Si les thalattosuchiens sont assez bien connus dans le Jurassique supérieur d'Europe et de Russie [20,25,26], ils montrent une moindre diversité taxonomique en Amérique du Sud et à Cuba [13,16,21,27]. Ils participent de la faune de reptiles marins peuplant le corridor caraïbe qui reliait l'archipel Européen et la marge Pacifique des Amériques depuis le Jurassique moyen [13,27]. Le Nord-Est du Mexique faisait partie du domaine marin à l'époque où furent déposés les sédiments deltaïques et de plate-forme interne de la formation La Casita et ceux de la plate-forme externe de la formation contemporaine La Caja [1,17]. Les assemblages de microfossiles et d'invertébrés semblent cependant indiquer l'isolement partiel du golfe du Mexique, à la fois de l'archipel Européen et du Pacifique, peut-être dû au soulèvement de la Floride [1,2]. Au cours du Jurassique supérieur, cette région était probablement divisée en bassins, en raison de l'activité tectonique régionale [18]. Les reptiles marins du Nord-Est du Mexique font l'objet de description systématique depuis peu [5,6,11]. Un premier membre mexicain du genre Geosaurus, G. vignaudi Frey et al., 2002 avait conforté l'hypothèse de l'isolement partiel du golfe du Mexique au Jurassique supérieur. Un nouveau spécimen du Tithonien inférieur du Nord-Est du Mexique, comprenant crâne et squelette post-crânien partiels, est rapporté ici à une nouvelle espèce du même genre et renforce encore cette hypothèse.

Late Jurassic thalattosuchians are well known from western Europe and Russia [20], [25] and [26]. They were comparatively seldom reported from South and Central America, where they have a lower taxonomical diversity [13], [16], [21] and [27]. Thalattosuchians are part of the marine reptile fauna that populated the Caribbean corridor linking the European Archipelago and the Pacific margin of the Americas since the Middle Jurassic [14] and [16]. Northeastern Mexico was part of the marine realm during the Late Jurassic when the deltaic and inner shelf conglomerates, sandstones and siltstones of the La Casita Formation were deposited; shales, siltstones and phosphorites characterize the more distal outer shelf sediments of the contemporaneous La Caja Formation [1] and [17]. However, according to the microfossil and invertebrate assemblages, the Mexican Gulf was at that time partly isolated from both the European Archipelago and the Pacific, with the Florida uplift possibly forming a barrier between the European Tethys and the Mexican Gulf [1] and [2]. Moreover, at the time that the La Casita/La Caja Formations were deposited, northeastern Mexico experienced large-scale tectonic activity and sea-level fluctuations, leading to irregular sea floor topography and causing variable restrictions or subdivision of basins [18].

Thalattosuchians and other marine reptiles have been known from the La Casita/La Caja Formations of northeastern Mexico for some time [3], [19], [22] and [24], but have only recently been formally described [5], [6] and [11]. The occurrence of at least one distinct species of Geosaurus (Geosaurus vignaudi Frey et al., 2002) in the Middle Tithonian of Puebla State (Central–East Mexico) appears to support the hypothesis that the Mexican Gulf was partially isolated during the Late Jurassic, as has been suggested on the basis of the invertebrate assemblages [1] and [2]. Recent field work in the Early Tithonian section of the La Caja Formation in southernmost Coahuila, northeastern Mexico (Fig. 1) yielded the partial cranium and postcranium of a thalattosuchian here referred to a new species of Geosaurus.

MUDE: Museo del Desierto, Saltillo, Mexico; SMNK: Staatliches Museum für Naturkunde Karlsruhe, Germany; SMNS: Staatliches Museum für Naturkunde Stuttgart, Germany.

Thalattosuchia Fraas, 1901 [9]

Metriorhynchidae Fitzinger, 1843 [8]

Geosaurus Cuvier, 1824 [7]

Geosaurus saltillense sp. nov.

Holotype. Partial cranium and postcranium, CEP1823, MUDE (Fig. 2 and Fig. 3).

Locus typicus. Sierra de Buñuelas, near Gomez Farías, Coahuila, Mexico (Fig. 1).

Stratum typicum. Early Tithonian (Upper Jurassic) section of the La Caja Fm.

Derivatio nominis. From the county of Saltillo, upon which the village of Gomez Farías depends.

Diagnosis. A member of the genus Geosaurus characterised by its extremely long and narrow supratemporal fenestra, the length of which reaches about 2.5 times its caudal width. The lateral margins of the supraorbital portion of the frontal are grooved, suggesting that this bone was excluded from the centre of the dorsal margin of the orbit by the prefrontal. The lateral margins of the intertemporal bar are straight, except for the narrow parietal table located in the middle area of the bar.

Preservation and taphonomy. Of the cranium of CEP1823, the frontal, intertemporal bar, occiput and quadrates, and the caudal-most portions of both supratemporal bars are preserved in articulation (Fig. 2 and Fig. 3). The rostral-most portion of the frontal (see § Description) was separated from the main portion of cranium by a 5- to 10-mm-thick vein of calcite and matrix, indicating that it was most likely broken prior to fossilisation. The occipital portion of the cranium suffered from dorsoventral compression, as the articular processes of the quadrates are now sub-horizontal. A pair of elements is tentatively identified as the rostral-most portion of the palatine, including the palatine part of the prefrontal pillar. Both palatines lack their caudal portion and were compressed dorsoventrally. The left one is preserved upside down upon the left lateral process of the frontal, the ascending process facing rostrally (Fig. 2 and Fig. 3). The right palatine is lying about 100 mm to the left of the main portion of the cranium, resting dorsal side up (Fig. 2). About 10 mm rostral to the right palatine fragment and in line with it, an 85-mm-long grooved element is identified as part of the (presumably right) nasopharyngeal duct. The surface of this element is very poorly preserved, and the bones that comprise it cannot be determined. Adjacent to the nasopharyngeal duct, some additional bone substance is preserved, not entirely freed from matrix, and cannot be further identified (Fig. 2). Inside and to the right of the right supratemporal fenestra, portions of the supra- and infratemporal bars and the caudal-most part of a quadratojugal are preserved, together with probable portions of the hyoid apparatus and possibly of a pterygoid (Fig. 3). A seemingly triradiate bone preserved lying dorsal to the right quadrate could represent a postorbital. However, in ventral aspect, it is covered by a long element that most likely is part of the mandible (Fig. 2 and Fig. 3B, see below); in dorsal aspect, it is obscured by matrix and remains of a rib and a seemingly flat bone, having the outline of two adjacent semicircles (Fig. 3A). A similarly looking element was identified in the same area of the holotype of G. vignaudi as a possible sclerotic plate [11]. However, as it is located, this element might also belong to the mandible (Fig. 2). The mandibular rami are preserved caudal to the main portion of cranium (Fig. 2); their surface is so heavily corroded that no individual bone can be identified with certainty, with the notable exception of the massive left articular and most likely the caudal-most portion of the articulated right angular and surangular (Fig. 2). No tooth, and indeed no alveolus, could be identified. The partly preserved atlas, the axis, ribs pertaining to the atlas and axis and the three articulated succeeding cervical vertebrae are preserved adjacent to the left articular. Of the rest of the postcranium, several thoracic ribs can be identified, but their numbers are uncertain; more corroded, intermeshed, unidentified bones or bone fragments are present (Fig. 2).

Either the frontal and parietal were fused, or their sutural contact was running where a break affects the intertemporal bar just rostral to the parietal table (Fig. 3). Frontal and nasals are possibly fused (see § Discussion). The ascending process of the palatines, if correctly identified, were not fused to the descending process of the prefrontals. Apparent fusion of the bones of the occipital face can possibly result from dorsoventral compression. The neurocorporal sutures of the preserved cervical vertebrae on the other hand are clearly visible. The individual age of the specimen is therefore difficult to assess, although it is probably safe to assume that it was not a fully mature adult [4].

CEP1823 was collected during a field campaign in March 2004 and mechanically prepared in the SMNK. The ventral side of the cranium was exposed when discovered, showing the olfactory tract; according to our observations of the local geology, this indicates that the cranium rested on its dorsal side. However, the dorsal portion of the caudal-most part of the parietal is weathered so that the spongiosa is exposed (Fig. 2 and Fig. 3A). The cranium therefore probably rested for some time on its ventral side. Several subcircular holes or deep depressions could be attributable to scavengers or represent bite marks (Fig. 3); the heavily corroded condition of the mandible and postcranial bones, though, suggests that these elements were dissolved. Several isolated bones from the same locality (MUDE undescribed specimens) lack a proper cortical layer and appear dissolved as well, as if having been exposed to an acidic environment. However, the surface of the main portion of the cranium of CEP1823 is smooth, showing no such trace of dissolution. The identical relative location of the articular in relation to the atlas/axis and cervical vertebral column in the holotype of G. vignaudi [11] and CEP1823 (Fig. 2) hints at a similar displacement of the neck and caudal portion of the mandible when still connected by muscles, ligaments and skin. The taphonomical history of the specimen is thus complex, combining agents mighty enough to tilt the skull, break or disarticulate and displace the brain case, temporal bars, prefrontals, rostrum and mandible, and still leave in place, e.g., the broken rostral-most portion of the frontal or portions of the supra- and infratemporal bars. Further geological and sedimentological studies of the finding site are necessary to understand the taphonomy of CEP1823.

Description. The rostral-most portion of the interorbital area is most likely formed mainly by the frontal (Fig. 2 and Fig. 3). The nasal, however, might form its lateral part; on the right side, an elevated linear area of bone running rostromedially on the dorsal surface could then mark the location of the suture with the frontal (Fig. 3A). In ventral aspect, no potential nasofrontal suture is visible (Fig. 3B).

No other suture is clearly visible on the cranium; when necessary in what follows, we designate the bones according to Fraas [10] and Vignaud [26].

In its narrower portion, both lateral surfaces of the frontal show conspicuous grooves extending caudally until the level of the rostral margin of the supratemporal fenestra; these grooves probably mark the contact of the frontal with the prefrontal. The ventral surface of the rostrolateral-most portion of the frontal on the main cranium fragment (Fig. 3B) exhibits a wrinkled area lateral to the olfactory bulb imprint on both sides, where most likely the prefrontal underlaid the frontal. It thus appears that the prefrontal formed the rostral and central portions of the dorsal margin of the orbit and extended caudally until level with the rostral margin of the supratemporal fenestra. The lateral extent of the prefrontal and its original dorsal exposure, however, cannot be determined. A square fragment of bone sitting in the right frontal lateral groove might represent part of the prefrontal (Fig. 3).

The lateral processes of the frontal form together an angle of 40 to 45°. The rostral margin of the supratemporal fenestra is rounded; at least in its rostral fourth, delimited medially by the intertemporal bar and laterally by the lateral process of the frontal, it is triangular (Fig. 3A). The caudal part of the supratemporal fenestra is subrectangular, the posttemporal bar being oriented laterocaudally, and the preserved portion of the supratemporal arch rostrally. A possible fragment of the proötic may be preserved in the caudomedial corner of the right supratemporal fenestra (Fig. 3). The length of the supratemporal fenestra is about 2.5 times its preserved maximal width (in its caudal portion, Fig. 3A). Owing to the breakage of the lateral portions of the lateral processes of the frontal and missing (or disarticulated) postorbitals, the actual extent of the supratemporal fossa can only be estimated: it might have occupied about half the surface, in any case about half the length measured along its midline of the supratemporal fenestra (Fig. 3A). The medial margin of both supratemporal fossae is genuine and clearest on the right; it smoothly curves caudomedially until it merges with the intertemporal bar. The intertemporal bar is presumably formed rostrally by the frontal and caudally by the parietal; it has straight, vertical margins, until the weak enlargement in its middle area forming an ill-defined parietal table. The margins of the intertemporal bar slightly converge caudal to the parietal table. The squamosal appears to form most of the caudal margin of the supratemporal fenestra.

The caudal-most portions of the supratemporal arches (presumably formed by the squamosal) show a sharp dorsal margin and a flat ventral articular surface for the quadratojugal. Of the infratemporal arch, only the caudal-most portion of a quadratojugal is identified, displaced to the right (Fig. 3). Its shaft emerges at low angle from its rectangular articular surface for the squamosal (Fig. 3B); it is sub-circular in cross-section, about 5 mm in diameter.

The osteological composition of the occiput is obscured by compression and/or fusion of the bones (see § Preservation); it generally agrees with the description by Vignaud [26] of undistorted, juvenile metriorhynchids. The parietal forms the median dorsal-most part of the occipital face, bordered lateroventrally by the exoccipital; the supraoccipital cannot be identified and was presumably fused to the adjacent elements (Fig. 3A). The ventral half of the occipital face comprises the basioccipital forming the occipital condyle, and the quadrates laterally (Fig. 3B). Few foramens of the occipital face are identifiable: the foramen magnum, the caudal aperture of the cranioquadrate canal and two of the foramina for cranial nerves IX, X and XI on the dorsal half of the occiput (Fig. 3A); the median intertympanic foramen on its ventral half (Fig. 3B). The occipital part of the basisphenoid cannot be distinguished from the basioccipital apophyses; the rostral portion of the basisphenoid, however, might be present close to its original position, i.e. ventral to the median portion of the intertemporal bar (Fig. 3B).

The articular condyle of the quadrate is transversely elongate, oval, concave in its middle portion, convex laterally and medially (Fig. 2 and Fig. 3).

Provided the palatines are correctly identified, their caudal process is not preserved (Fig. 2 and Fig. 3). Through comparison with the preserved dorsal width of the interorbital area and because the ventral surface of both palatines are identical in outline, the median margins might be close to original. They appear to diverge rostrally. The laterocaudal margin of the palatine (the mediorostral margin of the palatine fenestra) smoothly curves rostrolaterally (Fig. 3A). The ascending process appears to emerge just rostral to the rostral-most margin of the palatine fenestra. However, dorsoventral compression renders the actual, 3-dimensional morphology of the area uncertain. Both processes are about 20-mm long, and preserve lateral and medial crests that extend respectively laterally and medially and presumably merged with the dorsal surface of the ventral portion of the palatine.

As mentioned, the mandible is poorly preserved. The ventrally concave suture between the right angular and surangular might be preserved in lateral aspect (Fig. 2), although possibly not with its original outline, owning to extensive weathering. The articular as preserved appears to form about the lateral two-thirds of the glenoid fossa. The latter is slightly convex mediolaterally; a sharp crest, higher laterally, bounds it caudally and delimitates the rostral margin of the retroarticular process. The rostral margin of the glenoid is less sharply defined; it is rounded and the bone surface smoothly curves rostroventrally to form the rostral process of the articular. The articular forms only the dorsolateral-most part of the retroarticular process, which is broken caudally.

The atlas/axis complex and succeeding cervical vertebrae are similar in size and morphology to what was described for e.g. G. suevicus or G. vignaudi with similar cranium size [10] and [11].

Comparative palaeontology. CEP1823 is referred to the genus Geosaurus because of its gracile cranium lacking ornamentation; the 40 to 45° angle formed by the lateral processes of its frontal and its supratemporal arch, situated lower than the intertemporal bar [11] and [26].

Among the other members of the genus Geosaurus, G. giganteus (Soemmering, 1816), is poorly known by the fragmentary holotype cranium representing a massive form [26]. CEP1823 differs from G. gracilis (Meyer, 1830), G. araucanensis Gasparini & Dellapé, 1976 and G. vignaudi Frey et al., 2002 in its more gracile cranium and proportionately longer, narrower supratemporal fenestrae [10], [11], [13] and [26]. Geosaurus suevicus Fraas, 1901 was reconstructed by Fraas [10] as possessing long, oval supratemporal fenestrae looking similar to that of CEP1823. However, a crack runs caudoventrally through the caudal portion of the skull in the specimen upon which Fraas' reconstruction is based ([10]: specimen A, SMNS 9808, M.-C.B. pers. obs.); the extent and shape of the supratemporal fenestra cannot be judged properly. Based upon the location of the glenoid fossa on the occluding mandible, the length of the supratemporal fenestra in SMNS 9808 was close to that of G. vignaudi both relatively and absolutely.

Vignaud [26] showed that in thalattosuchians the width of the supratemporal fenestra compared to its length is subject to allometric growth during ontogeny. In all previously described species of Geosaurus, juveniles are characterised by a subcircular supratemporal fenestra, while it becomes oval in adults. CEP1823 therefore shows the adult configuration in this respect. Additionally, no interfrontal suture is present in CEP1823, as it is in many juvenile metriorhynchids and adult G. araucanensis as well [26]. Moreover, the preserved portion of cranium of CEP1823 is of a similar size as the equivalent portion of adult specimens of G. araucanensis and G. suevicus [10,13,26, and specimens in the SMNS, M.-C.B. pers. obs.], and of the adult holotype and only known specimen of G. vignaudi [11]. CEP1823 therefore cannot represent a juvenile of any of these taxa.

CEP1823 approaches G. araucanensis in the diverging mediorostral margins of its palatine, where a caudal process of the maxilla inserts in the latter taxon [13] and [26]. However, the palatal area is very poorly preserved or exposed in most other members of the genus [26].

The shape, extent and contact with the frontal of the prefrontal of G. araucanensis and other metriorhynchids were shown by Vignaud [26] to be variable independent of ontogenetic stage; however, the skull architecture and proportions, especially of the supratemporal fenestrae, are not significantly modified in relation to the variation of the cranial bones. Moreover, the orbital area and its three-dimensional anatomy in Geosaurus are poorly known, due to inadequately preserved or prepared specimens. The taxonomical importance of the lateral and caudal extent of the prefrontal dorsal to the orbit in CEP1823 therefore cannot be judged properly, but the shape and proportions of its cranium are significantly different from those of any other described thalattosuchian. For that reason, we feel confident in erecting a second species of the genus, G. saltillense, from the Tithonian of the Mexican Gulf.

Palaeobiogeography. Three species of the genus Geosaurus are described from the Early Tithonian of the European Archipelago, and on this basis the area has been proposed as the likely centre of origin for the genus [26]; a single species is described at present in the Early to Middle Tithonian of the Neuquén Basin of Argentina, while the Pacific margins yielded no other diagnostic remains (Fig. 4) [12], [13], [14], [15], [26] and [27]. G. vignaudi from the Middle Tithonian of the Mexican Gulf shows similarities with both G. suevicus and G. araucanensis [11] and [26], while G. saltillense represents a gracile, probably longirostrine form (on behalf of its long and narrow supratemporal fenestrae, see [26]) of unclear affinities at present. It adds to our knowledge of the diversity of the peculiar tetrapod assemblage of the Late Jurassic Mexican Gulf [6]. Further fieldwork will clarify the geological context and taphonomy of the find.