The first theropod tracks from the Middle Jurassic of Gansu, Northwest China: new and rare evidence of quadrupedal progression in theropod dinosaurs

The discovery of dinosaur fossils in Gansu Province can be traced to 1930–1931 when the Sino-Swedish Expedition collected the first dinosaur material from the Mesozoic basins of western Gansu (Bohlin 1953). Only one possible Late Jurassic sauropod was found in the central part of Gansu before this study (Young 1948). This was eventually attributed to the sauropod genus Mamenchisaurus (Young and Zhao 1972). All other records from this region, including abundant skeletons and tracks, are from the Lower Cretaceous rock units.

In Gansu, the Yanguoxia tracksites yielded the most abundant and diverse track records, including well-preserved sauropod, ornithopod, dromaeosaur, theropod, pterosaur, and bird tracks and trackways (Zhang et al. 2006; Xing et al. 2013a, 2016a, 2016b; Li et al. 2015). In the Honggu (Xing et al. 2014a), Zhongpu (Xing et al. 2014b, 2015), Guanshan (Xing et al. 2016a), and Changma (Xing et al. 2017) areas, some non-avian dinosaur and bird tracks were discovered.

The Institute of Vertebrate Paleontology, Gansu Agricultural University (IVPG) found an assemblage of tracks in the Baojishan Basin (Fig. 1) in year 2017. In the same fieldwork session, the IVPG also found a mamenchisaurid dinosaur, several large-sized theropod teeth and stegosaurian remains in the Baojishan Basin. In 2018, authors of this study investigated this tracksite again to make resin and 3D digit models. These tracks represent the only evidence of the Middle Jurassic small- to middle-sized theropod in Gansu, Northwest China.

The study area is in the northwestern Baojishan Basin, Pingchuan District of Baiyin City, Gansu Province. The Baojishan Basin is at the eastern end of Qilian Mountains. As a fault-bounded subsiding basin, it comprises a small part of the Hexi Corridor Basin which can be traced back to the Late Triassic and Jurassic. Its Middle Jurassic layers consist of two lithostratigraphic units, the Longfengshan Formation and the Wangjiashan Formation, in disconformable contact with the underlying Upper Triassic, and in conformable contact with the overlying Upper Jurassic (Fig. 2). The 607-m-thick Wangjiashan Formation was informally divided into two members based on differences in lithology and lithofacies (Li et al. 1982; Zhang 2009) (Fig. 2). The lower member is composed of straw-yellow sandstones (477 m thick) which were deposited mainly in a lake deltaic setting with its succession characterized by straw-yellow color. The coarse sandstones at the bottom gradually change to finer grained units with red mudstones interbedded in the upper part of this member. The upper member is 113 m thick, and comprises gray to greenish-gray siltstones and sandy mudstones, interbedded with thin beds of oil shale, mainly representing a lake environment.

Dinosaur fossils were recovered from sandstones in the upper part of the lower member (Fig. 2).

The oil shale member has yielded plant macrofossils, palynoflora and a lacustrine invertebrate assemblage, including conchostracans, charophytes, and bivalves. Based on this fossil material, the underlying straw-yellow sandstone member is thought to have been formed in the late Bathonian (Li et al. 1982; Du 1985; Zhang 2009). Similarities in the invertebrate and dinosaur assemblages (Li et al. 2011; Deng et al. 2017) imply some affinity between the Wangjiashan Formation in the Baojishan Basin and the Lower Shaximiao Formation in the Sichuan Basin.

Tracks preserved as convex hyporeliefs (i.e., natural casts) were digitally photographed (115–164 images of each) from various angles using a Nikon D5 with a 24 mm lens. Photographic jpg images were added to Agisoft Photoscan Professional Edition (Version 1.2.6 build 2038 64 bit) to generate scale-corrected models following the procedure adapted from Xing et al. (2018). The 3D models were automatically aligned to the centre of the Cartesian coordinate system using R Studio (Version 1.1.453; Lallensack 2018). The surface topography was visualised as false-colour elevation, contour maps, and ambient occlusion images using Paraview (Version 5.0.0 64 bit; Ahrens et al. 2005) and CloudCompare (Version 2.6.1 64 bit; www.cloudcompare.org) software.

Olsen (1980), Weems (1992), and Lockley (2009) suggested that theropod tracks could be differentiated by their mesaxony (M): how far the tip of the central digit (III) is projects anteriorly beyond the tips of the medial (II) and lateral (IV) digits, defining an anterior triangle. In most assemblages, the L/W ratios of the anterior triangle (used as an index of mesaxony in this paper) and of the whole track show a positive correlation, that is, narrow tracks are normally stronger in mesaxony than wide tracks.

Dinosaur tracks from Pingchuan tracksite were visible on bedding planes in three areas denoted here as assemblages A, B, and C covering the surface of 3.5 m², 0.27 m² and 0.34 m² respectively (Figs. 3, 4, 5, 6). Assemblages A, B, and C were located in different layers. A is in the relatively highest layer, B is about 30 cm below A, and C is the lowest, about 90 cm below A. All tracks were preserved in fine sandstones. Assemblages A and B yield many microbially induced sedimentary wrinkle structures and distinct mud cracks. Assemblage C has a lot of scratch marks parallel to the tracks.

Assemblage A has at least 37 tracks, most of which are oriented to the northwest while a minority point in the opposite direction, to the southeast. Quite a few tracks overlap. No clearly distinguishable trackway has been found. Measured tracks including PC-TI1–PC-TI14 are relatively well-preserved. These tracks can be divided into two types. Type A represents classic Grallator-like tridactyl theropod tracks, ranging from 11.0 cm to 19.6 cm in length and 1.4 to 2.4 in the L/W ratio. The divarication angles between digits II and IV range between 31° and 60°. The L/W ratios of the anterior triangle (mesaxony) are 0.32 to 0.76. Some footprints, such as PC-TI1 and PC-TI11, have elongated heels which are likely to be traces of the metatarsal-phalangeal pads. The best preserved track is PC-TI3, with a length of 12.6 cm, a width of 8.2 cm, a L/W ratio of 1.5, a divarication angle between digits II and IV of 46°, and a mesaxony measure of 0.43. Type B represents very wide tridactyl tracks and includes PC-TI6 and PC-TI8. For PC-TI6, the length is 12.1 cm, the L/W ratio is 0.8, the divarication angles range between digits II and IV is 104°, and the L/W ratio of the anterior triangle (mesaxony) is 0.40. For PC-TI8, the length is 13.8 cm, the L/W ratio is 0.7, the divarication angles range between digits II and IV is 90°, and the mesaxony is 0.28. Type B tracks are poorly preserved with unclear digit pads, and probably reflect extramorphological expressions of Type A tracks.

In addition to pes tracks Type A and Type B of the assemblage A (Type B most likely representing extramorphological variants of Type A), other tracks can be inferred to represent manus tracks. These have ichnotaxonomic significance, but are not given numerical or alphabetic labels, because we consider them to be associated with four tracks (PC-TI1, PC-TI5, PC-TI8 and PC-TI9) in the Type A group.

Assemblage B has at least 4 tracks oriented to the northwest, among which PC-TI15 is in the best preservation condition and in similar morphology to PC-TI3.

Assemblage C has at least 12 tracks, among which PC-TI16–PC-TI18 are better preserved. All of them lack metatarsophalangeal pad trace, so they only preserve traces of the main, distal part of digits II-IV. The lengths of PC-TI16–PC-TI18 are between 6.2 cm and > 8.8 cm (Table 1), measured from the traces of digit III, which are the deepest. The traces of digit II and IV are significantly shorter and much shallower (cf. Avanzini 1998). Some of the better preserved tracks, such as PC-TI16, show two observable digit pad traces associated with digit III. These tracks are frequently overlapped and both PC-TI17 and PC-TI18 overlap on another tridactyl track. All tracks point to northwest.

An interesting feature of the Pingchuan trackway is that tracks from three different levels all show similar orientation trends. The majority of mapped tracks (about 16 in number) from the largest assemblage A shows a northwest orientation with a lesser number (~ 11) oriented exactly to the opposite southeast direction. Meanwhile, in assemblages B and C all tracks are oriented to the northwest. This suggests there could have been some physical paleoenvironmental constraints, for example a convenient linear drainage that persisted for the duration represented by the deposition of the 90 cm sediment between levels A and C. A second alternative interpretation would be that theropod dinosaurs were moving as gregarious groups, in the same direction on three separate occasions (cf., Ostrom 1972). The third alternative explanation is that both above factors influence the trackmakers’ directions: groups of gregarious theropods moved in a landscape with physical feataures which encouraged movement in a preferred direction.

An important difference between tracks from the larger assemblage A and the smaller assemblage C is that the tracks from the smaller assemblage are much shorter and apparently digitigrade. Thus the difference in average length is 13.58 cm compared with 7.77 cm for assemblages A and C, whereas the width difference is only between 8.76 cm and 7.36 cm. This clearly indicates that the tracks in assemblage C appear more digitigrade, due to preservational or behavioral factors. We have also noted that digit III traces are deeper in assemblage C. This is a same pattern reported by Avanzini (1998) in analysis of the preservation of theropod tracks of 17.0 cm long, almost exactly the same length as tracks in the Pingchuan assemblages.

The Pingchuan theropod tracks can clearly be referred to as grallatorid in the broader sense of the Grallator-Eubrontes plexus (Lockley et al. 2013), and are here assigned to a new ichnospecies (G. pingchuanensis) of the former ichnogenus on the basis of a unique and diagnostic trackway pattern with novel implications for theropod behavior. All four known examples, in order of reporting (Connecticut, Poland, Utah and Gansu), are from the Lower-Middle Jurassic, and all represent Grallator or a closely related ichnogenus. This raises questions as to whether crouching theropod Grallator, or Grallator-like trackmakers engaged in gaits and postures that allowed registration of manus tracks. Since no trackway of a quadrupedal theropod has previously been reported, it is not known if the short step and inward rotation of the pes is a necessary corollary of the wide spacing of the manus traces and their positions relative to the pes. A functional locomotor/biomechanical analysis of potential trackmakers based on the known Lower Jurassic theropod skeletal reconstructions is outside the scope of this paper but could shed light on these questions.

It is well known that Grallator and Grallator-like tracks were widely distributed globally during the Early-Middle Jurassic. For example, Lucas (2007) considered the theropod tracks Grallator and Eubrontes (and Kayentapus) as indicative of the Early Jurassic biochron (cf. Lockley et al. 2013). Xing et al. (2013b, 2013c) described the medium-sized theropod tracks from the Middle Jurassic of Southwest China as Chongqingpus in Nan’an and Jinji tracksites in Chongqing region (Xing et al. 2013b) and as Kayentapus in Panxi region (Xing et al. 2013c). The latter assemblages are heavily theropod-dominated. Thus, the Pingchuan assemblages are similar to others of the Early-Middle Jurassic theropod track assemblages in China (Lockley et al. 2013).

The Pingchuan theropod tracksite in Gansu, Northewest China provides the fourth example (reporting after examples successively from Connecticut, Poland, Utah) of a theropod trackway with convincing examples of ectaxonic, tridactyl manus tracks, in this case forming a configuration of four consecutive manus pes sets. This example contrasts with others where only two tracks were recorded in association with traces indicating crouching behavior. All examples to date are the Early or Middle Jurassic in age. This unique example of a trackway of a theropod progressing quadrupedally is designated as Grallator pingchuanensis ichnosp. nov. Among the trackways, unique features are the evidence showing that the trackmaker’s forelimbs register a wide straddle gait. The trackmaker was walking slowly which made deep tracks over a soft substrate. This is the first evidence that theropods occasionally registered manus traces while moving, rather than only when crouching.