, mediolaterally compressed (eight mm at its thickest part). The obturator foramen is likely absent (Fig. 8G).

Femur. The femur of CCMGE 17–44/13328 is slender with proximal and distal ends only slightly expanded (Fig. 8A). Its proximodistal length comprises 0.74 of the humeral proximodistal length (0.67 in the holotype CMN 40608). The femur of CCMGE 17–44/13328 is very similar to that of the holotype, possessing flattened ventral process terminating proximal to the mid-point, and thereby being more prominent than that of A. lundi (Roberts et al., 2014). The dorsal process is less pronounced than the ventral process and shifted to the anterior edge of the femur. There are two distal facets, which are concave and poorly demarcated, forming a common distal groove for the epipodial elements (Fig. 8D). The fibular facet is slightly inclined posterodistally, whereas the tibial facet faces nearly distally.

Measurements: See Tables S1 and S2.

• Arthropterygius hoybergeti (Druckenmiller et al., 2012) comb. nov.

• (Figs. 9–12)

• v*2012 Palvennia hoybergeti Druckenmiller, Hurum, Knutsen & Narkem: 326, figs. 12–21

Holotype and only referred specimen: SVB 1451, see Table 1.

Emended diagnosis. A moderately large ophthalmosaurid (up to four m) distinguished from other species of Arthropterygius by the following unique character combination: basisphenoid longer anteroposteriorly than mediolaterally wide, with the widest part in the region of basipterygoid processes (unlike in A. volgensis); posterior foramen for internal carotid arteries opening on the posteroventral edge of the basisphenoid and forming a notch as in A. lundi (referred specimen SGM 1502) and unlike A. chrisorum; small basioccipital facet of the opisthotic (large in other known species of Arthropterygius); relatively large teeth with circular in cross-section roots and robust ridged crowns as in A. chrisorum but unlike gracile subtly ridged crowns of A. lundi; well developed deltopectoral crest (unlike in other species of Arthropterygius); anterodistal facet for the anterior accessory epipodial element sufficiently smaller than the radial facet, being thus relatively smaller than that in A. lundi and A. chrisorum.

Occurrence: Arthropterygius hoybergeti is known from the Slottsmøya Member of the Agardhfjellet Formation of Svalbard (type locality), where it was found most likely within the Dorsoplanites ilovaiskii Ammonite Biozone (lower Middle Volgian). Two specimens from the Volga Region (both found on the right bank of the Volga River near Gorodischi Village, Ulyanovsk Region) referred here to as A. cf. hoybergeti are corresponding to Dorsoplanites panderi Ammonite Biozone of Promza Formation.

Skull

Supratemporal. A posterodorsal portion of the right supratemporal is preserved (for the figure see Kasansky, 1903, Table 1; Fig. 10). The medial ramus is massive and mediolaterally short, it bears a concave facet for articulation with the parietal.

Parietal. The parietal is well preserved and similar to that of other Arthropterygius species. It possess a relatively elongated and slender supratemporal process (Fig. 16P). The posterodorsal surface of the supratemporal processes is rugose with the central ridge that contributed to a somewhat peg-and-socket articulation with the supratemporal (Fig. 16P). The medial articular facet is anteroposteriorly shortened; its surface is deeply ridged for a strong interdigitating articulation with the contralateral parietal. Posterior to the facet is a pronounced notch of finished ossification (Figs. 16P and 16S). Anteriorly, the parietal bears rugose facets for articulation with the frontal and postfrontal. The frontal facet completely occupies the anterior edge of the parietal, thus, the parietal unlikely contributed to the posterior border of the parietal foramen. Ventral surface of the element is divided into two areas: the deep and extensive impression of the cerebral hemisphere occupy more than a half of the anterior ventral surface (Fig. 16R, ich); posteriorly situated optic lobe impression, which is roughly circular in outline, occupies the rest of the element (Figs. 16R, iop). The dorsal surface of the parietal is convex and nearly horizontal along the midline in lateral view. There is no sagittal eminence.

Quadrate. The articular condyle of the quadrate is relatively reduced and dorsoventrally low compared to that of A. chrisorum. The articular and surangular bosses of the condyle are nearly equal in size (Figs. 16O and 16N). The articular boss is only slightly more pronounced ventrally, however, its ventral margin is gradually curved, but not V-shaped as in A. chrisorum. The facet for the quadratojugal is a small depression on the dorsal surface of the condyle (Fig. 16N). The quadrate foramen is shallow due to a reduction of the articular condyle and the occipital lamella (Fig. 16L). The occipital and pterygoid lamellae are slightly demarcated one from another forming an angle of c. 145°. A circular depression of the stapedial facet is located in the middle of the medial surface (Fig. 16L).

Basisphenoid. The basisphenoid is square in ventral view: its posterior and anterior ends are nearly equal in mediolateral length (Fig. 16A). The basipterygoid processes are reduced and faced anterolaterally. The basioccipital facet is a broad hexagonal irregularly pitted surface that faces posterodorsally. A pentagonal dorsal plateau is mediolaterally wide. The stapedial facet is oblique and relatively small (Fig. 16C). The anterior wall is high and vertical, even on the lateral sides. The dorsum sellae, located in the middle of the anterior surface, is smoothly bordered from the rest of the anterior wall (Fig. 16D). The impressions of a cartilaginous continuation of the crista trabecularis are well-pronounced (Fig. 16D). The posterior foramen for the internal carotid arteries opens posteroventrally, forming a medial notch of the posteroventral edge of the basisphenoid, as is CCMGE 3–16/13328, which may be due to the immaturity of these individuals.

Opisthotic. The paraoccipital process of the opisthotic is shortened and robust, however, this could be regarded as an immature condition as was discussed by Kear & Zammit (2014). The facet for the supratemporal is triangular in outline (Fig. 16H). The lateral muscular ridge is well pronounced. The trapezoid in outline stapedial facet is larger than the facet for the basioccipital, which is quadrant in outline with convex margin directed dorsolaterally (Figs. 16F and 16K). The stapedial facet bears a deep straight mediolateral groove either for VII or for IX nerve in its middle (Fig. 16K). The impressions for the semicircular canals of the otic capsule are deep and nearly equal in length as in other species of Arthropterygius. The impression of the posterior vertical semicircular canal is wider than that of the horizontal semicircular canal. The impression housing the posterior ampulla and the sacculus is expanded (Fig. 16J).

Articular. The articular is anteroposteriorly elongated and trapezoid in outline (Figs. 16T and 16U). It is similar to that of Arthropterygius lundi (Roberts et al., 2014) and PMO 222.669, the referred specimen of A. chrisorum, being more anteroposteriorly elongated than in holotypes of A. chrisorum and A. hoybergeti (cf. Fig. S4; Maxwell, 2010; Delsett et al., 2018).

Axial skeleton. The detailed description and measurements of the vertebral column (which is nowadays lost) were provided by Kasansky (1903).

Pectoral girdle. The preserved middle fragments of clavicles (Fig. 17H) have a morphology common of ophthalmosaurids: these are anteroposteriorly thin and dorsoventrally high elements, curving in dorsolateral direction. The clavicles are dorsoventrally high as in other species of Arthropterygius. The interclavicle (Figs. 17H and 17I) is a relatively large element, being approximately 2/3 of the coracoid length. Its posterior median stem is shaft-like, ventrally convex, and dorsally bearing a shallow trough. The scapula is incompletely preserved in two fragments. The acromial process of the scapula is large and flattened, anteroventrally curving at the anterior edge (Fig. 17G). The scapular shaft is mediolaterally compressed, as in other species of Arthropterygius and ophthalmosaurines Ophthalmosaurus icenicus and Acamptonectes densus (Fischer et al., 2012; Moon & Kirton, 2016). Both coracoids are well preserved, they are rounded in general outlines; however, their anteroposterior length slightly exceeds mediolateral width. The ventral surface of the element is slightly saddle-shaped (Fig. 17B), whereas the dorsal surface is nearly flat (Fig. 17A). The scapular facet is demarcated by an obtuse angle (160°) from the glenoid contribution. The medial symphysis is dorsoventrally thin, extending along anterior two-thirds of the coracoid, as in A. chrisorum and A. lundi (Roberts et al., 2014). The angle between articulated coracoids is close to 90° (Fig. 17E).

Femur. The only distal portion of the right femur is preserved (Figs. 17J–17M). Its distal facets are poorly ossified and slightly demarcated, thus it is even hard to say, whether two or three distal facets are present (Figs. 17J, 17K and 17M). The ventral process, located in the middle of the ventral surface is more prominent than the anteriorly shifted dorsal process (Fig. 17L).

Remarks. Kasansky originally identified the femur as a humerus, at the same time two broken pedicles of the neural arches were identified as femora (Kasansky, 1903).

The holotype and only known specimen KSU 982/P-213 is a juvenile individual, thereby the value of features used as diagnostic could be questioned. Indeed, a number of observed traits could be interpreted as juvenile conditions: reduced occipital lamella of the quadrate, minimally developed basipterygoid processes, and short paroccipital process of the opisthotic (see Kear & Zammit, 2014). However, a series of specimens of presumably different age classes available now for A. chrisorum allows supporting some of our conclusions. Although the relative development of the basipterygoid processes of the basisphenoid during the ontogeny is supported by our observations, we state that the general ventral (or dorsal) outline of the basisphenoid is stable between all the age classes. Kear & Zammit stated that in the in utero P. australis “the basipterygoid processes are minimally developed, giving the basisphenoid a much narrower anterior profile when compared with those of adults” (Kear & Zammit, 2014: 77). Based on this, they concluded that for characters dealing with a shape of basipterygoid processes, that is, Maxwell (2010: char. 11) and Fischer et al. (2011: char. 17 and 2012: char. 16), foetal individual scores differently than mature ones. However, this is a subjective observation, as both foetal and mature P. australis, regardless the state of development of basipterygoid processes, preserve generally “pentagonal” (or, it is better to say, trapezoidal) ventral outline of the basisphenoid with anterior region markedly wider than the posterior part. This is clearly seen from the Fig. 5M of Kear & Zammit (2014). In contrast, taxa with “square” ventral outline of the basisphenoid always have the same width of anterior and posterior basisphenoid (Nikolay G. Zverkov, 2015–2018, personal observation). All specimens of Arthropterygius chrisorum have basisphenoid, that is, mediolaterally wider anteriorly than posteriorly. Indeed, the small (juvenile) CCMGE 3–16/13328 has narrower anterior profile when compared with those of large individuals CCMGE 17–44/13328 and CMN 40608 (Fig. 18), supporting the observation of Kear & Zammit (2014); still the anterior region of the basisphenoid of small individual CCMGE 3–16/13328 is wider than the posterior region (Fig. 18B). In contrast, the posterior region of the basisphenoid of KSU 982/P-213 is wider than the anterior region (Fig. 18R); although CCMGE 3–16/13328 and KSU 982/P-213 represent presumably close ontogenetic stages (basisphenoid and quadrate of KSU 982/P-213 are slightly smaller, whereas coracoid is bigger than those of CCMGE 3–16/13328). Another marked difference of CCMGE 3–16/13328 and KSU 982/P-213 is the shape of the condyle of their quadrates. Whereas CCMGE 3–16/13328, CCMGE 17–44/13328, and PMO 222.669, regardless differences in size, have similar shape of the condyle, KSU 982/P-213 differs in having less dorsoventrally high condyle with gradually curving (not V-shaped) ventral margin. This suggests that the shape of the quadrate could also be regarded as interspecifically and ontogenetically stable feature. Thereby we conclude that at the current state of knowledge, A. volgensis should be regarded as a distinct valid species of Arthropterygius rather than a synonym of other known species of the genus or a nomen dubium.

• Measurements. See Kasansky (1903) and Table S4.

Phylogenetic analysis

Our analysis of the full dataset recovered ten most parsimonious trees of 310 steps with the consistency index (CI) = 0.416 and retention index (RI) = 0.662. The strict consensus (length of 321 steps; CI = 0.402; RI = 0.642) is relatively well resolved, however, supports for relationships within Ophthalmosauridae are still low (Fig. 19). Despite the modifications of the original matrix, the recovered topology is nearly identical to that of Zverkov & Efimov (2019), except for minute changes in relations of derived-most platypterygiines that are even more badly resolved. A clade that includes species of Arthropterygius (“A” in Fig. 19) is recovered as the sister group to Platypterygiinae. Sister relations of Arthropterygius and platypterygiines are supported by two synapomorphies: “T”-shaped prootic osseous labyrinth (49.0→49.1) and absence of the obturator foramen in the ischiopubis (98.1→98.2).

Only two most parsimonious trees (length of 300 steps, CI = 430, RI = 662) were recovered by the pruned analysis. In the strict consensus tree (length of 302 steps, CI = 425, RI = 656; Fig. 19B), Platypterygiinae is relatively better resolved. Surprisingly, Caypullisaurus is found as a sister, not to Grendelius, but to Leninia (based on two non-unique synapomorphies: presence of prefrontal dorsomedial expansion (16.0→16.1), and squared sqamosal (34.1→34.0). However, the relations of derived platypterygiines is not a focus of the current paper. Concerning this manuscript, Arthropterygius clade is recovered as a sister group to ophthalmosaurines. These two groups from a clade with low support, and share three synapomorphies (presence of the lateral “wing” of the nasal (14.0→14.1); absence of supratemporal-postorbital contact (27.1→27.0); and circular shape of the basioccipital condyle (43.1→43.0).

The Arthropterygius clade is supported by nine autapomorphies: posterior position of the foramen for internal carotid arteries (unique, 40.1→40.2); dorsally facing basioccipital facet of the basisphenoid (non-unique, 41.0→41.1); raised opisthotic facet of the basioccipital (non-unique, 46.0→46.1); anteriorly shifted stapedial and opisthotic facets of the basioccipital (unique, 47.0→47.1); gracile stapedial shaft (non-unique, 52.0→52.1); robust clavicles (unique, 78.0→78.1), ulnar facet/radial facet ratio less than 0.83 (unique, 84.0→84.1); weak quadrate condyle (non-unique, 110.0→110.1); angle between the articulated coracoids less than 110° (unique, 111.0/1→111.2).

In both the full and pruned analyses the Arthropterygius clade has a comparatively high Bremer support values (four and five), Bootstrap and Jackknife (more than 80), thus being the best supported clade in our analyses (Fig. 19). The result further augment our taxonomic decision, leaving no substantial reasons to consider taxa within the Arthropterygius clade as representatives of separate genera.

Multivariate analysis of ophthalmosaurid humeral morphology

The first two axes describe over 60% of the total variance (38.12% and 21.9%, respectively). All variables show low loadings on PC1 (≥0.50; <0.50); among them better pronounced are humeral distal expansion (variable 2: −0.46), humeral proximodistal proportionality (variable 4: 0.43), humeral distal compression (variable 6: −0.4), relative size of faae (variable 7: −0.42), as well as relative dorsoventral width of ulnar and radial facets (variable 9: 0.32) and an angle between these facets (variable 10: 0.28). For the PC2 highest positive loadings are shown by variables 1 (0.58), 4 (0.36), 8 (0.37) and negative loadings by variables 9 (−0.33) and 10 (−0.42). Thereby PC2 characterize humeral proximal expansion, humeral proximodistal proportionality, relative anteroposterior (v8) and dorsoventral (v9) width of ulnar, and radial facets as well as angle between these facets (v10). The distribution of variable loadings could be found in Table S9.

Considering low sampling for some of the taxa in our analysis, it is hard to say with confidence if the absence of marked morphospace overlap between the Jurassic ophthalmosaurid taxa is a true condition, or it is biased by the sampling. Whether or not, it is clear that most of the Late Jurassic ophthalmosaurid genera are well separated, forming four clusters: Brachypterygius–Grendelius cluster (high values on PC1 and low values on PC2), Ophthalmosaurus cluster (low values on both PC1 and PC2), Arthropterygius cluster (low values on PC1, high values on PC2) and Undorosaurus cluster (moderate values on both the axes, so that its morphospace is between the others) see Fig. 20.

Our PCA demonstrate a relatively wide morphospace occupation for species of Arthropterygius, which is mostly due to Arthropterygius cf. hoybergeti, having humeri that are morphologically closer to “moderate” ophthalmosaurid condition and thereby falling closer to Undorosaurus gorodischensis and Platypterygius hercynicus (Fig. 21). A. lundi falls close to A. chrisorum (Fig. 20).

Undorosaurus gorodischensis morphospace is separated from other species of Undorosaurus by the first principal component axis, as U. nessovi and U. trautscholdi demonstrate low positive and negative values on PC1, and more pronounced negative value on PC2 in case of U. nessovi. In general morphology, U. gorodischensis have anteroposteriorly elongated humeral proximal end, that is, of roughly oval outline, whereas U. nessovi and U. trautscholdi are characterized by a nearly circular outline of the humeral proximal end, which is depicted by PC1 partially responsible for humeral proximodistal proportionality, in this regard U. nessovi is closer to Ophthalmosaurus icenicus morphospace (Fig. 20).

Several derived Cretaceous platypterygiines, added to our analysis, occupy different parts of the morphospace also demonstrating the potential of humeral morphology for distinguishing Cretaceous ichthyosaurs. At the same time, they occur in the morphospace of Middle to Late Jurassic ophthalmosaurids. Platypterygius australis is found within the morphospace of O. icenicus, P. americanus is found in morphospace of Brachypterygius–Grendelius and Undorosaurus and P. hercynicus shows high positive values on PC2 and cross the morphospace of Undorosaurus and Arthropterygius. Only Plutoniosaurus and Sveltonectes occupy the new morphospace with no overlap with Jurassic ophthalmosaurids.

The interesting result of our analysis is that in some ophthtalmosaurid individuals left and right humeri can fall wider to each other than to humeri of other specimens of the species and even to other species and genera, indicating the presence of a pronounced humeral asymmetry in ophthalmosaurids. The most outstanding specimen with humeral asymmetry in our analysis is Platypterygius hercynicus. Although the asymmetry is a normal condition in all bilaterally symmetric organisms, and a marked example of natural humeral asymmetry was recently described for Undorosaurus (Zverkov & Efimov, 2019), it is more likely that pronounced asymmetry, like that in P. hercynicus, could be explained by artefacts of preservation and/or pathologies. This assumption is consistent with the recent study of the effect of taphonomy on asymmetry in fossil organisms (Hedrick et al., in press).

New concept of Arthropterygius: discussion of taxonomic decisions

New observations and results of our phylogenetic analysis, in our opinion, do not allow to further support the generic validity of ichthyosaurs from Arthropterygius clade. Within the clade, A. volgensis and A. lundi are recovered in a polytomy at the base. No autapomorphies are found for A. volgensis and a single non-unique autapomorphy, long anterior process of the maxilla (10.0→10.1), characterises A. lundi. The smaller clade of A. hoybergeti and A. chrisorum is characterized by two non-unique autapomorphies: pronounced striation of the crown (1.1→1.0) and well-developed occipital lamella of the quadrate (36.1→36.0). No autapomorphies are found for A. hoybergeti, whereas there are two for A. chrisorum: reduction of the angular process of the quadrate (109.0→109.1) and a very large anterior accessory epipodial facet of the humerus (112.1→112.2). In our opinion, none of the autapomorphies characterising the taxa within the discussed clade is sufficient for applying an alternative taxonomic context with Janusaurus and Palvennia being regarded as valid genera, sister to a more derived Arthropterygius.

The results of PCA shows that the morphospace occupied by ichthyosaurs of Arthropterygius clade is only slightly larger than the morphospace occupied by Ophthalmosaurus icenicus solely. At the same time, no sufficient overlap with other ophthalmosaurid clusters is observed, which could also be considered as support of the new taxonomic context for Arthropterygius.

On Fig. 21, we summarized the importanl overlapping skeletal elements of A. chrisorum (holotype, CMN 40609, and the referred specimens from Franz Joseph land), “Keilhauia nui” (holotype, PMO 222.655), “Palvennia” hoybergeti (holotype, SVB 1451, and referred specimen PMO 222.669), “Janusaurus” lundi (PMO 222.654), and “Ichthyosaurus” volgensis (KSU 982/P-213) and compared them with other well-known contemporary taxa. Following is consideration of taxonomic attribution of specimens referred to as Arthropterygius in the present contribution.

In the original description of Palvennia hoybergeti (Druckenmiller et al., 2012), the only considered overlapping elements with the holotype of A. chrisorum (CMN 40608) were the basioccipital, atlas-axis complex and the anterodistal fragment of the humerus. Only the basioccipital was compared with that of A. chrisorum and their similarity was noted: “in posterior articular view the basioccipital of Palvennia is nearly oval in outline with only the condyle visible, more similar to Arthropterygius (Maxwell, 2010) and Platypterygius australis (Kear, 2005)” (Druckenmiller et al., 2012: 337). As a difference between the two, the following feature was proposed: “in Arthropterygius, the anterior face of the basioccipital possesses a notochordal pit and a distinct basioccipital peg, both of which are absent in Palvennia” (Druckenmiller et al., 2012: 337). Based on our observations on the holotype of Palvennia hoybergeti (SVB 1451) we are unable to support the latter conclusion. An anterior protrusion of the basioccipital under the floor of the foramen magnum interpreted by Maxwell (2010) as an “incipient basioccipital peg” is also present in P. hoybergeti (SVB 1451) and J. lundi (PMO 222.654) although in a slightly lesser degree (Nikolay G. Zverkov, 2017, personal observation). This anterior protrusion was reported for some other ophthalmosaurids (see Moon & Kirton, 2016: 41). Although this structure is a vestige of a basioccipital peg, the condition observed in Arthropterygius could not be considered as a plesiomorphic state, as was supposed and coded in some previous works (Fischer et al., 2011, 2012). When present, the basioccipital peg is large and tapered, terminating in a pointed tip, but not in a flattened pitted facet, as it does in Arthropterygius and some other ophthalmosaurids (see Discussion of this structure in Moon & Kirton (2016: 41)). It is hard to verify the absence or presence of a notochordal pit on the anterior surface of the discussed protrusion in SVB 1451. It was described as absent in this specimen by Druckenmiller et al. (2012: 330), however, we were unable to verify this because of preservation of the anterior region in the latter. Furthermore, the value of this character (i.e., presence/absence of the anterior notochordal pit) is somewhat uncertain, as for the taxa where more than a couple of specimens is know, for example, Ophthalmosaurus icenicus, both the conditions could be observed (Nikolay G. Zverkov, 2018, personal observation on O. icenicus specimens in CAMSM and NHMUK).

The extracondylar area of the basioccipital of A. chrisorum CMN 40608 is extremely reduced and completely unseen in posterior view, as in P. hoybergeti (SVB 1451) and Janusaurus lundi (Roberts et al., 2014; although this element in the latter is poorly preserved). However, the extracondylar area in these specimens is relatively anteroposteriorly wide in lateral view, unlike that of Grendelius spp. (Fig. 21; McGowan, 1976; Zverkov, Arkhangelsky & Stenshin, 2015). Maxwell (2010) has interpreted a part of the extracondylar area as a stapedial facet, probably due to poor preservation of CMN 40608. The stapedial facet of CMN 4060 faces anteriorly and is practically unseen in lateral view (cf. Maxwell, 2010, fig. 2D). This feature is present in all the specimens referred to as Arthropterygius in current contribution, for which the basioccipital is available (Fig. 21; CMN 40608, SVB 1451, PMO 222.654, PMO 222.669) and is currently unknow for any other ichthyosaurs. Thus, we consider this an autapomorphy of Arthropterygius.

Recently referred to as Palvennia hoybergeti PMO 222.669 shares all features of A. chrisorum. However, Delsett et al. (2018) provided a brief comparison of PMO 222.669 and A. chrisorum (holotype CMN 40608), mainly in their emended differential diagnosis of P. hoybergeti. According to those comparisons, P. hoybergeti differs from A. chrisorum in the following features: anterior face of basioccipital lacks notochordal pit and basioccipital peg (see Comments above); rectangular in outline articular with a slight constriction at the anteroposterior midpoint; longer and narrower anterior notch of the coracoid; proximodistally shorter dorsal process of the humerus (this could be explained by ontogenetic variation, see Discussion of ontogenetic changes below); not as convex articular faces of epipodial elements (also could be due to ontogenetic and interspecific variation; see below). We suppose that none of these differences is sufficient enough to distinguish the species.

The overlapping elements between PMO 222.669 and the holotype of A. chrisorum (CMN 40608) are the basioccipital, basisphenoid, articular, atlas-axis complex, and other vertebrae, pectoral girdle elements, and forelimbs. We suppose that this is sufficient overlap to consider the potential affinity of the specimen. For the comparison of these specimens especially valuable are basicranial elements and the humerus that bear a number of autapomorphic traits (Fig. 21). Basioccipitals of PMO 222.669 and CMN 40608 are very similar and have a feature not observed in most of the other ophthalmosaurids: stapedial and opisthotic facets of the basioccipital shifted anteriorly and poorly visible in lateral view (this condition is shared only with type specimens of Palvennia hoybergeti (SVB 1451) and Janusaurus lundi (PMO 222.654), Nikolay G. Zverkov, 2017, personal observation). The basisphenoids of these specimens (i.e., PMO 222.669 and CMN 40608) have a foramen for the internal carotid arteries opening posteriorly and not visible in ventral view, being separated from the ventral surface by a thin shelf. This condition is known exclusively for A. chrisorum and is autapomorphic. The humeri of PMO 222.669 and CMN 40608 bear a strong ventral skew between the radial and ulnar facets with the dorsoventral thickness of the radial facet greatly exceeding that of the ulnar facet (autapomorphic feature of A. chrisorum). The proximal ends of the discussed humeri are strongly anteroposteriorly elongated with reduced deltopectoral crest shifted to its anterior edge (among other ophthalmosaurids, similar condition is found in Undorosaurus gorodischensis; Zverkov & Efimov, 2019). The facet for the anterior accessory epipodial element of the humerus is comparatively large, as wide as, and close in size to the radial facet (this is a non-unique autapomorphy of A. chrisorum, wich also occurs in mid-Cretaceous “Ophthalmosaurus” cantabrigiensis Lydekker, 1888, Maiaspondylus lindoei Maxwell & Caldwell, 2006, and “Platypterygius ochevi” Arkhangelsky et al., 2008, but not in any other ophthalmosaurid, including the holotype of Palvennia hoybergeti).

Indeed, PMO 222.669 and the holotype of Palvennia hoybergeti (SVB 1451) have a very similar morphology of the dermatocranium, including one feature that was considered as an autapomorphy of P. hoybergeti—“a very large pineal foramen” (Delsett et al., 2018: 11). However, the frontal region and consequently the shape of the parietal foramen is still unknown for a number of Late Jurassic ichthyosaurs, including Brachypterygius extremus (Boulenger, 1904), Nannopterygius enthekiodon (Hulke, 1871), Grendelius spp. (Zverkov, Arkhangelsky & Stenshin, 2015), Undorosaurus spp. (Zverkov & Efimov, 2019), and A. chrisorum (i.e., its holotype). We cannot exclude the presence of this trait in any of the listed taxa. In this regard, the diagnostic value of this feature is reduced. At the same time, there is a number of differences between PMO 222.669 and the holotype of P. hoybergeti (SVB 1451). These are: basisphenoid posterior foramen for the internal carotid arteries separated from the ventral surface by a thin shelf in PMO 222.669 (shifted to the posteroventral edge of the element in SVB 1451, see Description above); extremely shortened and robust paraoccipital process of the opisthotic (relatively elongated and dorsoventrally compressed in SVB 1451; see Description); slender distal stapedial process (expanded in SVB 1451, see Description above); markedly less expanded mediolaterally anteromedial tongue of the supratemporal (see Description); reduced deltopectoral crest of the humerus shifted to its anterior edge (well pronounced in SVB 1451, see Description); large semicircular facet for the anterior accessory epipodial element of the humerus (comparatively small and anteriorly tapered in SVB 1451; see Fig. S7A); large and rounded in outline anterior accessory epipodial element (this element is relatively small in SVB 1451, semicircular in outline, with a nearly straight anterior margin, see Description above).

From our observations on PMO 222.669 (NGZ), we have not found any additional differences in overlapping material with both the holotype of A. chrisorum (CMN 40608) and the other specimens referred to as A. chrisorum herein (especially, CCMGE 3–16/13328 and 17–44/13328). Thus, we refer PMO 222.669 to as Arthropteryguis chrisorum.

The specimens from Franz Joseph Land (CCMGE 3–16/13328 and 17–44/13328) are referred to as A. chrisorum based on exact the same grounds as PMO 222.669 (see above), except for the features of the basioccipital, that is, lost in both. In addition to this, a femur, present in CCMGE 17–44/13328, is well consistent with that of the holotype of A. chrisorum (see Description above).

The humeral morphology of SGM 1573 is typical of A. chrisorum with the presence of all autapomorphic traits (listed above). Thus, we consider this sufficient for the referral of SGM 1573 to A. chrisorum.

The referred specimens of A. chrisorum provide additional overlapping elements: supratemporal, postfrontal, jugal, parietal, quadrate, opisthotic, stapes, interclavicle, clavicle, and scapula. Some of these elements also bear autapomorphic traits.

The supratemporal of PMO 222.669 produce a wide anterior projection—a tongue covering the postfrontal with the formation of a marked facet (this could be observed in the isolated postfrontal of CCMGE 17–44/13328). Among ophthalmosaurids, this condition is found only in Athabascasaurus bitumineus, “Palvennia” hoybergeti and “Janusaurus” lundi (see Description above).

The jugal is strongly bowed ventrally in CCMGE 17–44/13328 and in PMO 222.669. Among ophthalmosaurids, this condition is shared exclusively with “Palvennia” hoybergeti and “Janusaurus” lundi (Fig. 21).

The parietal of PMO 222.669 is characterized by the extremely anteroposteriorly shortened medial symphysis, that is, posteriorly restricted by a pronounced excavation and notch. This morphology became clear after the additional preparation that was performed by the enquiry of NGZ, as in the original study of the specimen (Delsett et al., 2018) the interparietal region was described before a preparation. The revealed morphology of the interparietal contact is uniquely shared by PMO 222.669 and holotypes of “Ichthyosaurus” volgensis, “Palvennia” hoybergeti, and “Janusaurus” lundi (Fig. 21). We consider this feature as an autapomorhy of Arthropterygius (sensu herein).

The stapes of PMO 222.669 is characterized by a very slender lateral process. Among other ophthalmosaurids, this condition is observed only in “Palvennia” hoybergeti, “Janusaurus” lundi, and Acamptonectes Fischer et al., 2012. There is a number of differences between PMO 222.669 and Acamptonectes both cranial and postcranial (cf. Fischer et al., 2012), still, the presence of this trait in PMO 222.669 (referred to as A. chrisorum herein) and in “Palvennia” hoybergeti and “Janusaurus” lundi could also be used as a non-unique autapomorphy of Arthropterygius (in its sense in the current paper).

The complete interclavicles are known for SGM 1573 and the holotype of “Janusaurus” lundi. Although there are differences between the two (see Description above), both bear a bulge in the middle of the interclavicle posterior median stem (see Description). This feature is unique for ophthalmosaurids and considered as autapomorhy of Arthropterygius herein.

The clavicles are known for PMO 222.669, CCMGE 17–44/13328, “Palvennia” hoybergeti (SVB 1451) and “Janusaurus” lundi (PMO 222.654). In all of these specimens the clavicle is very dorsoventrally high (medial dorsoventral height to anteroposterior length ratio greater than 0.2) and extremely robust compared to any other Jurassic ophthalmosaurid (Fig. 21). We consider this feature as an autapomorhy of Arthropterygius (in its sense in the current paper).

The coracoids in all the specimens, where available (see Table 1), are comparatively large (the scapular proximodistal length is less than the coracoid anteroposterior length). In their general shape, the coracoids are similar to those of Ophthalmosurus (Moon & Kirton, 2016). Although it is not assessable in all specimens (due to preservation), when visible, a significant angle close to 90–100° could be observed between the articulated coracoids (CCMGE 3–16/1332, PMO 222.654, KSU 982/P-213). This configuration is unknown for other ophthalmosaurids and thus considered autapomorhic of Arthropterygius.

Recently erected from the Berriassian of Svalbard Keilhauia nui is also referable to Arthropterygius, however, only in open nomenclature. The holotype and only known specimen of this taxon is a poorly preserved skeleton of a small individual that was considered to be of “late juvenile to adult ontogenetic stage” (Delsett et al., 2017: 14). From our observations of the holotype (PMO 222.655), we are unable to confirm any of the evidences proposed by Delsett et al. (2017) as supporting maturity of PMO 222.655. The proximal portion of the humerus of PMO 222.655 is heavily weathered and its posterior portion is broken so that it is impossible to say something regarding its natural shape and its value for identification of maturity. The same concerns a texture of the humeral shaft, which along with other skeletal elements of PMO 222.655 is poorly preserved, weathered, and partially covered by matrix along with products of pyrite decay. It is unclear what Delsett et al. (2017) meant under the degree of ossification that “(when it is possible to observe) resembles mature finished bone,” as, in our opinion, all the available surfaces are too pooly preserved. Furthermore, the facets of appendicular elements are poorly demarcated from each other, the epipodial and authopodial elements are rounded and very loosely arranged and the ventral margin of the ischiopubis bears an excavation along its ventral margin, which could be a result of an extensive cartilaginous continuation of the element. The listed traits could be considered as arguments for the immaturity of PMO 222.655, although could be alternatively interpreted as a result of poor preservation. A natural shape of the ischiopubis is unclear because its proximal portion is partially eroded and diagenetically compressed. PMO 222.655 is generally similar to CCMGE 3–16/13328 being only slightly smaller in size, and has a number of features that are diagnostic of Arthropterygius: the humerus of PMO 222.655 has a ventral skew between the radial and ulnar facets, its ulnar facet:radial facet dorsoventral width ratio is less than 0.8; the facet for anterior accessory element is nearly as large as the radial facet (a diagnostic feature of A. chrisorum); the clavicle of PMO 222.655 is relatively large and robust. Judging from the field photographs (J. Hurum, September 2017, personal communication), the coracoid was originally longer anteroposteriorly than mediolaterally wide and extremely similar to that of CCMGE 3–16/13328, its current shape is likely due to conservation processes. The ischiopubis of PMO 222.655 is plate-like and lacks an obturator foramen similarly to the condition observed in J. lundi (PMO 222.654) and several indeterminate ophthalmosaurids from Svalbard (Delsett et al., 2017), and unlike any other known ophthalmosaurid. What concerns the ilium of 222.655, its expanded dorsal portion is an important character that probably demonstrates a juvenile condition of what in A. lundi (PMO 222.654) developed in an “anteromedial process” and posteriorly curved end (Roberts et al., 2014). Thus, expanded dorsal portion of the ilium and plate-like ischiopubis with no obturator foramen could also be generic features of Arthropterygius (in its sense in the current paper).

Taking into account all the arguments above, we consider “Keilhauia nui” as a nomen dubium and identify its type specimen as Arthropterygius sp. juv. cf. A. chrisorum.

Ontogenetic changes, intra- and interspecific variation in Arthropterygius

Owing to new specimens of A. chrisorum, we can now make some observations on the ontogenetic changes and variation in this taxon.

In general, changes in morphological proportions during growth of A. chrisorum are consistent with those observed in other ichthyosaurs (Huene, 1922; McGowan, 1973b; Deeming et al., 1993). Having largely incomplete specimens (Table 1; Fig. 22), we are unable to assess the growth of the whole skull and the whole body. Thereby we compared selected cranial and postcranial elements (Fig. 23). The growth of elements of the skull base and occiput of A. chrisorum is more or less isometric compared to each other. The same concerns the growth of elements of the appendicular skeleton (Fig. 23A). At the same time, the growth rates differ between the skeletal regions.

Relative anteroposterior length of the basisphenoid and the humerus is among the few ratios that could be calculated for A. chrisorum in order to compare the growth of the cranial and postcranial skeleton. In juvenile CCMGE 3–16/13328 this ratio is 0.58, in young adult CCMG E 17–44/13328—0.42, and in mature individual CMN 40608—0.35; thus we observe typical negative allometry. It is not surprising that the growth of the cranial elements is negatively allometric relative to the growth of the appendicular elements. Interesting is that growth of the appendicular skeleton is somewhat positively allometric relative to that of the axial skeleton (Fig. 23A), whereas for Ichthyosaurus and Stenopterygius this reported as being isometric (McGowan, 1973b).

Judging from the available cranial elements, the general morphology and proportions of the occipital region have not undergone dramatic changes with age. Despite differences in size CCMGE 3–16/13328, CCMGE 17–44/13328, and PMO 222,669 have a peculiar shape of the quadrate condyle: it is dorsoventrally high with a V-shaped ventral margin of the articular boss. Furthermore, the quadrate do not develop the anterior protrusion with age. In all specimens of A. chrisorum, the basisphenoid is trapezoidal in ventral outline, being mediolaterally wider anteriorly than posteriorly. The juvenile CCMGE 3–16/13328 has a narrower anterior profile when compared to those of adults CCMGE 17–44/13328, PMO 222,669, and CMN 40608 (Figs. 23B–23E), supporting observations of Kear & Zammit (2014) on Platypterygius australis. The only marked difference of the basisphenoids is the relative position of the posterior foramen for the internal carotid arteries, which is still exposed ventrally in juvenile CCMGE 3–16/13328, but already separated by a grown shelf in young adults PMO 222,669 and CCMGE 17–44/13328 (Figs. 23B–23D).

All the specimens of A. chrisorum have concave humeral distal facets and convex proximal articular facets of the epipodial element. A tendency for deepening of humeral distal facets with age could be observed, however, it is non-uniform. Although the old adult CMN 40608 has very deeply concave facets (Maxwell, 2010), comparable in size SGM 1502 has less concave facets and consequently should have had less convex proximal surfaces of the epipodial elements. Considering this variation and the fact that after the publication of Maxwell (2010) humerus-epipodial peg-and-socket articulation was reported for other ophthalmosaurids (Zverkov et al., 2015), we suggest that “proximal surface of zeugopodial elements angular in outline for articulation with humerus” (Maxwell, 2010: 404) cannot be further considered as a diagnostic character of Arthropterygius.

It is interesting that there are no marked differences in humeral morphology between the juvenile and adults. The marked change is the angle between the radial facet and facet for the anterior accessory epipodial element that became less pronounced with age (Figs. 23M–23Q). The same concerns the angle between the ulnar and radial facets. The absence of marked ontogenetic changes in relative size and shape of the humeral distal facets supports their diagnostic value. Thus, the features related to humeral distal facets could likely be used for diagnosing specimens of Arthropterygius irrespective of their maturity.

As in case of Undorosaurus (Zverkov & Efimov, 2019) and Grendelius (Zverkov, Arkhangelsky & Stenshin, 2015), species of Arthropterygius could be potentially distinguished based exclusively on humeral morphology, which was already demonstrated above by the results of PCA. Especially valuable is the outline of the humeral proximal end—each of these genera has species with anteroposteriorly elongated humeral proximal ends (Grendelius zhuravlevi, Undorosaurus gorodischensis, A. chrisorum) and those with isometric proximal ends (G. alexeevi, U. nessovi, U. trautscholdi, A. lundi). We cannot exclude the possibility that some of these species may actually represent males and females, thus demonstrating sexual dimorphism, differing in limb morphology in a way, similar to that hypothesized for Triassic ichthyopterygians Chaohusaurus and Shastasaurus (Shang & Li, 2013; Motani et al., 2018). However, given other existing differences (especially cranial) between the discussed species, and considering that in some genera more than one species with either elongated or isometric humeral proximal end could be present, it is impossible to say, which of the species are representing sexual morphs of the same species and which of them are morphs of other species. Thereby, in the current state of knowledge, we prefer to retain all the “morphs” as separate species.

Palaeobiogeographic implication of Arthropterygus

After the discovery of Arthropterygius in Argentina (Fernández & Maxwell, 2012), this taxon, even being known from a couple of specimens, has already raised a question regarding the cosmopolitan distribution of ichthyosaurs (Fernández & Maxwell, 2012; Zverkov et al., 2015). New discoveries further support the idea that many ophthalmosaurids have had a widespread distribution. For more details on palaeobiogeography of Boreal ichthyosaurs, we direct the reader to (Zverkov et al., 2015; Zverkov & Efimov, 2019; Arkhangelsky et al., in press).

Given the new data, Arthropterygius seem to be very common ichthyosaurs for the Volgian (especially middle Volgian): A. chrisorum is found in Arctic Canada, Svalbard, and Volga Region, thus indicating a wide distribution of this species across the Arctic basins and Middle Russian Sea. The same concerns Arthropterygius hoybergeti and A. lundi, which are both known from Svalbard and Volga Region (Fig. 24A). Additionally A. lundi is known from the Timan-Pechora, thus unambiguously demonstrating that the Mezen-Pechora Strait connecting the Middle Russian Sea with Arctic Basins was used as a passage during this time.

From the Middle Volgian Virgatitus virgatus Chron the unifying element of the Middle Russian Sea and Arctic basins is Undorosaurus gorodischensis (Zverkov & Efimov, 2019), whereas Arthropterygius are currently unknown in the Middle Russian Sea from this time interval (Fig. 24B), but they still existed at high latitudes during the late Volgian and Ryazanian (latest Tithonian and Berriassian), thus being among the few ichthyosaur taxa that are recognized in the Berriassian.

Significance of the new finds and further perspectives in the study of ophthalmosaurids

The Beriassian fossil record of marine tetrapods is scarce and patterns of faunal turnover during the Jurassic–Cretaceous transitional interval are non-uniform (Benson et al., 2010, 2013; Fischer et al., 2012; Benson & Druckenmiller, 2014; Tennant et al., 2017; Zverkov et al., 2018). It has already been suggested that ichthyosaurs survived the Jurassic–Cretaceous transition relatively unscathed (Fischer et al., 2012, 2013). However, Berriassian ichthyosaur record is still poor (Fernández & Aguirre-Urreta, 2005; Fernández, 2007a; Ensom et al., 2009; Fischer et al., 2012; Green & Lomax, 2014; Delsett et al., 2017). As was discussed above, it is better to consider “Keilhauia nui” from the Berriassian of Svalbard as a nomen dubium. In this regard, only one Berriassian ichthyosaur, Caypullisaurus bonapartei from the Neuquen Basin of Argentina, is hitherto to be recognized at the species level (Fernández, 2007a), demonstrating that this Tithonian species sucsessfully crossed the Jurassic–Cretaceous boundary. Discovery of A. chrisorum in the Berriassian of Franz Joseph Land provides the second ophthalmosaurid species that unambiguously crossed the Jurassic–Cretaceous boundary, further argument that this transition had minimal (if some) effect on ichthyosaurs.

A discrete character of the fossil record of ophthalmosaurids (Cleary et al., 2015) has led to certain problems in the study of this group. The only more or less thoroughly investigated ophthalmosaurids to date are Callovian Ophthalmosaurus icenicus (Andrews, 1910; Appleby, 1956; Kirton, 1983; Moon & Kirton, 2016) and Albian Platypterygius australus (Wade, 1984, 1990; Kear, 2005; Zammit, Norris & Kear, 2010; Kear & Zammit, 2014). Other ophthalmosaurids are incomparably poorly known either due to a small sample size or because of fragmented and/or poor preservation. In such conditions, it is hardly possible to develop a strong phylogenetic hypothesis for ophthalmosaurids. The continuing replenishment of the ophthalmosaurid taxon list by new poorly known and difficult to compare (but having withal a number of autapomorphies) taxa do not make this task easier. The fair attempt to consider all the known ophthalmosaurid taxa and all the proposed phylogenetic characters results in the extremely poorly resolved Ophthalmosauridae (Moon, 2019).

Recently Massare & Lomax (2018) demonstrated the effect of large sample sizes on the identification of taxonomically distinct morphological characters in Ichthyosaurus. This is what is actually needed for ophthalmosaurids. In this regard, Late Jurassic to Early Cretaceous formations of Arctic, considering the abundance and exceptional preservation of marine reptiles (Delsett et al., 2016; Nikolay G. Zverkov, 2015–2016, personal observation), have great perspectives for collection of a large sample size, comparable to those of the Lias Group and Posidonia Shale lagerstätten of Western Europe.