On some anthracotheriid (Artiodactyla, Mammalia) remains from northern Greece: comments on the palaeozoogeography and phylogeny of Elomeryx

Abstract

A few isolated mammal teeth from some Greek coal samples of unknown origin represent a small bothriodontine anthracotheriid, ascribed to Elomeryx. Analysis of the coalification stage of the fossil-bearing coal samples indicates the Lower Miocene Moschopotamos coal pits (Katerini Basin) as the most probable site of origin. The studied teeth are metrically close to E. crispus from Western Europe but share dental apomorphies with E. borbonicus, and E. japonicus and along with Elomeryx material from some Greek and Turkish sites pose a number of systematic, biochronologic, zoogeographic and phylogenetic questions. In the light of new evidence it seems that a small but advanced Elomeryx spanned the Oligo-Miocene boundary of S. Balkans. Furthermore, a revision of the old southern Balkan record together with a parsimony analysis suggest that Bakalovia is a first stage within the evolutionary history of Elomeryx, which complicated phylogeography is further discussed.

Introduction

Anthracotheriidae, a primitive and extinct group of Palaeogene to Pliocene artiodactyls known from the Old World and North America, is central to the discussion about the origin of hippos (e.g., Boisserie et al. 2005; Pickford 2008). Varying greatly in size, anthracotheriids are characterized by a long and narrow amastoid skull, orbits behind the M2 in derived taxa, heavy zygomatic arches, open post-orbital bar, narrow palate, complete dental formula, brachyodont and bunoselenodont molars with diverse trends in molar selenodonty, sexually dimorphic incisors and canines, upper molars with four or five cusps without hypocone but with a well-developed metaconule, enlarged third molars and a well-developed hypoconulid on the m3 (Coombs and Coombs 1977; Lihoreau and Ducrocq 2007). Botriodontine anthracotheriids are widely distributed in the Old World and North America, known by several genera and species (Lihoreau and Ducrocq 2007). Among them, Elomeryx Marsh, 1894 is the most widespread taxon but its intra- and intergeneric phylogenetic relationships are not yet fully understood (Hellmund 1991; Ducrocq and Lihoreau 2006; Lihoreau and Ducrocq 2007; Lihoreau et al. 2009; Tsubamoto and Kohno 2011).

Lüttig and Thenius (1961) were probably the first to record the presence of Elomeryx in the Greek part of Thrace (Fig. 1b). Over the next few years Ozansoy (1962), and Nikolov (1967) described more material of the same genus from Eastern (Turkish) Thrace, and Bulgaria, respectively (Fig. 1b). Nikolov and Heissig (1985) proposed later ascribing the Bulgarian sample from Burgas to the new genus Bakalovia. Since the 1960s no more Elomeryx material has been reported from the area, but recently a mandible of Elomeryx from Tozakli, Eastern Thrace, was briefly discussed by Islamoğlu et al. (2010) (Fig. 1b).

Fig. 1

Greek Palaeogene–early Miocene coal deposits (a) and local geographic distribution of bothriodontine anthracotheriids (b). 1 Orestias, 2 Alexandroupolis, 3 Kotili, 4 Paranesti, 5 Moschopotamos, 6 Middle Hellenic trough, 7 Ioanian flysh (Zagoria), 8 Aliveri-Kymi, black pentagon Brachyodus, Kalimeriani, Euboia; asterisk Bakalovia, Burgas, Bulgaria; squares Elomeryx from a Chandras, Greece, b Haskoy, Turkey, c Sigircili, Turkey, d Tekirdrağ, Turkey, e Tekirdrağ-Malkara, Turkey, f Tozakli, Turkey, x possible provenance of studied sample: Moschopotamos coal pits (data from Christanis 2004; Islamoğlu et al. 2010; Lüttig and Thenius 1961; Melentis 1965; Nikolov 1967; Ozansoy 1962; Papanicolaou et al. 2004)

The 2009 reorganization and updated cataloging of the collections of the Museum of Geology and Palaeontology of the Geological Department of the Aristotle University of Thessaloniki (LGPUT) revealed a number of coal blocks containing isolated teeth of a bothriodontine anthracotheriid described and discussed here. Although small, this fossil sample is significant in view of the extreme rarity of pre-middle Miocene large mammals in Greece and the Southern Balkans and the inadequate knowledge of SE European anthracotheriids.

Unfortunately, the fossil-bearing coal blocks lack geographic and/or stratigraphic labels, requiring an indirect inquiry of their origin. Coal deposits are widely distributed in Greece. Most of them occur in latest Neogene–Quaternary stratigraphic sequences that undoubtedly postdate the known time-range of Eurasian anthracotheriids. Palaeogene–middle Miocene Greek coal deposits are rare, being known mainly in western Thrace-Eastern Macedonia (1–4 in Fig. 1a), Katerini District (5 in Fig. 1a), Central Euboea (8 in Fig. 1a), and Western Greece (6, 7 in Fig. 1a) (Christanis 2004; Papanicolaou et al. 2004). Apart from Western Greek coal locations that are related to syn- and post-orogenic processes and marine–brackish environments, all the other mentioned sites could be considered as possible provenance areas of the studied material.

Materials and methods

In order to assess the possible origin of the fossil material (‘unknown’ coal sample #1), a random coal sample was collected from four coal deposits suspected to have yielded the fossils, namely Avantas (Eastern Greek Thrace; sample #2), Xanthi (Western Greek Thrace; sample #3), Siderokastro (Eastern Macedonia; sample #4) and Moschopotamos (Katerini Basin; sample #5). As the petrological, chemical and mineralogical compositions of coal vary strongly from place to place within one deposit, the comparison among various coal samples cannot depend on the maceral, geochemical and mineralogical analyses, respectively. In contrast, the huminite/vitrinite reflectance provides a tool for measuring the coal rank, i.e., the coalification (sensu maturity) stage a coal has reached. The rank within a coal deposit varies slightly or usually not at all. As the measurement is carried out on a single maceral group (huminite), particularly on one specific maceral only, namely eu-ulminite B, petrographic, chemical or mineralogical variability is avoided (Taylor et al. 1998).

All the samples were dried at 105 °C in order to remove residual moisture and then crushed to −1 mm and homogenized. Polished blocks were prepared according to International Standards (ISO 7404-2 2004a): the crushed samples were firstly embedded with epoxic resin (Epofix of STRUERS^®) in cylindrical mounts and the prepared blocks were ground using a series of grinding papers (no. 320, 500, 800, 1200, and 2400), then polished using diamond pastes (∅ < 3 μm and ∅ < 1 μm) and finally, alumina suspension (∅ < 0.06 μm). The blocks were examined under a LEICA DMRX coal-petrography microscope equipped with an MPV photometer in oil immersion and 500× total magnification. Random huminite reflectance (R _r) was measured on eu-ulminite B according to ISO 7404-5 (2004b).

Dental terminology follows Coombs and Coombs (1977) and Lihoreau and Ducrocq (2007: fig. 7.1). The taxonomy follows Hellmund (1991) with the exception of Brachyodus cluai Deperet, 1906 that is considered to be a distinct species of Elomeryx (in agreement with Ducrocq and Lihoreau 2006). All measurements are in mm.

The studied material is compared with E. crispus samples from: Möhren 13, Möhren 9 and Grafenmuhle 10 stored in the Bayerische Staatssammlung für Paläontologie und Ηistorische Geologie, München (BSPM); E. borbonicus samples stored in the Centre de Conservation du Musée des Confluences, Lyon (CCMCL) and in the Laboratoire de Géology, Université Claude-Bernard Lyon I (GUL); the cast of PIUW 16397 (here referred to as the Chandras palate) described by Lüttig and Thenius (1961) and stored in the Department of Paleontology, University of Vienna (PIUW); Elomeryx sample from Sigircili, Eastern Thrace stored in BSPM; Elomeryx material from Tekirdrağ-Malkara, and Tozakli Eastern Thrace stored in the Paleontological Department of Natural History Museum of the Mineral Research and Exploration Institute of Turkey, Ankara (MTA).

The phylogenetic relationships among Elomeryx species and related taxa discussed in this paper were obtained through a parsimony analysis based on 23 binary and 7 multistate dental, and mandible, characters, 19 of them adopted from Lihoreau and Ducrocq (2007) (Electronic Supplementary Material, Appendices I and II). The branch and bound algorithm and Fitch optimization criteria have been followed using PAST ver. 2.12 free software (Hammer et al. 2001). The cladogram was rooted to Siamotherium Suteethorn et al., 1998, regarded as the most primitive member of the family Anthracotheriidae (Lihoreau and Ducrocq 2007).

Origin and age of the material

The results of the reflectance measurements are presented in Table 1. The Oligocene Xanthi (sample #3) coal deposits display higher mean random reflectance values (0.7 %) than the unknown one (sample #1; 0.39 %). High mean reflectance values are also reported for Upper Paleogene coals from Turkey and the Balkans (Belkin et al. 2010; Karayigit et al. 2002), whereas mean reflectance values of the rich Upper Miocene to Pliocene Greek coals are usually below 0.35 % (Christanis 2004). Both samples picked up from the Oligocene Avantas (sample #2) and the Oligo-Miocene Siderokastro (sample #4) coal deposits show minimum reflectance values close to the maximum value of the unknown sample (sample #1), but they exhibit a significantly higher coalification stage on the average (Table 1). The reflectance of the unknown (#1) sample is absolutely similar to the ranges reported for Lower Miocene coals from Serbia, and Turkey (Demirel and Karayigit 1999; Gürdal and Bozcu 2011; Karayigit 2005; Žitović et al. 2005, 2010), and identical to that from Moschopotamos sample (sample #5 in Table 1) that proved to have reflectance values ranging from 0.35 to 0.45 %, in agreement with the data provided by other researchers (Kalkreuth et al. 1991; Kotis 1997; Papanicolaou 1994; Papanicolaou et al. 2000). Thus, Moschopotamos seems to be the most probable site of origin of the unknown coal sample and consequently of the fossils in question.

Table 1 Results of the random mean reflectance measurements (R _r,oil in %)

The Moschopotamos coal deposit is located 15 km west of the city of Katerini, N. Greece (5 in Fig. 1a; 40^o19′20″N, 22^o18′51″E). The lignite derived from helophytic reed-sedge and tree vegetation and it belongs to the Moschopotamos lithostratigraphic formation (Fm) of Katerini Basin that is exposed around the homonymous village and represents a low deltaic depositional environment (Kotis 1997; Sylvestrou 2002 and literature therein). Moschopotamos Fm consists of clays, sands, silts, and marly silts with lignite intercalations few cm to 1 m thick. Biochronological data based on pollen and micromammals indicate that the Moschopotamos Fm is most likely to be of late Lower Miocene age (Benda and Meulenkamp 1979; Ioakim and 1986; Kotis 1997; Sylvestrou 2002).

Systematic palaeontology

• Class: Mammalia Linnaeus, 1758

• Order: Artiodactyla Owen, 1848

• Family: Anthracotheriidae Leidy, 1869

• Subfamily: Bothriodontinae Scott, 1940

• Genus: Elomeryx Marsh, 1894

• Type species: Elomeryx armatus Marsh, 1894

• Diagnosis: in Lihoreau and Ducrocq (2007: 95)

Elomeryx sp.

Studied material

Upper right canine, LGPUT MP35; right P3, LGPUT MP36; right P4, LGPUT MP37; right m1 or m2, LGPUT MP38; right m3, LGPUT MP39.

Comparative description

Nearly all fossil specimens extracted from or still in the coal samples belong to an extremely deformed and crashed mandible and maxillae of a single old male individual, judging from the size and shape of the canine and the wear stage of the m3. The low-crowned cheek teeth are of bunoselenodont type with finely wrinkled enamel and rather weak cingula. P4 have well-developed cingula and the distal crest of the lingual cusp does not reach the distal cingulum; the buccal cuspids of the m1 or m2, and the m3 are highly crescentic and wider than the lingual ones; the hypoconulid of the m3 forms a transversely compressed loop that is rather narrow and buccally displaced; the prehypocristid (cristid obliquid) of the m3 is strong, descending on the distal trigonid wall but without reaching the lingual wall of the tooth; four cristids issue from the m3 metaconid. This set of features suggests Elomeryx as the most likely generic affiliation (Hellmund 1991; Lihoreau and Ducrocq 2007; Lihoreau et al. 2009; Tsubamoto and Kohno 2011). European species of Elomeryx include E. crispus known from the latest middle Eocene to early late Oligocene, E. borbonicus known from the late Oligocene–early Miocene, and E. cluai (considered by Hellmund 1991 to be subspecies of E. crispus) known exclusively from the late early Oligocene of Spain (Hellmund 1991; Ducrocq and Lihoreau 2006; Lihoreau et al. 2009).

The thin enameled, pointed and caniniform upper canine is compressed transversally, and has an elongated tear-shaped cross section with a rather sharp and finely serrated distal edge (Fig. 2). This condition differs from the one usually seen in E. crispus and E. cluai and is more similar to the upper canine morphology reported for E. borbonicus (Hellmund 1991). Original comparison of LGPUT MP35 with the male canine of GUL FSL8514 of E. borbonicus from St. Henri, France (Geais 1934), shows close morphological resemblance with similar flattening (TDb/APDb × 100 = 57.7 and 53.6 %, respectively) and backward curvature; nevertheless, the Greek specimen is about 30 % smaller in linear measurements (Table 2). Two upper male canines of E. crispus from Möhren 13 (namely, BSPM1529, and BSPM1480) are similar in size with LGPUT MP35 (Table 2) but less compressed transversally (TDb/APDb × 100 = 74.5 and 64.0 %, respectively), and not serrated distally. Hellmund (1991) indicated male upper canine serration as a diagnostic feature of E. borbonicus, whereas Lihoreau and Ducrocq (2007) designate this character as diagnostic at the generic level. Although the total number of canine specimens per species is rather inadequate for definite conclusions, we are inclined to recognize this feature as characteristic of advanced Elomeryx species.

Fig. 2

Upper canine LGPUT MP35 in a mesial and b lingual view. APDb anteroposterior diameter at the base of the crown, TDb (not shown) buccolingual diameter at the base of the crown (obtained perpendicularly to APDb), Lext length of the canine crown along its anterior surface, Hint distal height of the canine from the tip to the base of the crown

Table 2 Metrical comparison of the upper canine of E. crispus, E. borbonicus and Elomeryx from Moschopotamos

The P3 (Table 3) is small for E. borbonicus (P3 length > 15.7; Lihoreau et al. 2009) and slightly narrower than that of E. crispus (P3 width > 10.6; Lihoreau et al. 2009), metrically approaching the Spanish E. cluai (Lihoreau et al. 2009: tab. 1). The tooth is heavily worn but it shows a typical disto-lingual talon. In occlusal view the tooth outlines a right triangle with the right angle at disto-lingual position. The tooth looks more symmetrical than those of E. crispus figured by Hellmund (1991) and Lihoreau et al. (2009) and seems to be rather closer to the symmetrical P3 morphology seen in E. armatus (Hellmund 1991: Pl. 1, fig. 1) but intraspecific variation may be great. The buccal cingulum is weak and the distal and palatinal roots are fused together, a condition also seen on some specimens of E. crispus (Hellmund 1991).

Table 3 Teeth measurements of Elomeryx from Moschopotamos, Greece

The P4 is metrically similar to that from the Chandras palate (Lüttig and Thenius 1961), i.e., slightly shorter and significantly narrower than the P4 of E. borbonicus, and within the size ranges of both E. crispus and E. cluai (Fig. 3; Table 3). The cingulum of the P4 is strong and the distal crest of the lingual cusp does not reach the distal border of the tooth as in all Elomeryx species (Lihoreau et al. 2009) (Fig. 4). 82 % of the P4 of E. crispus lack lingual crests on the lingual cusp, whereas 16 % of the specimens bear a disto-lingual crest (Hellmund 1991). In the studied sample the lingual cusp of P4 bears a rather strong crest trending mesio-lingually to lingually (as in some specimens of E. cluai and E. borbonicus according to Hellmund 1991; Schaub 1948), whereas another weak disto-lingual crest is present, though weakly developed (Fig. 4). The mesial border of the P4 is rather straight instead of concave and therefore more similar to E. borbonicus than to E. crispus.

Fig. 3

Metrical comparison of P4 of the European species of Elomeryx. Cross LGPUT MP37, squares Elomeryx palate from Chandras, Greek Thrace (data from Hellmund 1991; Lihoreau et al. 2009; Lüttig and Thenius 1961)

Fig. 4

Morphological comparison of P4, LGPUT MP37. a Occlusal morphology of LGPUT MP37; b, c, f P4 morphotypes of E. borbonicus; d P4 morphotype of E. borbonicus and advanced E. crispus; e P4 morphotype of E. cluai (data from Hellmund 1991)

The m1 or m2 is slightly larger than that of E. cluai, and within the size range of both E. borbonicus and E. crispus (Table 3; metrical comparison with data provided by Hellmund 1991). The tooth is strongly worn prohibiting detailed morphological observations. A weak mesial and a less developed distal cingulum are still present. The posthypocristid is rather short and thus the posterior lobe appears open disto-lingually but this may be only due to the advanced wear stage.

The m3 is larger than that of E. cluai and shorter than that of E. borbonicus, being well within the reported size range of E. crispus (Fig. 5; Table 3). The hypoconulid of LGPUT MP39 is bicuspid like that of E. crispus from Detan Dvérce (Czech Republic) illustrated by Hellmund (1991: Pl. 4, fig. 6), and much less compressed bucco-lingually than that of E. borbonicus. Similar to E. borbonicus and advanced (sensu Hellmund 1991: fig. 7) E. crispus the posterior lobe shows a characteristic “Y” pattern resulting from the crossing of the prehypocristid with the preentocristid, whereas a direct mesial entoconid–hypoconid connection is absent (Fig. 5). The postectometacristid of LGPUT MP39 is less prominent than in E. crispus (e.g., Hellmund 1991, Pl. 4, figs 5–8) and comparable to that of E. borbonicus. LGPUT MP39 also shows a weak and short posthypocristid, linked with the prehypocristulid (or pre-hypoconulid cristid) but also with the entoconid through a weak accessory transverse crest (Fig. 5). This character is unlike E. crispus and E. borbonicus, both of which exhibit a strong crest directly connecting the hypoconid with the entoconid distally, and intermediate between Elomeryx and Bakalovia (Nikolov 1967; Nikolov and Heissig 1985; Hellmund 1991), the latter lacking the junction between the posthypocristid and postentocristid in the lower molars. The specimen LGPUT MP39 is close dimensionally to Bakalovia palaeopontica and smaller than Bakalovia astica which, however, might show an incipient transverse crest on the entoconid of some m3 (Nikolov 1967: pl. 18, fig. 3).

Fig. 5

Metrical and morphological comparison of m3, LGPUT MP39. Circles represent mode values, dashed lines represent measurements of LGPUT MP39, A occlusal and B buccal views of LGPUTh MP39. a Occlusal basic morphological pattern of LGPUT MP39 (reversed), b m3 morphotype of E. crispus, c m3 morphotype of advanced E. crispus, d m3 morphotype of E. borbonicus, e m3 morphotype of Bakalovia (data from Hellmund 1991)

Phylogenetic relationships

The cladistic analysis carried out by Lihoreau and Ducrocq (2007: fig. 7.2) failed to adequately resolve the phylogenetic relationships among Elomeryx species and its allies, due to the morphological stability at the genus level, as well as to the high degree of homoplasies among anthracotheriids. The parsimony analysis conducted here is principally based on a subset of the data matrix used by the above authors (Lihoreau and Ducrocq 2007: table 7.1) with minor modifications and addition of 11 dental characters (ESM Appendices I and II), as well as with the introduction of three more taxa: Bakalovia Nikolov and Heissig, 1985 from Bulgaria (Nikolov 1967), Elomeryx japonicus (Matsumoto in Tokunaga, 1925) from Japan (Tsubamoto and Kohno 2011) and Elomeryx from Moschopotamos, Greece (ESM Appendix I). The analysis provided three maximally parsimonious trees 56 steps long (Fig. 6). As in Lihoreau and Ducrocq (2007: fig. 7.2), the monophyly of Elomeryx is demolished. All trees support Bakalovia as the sister group of Elomeryx plus advanced bothriodontines (node 2 in Fig. 6). The Miocene Indo-Pakistani Sivameryx appears as the sister group of a clade incorporating Elomeryx species and Arretotherium (node 3 in Fig. 6). Nevertheless, some character changes interpreted as homoplasies in the current analysis [i.e., 27(1), d(1), f(1) and j(2)] might indicate closer relationships between Sivameryx and advanced Elomeryx species (e.g., Lihoreau 2003). Based on a single reversal [12(0)] Arretotherium appears as the sister group of Elomeryx armatus; this North American clade is branching next to E. crispus and appears as the sister group of an unresolved clade incorporating advanced Eurasian Elomeryx species (E. borbonicus, E. japonicus, and Elomeryx from Moschopotamos) (node 6 in Fig. 6).

Fig. 6

Strict consensus of three most parsimonious trees (Length: 56; ECI: 0.64; ERI: 0.84) from analysis of the data matrix in ESM Appendix I. Character changes supporting nodes: node 1 16(1), 17(1), 34(1; reversed in Bakalovia), 38(1); node 2 7(1), 8(1), 12(1), 24(1), 29(1), 38(2); node 3 26(1), b(1); character changes 17(2), 18(1; reversed in Arretotherium), and 39(1) may further support node 2 or node 3; node 4 23(1), 35(1), g(1; reversed in Elomeryx from Moschopotamos); node 5 27(1), a(1), i(1), k(1), j(2); node 6 6(1), 9(1), b(2)

Discussion and concluding remarks

The Paleogene–early Miocene Thrace and adjacent areas of the Southern Balkans represented a suitable environment for anthracotheriids, especially bothriodontines: Brachyodus, Bakalovia and Elomeryx have a sporadic but certain fossil record here (Fig. 1b), whereas Elomeryx appears diachronously. Hellmund (1991) attributed most of the old Greek and Turkish Elomeryx material to Elomeryx crispus. Reexamination of the Chandras palate (PIUW 16397), originally described by Lüttig and Thenius (1961: fig. 1), leaves no doubts about its generic attribution (typically pentacuspidate molars with small paraconule almost equal in size to the protocone; protocone with a single preprotocrista and double postprotocrista; preprotocrista in contact with the paraconule; postparacristule in contact with the buccal postprotocrista). Nevertheless, some particular features, as well as the tooth proportions, make the species affiliations more delicate. The metaconule displays three crests, the strong mesio-lingual one present in E. cluai and E. crispus but not in E. borbonicus. The less worn M2 and M3 show a weak lingual postprotocrista that does not reach the mesio-buccal crest of the metaconule (or premetacristule) but such a join is clear on the M1 (Lüttig and Thenius 1961: fig. 1). Thus, the transverse valley is continuous on the M2 and M3, as in E. borbonicus, but blocked lingually in the M1, as in E. crispus. The postprotocrista is buccally directed and connected with the postparacristule. The P4 is similar to that from Moschopotamos, but with an additional tubercle emerging from the mesial cingulum (like in types “B”, “J”, “L” and “M” of Hellmund 1991: text-fig. 3). The M2 and M3 are comparable in size to those of E. crispus from Detan Dvérce, Czech Republic, which is considered to be one of the earliest known and smaller specimens of this species (comparison based on data provided by Lihoreau et al. 2009: table 1). Lüttig and Thenius (1961) supposed the Chandras Elomeryx palate is of early Oligocene age, based on its similarities with west European samples of this genus. However, Kopp (1965) summarized local lithostratigraphic and mollusk biochronological evidence and showed that the coal deposits in which the specimen was found are most likely of upper Oligocene age (Fig. 7).

Fig. 7

Chrono-spatial distribution of Eurasian Elomeryx species (based on data by Ducrocq and Lihoreau 2006; Hellmund 1991; Lihoreau et al. 2009; Made 1996; Tsubamoto and Kohno 2011; Sach and Heizmann 2001; Scherler et al. 2011; this study)

The Elomeryx record from Turkish Thrace (Ozansoy 1962) seems to cover almost the entire Oligocene (Fig. 7). The lower coal levels (levels III–V in Lebküchner 1974), where E. cf. crispus is associated with Anthracotherium praealsaticum Ozansoy, 1963 and Anthracotherium cf. monspialense de Zigno, 1882, have been correlated with the early–middle Oligocene with the use of micromammals (Ünay-Bayraktar 1989). The upper coal levels (levels VI–X in Lebküchner 1974), where E. cf. crispus is associated with Anthracotherium magnum, are certainly younger and most likely correspond to an early late Oligocene age (MP25–MP27) (Ünay-Bayraktar 1989; Ruckert-Ulkumen 1992). Direct comparison with Elomeryx material described by Ozansoy (1962; see also Lebküchner 1974) was unfortunately limited. A right toothrow with M1–M3 from Tekirdrağ-Malkara in the MTA, illustrated by Ozansoy (1962: pl. II, figs 1, 2; photos kindly provided by E. Albayrak), shows a typical E. crispus morphology with a third marked crest on the lingual side of the metaconule and, different from the Chandras palate, a clear join between the lingual postprotocrista and the premetacristule.

The Sigircili-Edirne Elomeryx material discussed by Hellmund (1991) and the Tozakli mandible described by Islamoğlu et al. (2010) come most probably from coal seams of the Danişman Formation, Eastern Thrace of upper Oligocene age, probably younger than 26–27 Ma (Fig. 7) (Siyako and Huvaz 2007: figs 1, 5; Islamoğlu et al. 2010; Islamoğlu pers. comm. 2011). Some Sigircili-Edirne specimens in BSPM allocated by Hellmund (1991) to E. crispus, exhibit an advanced lower molar morphology with “Y crest pattern”, strong hypoconid–entoconid distal connection and strong premetacristid in contact with the preprotocristid (type “b” of Hellmund 1991: text-fig. 6). The Sigircili P4 lacks the weak disto-lingual crest seen in Moschopotamos and Chandras, the transverse valley of the M3 is blocked lingually (Hellmund 1991: Pl. 7, fig. 6) and its lingual postprotocrista is strong, differing from the Chandras palate. Reexamination of the Tozakli mandible (stored in MTA; photos and measurements kindly provided by E. Albayrak) ascribed to E. borbonicus by Islamoğlu et al. (2010) shows that the teeth fall dimensionally within the lower limits of the E. borbonicus size range. Although the “Y” pattern between prehypocristid–preentocristid is hardly observable due to the state of preservation, the molars lack a direct hypoconid–entoconid mesial connection, but they do show a clear premetacristid and a rather strong protoconid–metaconid mesial connection.

In the light of new evidence on the origin and expansion of Elomeryx (see discussions in Ducrocq and Lihoreau 2006; Tsubamoto and Kohno 2011), SE Europe seems to represent a key area for the understanding of the zoogeography and evolution of the genus. Previous fossil evidence indicates that Elomeryx originated in East Asia during the late middle Eocene, with the most primitive known representative “Bothriodon tientongensis Xu, 1977” coming from the Bose Basin of southern China (Ducrocq and Lihoreau 2006). Until the end of the Eocene, Elomeryx spread into western Europe, leading to E. crispus, and slightly later entered North America, leading to E. armatus (Ducrocq and Lihoreau 2006; Lihoreau and Ducrocq 2007). Unfortunately, the great chrono-spatial gap between the East Asian and the West European Eo-Oligocene Elomeryx record (Fig. 7; Ducrocq and Lihoreau 2006: fig. 4) prohibit more comprehensive phylogeographic interpretation (e.g., Lihoreau and Ducrocq 2007), whereas the position of the closely related but poorly known Balkan genus Bakalovia in the evolutionary history of Elomeryx remained uncertain.

Through comparisons of the lithological sequence of the Bakalovia type locality provided by Nikolov (1967) with the updated stratigraphy of the Burgas Palaeogene deposits (Juranov 1992; Kostova and Markova 2005), it appears that the fossils come from the lower coal level of the Ravnec Formation of most likely late middle Eocene age (Bartonian, MP16). This may imply that Bakalovia slightly predates the oldest records of E. crispus at La Débruge, France (MP18), and Detan Dvérce, Czech Republic (MP21) (de Bonis 1964; Fejfar and Kaiser 2005; Lihoreau et al. 2009) (Fig. 7). The absence of a distal entoconid–hypoconid connection (character 23 in ESM Appendix I), the three cristids issued from the hypoconid (character 26), the preentocristid position (character 27), the absence of diastemata between c-p1 and p1–p2 (characters 34 and 35, respectively) and the simple lower premolar morphology (character “b” in ESM Appendix I; see also Hellmund 1991: text-fig-2; but note that the author mistakenly illustrated the buccal instead of the lingual side of Bakalovia lower premolars) of Bakalovia should be regarded as primitive features within the Elomeryx lineage, a scheme additionally supported by the present parsimony analysis (Fig. 6) and the revised chronological data. Interestingly, “Bothriodon tientongensis Xu” from southern China seems more advanced than Bakalovia in the stronger distal entoconid–hypoconid connection on the m3 (compare Nikolov 1967: pl. 17, fig. 1 and pl. 18, figs 2, 3 with Ducrocq and Lihoreau 2006: fig. 1H), the better developed talonid of the p3 and p4 and the stronger p4 paraconid (compare Hellmund 1991: pl. 12 with Tsubamoto and Kohno 2011: fig. 5). Such evidence would therefore place the origin of Elomeryx more westward than previously thought, around the current Black Sea region, from where E. crispus spread into Europe and E. cf. crispus reached the Far East at early late Eocene times (Fig. 8). A branch of the Asian E. cf. crispus might have dispersed to N. America at the basal Oligocene, leading to E. armatus and its relatives, while another branch might have led to E. japonicus (Fig. 8).

Fig. 8

Two evolutionary scenarios (a and b) of Elomeryx species, correlated with their geographical distribution, here extrapolated on a simplified Oligocene palaeogeographic map. Scenario a in accordance with present parsimony analysis; scenario b in accordance with the traditional European E. crispus-to-E. borbonicus concept. 1 Bakalovia, Bulgaria; 2 E. crispus, CW Europe; 3 E. cluai, Spain; 4 E. borbonicus, CW Europe; 4a E. borbonicus minor CW Europe; 4′ E. cf. borbonicus from Pakistan, Kazakhstan, and Georgia (Ducrocq and Lihoreau 2006); 5 Elomeryx sp. from Moschopotamos Greece; 6 E. cf. crispus from Bose Basin, China; 7 E. japonicus, Japan; E. armatus, N. America

The Japanese species, originally attributed to Brachyodus, has been discussed by Ducrocq and Lihoreau (2006) and was more recently revised by Tsubamoto and Kohno (2011). E. japonicus is the oldest record marking the shift from the small to the large and more advanced dentally Eurasian Elomeryx. The single known mandibular fragment is indirectly aged by the fission track method at about 29 Ma (that is equivalent to Rupelian, MP25), that is just before the emergence of the European E. borbonicus, first occurred at St. Henri, France (late Oligocene, MP26) (Fig. 7). According to its revised diagnosis, E. japonicus differs from other Elomeryx species in having a wider p4 talonid with better developed distobuccal cingulum (Tsubamoto and Kohno 2011: 119). Apart from the latter character, E. japonicus is rather similar in size (Tsubamoto and Kohno 2011: table 1) and morphology to E. borbonicus (Tsubamoto and Kohno 2011: fig. 4; note the “Y” crest pattern on molars and strong p4 paraconid and talonid) and as the present phylogenetic analysis implies (Fig. 6), they could be closely related. The large late Oligocene Elomeryx sp. from Nabeshima Island, Japan (Ducrocq and Lihoreau 2006; Tsubamoto and Kohno 2011) may also provide evidence for such a relationship, but morphological data are limited. Ducrocq and Lihoreau (2006) further recognized E. borbonicus in several central Asian sites of Oligo-Miocene age, widening the geographic distribution of this species from CW Europe to Pakistan via Georgia and Kazakhstan (Fig. 7). Although E. borbonicus is usually regarded to be a direct descendant of the European E. crispus (e.g., Hellmund 1991: 77; Lihoreau et al. 2009), a scenario that cannot be easily excluded by the available data (Fig. 8b), the parsimony analysis provided here (Fig. 6) and that of Lihoreau and Ducrocq (2007: fig. 7.2) may well allow alternative interpretations (see also Tsubamoto and Kohno 2011: fig. 7). Thus, an Asiatic origin and westward dispersal of E. borbonicus related to the late Oligocene warming (Zachos et al. 2001) cannot, in our opinion, be excluded (Fig. 8a).

In western Europe, E. borbonicus replaced E. crispus around 27 Ma ago (Lihoreau et al. 2009; Scherler et al. 2011) and probably survived until the late early Miocene (Fig. 7), as it has been reported from the MN3–MN4 site of Becken in the Swiss Molasse Basin and from the MN4 site of Eggingen-Mittelhart 3 in S. Germany (Hellmund 1991; Lihoreau et al. 2009; Made 1996; Sach and Heizmann 2001; Scherler et al. 2011). There is also evidence that west European E. borbonicus decreased in overall size and reduced its premolar length during the early Miocene, leading to E. b. minor (Depéret, 1906) (e.g., Schaub 1948; Made 1996). At the same time, the Greek record and at least some Eastern Thrace samples indicate that another advanced Elomeryx species, dimensionally close to E. crispus or even to the smaller Iberian E. cluai, spanned the southern Balkan Oligo-Miocene boundary. The early Miocene Elomeryx sample under study shares with advanced Elomeryx species the compressed and distally serrated upper canines [character states 6(1) and a(1) in Fig. 6] and the “Y” crest pattern on the posterior lobe of the m3 [character state 27(1) in Fig. 6], but it still retains some primitive features, such as the bucco-lingually uncompressed m3 hypoconulid, the weak posthypocristid and the strong mesio-lingual entocristid [character states f(0), g(0) and h(0), respectively], the latter character seen also in E. japonicus but not in E. borbonicus.

Because of missing data, any hypothesis about the origin of the Balkan species and its taxonomy is rather fragile, and depends largely on the preferred scenario of origin of E. borbonicus. Following the traditional concept of E. crispus-to-E. borbonicus evolutionary scheme (Hellmund 1991; Lihoreau et al. 2009), the advanced Elomeryx from the southern Balkans, should represent a grade or subspecies within this lineage (Fig. 8b); Hellmund (1991) already recognized an “advanced E. crispus” in Europe that shares with E. borbonicus, E. japonicus and the Balkan species the derived “Y crest pattern” on the lower molars. Assuming, however, that E. borbonicus descended from an Asian E. japonicus-like stock, in accordance with the present cladistic analysis, the advanced Elomeryx from the southern Balkans would have a similar biogeographic history; it could have split off from this lineage possibly before the emergence of E. japonicus and dispersed westwards at late Oligocene times (Fig. 8a). Within the current systematic status of European E. crispus (sensu Hellmund 1991), both evolutionary schemes imply the parallelophyletic development of similar dental features in the Asian and European Elomeryx lineages.