Only the following four Rhinocerotidae specimens have been collected at Sahabi. •

A large-sized M2 was initially assigned to the American genus Teleoceras (D’Erasmo, 1954) and later referred to the giant rhinoceros Baluchitherium or Indricotherium (= Paraceratherium) (Hooijer, 1968 and Savage, 1971). The M2 collected by Petrocchi (1951) between January and February 1938 and published by D’Erasmo (1954) is currently missing. Our attempt to locate the specimen while revising the collection stored at the Rome Civic Zoology Museum, at the Museum of Paleontology of the University of Naples (where D’Erasmo was active, and where a small collection of marine vertebrates from Sahabi is still kept), and the Sapienza University in Rome did not allow us to recover the specimen. Thus, all our morphometric and morphological considerations are limited to the evaluation of D’Erasmo's figures.

The specimens were morphologically compared with the rhinocerotid material collected from selected Miocene African and European localities. The comparisons were based on direct observation of the material housed in several museums and institutions, as well as on data from literature. The dental terminology follows Antoine (2002); the morphometric methodology follows Guérin (1980). The suprageneric classification follows Antoine (2002).

Family Rhinocerotidae Gray, 1821

Referred material: 96P16B, a second phalanx of the middle digit.

Description and comparison: According to Heissig (1982), the size of the specimen is comparable to that of Ceratotherium neumayri. However, phalanges of Ceratotherium douariense (= Diceros douariensis, see Geraads, 2010) and other late Miocene rhinoceroses are unknown. An attribution to the family level seems to be appropriate due to the absence of detailed morphological and morphometric studies on Neogene rhinocerotid phalanges.

Subfamily Rhinocerotinae Gray, 1821

Aceratheriini Dollo, 1885 vel Teleoceratina Hay, 1902

Referred material: A partially preserved mandible currently lost.

Description and comparison: Petrocchi did not figure the remains and only reported a very short description, giving some details about the state of preservation. According to Petrocchi (1951), the horizontal ramus of the mandible preserved a large-sized incisor. Unfortunately, no information about the orientation of i2s (parallel or diverging rostrally) was reported by Petrocchi. During the latest Miocene, large-sized incisors are only documented in aceratheres (Aceratheriini; e.g., Chilotherium) and brachypotheres (Teleoceratina; e.g., Brachypotherium). Dicerotines (Rhinocerotina) lack well-developed and large lower incisors.

Tribe Rhinocerotini Gray, 1821

Subtribe Teleoceratina Hay, 1902

genus Brachypotherium Roger, 1904

Brachypotherium lewisi Hooijer & Patterson, 1972

(Fig. 2A)

Referred material. One second upper molar figured by D’Erasmo (1954, figs. 1–3), currently missing.

Description and comparison. The upper molar from Sahabi was originally described and compared by D’Erasmo (1954). The greatest width of 99 mm reported by this author led Savage, 1967 and Savage, 1971 and Hooijer (1968) to consider this tooth as belonging to a baluchithere (Indricotherium or Baluchitherium = Paraceratherium, family Hyracodontidae, subfamily Indricotheriinae). Indricotheriinae occurred from the middle Eocene to the late Oligocene in Eurasia and this group seemingly never reached the Afro-Arabian continent (Antoine et al., 2008 and Antoine et al., 2004 and references therein). The greatest width of Sahabi's tooth is most probably overstated by differences in the method of measuring; considering the size reported in the original figure, the greatest width of the tooth is approximately 91 mm, whereas the length of 78 mm seems to be correct (Table 1). As correctly depicted by D’Erasmo (1954), the tooth displays a large and robust crochet, a prominent antecrochet, and a buccal cingulum in the posterior side of the ectoloph, which is rather flat. Protoloph and metaloph are relatively parallel (Fig. 2A). The author also reported the presence of a tubercle at the entrance of the medisinus and a narrow and buccal-lingually elongated postfossette.

Contrary to Sahabi's tooth, the M2 of Paraceratherium lacks a crochet, and the buccal cingulum is strong and continuous (e.g., Antoine et al., 2004, fig. 3A; Fooster-Cooper, 1924: fig. 11). The European Brachypotherium brachypus (which is doubtfully distinct from B. goldfussi) shows a well-developed crochet and an antecrochet on M2 (Cerdeño, 1993) as found in the studied specimen, but the dimensions are considerably smaller compared with those of the Sahabi tooth (Table 1). Late Miocene Brachypotherium sp. from Bulgaria (Geraads and Spassov, 2009; pl. I, fig. F) has a weak and double crochet, a less prominent antecrochet than in the studied specimen and a backwards-directed metaloph. The dimensions of the M2 of B. perimense from Asia reported by Heissig (1972) are slightly smaller than those of the specimen from Sahabi (Table 1). The M2 figured by Heissig (1972: Pl. 10, fig. 9) displays a longer crochet, a shorter metaloph, and a backward directed protoloph and metaloph with respect to Sahabi's tooth. The latter features are present on the worn-out M2 from the Siwaliks of Burma (Lydekker, 1884: pl. 1, fig. 5); this tooth also displays lingual pillars at the entrance of the medisinus, similarly to the specimens from Punjab (Lydekker, 1881: pl. 2, fig. 1; pl. 3, fig. 3) and Perim Island (Lydekker, 1881: pl. 2A). The worn-out M2 of B. perimense figured by Colbert (1935: fig. 89) differs from the studied specimen by having a constricted metaloph, a larger postfossette and by being smaller (L = 69 mm; B = 78 mm; Colbert, 1935: p. 198); nevertheless, the specimen AM19470 from Pakistan (Colbert, 1935: fig. 88) has larger size (L = 87 mm; B = 94 mm).

Compared with the M2 from Sahabi, the middle Miocene B. minor (Geraads and Miller, 2013) has smaller crochet and antecrochet, a slightly concave posterior part of the ectoloph and smaller dimensions (Table 1). The M2 of ‘Aceratherium’ campbelli from the early Miocene of Jebel Zelten (Libya) has a smaller size, smaller crochet and antecrochet with respect to the Sahabi tooth and a slightly concave posterior part of the ectoloph. A weak and reduced labial cingulum occurs on the posterior side of the ectoloph on M2 of B. snowi (Hamilton, 1973). In Brachypotherium heinzelini from Sinda (Zaire), the M2 displays a rather flat ectoloph, the crochet and the antecrochet are well developed and the lingual cingulum is reduced to a tubercle (Hooijer, 1963; pl. VIII, figs. 4, 6). A labial cingulum occurs on the posterior half of the ectoloph on the molars of B. lewisi from Lothagam (Kenya) (Hooijer and Patterson, 1972: p. 5); the M2 of this species displays crochet and antecrochet, a lingual cingulum reduced to a tubercle at the entrance of the medisinus, and a flattened ectoloph profile behind the paracone style. The anterior width of M2 of B. lewisi from Lothagam (KNM LT 100) reaches ca. 90 mm (Hooijer and Patterson, 1972: p. 13); Harris and Leakey (2003) also reported an anterior width of more than 90 mm (Table 1).

Subtribe Rhinocerotina

Genus ‘Diceros’ sp.

(Fig. 2B)

Referred material: 445P34A, a second upper premolar figured by Bernor et al. (1987, fig. 15).

Description and comparison: the tooth is very worn; the anterior width is less than the posterior one, resulting in a trapezoidal shape of the occlusal surface of the tooth (Fig. 2B). The buccal profile of the tooth displays a concavity in the middle; the postfossette is circular, the medisinus is buccal-lingually elongated and does not reach the lingual side of the tooth at this stage of wear (Fig. 2B). A lingual cingulum is expressed by an enamel fold on the lingual side of the tooth; a prefossette is evident on the anterior side of the protoloph. Protocone and hypocone are similar in size and fused at this stage of wear; every constriction on the metaloph and protoloph is absent.

Worn P2s of Ceratotherium simum have a trapezoidal shape (e.g., Fig. 3A), but the lingual length is shorter than the buccal one; the medifossette is usually evident also at an advanced stage of wear, the medisinus is generally elongated anterior-posteriorly, protocone and hypocone appear completely fused, and the lingual cingulum is absent. Worn-out P2s of Diceros bicornis have a rectangular shape (e.g., Fig. 3B–E), with the anterior width slightly smaller than the posterior one. Protocone and hypocone appear separated also in an advanced stage of wear; the postfossette is subcircular and the medisinus is buccal-lingually elongated. A small prefossette is sometimes present as well as the lingual cingulum (Fig. 3B–E). Worn-out P2s of Ceratotherium efficax (= Ceratotherium mauritanicum according to Geraads, 2005) differ from the Sahabi specimens by having an anterior-posteriorly elongated medisinus, a lingual-distally directed protoloph, and protocone and hypocone completely fused (selected tooth reported in Fig. 3F). The P2s of Ceratotherium douariense from Douaria (Guérin, 1966; figs. 6, 7) are less worn than that from Sahabi; they have a rectangular shape, a subcircular postfossette, a buccal-lingually elongated medisinus and a lingual cingulum as found in the studied specimen. Contrary to the studied P2, however, protocone and hypocone are separated on the Douaria's specimen, apparently, until their bases; the anterior width is larger than the posterior one, the protoloph is narrower and the metaloph is constricted in respect to Sahabi's tooth. The latter is slightly larger than the P2s from Douaria (Table 1). The P2 of C. douariense from the Middle Awash (Giaourtsakis et al., 2009; fig. 14.2 A) is less worn than that from Sahabi and displays a circular postfossette, a prefossette, a lingual cingulum, a buccal-lingually elongated medisinus. However, the Middle Awash tooth differs from that from Sahabi by having a crochet, a constricted metaloph, a protocone and a hypocone well separated until their bases and a narrower protoloph. The P2 from the Middle Awash is larger than the type material from Douaria and slightly wider than that from Sahabi (Table 1). A relatively worn P2 of C. neumayri from Mytilinii (Giaourtsakis et al., 2009; pl. IV, fig. 1) displays protocone and hypocone completely fused, similar to C. simum, lingual length shorter than buccal length, and a lingual-distally directed protoloph. Protocone and hypocone are separated until the dorsal surface of the lingual cingulum on P2 of C. neumayri from Karacaşar (Turkey), which also displays a constricted metaloph, a narrow protoloph, and protocone smaller than hypocone (Antoine et al., 2012; fig. 3.b). The latter two characters are also evident of P2 from Akkaşdaği (Turkey) as well as a lingual length smaller than the buccal one (Antoine and Saraç, 2005; fig. 2A); all of them can be recognized on a worn-out P2 from Mytilinii.

Late Miocene Rhinocerotidae remains are relatively scarce and poorly represented in North Africa and, excluding Sahabi, they have been collected in other five localities: •

Douaria (Tunisia, considered latest Miocene in age by Geraads, 2010, whereas Guérin, 2000 suggested an age around 9.5 Ma), which yielded a partial adult skull with associated mandible and a partial juvenile skull assigned to C. douariense;

Ceratotherium douariense is doubtfully distinct from C. neumayri according to Geraads (2010) and the paratype of the Douaria's rhinoceros, a partial juvenile skull, morphologically resembles a non-dicerotine taxon, probably a teleoceratine (Geraads and Miller, 2013). The systematic affinities of the partial juvenile skull from Bou Hanifia, at first referred to the genus Dicerorhinus, are still doubtful (Geraads, 1986 and Geraads, 2010). Within this framework, any contribution on late Miocene Northern African Rhinocerotidae would be helpful to better understand the biogeographic patterns and the affinities between the faunal assemblages.

Unfortunately, most of Petrocchi's Sahabi collection is missing. This is, for instance, the case of rhinoceros material collected during the 1930s. A mandible with a large-sized lower incisor would belong to either a brachypothere or an acerathere although the latter group is recorded only doubtfully in northern Africa (one uncertain record from Tizi N’Tadderht). The M2 from Sahabi described by D’Erasmo (1954) concurs well, morphologically and dimensionally, with a referral to B. lewisi from Lothagam. The tooth also resembles in morphology the M2 of B. heinzelini published by Hooijer (1963) as Aceratherium cf. tetradactylum and was later included within the new species of the brachypothere he established (Hooijer, 1968). Geraads (2010) and Geraads and Miller (2013) recently recommended restricting the nomen B. heinzelini only to the material collected from the type locality of Sinda 15 (Democratic Republic of the Congo) and reported the probable synonymy of B. heinzelini to B. lewisi. The morphological similarities between the M2s of the two species and the dimensions of B. heinzelini, which falls within the range of B. lewisi, would support this hypothesis, even if the material from Sinda is too scarce for any conclusion.

The P2 from Sahabi resembles Diceros, sharing some characters with Diceros bicornis, which never occurred north of the Sahara area. The tooth from Sahabi differs from worn-out P2s of C. simum, C. efficax, and C. neumayri. The presence of the eastern Mediterranean rhinoceros C. neumayri in northern Africa is therefore not confirmed. Considering the ambiguous generic status of several African dicerotines (alternatively assigned to Ceratotherium or Diceros), a provisional attribution to “Diceros” sp. is here proposed for the Sahabi dicerotine, pending the discovery of more preserved material.

A Rhinocerotidae association similar to that reported at Sahabi, with Brachypotherium and a dicerotine, is at present recorded only in Kenya, at Lothagam, upper and lower Nawata, and at Mpesida (Geraads, 2010, Harris and Leakey, 2003 and Hooijer, 1973), suggesting a strict East African affinity of the rhinoceros assemblage from Sahabi.