Introduction

The pterosaur assemblage from the Romualdo and Crato formations (Araripe Basin, northeastern Brazil) has attracted the attention of researchers primarily due to the exceptional preservation and diversity of its taxa. On some rare occasions, however, what actually claims most attention is the anatomical extremism of some species. One of the best examples is the pterosaur Tupandactylus imperator (Campos and Kellner, 1997). When it was first described (referred, at the time, to the genus Tapejara), this bizarre pterosaur was enigmatic for the huge size of its cranial median crest which comprised approximately 5/6 of the total lateral area of the skull (Campos and Kellner 1997).

Campos and Kellner (1997) placed T. imperator within Tapejaridae, a clade of pterodactyloid pterosaurs supported by five synapomorphies: (1) comparatively large nasoantorbital fenestra, forming 45% or more of the skull length; (2) long median crest, formed mostly by the premaxillae, arising from the rostral terminus of the skull and extending posteriorly; (3) thin subvertical lacrimal process of the jugal; (4) small reverse pear-shaped orbit; and (5) a broad and well-developed tubercle at the ventro-posterior margin of the coracoid (Kellner 2004). Monophyly of the Tapejaridae sensu Kellner has been recently questioned (Unwin and Lü 1997; Unwin 2003; Martill and Naish 2006; Lü et al. 2006).

After the first description of T. imperator, further cranial remains have been referred to this taxon (Fig. 1). Frey et al. (2003) documented an incomplete skull housed at the collection of the Staatliches Museum für Naturkunde, Karlsruhe, Germany (SMNK PAL 2839) (Fig. 1C). The proximal portion of the lower jaw is preserved, though this was not mentioned in the description. Unwin and Martill (2007) illustrated a more complete, well-preserved specimen without lower jaw, presently housed in a private collection (Fig. 1D). At least two more, hitherto undescribed specimens are housed in private collections. Only one of them has associated lower jaws (David Martill, personal communication 2010).

All the fossil remains attributed to T. imperator are restricted to the biomicritic laminated limestone of the Crato Formation (?Aptian) in the Araripe Basin (Campos and Kellner 1997; Frey et al. 2003; Unwin and Martill 2007). This stratigraphic unit, interpreted as deposited in a low-energy lacustrine environment, is known for its abundant and diversified fossil biota which is composed by plant remains, arthropods, fishes, anurans, lizards, crocodylomorphs, pterosaurs, and birds (Neumann and Cabrera 1999; Viana and Neumann 2002; Heimhofer and Martill 2007).

Fig. 1.

The four known specimens of tapejarinid pterosaur Tupandactylus imperator (Campos and Kellner, 1997) from the Crato Formation of the Araripe Basin, Brazil. A. CPCA 3590, the specimen described herein. B. MCT 1622-R, the holotype, described by Campos and Kellner (1997). C. SMNK PAL 2839, illustrated by Frey et al. (2003). D. The specimen housed at a private collection, illustrated by Unwin and Martill (2007). Scale bars 10 cm.

The laminated limestone of Crato Formation is intensely exploited and, usually, new fossils are discovered during commercial prospecting (Andrade 2007). Thus, the specimens lack their stratigraphic context and are often damaged. In the year 2009, at the Mina Triunfo quarry (city of Nova Olinda, Ceará province, UTM 24M-0423025E/9212692N) (Fig. 2), a pterosaur skull and lower jaw were found. The new specimen is here identified as T. imperator based on the presence of a spine-like dorso-posteriorly oriented process and an anteriorly projecting convex blade on the premaxillae (see below). Likely the specimen was complete at the time of the finding. Due to the activities of quarry workers, however, the occipital and temporal portions of the skull were lost. This new material, although incomplete, is the best-preserved T. imperator known so far. It allows, as well, a better comprehension of its phylogenetic relationships and the possibility of testing the position of this species within Tapejaridae.

Institutional abbreviations.—CPCA, Centra de Pesquisas Paleontológicas da Chapada do Araripe (Departamento Nacional de Produção Mineral, Crato, Brazil); MCT, Museu de Ciências da Terra (Departamento Nacional de Produção Mineral, Rio de Janeiro, Brazil); SMNK, Staatliches Museum für Naturkunde (Karlsruhe, Germany).

Systematic paleontology

Fig. 2.

Location map of the Mina Triunfo quarry, Ceará Province, northeastern Brazil.

Fig. 3.

Tapejarinid pterosaur Tupandactylus imperator (Campos and Kellner, 1997), specimen CPCA 3590 from the Crato Formation (?Aptian), Brazil. A. Specimen in right lateral aspect. C. Counterpart of the specimen. Photographs (A, C) and drawings (B, D).

Fig. 4.

Tapejarinid pterosaur Tupandactylus imperator (Campos and Kellner, 1997), Crato Formation (?Aptian), Brazil. Soft-tissue preservation in CPCA 3590. A. Probable pycnofibres associated with the left mandibular ramus (indicated by the arrows). B. Pattern of subvertical parallel fibers that compose the soft-tissue component of the cranial crest. C. Probable rhamphotheca associated with the tip of the premaxillae (indicated by the arrow). D. Probable rhamphotheca associated with the anterior end of the dentary (indicated by the arrow). Scale bars 20 cm.

Discussion and comparisons

CPCA 3590 is referred to Tupandactylus imperator based on the presence of the following diagnostic anatomical features: (i) a very large nasoantorbital fenestra, comprising most of the preserved length of the skull; (ii) rostral end of premaxillomaxilla turned ventrally; (iii) cranial crest supported anteriorly by a spine-like and slightly caudally-inclined supra-premaxillary process; and (iv) premaxillae bearing an anteriorly projecting convex blade. The two latter features are exclusive for T. imperator. The main autapomorphy distinguishing T. imperator from other tapejarids is a caudally orientated occipital process that reaches nearly the size of the rest of the skull measured from the anterior tip of the premaxilla to the posterior extremity of squamosals (Campos and Kellner 1997; Kellner and Campos 2007). This feature, however, is not preserved in CPCA 3590. Tapejara wellnhoferi also has an occipital process, mainly formed by the frontoparietals with the contribution of the supraoccipital ventrally (Wellnhofer and Kellner 1991). Although the real extension of this process cannot be inferred from the known specimens of T. wellnhoferi, it is unlikely that it reached the dimensions observed in T. imperator, which may reflect ontogenetic differences among specimens. An occipital process also occurs in the Chinese tapejarids from the Jiufotang assemblage (Aptian, western Liaoning province). Sinopterus dongi has a short frontoparietal posterior process that differs from T. imperator both in size and morphology; in S. dongi the structure is curved dorsally (Wang and Zhou 2003). The conditions for Sinopterus jii and “Huaxiapterus” corollatus remain unknown (Lü and Yuan 2005; Lü et al. 2006). “Huaxiapterus” benxiensis has a long occipital process that, like in S. dongi, curves dorsally (Lü et al. 2007).

As observed in other T. imperator specimens and in SMNK PAL2344 (the best-preserved Tupandactylus navigans known so far), in CPCA 3510, the posterior extension of the premaxillae articulates with nasals and frontoparietals. This contrasts with the condition observed in other tapejarinid pterosaurs such as T. wellnhoferi and S. dongi, where there is a short space between these bones and the posterior extension of the premaxilla runs parallel to, but not in direct contact with, nasals and frontoparietal (Wellnhofer and Kellner 1991; Wang and Zhou 2003). It remains unclear whether or not this condition is due to the ontogenetic stage of the specimens, since adult individuals of T. wellnhoferi and S. dongi remain unknown.

Fig. 5.

Variation of mandibular morphology in Tapejaridae. A. Tupandactylus imperator (Campos and Kellner, 1997). B. Tapejara wellnhoferi Kellner, 1989. C. Sinopterus dongi Wang and Zhou, 2002. D. Tupuxuara leonardii Kellner and Campos, 1994. Not to scale.

Following Frey and Martill (2003), the main differences between T. navigans and T. imperator are the lack of an occipital spine in the former and the inclination of the leading edge of the soft-tissue crest, which stands vertical to the long axis of the skull in T. navigans whereas in T. imperator it is inclined caudally. This difference in the inclination of the supra-premaxillary process is well marked in T. imperator by an anteriorly projecting convex blade on the premaxilla. This projection is, therefore, diagnostic of T. imperator. It is also noteworthy that the skull of T. imperator is more flat than that of T. navigans, with a length-height ratio of about 3.6 for T. imperator and 2.3 for T. navigans (whereby the length is measured from the tip of the premaxilla to the caudal margin of the squamosal and the height from the quadrate condyle to the dorsal margin of the premaxilla dorsal to the orbits). Provided the holotype of T. navigans completely preserves the occipital region, the presence or absence of an occipital process is sufficient to distinguish these two species. However, the referred specimen of T. navigans (SMNK PAL 2343) is abraded in occiput and in the holotype (SMNK PAL 2344) the posterior extension of the frontoparietals, where the occipital spine would be presumably located, is eroded. The integrity of the holotype of T. navigans has been previously questioned by Kellner (2007). Although it is improbable that the differences between the two taxa could be interpreted as ontogenetic variation (Frey and Martill 2003), we do not discard the possibility of sexual dimorphism, considering the fact that most of the differences between the two species are associated with the cranial median crest, a structure probably related to sexual display and, therefore, most likely to be sexual dimorphic (Elgin et al. 2008). The hypothesis that sexual dimorphism in Pterosauria is associated with cranial crests has been proposed previously by Bennet (1992). Nevertheless, it was later questioned by Kellner and Tomida (2000). In the specific case here discussed, the issue can be only solved in the light of new specimens.

CPCA 3590 is the first known specimen of T. imperator that preserves a nearly complete mandible. There is sparse information about one other specimen with a complete mandible which is presently housed in a private collection (David Martill, personal communication 2010). Specimen SMNK PAL 2839 preserves a fragmentary but yet undescribed lower jaw and a pair of hyoids. Although not much can be said about the lower jaw of CPCA 3590 due to the lack of sutures between bones, its structure is similar to that of T. wellnhoferi in the presence of a very deep ventral median crest in the symphyseal area. In CPCA 3590, however, the crest is deeper and more asymmetrically rounded than that of T. wellnhoferi (Fig. 5B). Furthermore, the deepest part of the mandibular crest in CPCA 3590 lies more anteriorly when compared with T. wellnhoferi, which has a pointed set-off symphyseal beak anterior to the crest (Wellnhofer and Kellner 1991). Jiuofutang Formation tapejarids such as S. dongi, Sinopterus jii, “H”. corollatus, and “H”. benxiensis all show a low, blade-like crest comparatively longer than that of CPCA 3590 and T. wellnhoferi with respect to the mandibular length (Wang and Zhou 2003; Lü and Yuan 2005; Lü et al. 2006, 2007; Fig. 5). In this respect, the lower jaws of the Chinese tapejarids are more comparable with the ones of some azdharchoid pterosaurs such as Bakonidraco galaczi Ösi, Weishampel, and Jianu, 2005. In CPCA 3590 the dorsal margin of the mandible has a gentle anterior concavity following the ventrally turned rostral portion of premaxillomaxilla. This concavity is better developed in T. wellnhoferi (Wellnhofer and Kellner 1991). The lower jaw of CPCA 3590 bears a ventrally sloping retroarticular process similar to the one of T. wellnhoferi, although smaller.

The filamentous structures associated with the left mandibular ramus resemble the pycnofibres described by Kellner et al. (2010). Pycnofibres are hair-like structures that covered most of the pterosaur body. The composition of these fibers is unknown but is likely that they were mostly composed by keratin (Kellner et al. 2010). Pycnofibres associated with the mandibular rami could suggest that, in T. imperator, the rhamphotheca was restricted to the symphyseal region of the lower jaw.

Structure and morphology of the cranial crest

A soft-tissue cranial crest showing a pattern of sub-parallel vertical fibers has been already reported for Tupandactylus navigans. In this pterosaur, however, these fibers gently curve rostrally (Frey and Martill 2003), while in CPCA 3590 they curve caudally.

In the holotype of Tupandactylus imperator (MCT 1622R) the soft tissue crest is preserved as an impression of a triangular patch dorsal to the bony crest (Campos and Kellner 1997). For that reason T. imperator has been repeatedly depicted with a “sail-shaped” median crest with a deeply concave caudal margin. However, considering the bad preservation of MCT 1622-R, it is unlikely that the crest morphology as preserved represents its actual shape in life. Although incomplete, the cranial crest of CPCA 3590 (the best-preserved known so far) shows a dorsal margin posterior to the suprapremaxillary ossification describing a curvature that suggests that the dorsoposterior contour of the crest must have been convex rather than concave (Fig. 6). This convex margin of the cranial crest is consistent with the condition observed in the specimen illustrated by Unwin and Martill (2007: fig. 17.9; at time of publication stored in a private collection) as well as with the morphology proposed by Frey et al. (2003: fig. 10). A fibrous crest supporting a soft-tissue element, both anchored in the bony premaxillary crest, as illustrated by Frey et al. (2003) could not be observed in CPCA 3590. Here, the fibers of the soft-tissue crest remain uniform throughout the extension of this structure. Some differences in the color pattern in different portions of the crest are likely a result of differential oxidation. Because only the anterior part of the crest is preserved, nothing can be said about density variations in the posterior part of this structure.

Fig. 6.

Cranial crest morphologies reconstructed for tapejarinid pterosaur Tupandactylus imperator (Campos and Kellner, 1997). A. “Sail-shaped” morphology, suggested by the holotype (MCT 1622-R). B. Rounded morphology, suggested by CPCA 3590 and the specimen stored in a private collection.

It was proposed that the function of the cranial crest of T. imperator was related to a self-adjustment rudder system (Frey and Martill 2003). Turns of the head at low flight speeds would result in an automatic readjustment into the wind direction, like a weather vane. However, experimental data corroborating this hypothesis are still unpublished. Cranial sagittal crests are extremely frequent among Pterodactyloidea and, as recent finds suggest, were also present in a considerable number of long-tailed pterosaurs (Czerkas and Ji 2002; Dalla Vecchia et al. 2002; Carpenter et al. 2003; Stecher 2008). The great variety and distribution of cranial crests between pterosaur clades, culminating in the huge bony crest showed by Thalassodromeus sethi (Kellner and Campos 2002), suggest that they were linked to either a successful functional or reproductive strategy such as specific recognition or sexual display. Pterosaur neuroanatomy indicates that these animals were mostly visual-based (Witmer et al. 2003), a fact that also corroborates a sexual display function for the headcrests. A thermoregulatory function was proposed for the cranial crests of Tapejara wellnhoferi (Kellner 1989) and T. sethi (Kellner and Campos 2002). This was mostly based on branching channels on the bony surface of the crests, interpreted by these authors as indicative of the presence of blood vessels. Though the huge bony crest of T. sethi could actually work as a heat-dissipation structure, this function would probably be better performed by the patagia, the same being valid for T. wellnhoferi. There is no sign of similar channels in CPCA 3590 but this could be influenced by the laterally compressed preservation of this specimen.

Fig. 7.

Most parsimonious tree of 37 steps recovered in the cladistic analysis. 1, Azhdarchoidea; 2, Azhdarchidae; 3, Tapejaridae; 4, Thalassodrominae; 5, Chaoyangopterinae; 6, Tapejarinae.

Phylogenetic relationships of Tupandactylus imperator

The completeness of the new specimen of Tupandactylus imperator described here, together with previously described remains, allow testing the phylogenetic relationships of Tupandactylus imperator through a comprehensive cladistic analysis.

Kellner (2003) was the first to include Tupandactylus imperator in a phylogenetic analysis (Tapejara imperator at that time), recovering a close relationship among Tapejara, Tupandactylus, and Tupuxuara. However, this work and further analyses (e.g., Kellner 2004; Wang et al. 2005, 2008; Martill and Naish 2006; Lü et al. 2010) used a reduced taxa sampling of Tapejaridae. The analysis presented here includes all described species of Tapejaridae sensu Kellner and Campos (2007), except Tupuxuara longicristatus Kellner and Campos, 1998, due to its incompleteness. It also includes three species that comprise the Chaoyangopteridae sensu Lü et al. (2008). The analysis has the objective to test the monophyly of Tapejaridae and the relationships of tapejarid taxa.

The dataset is based on 19 taxa and 22 characters. Three Dsungapteridae taxa were selected as successive outgroups, and the ingroup includes Azhdarchidae, Tapejaridae sensu Kellner and Campos (2007) and Chaoyangopteridae sensu Lü et al. (2008) taxa (see Appendix 3). Characters are from Kellner (2003), Martill and Naish (2006), Lü et al. (2008) and some are new (see Appendix 2).

This phylogenetic dataset was analyzed with equally weighted parsimony using TNT v. 1.1 (Goloboff et al. 2003). The low number of taxa allowed us to perform an exact search strategy using the “implicit enumeration” option. This analysis resulted in only one most parsimonious tree (Fig. 7) of 37 steps (CI = 0.83, RI = 0.9).

Fig. 8.

Biochronology of Azhdarchoidea, based on the most parsimonious tree recovered by our analysis and recorded temporal range.

The most parsimonious tree of this analysis depicts Tapejaridae as a monophyletic group, supported by three synapomorphies (see Appendix 4 for list of apomorphies), in agreement with previous works (Kellner 2004; Wang et al. 2005, 2008, 2009; Andres and Ji 2008). Also, Thalassodrominae Kellner and Campos, 2007 (Thalassodromeus sethi and Tupuxuara) and Tapejarinae (Tapejara wellnhoferi, Tupandactylus, “Huaxiapterus”, and Sinopterus) were recovered as monophyletic groups (Fig. 8).

Andres and Ji (2008) recovered a monophyletic clade composed by the Chaoyangopteridae sensu Lü et al. (2008). In the analysis proposed by the authors, this clade has a sister-group relationship with the Azhdarchidae.

According to our analysis, the Chaoyangopteridae sensu Lü et al. (2008) (Chaoyangopterus Wang and Zhou, 2006, Jidapterus Dong, Sun, and Wu, 2005, and Shenzhoupterus Lü, Unwin, Xu, and Zhang, 2008) are a clade within Tapejaridae and form a sister-group relationship with Tapejarinae. To avoid nomenclatural incoherence, we suggest name change to Chaoyangopteridae, being henceforth called Chaoyangopterinae (Lü et al. 2008), a clade defined as all pterosaurs more closely related to Chaoyangopterus zhangi than to Tapejara wellnhoferi. In our analysis, the clade is supported by a dorsal margin of nasoantorbital opening bounded by a slender bar (character 8). Eopteranodon lacks a more accurate description and illustration in order to provide a precise attribution to Chaoyangopterinae. The position of Eoazhdarcho remains dubious because cranial remains referred to this species are still unknown. It is noteworthy that, mainly due to the limited anatomical information provided at the original descriptions of most Chaoyangopteridae (sensu Lü et al. 2008) taxa, the phylogenetic relationships of these enigmatic pterosaurs are still controversial.

It is noteworthy that our analysis is congruent with the proposition of Kellner and Campos (2007) that “Huaxiapterus” jii should be renamed to Sinopterus jii. Therefore, as “H”. jii is the type species of the genus Huaxiapterus, “Huaxiapterus” corollatus, and “Huaxiapterus” benxiensis still need a new generic name (Kellner and Campos 2007).

Lü et al. (2010) also recovered a monophyletic Tapejarinae, but all taxa laid in a polytomy. The analysis presented here recovered a monophyletic Tapejarinae, positioned as the sister group of Chaoyangopterinae. The Brazilian tapejarines form a monophyletic group, whilst the Chinese tapejarines form a paraphyletic group. It suggests that the Tapejarinae originated in China and then dispersed to South America.

The monophyly of Tapejaridae

When Kellner (1989) described Tapejara wellnhoferi, this new pterosaur species was, together with Tupuxuara longicristatus (a taxon named one year before: Kellner and Campos 1988), placed in a new clade of pterodactyloid pterosaurs: the Tapejaridae Kellner, 1989. Some years later, some other pterosaur species such as Tupandactylus imperator, Tupandactylus navigans, Thalassodromeus sethi, Tupuxuara leonardii, as well as a series of toothless pterosaurs from the Jiufotang Formation (China) were positioned within this clade. According to Kellner (2004), the Tapejaridae clade is supported by five synapomorphies (listed in the introduction of the present paper).

Unwin (2003) proposed a new taxon (Neoazhdarchia) including Tupuxuara longicristatus, Quetzacoatlus northropi, their most recent common ancestor and all its descendents. This clade was supported by the presence of a notarium and the loss of contact between metacarpals I–III and the syncarpal. After Unwin (2003), a notarium is reported for azhdarchids and Tupuxuara but is absent in Tapejara. According Unwin (2003), in Tapejara at least one of these metacarpals retains the contact with the syncarpal. An additional character supporting Neoazhdarchia would be a long rostrum, measured from the anterior margin of the orbit to the anterior end of the premaxillae, with more than 88% of the total skull length, contrasting with Tapejara which, after Unwin (2003), has a rostrum of only 70% of the total skull length. Therefore, following Unwin and Lü (1997) and Unwin (2003), the Tapejaridae sensu Kellner would be paraphyletic with respect to Azdharchidae. This was further supported by Martill and Naish (2006) and Lü et al. (2006).

The analyses of Kellner (2003, 2004), Wang et al. (2005, 2008, 2009) and Andres (2008) supported monophyly of the Tapejaridae sensu Kellner and Campos, 2007. According to Kellner and Campos (2007) the loss of contact of metacarpals I–III with the syncarpal is not conclusive for Tupuxuara due to the fact that these structures are not completely preserved in any known specimen of this taxon. Kellner (2004) argued that the presence of a notarium is a dubious feature because this structure is only observed in ontogenetically mature individuals. With respect to the comparative length of the rostrum, Kellner (2004) argued that, although there is a difference between the taxa, the basic configuration and proportions of the skull in Tupuxuara and azhdarchids are distinct, with Tupuxuara having a larger nasoantorbital fenestra and a shorter rostrum (anterior to the nasoantorbital fenestra) in respect to the total length of the skull. This configuration is also observed in Thalassodromeus.

It is also noteworthy that in most of the Brazilian tapejarines postcranial elements are missing (e.g., Tupandactylus imperator and Tupandactylus navigans) or are only known from immature individuals (e.g., Tapejara wellnhoferi; Wellnhofer and Kellner 1991; Kellner 2004). Some of the Chinese Tapejarinae are also known only from juveniles (Kellner and Campos 2007). Therefore, the absence of structures such as a notarium could be biased by ontogenetic factors. Besides this, the condition of the contact between metacarpals I–III and the syncarpus cannot be accessed. According to Lü et al. (2003) “Sinopterus gui” has a notarium. However, Kellner and Campos (2007) stated that in “S. gui” a notarium is in fact absent and, based also on other anatomical and biometrical features such as the proportions between femur and tibia, synonymized this species with Sinopterus dongi.

In our phylogenetic analysis the Tapejaridae sensu Kellner is supported by three synapomorphies: (1) nasoantorbital fenestra forming 45% or more of the skull length measured from the tip of premaxilla to the caudal terminus of the squamosal (2) orbit reverse pear-shaped and (3) very thin sub-vertically orientated lacrimal process of the jugal.

It is noteworthy that a “long sagittal crest made mostly by the premaxillae that starts at the rostral end of the skull and extends posteriorly”, advocated by Kellner (2003, 2004) and Kellner and Campos (2007) as a synapomorphy of Tapejaridae, is broad enough to agglutinate two very distinct morphologies: In Tapejarinae, the premaxillary crest has a constriction dorsal to the nasoantorbital fenestra, whereas in thalasso-drominids this structure remains high dorsal to the nasoantorbital fenestra and extends posteriorly together with the frontoparietals forming a high laterally compressed post-occipital crest.

Although it is possible that a broad tubercle situated at the ventroposterior margin of the coracoid is indeed a synapomorphy of Tapejaridae, as was proposed by Kellner (2003, 2004) this character was so far observed only in T. wellnhoferi and T. leonardii. The presence of a coracoid tubercle among the Chinese tapejarids could be only confirmed through a closer examination of the specimens.

Conclusions

CPCA 3590 has sufficient diagnostic features that allow its accurate identification as Tupandactylus imperator. The specimen constitutes the best-preserved T. imperator known so far. It preserves associated soft tissues such as a soft-tissue headcrest, a ramphotheca and, probably, pycnofibres. Probably the most interesting feature of the new specimen is the associated lower jaw, which has a morphology similar to the one observed for Tapejara wellnhoferi. A revised diagnosis is proposed for the species, with new characteristics related to the lower jaw and the sagittal cranial crest.

The cladistic analysis showed that Tupandactylus imperator, together with Tupandactylus navigans, Tapejara wellnhoferi, Sinopterus and “Huaxiapterus” forms a monophyletic group, the Tapejarinae, in agreement with Kellner (2007). Thalassodrominae and Tapejaridae are also considered monophyletic taxa.