Among the fossil members of the giant panda subfamily of ursid carnivorans, Ailuropodinae, one group of species is of giant size, those of Indarctos. Indarctos species have some dental resemblances to and may be closely related to Agriotherium, although there are other clear differences between these taxa, and no known species has definitive shared derived traits that could link these two genera. Here we describe a rich suite of fossil material from both North America and eastern Asia, all belonging to a new genus, Huracan, which possesses characters shared with both Agriotherium and Indarctos but also has diagnostic autapomorphies. The new taxon was distributed widely in the Holarctic during the latest Miocene, including at least four species: the type species Huracan schneideri (previously Agriotherium schneideri) from the latest Hemphillian (Hh4) and possibly early Blancan North American Land Mammal “Ages” (NALMAs), North America; H. coffeyi from the early Late Hemphillian (Hh3) NALMA, North America; H. qiui, sp. nov., from the Baodean Asian Land Mammal “Age” (ALMA), northern China; and H. roblesi from the MN13 zone (latest Miocene–earliest Pliocene) of Spain. Huracan is the nearest sister taxon to Agriotherium, the latter herein considered to be an ailuropodine (in the tribe Agriotheriini) rather than a hemicyonid, and the common ancestor of both genera evolved from Indarctos (with resultant paraphyly of that taxon) or another Indarctos-like ailuropodine bear, likely in eastern Asia. The dentitions of Huracan and Agriotherium both are more specialised for carnivory than most Indarctos species, indicating a radiation of diverse ecological carnivores earlier in the history of the later-diverging, highly specialized herbivores in the giant panda lineage. Their postcranial morphology suggests that species in both genera (Huracan and Agriotherium) were more cursorial than species assigned to Indarctos, and thus well adapted to more open habitats. These derived traits may explain the worldwide replacement of Indarctos species by Huracan and Agriotherium species during the latest Miocene, in response to significant global cooling and expansion of C4 grasslands that occurred at that time.
INTRODUCTION
The giant panda Ailuropoda is a now monospecific extant genus, feeding nearly exclusively on bamboo (Gao, 1987; Li et al., 2006). The subfamily Ailuropodinae, however, is generally believed to contain at least one other taxon of giant size, the omnivorous-carnivorous Indarctos (Abella et al., 2012; Qiu et al., 2014). Traditionally considered a close relative of Indarctos, more recently Agriotherium has been considered by some to be the latest representative of Hemicyoninae rather than an ailuropodine bear (Morales et al., 2005; Abella et al., 2014). Traditionally, Indarctos and Agriotherium are viewed as clearly distinct genera, with their differences summarized by several authors (Frick, 1926; Qiu and Schmidt-Kittler, 1983; Petter and Thomas, 1986). However, as noticed by Hunt (1998), the sole North American species of Agriotherium known at that time has some characters more frequently observed in species of Indarctos than in Old World species of Agriotherium, e.g., three cusps in the metaconid-entoconid complex of the lower carnassial. This observation was further supported by Jiangzuo and Flynn (2019) when they described a new species of North American Agriotherium, A. hendeyi. They found that the lower dentition of North American “Agriotherium” material also differs markedly from typical Agriotherium, and is closer to Indarctos in many aspects, especially in m1 features. The distinctive nature of the North American “Agriotherium” species, with a mix of characters traditionally viewed as characteristic of two distinct genera, provides an opportunity to test potential interrelationships of Indarctos and Agriotherium and their potential alternative placements as ailuropodines versus hemicyonines, and to clarify the past taxonomic diversity and range of locomotor and dietary adaptations of the giant panda lineage Ailuropodinae. Furthermore, long-term expeditionary research in the Linxia Basin of Gansu Province in northern China over the past 40 years has yielded an extremely abundant Late Miocene fauna, including many skulls of bears (Qiu et al., 2014). One of them, from the Wangjiashan locality, Baodean ALMA, has an upper dentition that is similar to those of the North American species previously considered to represent Agriotherium (A. schneideri) but herein recognized as the genotype species of the new genus Huracan. The new material of North American Huracan and the material from Asia (Gansu) each are named and described in this study.
FIG. 1.
Map of localities with specimens assignable to species of Huracan. Huracan coffeyi, Hh3, stars: 1. Coso Formation; 2. Hay Ranch Formation; 3. Big Sandy Formation; 4. Quiburis Formation, 5. Ogallala Formation; 6. Snake Creek Formation. Huracan schneideri, Hh4, squares: 7. Eden Formation; 8. Bone Valley Formation; 9. Yepomera; 10. Rinconada; 11. Tehuichila. Huracan cf. H. schneideri, Blancan NALMA, diamonds: 12. Glenns Ferry Formation; 13. Ringold Formation. Huracan qiui, Baodean ALMA, triangle: 14. Wangjiashan. Huracan roblesi, MN13, white circle: 15. Venta del Moro. ?“Huracan” punjabiensis, Dhok Pathan, black circle: 16. Hasnot.
GEOLOGICAL BACKGROUND
The type specimen of “Agriotherium” schneideri is from the Palmetto Fauna (Upper Bony Valley), Hardee and Polk counties, Florida, a typical Latest Hemphillian (Hh4) NALMA fauna (Tedford et al., 2004). The holotype mandible was the main material known from this locality (Merriam et al., 1916), but several new specimens were collected by the Florida Museum of Natural History (University of Florida) and are described here. The majority of the material described in this study comes from the Quiburis Formation (Pima County), the Big Sandy Formation (Mohave County) of Arizona, and the Optima Local Fauna, Ogallala Group near Guymon (Texas County) of Oklahoma, all of early Late Hemphillian age (Hh3; see fig. 1). The material from these localities is very abundant but most specimens are heavily crushed. The Asian material from Wangjiashan, Linxia Basin, Gansu Province, north China, belongs to the Qingbushan Fauna, from the upper part of the Late Miocene deposits, Baodean ALMA (Deng, 2004; Deng et al., 2013).
MATERIALS AND METHODS
The specimens described in this study mainly come from the Frick Collection of Fossil Mammals in the American Museum of Natural History (AMNH), New York, and several others from the Florida Museum of Natural History, University of Florida, Gainesville, and the Hezheng Paleozoological Museum, Hezheng, Gansu, China. Fossil material for comparison comes from the AMNH; Smithsonian National Museum of Natural History, Washington DC; Natural History Museum of Los Angeles County, Los Angeles (including La Brea fossils of the Hancock Collection); University of California Museum of Paleontology, Berkeley, CA; Institute of Vertebrate Paleontology and Paleoanthropology, Chinese Academy of Science, Beijing; and British Museum of Natural History, London. The extant mammal collections studied for comparison come from the AMNH; Smithsonian National Museum of Natural History, Washington DC; and Museum of Comparative Zoology, Harvard University, Cambridge, MA.
FIG. 2.
Skull measurements of bears, modified from Baryshnikov (2007): Cranium: L1, total length; L2, condylobasal length; L3, basal length; L4, neurocranium length; L5, viscerocranium length; L6, rostrum length; L7, palatal length; L8, toothrow length, C1-M2; L9, toothrow length, P4-M2; L10, zygomatic width; L11, neurocranium (braincase) width; L12, postorbital constriction width; L13, interorbital width; L14, condylar width; L15, mastoid width; L16, minimal palatal width; L17, greatest palatal width; L18, rostrum width (at canines); L19, greatest diameter of orbit; L20, cranial height; L21, toothrow diastema length, C-P4. Mandible: L1 – total length; L2- total length2 (to the posterior border of the coronoid process); L3, toothrow length, c1-m3; L4, toothrow length, p4-m3; L5, toothrow diastema length, c-p4; L6, symphysis length; L7, 8, mandible height and width behind the canine; L9, 10, mandible height and width at p2 (or middle of canine-p4 diastema); L11, 12, mandible height and width behind m1; L13, 14, mandible height and width behind m3; L15, masseteric fossa length; L16, coronoid process height; L17, mandibular condyle width.
Skull measurements follow Baryshnikov (2007), with modification (see fig. 2). Dental measurements follow Jiangzuo et al. (2018), adding one measurement (talonid length [of the m1 and m2]; see fig. 3). Anatomical terminology follows Davis (1964), Qiu et al. (2014) and Jiangzuo et al. (2019) with slight modification (see fig. 4 for dental terminology).
The phylogeny was performed using the maximum parsimony method, in software TNT1.6 (Goloboff et al., 2008; Goloboff and Catalano, 2016), implementing both traditional settings and implied weighting (Goloboff et al., 2018) approaches. Parts of figure plots were made in the software package ggplot2 (Wickham, 2016) in R (R Development Core Team, 2016).
INSTITUTIONAL ABBREVIATIONS
AMNH FM Fossil Mammal collection, Division of Paleontology, American Museum of Natural History, New York
AMNH F:AM Frick Collection (Fossil Mammals), Division of Paleontology, American Museum of Natural History, New York
AMNH M Mammalogy collection, American Museum of Natural History, New York
BMNH Fossil Mammal collection, Natural History Museum (British Museum of Natural History), London
HMV Hezheng Paleozoological Museum, Hezheng, China
IGM Instituto de Geología de México, Mexico City, Mexico
IVPP Institute of Vertebrate Paleontology and Paleoanthropology, Chinese Academy of Sciences, Beijing
LACM Natural History Museum of Los Angeles County, Los Angeles, California
LACMHC Natural History Museum of Los Angeles County, Hancock Collection, Los Angeles
MCZ Museum of Comparative Zoology, Harvard University, Cambridge, MA
PQ-L Quaternary Palaeontology (Langebaanweg), Iziko South African Museum, Cape Town
UCMP University of California Museum of Paleontology, Berkeley, CA
UF Florida Museum of Natural History, University of Florida, Gainesville, FL
UF/FGS The Florida Greek Standards Accreditation Program of Florida Museum of Natural History, University of Florida, Gainesville, FL
UNSM University of Nebraska State Museum, Lincoln, NE
USNM Smithsonian National Museum of Natural History, Washington, DC
OTHER ABBREVIATIONS
ALMA Asian Land Mammal “Age”
C/c Upper/lower canine
DH distal height (of limb bone)
DW distal width (of limb bone)
H height
Hh2 Hemphillian stage 2, late Early Hemphillian North American Land Mammal ‘Age’ (NALMA)
Hh3 Hemphillian stage 3, early Late Hemphillian NALMA
Hh4 Hemphillian stage 4, latest Hemphillian NALMA
I/i upper/lower incisor
ip index of plumpness
L length
M/m upper/lower molar
MN European “Mammal Neogene” zone
NALMA North American Land Mammal “Age”
P/p upper/lower premolar
PW proximal width (of limb bone)
SYSTEMATICS
Order Carnivora Bowdich, 1821 (phylogenetic definition in Flynn et al., 2020)
Family Ursidae Batsch, 1788
Subfamily Ailuropodinae Grevé, 1892
Tribe Agriotheriini Kretzoi, 1929
Huracan, gen. nov.
Etymology: Name after Hunraqan or Hurakan (Huracan in Carib [Karìna auran] or English, Huracán in Spanish), the Maya and Taíno god of wind and storm; in reference to the cursorial running adaptation and formidable power of the animal. Grammatically, this is the masculine gender in Spanish.
Type species: Agriotherium schneideri Sellards, 1916.
Included species: Huracan schneideri (Sellards, 1916) and Huracan coffeyi (Dalquest, 1986) from North America; Huracan qiui, sp. nov., and Huracan sp. (Yushe Basin) from East Asia; Huracan roblesi from Europe; and questionably ?“Huracan” punjabiensis (Lydekker, 1884) from South Asia.
Diagnosis: Large to huge ailuropodine bear, moderately hypercarnivorous in dental features, and exhibiting substantial sexual dimorphism. Cranial condylobasal length 380–515 mm. Mandible often with premasseteric fossa. I1/I2 with large posterocentral accessory cusp and reduced posterolateral cusp; I3 and canines large and robust; anterior premolars (P1–3/p1–3) present, occasionally lost in numbers in Huracan schneideri, H. roblesi, and Huracan coffeyi; P4 with large parastyle and variably (intraspecifically) subdivided hypocone; M1 slightly trapezoidal in shape; M2 equals M1 in length, each with slightly developed talon; p4 with distinct anterior and posterior cingulid cusps, but lacks medial ridge; and m1 with elongate paraconid, and distinct metaconid and mesoconid.
Differential diagnosis: Huracan differs from Indarctos, a paraphyletic assemblage of more basal taxa representing the nearest outgroups to a clade of the new genus plus Agriotherium, by having generally larger size (except I. oregonensis of similar size), upper I1/I2 with large posterocentral accessory cusp, P4 with larger parastyle, M2 with distinctly shorter talon, p4 with stronger anterior and posterior cingulid cuspids, m1 with more elongated paraconid, and presence of mandibular premasseteric fossa; Huracan differs from Agriotherium, its sister group, by generally larger size, and with upper I1/I2 with large posterocentral accessory cusp, P4 with more frequently subdivided hypocone, M1 more elongated, M2 with generally longer talon, p4 often without distinct medial ridge, and m1 with presence of large metaconid and two smaller entoconids.
Huracan schneideri (Sellards, 1916)
Agriotherium schneideri: Sellards, 1916: 98.
Hyaenarctos gregoryi: Frick, 1921.
Hyaenarctos gregori: Frick, 1926: 74.
Agriotherium schneideri: Stock, 1950: 2.
Agriotherium schneider:- Miller et al., 1996: 570.
Agriotherium schneideri (in part): Hunt, 1998: 184.
Type specimen: USNM 8838, mandible from Upper Bone Valley, Florida (fig. 5).
Hypodigm: In addition to the holotype, the following specimens from Upper Bone Valley are also referred to this taxon—UF 18528 (isolated I1 or I2), UF 206887, UF/FGS 3892, UF 21227 (three M1s), UF 203158, UF 133944 (two M2s), UF 47482 (p4), UF 17485, UF 21226 (two m2s).
Emended diagnosis: Large ailuropodine bear with distinct hypercarnivorous features. Mandible with premasseteric fossa always present; anterior premolars single rooted; P4 longer than M1, with variably subdivided hypocone; M2 anteroposteriorly short; p4 with weak anterior and posterior cingulid cusps and medial ridge; m1 with elongated paraconid, distinct metaconid, moderate mesoconid, and weak posterior entoconid.
Differential diagnosis: Differs from Huracan coffeyi in H. schneideri by having smaller size, wider M2 with proportionally shorter talon, p4 with weaker anterior accessory cusp and stronger medial ridge, m1 with weaker mesoconid and smaller posterior entoconid, m2 with weaker and more posteriorly located metaconid and poorly subdivided entoconid; differs from Huracan qiui in H. schneideri by having distinctly shorter M1 and M2.
Occurrence: Known from the latest Hemphillian NALMA (Hh4), from the Upper Bone Valley in Polk and Hillsborough Counties of Florida, the site of the holotype of “A.” schneideri (now the holotype specimen of the genotype species, Huracan schneideri). At that time, species of Huracan occurred in North America in the southern part of the United States, and in Yepomera in Chihuahua and Rinconada in Guanajuato, both in Mexico (Stock, 1950; Miller and Carranza-Castaóeda, 1996). A single lower m1 was found in Tehuichila in Hidalgo, Mexico, but it is of uncertain age (Frick, 1926). Hyaenarctos gregoryi was erected based on several isolated teeth from the Mount Eden local fauna, California (Frick, 1921), of Hh4 age (Albright, 2000). These teeth are either broken or heavily worn, so their taxonomic assignment is unclear, but judging from the wide M1, these teeth could represent H. schneideri. A mandible that was recently found from Sedgwick County, Kansas also belongs to this species (Everhart and Hawkins, 2020).
The absence of records in the more northern part of North America is probably biased by the paucity of fossil localities of similar age. Blancan records, consisting of a single M2 from the early Blancan Ringold Formation in Washington (Martin, 2013) and a broken mandible from the Middle Blancan Glenns Ferry Formation that produces the Hagerman local fauna in Idaho (Samuels et al., 2009), are insufficient to definitively diagnose the specimens to species, and thus are here regarded as Huracan cf. H. schneideri.
Description: New material of H. schneideri from the type locality of Upper Bone Valley, Florida is mainly represented by several isolated teeth. Nevertheless, these teeth show stable differences relative to those of Huracan coffeyi from Hh3.
Upper dentition: A single upper incisor (UF 18528) represents I1 or I2. The medial ridge is subdivided. A lateral ridge of the main cusp is slightly subdivided.
Three isolated M1s are added to the hypodigm for this species. UF 206887 (fig. 6: E5) is square in shape. M1 width across the metacone is significantly wider than that across the paracone. The most striking character of M1 is its strongly concave anterior border. The parastyle and metastyle are probably absent (but cannot be determined with certainty due to wear). The protocone is subdivided. UF/FGS 3892 and UF 21227 (fig. 6: E3) are similar in morphology and degree of wear. Both of these teeth are more heavily worn than is UF 206887. The width across the metacone is only slightly wider than that across the paracone. The concavities in the anterior and buccal borders of UF/FGS 3892 and UF 21227 are weaker than in UF 206887.
Two M2s are preserved. UF 203158 (fig. 6: E4) is large and wide. The talon is nearly absent, the buccal cingulum is weak, and there is a small transverse ridge between the paracone and protocone. UF 133944 (fig. 6: E2) represents a much smaller M2 in a maxillary fragment. The talon is stronger in this tooth than in UF 203158, its lingual border is rounded, and the metaconule is not distinguishable from the protocone.
Lower dentition: The p4 UF 47482 (fig. 7: A2) has a triangular occlusal shape. A buccal concavity is present. The anterior accessory cusp is present but weak, whereas the posterior accessory cusp is well developed. A distinct medial ridge is present on the lingual side of the main cusp. The m1 is robust. A cingulid is only present on the buccal and lingual sides of the paraconid of m1. A strong buccal concavity is present between the trigonid and talonid, whereas only weak lingual concavities are present between the paraconid and protoconid and between the protoconid and metaconid. The mesoconid is present, but not well distinct. At the posterolingual side of the tooth, there are three cusps, representing the metaconid and a subdivided entoconid. The posterior entoconid is distinctly smaller than the anterior one. Two m2s are known. UF 17485 (m2 cast) also is robust. Its metaconid is located slightly posterior to the protoconid, and on its talonid the entoconid is small and not subdivided. UF 21226 (fig. 7: A4) is unusual in having a more posteriorly located and much weaker metaconid, and its entoconid also is very small and not subdivided.
FIG. 5.
Holotype specimen of Huracan schneideri, USNM 8838, in A. lateral; B. medial; and C. dorsal views.
Huracan coffeyi (Dalquest, 1986)
Agriotherium schneideri: Savage, 1941: 696.
Agriotherium sp.: Harrison, 1983: 22.
Agriotherium schneideri (in part): Hunt, 1998: 184.
Agriotherium” schneideri: Jiangzuo et al., 2021.
Type specimen: MWSU 12147, a partial mandible preserving c-m3, from Coffee Ranch.
Referred specimens: In addition to the holotype, and specimens listed in Frick (1926), Harrison (1983), and Dalquest (1986), the following specimens also now are referred to this species. Cranium: Guymon Formation: AMNH F:AM 30399 (with associated mandible, shortly mentioned by Frick, 1937), AMNH F:AM 50030; Quiburis Formation: AMNH F:AM 49372, AMNH F:AM 76001, AMNH F:AM 76005, AMNH F:AM 76006, AMNH F:AM 76007. Mandible: Coffee Ranch: AMNH F:AM 23364, AMNH F:AM 23368; Guymon Formation: AMNH F:AM 50032, AMNH F:AM 50033; Quiburis Formation: AMNH F:AM 76005, AMNH F:AM 76008–10, AMNH F:AM 76012, AMNH F:AM 76012a, AMNH F:AM 76020, AMNH F:AM 76031, AMNH F:AM 144551. Isolated teeth or jaw fragments: Coffee Ranch: AMNH F:AM 23365, AMNH F:AM 23367, AMNH F:AM 23369a, AMNH F:AM 23369b, AMNH F:AM 23369c, AMNH F:AM 23370, AMNH F:AM 23371, UCMP 30517, UCMP V24025, UCMP V24027, UCMP V24026; Guymon Formation: AMNH F:AM 50034, AMNH F:AM 50034a,, AMNH F:AM 50035–50037, AMNH F:AM 50037a, AMNH F:AM 50038, AMNH F:AM 50039; Quiburis Formation: AMNH F:AM 145923, AMNH F:AM 76017, AMNH F:AM 76005, AMNH F:AM 76006, AMNH F:AM 76007, AMNH F:AM 49372, AMNH F:AM 76035, AMNH F:AM 76013, AMNH F:AM 76032, AMNH F:AM 76019, AMNH F:AM 76021, AMNH F:AM 76020, AMNH F:AM 76033, AMNH F:AM 76024, AMNH F:AM 76034, AMNH F:AM 76023, AMNH F:AM 76011, AMNH F:AM 76016, AMNH F:AM 76020, AMNH F:AM 76018C, AMNH F:AM 76018D, AMNH F:AM 76015; AMNH F:AM 146568; AMNH F:AM 146569; AMNH F:AM 146468–502; AMNH F:AM 146570; AMNH F:AM 146571; Big Sandy Formation: AMNH F:AM 49380, AMNH F:AM 49382, AMNH F:AM 49376, AMNH F:AM 49375, AMNH F:AM 49379, AMNH F:AM 49378. Humerus: Coffee Ranch: UCMP31818, Quiburis Formation: AMNH F:AM 76106, AMNH F:AM 76070, AMNH F:AM 76071, AMNH F:AM 76072, AMNH F:AM 76073, AMNH F:AM 146513. Radius: Ash Hollow Formation: UNSM 76013; Quiburis Formation: AMNH F:AM 146513, AMNH F:AM 76078, AMNH F:AM 76077, AMNH F:AM 76008, AMNH F:AM 146512, AMNH F:AM 146456; Guymon Formation: AMNH F:AM 50042. Ulna: Quiburis Formation: AMNH F:AM 76082, AMNH F:AM 76081, AMNH F:AM 76080, AMNH F:AM 76079, AMNH F:AM 76008, AMNH F:AM 146511, AMNH F:AM 146457; Guymon Formation: AMNH F:AM 50042. Carpal bones: Guymon Formation: AMNH F:AM 146450 (scapholunar); Coffee Ranch: AMNH F:AM 146452 (cuneiform), AMNH F:AM 112588 (pisiform); Quiburis Formation: AMNH Red (Redington)117-1831 (pisiform). Femur: Quiburis Formation: AMNH F:AM 76090, AMNH F:AM 146513, AMNH F:AM 146503, AMNH F:AM 76105, AMNH F:AM 76085, AMNH F:AM 76086. Tibia and Fibula: Quiburis Formation: AMNH F:AM 146513, AMNH F:AM 146514, AMNH F:AM 76092, AMNH F:AM 76090, AMNH F:AM 76087, AMNH F:AM 76097, AMNH F:AM 76096, AMNH F:AM 76095, AMNH F:AM 76008, AMNH F:AM 76014, AMNH F:AM 76014. Astragalus: Quiburis Formation: AMNH F:AM 76103, AMNH F:AM 146508, AMNH F:AM 76102, AMNH F:AM 146509, AMNH F:AM 76014, AMNH F:AM 76014; Coffee Ranch: AMNH F:AM 146454. Calcaneum: Quiburis Formation: AMNH F:AM 146510, AMNH F:AM 76097, AMNH F:AM 76008, AMNH F:AM 76014; Big Sandy Formation: AMNH F:AM 146467; Coffee Ranch: UCMP31837. Mesocuneiform: Coffee Ranch: AMNH FM 146707. Cuboid: Quiburis Formation: AMNH F:AM 76014, AMNH F:AM 146459; Coffee Ranch: UCMP30519. Metapodial: MC1, Big Sandy Formation, AMNH F:AM 146563, AMNH F:AM 146564, AMNH F:AM 146458. MC2, Quiburis Formation, AMNH F:AM76084. MC4, Quiburis Formation, AMNH F:AM 76101, AMNH F:AM 76008, AMNH F:AM 146505, AMNH F:AM 146564. MC5, Quiburis Formation, AMNH F:AM 76100, AMNH F:AM 69325A, AMNH F:AM 69327. MT1, Quiburis Formation, AMNH F:AM 76099, AMNH F:AM 76014, Big Sandy Formation, AMNH F:AM 146564. MT2, Quiburis Formation, AMNH F:AM 146513, AMNH F:AM 76014, AMNH F:AM 146460. MT3, Quiburis Formation, AMNH F:AM 146513, AMNH F:AM 76014. MT4, Quiburis Formation, AMNH F:AM 146513, AMNH F:AM 76014, Big Sandy Formation, AMNH F:AM 146466. MT5, Quiburis Formation, AMNH F:AM 76029, AMNH F:AM 76030, AMNH F:AM 76087, AMNH F:AM 76014.
Emended diagnosiS: Huge ailuropodine bear with hypercarnivorous craniodental features. Cranial condylobasal length 380–515 mm. Mandible often with premasseteric fossa; anterior premolars single-rooted (P3/p3 occasionally double rooted); P4 longer than M1, variably with subdivided hypocone; p4 with distinct anterior and posterior cingulid cusps but no medial ridge; m1 with elongated paraconid and distinct metaconid, mesoconid and distal entoconid; humerus lacking entepicondylar foramen.
Differential diagnosis: Differs from H. schneideri in having slightly larger dental size; narrower M2 with proportionally longer talon; p4 with less distinct anterior accessory cusp and weaker medial ridge; m1 with stronger mesoconid and larger posterior entoconid; m2 with stronger metaconid and better subdivided entoconid. Differs from Huracan qiui in having more reduced P4 anterior hypocone, shorter M1 and M2.
Occurrence: Known from the late Hemphillian NALMA (Hh3). In the Great Plains of the U.S., this species has been found at USGS Locality Ml430 in the Coso Formation in Inyo County, California; the Carlin High Quarry in the Hay Ranch Formation in Elko County, Nevada; the Clay Bank and Bird Bone Quarries in the Big Sandy Formation in Mohave County, Arizona; the Old Cabin, Redington, and Camel Canyon quarries in the Quiburis Formation in Pima and Pinal Counties, Arizona; and at Coffee Ranch, Guymon-Optima quarries, Axtel Quarry, Edson Quarry, Lost Quarry and the nearby UNSM Locality SM-101 in the Ogallala Formation in Texas and Oklahoma (Savage, 1941; Harrison, 1983; Dalquest, 1986; Hunt, 1998; Tedford et al., 2004). An isolated M2 from the Johnson Member of the Snake Creek Formation in Sioux County, Nebraska is also from Hh3 (R. Tedford, in note associated with this specimen in the AMNH collections).
Description: This species was previously known only by isolated teeth and several mandibles. Additional study of the extensive AMNH Frick Collection of Fossil Mammals brings many new specimens of this species to light, including seven crania, 14 mandibles, and numerous isolated teeth and postcranial bones. Unfortunately, nearly all the crania were found in Arizona and Oklahoma from sites in which most fossils are crushed. The crania and mandibles are generally large in size and vary substantially in morphology. Complete crania permitting measurement of total length and condylobasal length are rare, but those have an average of 490 mm for cranial total length (n = 2) and 440 mm for condylobasal length (n = 4; table 5). However, complete mandibles are abundant, with a mean length of 318.5 mm (n = 8; table 11). Based on the ratios of 1.47 and 1.37 for cranial total length/mandibular length and condylobasal length/mandibular length, respectively, from an individual associated cranium and mandible of AMNH F:AM 76005, the range of cranial total lengths and condylobasal lengths in this species are estimated at 411–553 mm and 383–515 mm, respectively. The large size range and bimodal distribution of mandibular size suggest considerable sexual dimorphism in this species.
The only skull that is not severely compressed is AMNH F:AM 49372 from the Camel Canyon Quarry, Quiburis Formation, Arizona (fig. 8). Our description of the cranium thus is based primarily on this specimen, although some of the other crushed crania are informative and figured for comparison (e.g., AMNH F:AM 76015, fig. 9B).
Dorsal view: The rostrum is robust, with a distinct postcanine constriction. The forehead is rather wide and the postorbital process of the frontal is moderately developed. The temporal ridges unite into the sagittal crest in the middle (anteroposteriorly) part of the braincase. The zygomatic arch reaches its widest point at the level of the glenoid fossa. The mastoid process is well developed and extends far laterally, surpassing the lateral extent of the squamosal shelf.
Lateral view: The cranium in AMNH F:AM 49372 is dorsoventrally compressed postmortem. What can still be seen is that the anterior border of the orbit lies at the level of the M2 metacone. The mastoid process extends farther ventrally than does the postglenoid process. The paroccipital process is weak and not posteriorly extended. The squamosal shelf is not dorsally convex. The sagittal crest is strongly developed at the posterior part of the braincase.
Ventral view: The anterior border of the rostrum is arched. The palate is not flat, as it is distinctly convex at the level of canine and concave at the level of M2. This pattern is very similar to that of the extant sloth bear Ursus ursinus. The palatine fissure is small, located at a level between I3 and the canine. The anterior palatine foramen is located at a level between M1 and M2; the rostrum also reaches its widest point at this level. The anterior end of the zygomatic arch is located at the M2. The ventral border of the bony choana is located at the level of the M2 posterior border. The postglenoid process is moderately developed, whereas the preglenoid process is weak. The mastoid process is greatly expanded laterally. The paroccipital process is located at the base of the mastoid process, lateral to the auditory bulla. The bulla is poorly preserved, but the external meatus is well developed and the bulla is slightly inflated in the middle part. The postglenoid foramen is located at the level of the postglenoid process, distinctly medial to the level of the paroccipital process.
Mandible: No complete and well-preserved mandibles are known. Our mandible description is mainly based on three of the most complete specimens (AMNH F:AM 76005, AMNH F:AM 76010, and AMNH F:AM 144551) from the Quiburis Formation (fig. 10).
AMNH F:AM 76005 is a laterally crushed skull with associated mandible. The mandible preserves a nearly complete contour, though details cannot be clearly seen. The two most distinctive characters are the presence of a well-demarcated premasseteric fossa and the high positions of the mandibular condyle and angular process. The premasseteric fossa is rounded and confined to the posterior part of the horizontal ramus. The mandibular condyle is located significantly higher than the toothrow, and the angular process is located slightly lower than the toothrow. The coronoid process is well developed and posteriorly inflected.
The horizontal ramus of AMNH F:AM 76010 is in good condition, but the ascending ramus is broken. The horizontal ramus of AMNH F:AM 76010 has a uniform height along its A-P length, whereas in AMNH F:AM 76005 the anterior part of the horizontal ramus is lower. The most noteworthy difference between AMNH F:AM 76010 and AMNH F:AM 76005 is the total absence of a premasseteric fossa in AMNH F:AM 76010. This character is variable within the species, as the premasseteric fossa is present in 16 in 18 specimens examined, and in seven it is only weakly developed (it is absent in the other two). There are three mental foramina. The anterior is the largest and is located at the level of the p1; the two posterior ones are nearly the same size and located close to each other below and between p2 and p3. The mandibular condyle and angular process are elevated in position relative to the mandibular corpus in AMNH F:AM 76010 and AMNH F:AM 76005. AMNH F:AM 144551 is generally similar to AMNH F:AM 76010, but there is a weak premasseteric fossa present in AMNH F:AM 144551.
Dentition: Our description is based on a summary of all the specimens we examined, with intraspecific variation noted where applicable. The teeth of AMNH F:AM 49372 are lost or worn.
I1 and I2 are represented mainly by isolated teeth (fig. 11). I1 and I2 are not distinguishable from each other due to their generally similar morphology and the lack of directly associated specimens within a single skull for intraindividual comparison. In I1/2 the posteromedial accessory cusp is strongly developed, and a very large accessory cusp (here termed the posterocentral accessory cusp) is separated from the posteromedial accessory cusp. The posterolateral accessory cusp is small in size, is much weaker than the other accessory cusps, and is ridgelike. The main cusp is often subdivided on the lateral side, but this character seems to be variable intraspecifically. I3 is distinctly larger than I1 and I2. Both medial and lateral ridges are present. The medial cingulum is strongly developed, and there is a small notch between the medial ridge and medial cingulum, and a distinct groove along the central line of the medial cingulum, as in Indarctos.
P1–P3 are small, buttonlike, and always present. P1 is single-rooted. P2 is mostly single rooted, but occasionally with fused two roots (in one of five specimens). P3 is more variable, with a single root most frequently observed (two of four cases), but with some specimens having two partially to fully separated roots (one separated, and one partially fused).
The P4 parastyle is large (fig. 6). The P4 buccal contour is generally straight or shows little concavity. The lingual lobe is moderate in size, and the hypocone is variable in its morphology, typically subdivided into two cusps, with the anterior one often smaller. M1 varies in shape from square to slightly trapezoidal. The width across the metacone is always wider than that across the paracone. The anterior border of M1 is concave in most specimens. The parastyle and metastyle are often absent, but can be present and distinct from each other in some specimens. The protocone is occasionally weakly subdivided into two cusps. The direction of the postprotocrista most often diverges from the paracone-metacone axis, but sometimes parallels this axis. A small transverse ridge between the paracone and protocone is often present. A medial ridge of the metaconule is mostly present. M2 always has a talon, but it varies in size. The talon can be moderately elongated (e.g., AMNH F:AM 76006, fig. 6: B1) or very short (e.g., AMNH F:AM 49372, fig. 6: B3); the latter condition is not much different from that of Agriotherium. The buccal concavity between the paracone and metacone is always present. A small medial branch of the anterior ridge of the paracone (RPa1.2 of Jiangzuo et al., 2019), a small transverse ridge between the paracone and protocone, and a medial ridge of the metaconule are all only occasionally present.
The i1 and i2 are simple, dominated by a single main cusp. The i3 bears a large lateral accessory cusp.
Like the upper premolars, lower p1–p3 are generally single-rooted, and always rounded and buttonlike in shape. The p1 and p3 are always present, whereas p2 is occasionally absent (in two of 16 specimens). The p3 is occasionally double rooted (two of 10 specimens), but otherwise is typically single rooted.
The p4 has a triangular occlusal outline, with a buccal concavity in most specimens (fig. 7: B1, 5, 8, 12). A distinct anterior accessory cusp is nearly always present (20 of 22 specimens) and can be moderate to strong. The hypoconid (posterior cingulid cusp) also is generally present. A medial ridge of the main cusp is generally absent, but indistinctly occurs in some specimens (3 of 21 specimens). The m1 is elongate, especially so with its anteriorly directed paraconid; a distinct lingual concavity often occurs between the paraconid and protoconid. There are always three successively smaller cusps, from anterior to posterior, on the posterolingual side of m1, representing the metaconid and subdivided entoconid, a typical Indarctoslike pattern (Qiu and Tedford, 2003; Jiangzuo and Flynn, 2019). A well-developed mesoconid is situated between the protoconid and hypoconid, again similar to the condition in Indarctos. The m2 has a massive trigonid that is wider than the talonid. The m2 metaconid is always well developed, and often is similar in height to that of the protoconid. An m2 premetaconid is sometimes present; the two m2 entoconids (subdivided entoconid) are always well separated. The m3 is oval to triangular in shape, sometimes with a metaconid present.
Postcranial skeleton: A large number of postcranial bones of this species are known from the Quiburis Formation. However, nearly all these bones are crushed, limiting the anatomical information that can be extracted from them.
Scapula: The scapula is extremely rarely preserved in this species—only a broken distal scapula fragment association with a radius and ulna, AMNH F:AM 50042, is known from the Guymon Formation, Texas. The distal articular facet for the humerus is elongated anteroposteriorly.
Humerus: Five nearly complete humeri are present. The best preserved is UCMP 31818 from Coffee Ranch, Texas, with a nearly undeformed distal part but crushed proximal portion (fig. 12). The deltoid crest is well developed and extends distally along 2/3 the length of the humerus. The distal part of the humerus is weakly expanded. The trochlea is weakly undulating on its distal border, whereas there are two concavities on the proximal border. The epicondyles are very weak for an ursoid. The lateral border of the ectepicondyle ridge is slightly folded medially. The entepicondylar foramen is absent. In distal view, both the entepicondyle and ectepyle are well developed and oriented perpendicularly to the trochlea, forming the lateral wall of the olecranon fossa. AMNH F:AM 76070 also is well preserved, with moderate dorsoventral compression. Its morphology is generally similar to that of UCMP 31818, although the lateral border of its ectepicondyle ridge is distinctly medially folded. Other humeri in this species have similar morphologies. None of the humeri preserve an undeformed proximal end, but it is probably anteroposteriorly elongated, judging from the corresponding facet of the scapula.
Radius: The only undeformed radius of this species, UNSM 76013, comes from the SM-101 locality, Sherman County, Nebraska (fig. 13). This bone is slightly curved. The proximal facet for the humerus is oval in shape, with a slightly concave anterior border. A proximal tuberosity is present near the proximal facet, and the distal facet for articulation with the scapholunar is robust. There are three muscle tendon grooves on the distal end of the radius (Evans and de Lahunta, 2013). The groove for the extensor digitalis communis is the deepest, as in Indarctos, but distinct from Ursus and Ailuropoda in which this groove is wide and shallow. All other radii are either broken or compressed, but their general morphology does not differ from that of UNSM 76013.
Ulna: The best-preserved ulna, AMNH F:AM 50042, comes from the Guymon Formation (fig. 14). The shaft is slightly crushed, but the proximal and distal ends are largely undeformed. The olecranon process is moderately developed, with a strongly medially expanded process; its anterior boundary with the anconeal process is unclear. The coronoid process of the ulna is well developed and wide. A fossa distal to the coronoid process is present but shallow. Other ulnae, though crushed or incomplete, show similar morphologies.
Carpal bones: Carpal bones of this species are rare. Only the scapholunar, cuneiform, and pisiform are represented. AMNH F:AM 146450 is the best-preserved scapholunar, though its lateral border is broken. The bone is massive, with a strongly developed ventromedial process; the distal end of this process bends laterally. In distal view, the articular facet for the trapezium and trapezoid is not ventrally extended. The only complete cuneiform is AMNH F:AM 146452 from Coffee Ranch. This bone has a rhomboid shape. The medial facet for the scapholunar is small, whereas the lateral facet for the pisiform is well developed. Two nearly undeformed pisiforms are preserved. AMNH F:AM 112588 from Coffee Ranch is massive. The proximal facets for articulation with the cuneiform and ulna are asymmetric in size. AMNH Red. (Redington) 117-1831 from the Quiburis Formation is smaller than the Coffee Ranch specimens, but similar in morphology.
Femur: All femurs are heavily crushed. Nevertheless, some characters can be observed (fig. 15). The femur is generally short and the greater trochanter is well developed. The femoral head is slightly higher than the greater trochanter, as also can be seen in the undeformed femur figured by Harrison (1983). The greater trochanter of AMNH F:AM 146503 is similar in height to the head of the femur, reminiscent of the condition in felids, although deformation may account for its unusual morphology relative to other specimens. The only large felid from the Quiburis Formation is Amphimachairodus coloradensis, but it does not seem possible for this species to have a femur that reached such an extremely large size (475 mm), so it is unlikely that this bone represents a felid. Its shaft is elongated. The distal articular facets are heavily crushed, but the medial border is clearly substantially higher than the lateral border. The undeformed femur (L= 412.4 mm) from the Edson Local Fauna (Harrison, 1983) is distinctly smaller than those from the Quiburis Formation (L= 454.12–516.28 mm), suggesting possible intraspecific geographic variation.
Tibia and Fibula: All tibias are heavily crushed and few distinct characters can be seen. The tibial tuberosity seems to be moderately developed. Only a broken fibula attached to a tibia (AMNH F:AM 146513) is known; on it, the distal fibular facet for the astragalus is well developed.
Astragalus: Several astragali are preserved, but all are crushed or broken. The neck is moderately long. The facet for the cuboid is present, but seems to be small.
Calcaneum: The best-preserved calcaneum UCMP 31837, which is undeformed, is from Coffee Ranch (fig. 16). This calcaneum is massive and deep. Its sustentacular process is relatively small, and its distal border is widely separated from the distal articular facet. The distal facet for the cuboid is anteroposteriorly narrow. Other calcanea are consistent in these characters.
Mesocuneiform: Only one mesocuneiform, from Coffee Ranch (AMNH FM 146707), is preserved. Both the proximal and distal articular facets are triangular in shape. Its medial border is thicker than the lateral border, and the facets for the entocuneiform and ectocuneiform both are well developed.
Cuboid: The only undeformed cuboid, UCMP 30519, is from Coffee Ranch. The bone is relatively elongate. The proximal and distal facets form a small acute angle. The proximal facet for the calcaneum is transversely elongated. Two smaller facets lie medial to the calcaneal facet. The dorsal one is larger, representing the facet for the astragalus, and the ventral one is smaller, representing the facet for the navicular. The medial facet for the ectocuneiform is small and less distinct than the other cuboid facets. The distal facet for MT4 and MT5 is triangular in shape, and the part articulating with MT5 is rather small. The ventral process of the cuboid is massive, as in all bears.
Metapodials (metacarpals/metatarsals): The metapodial bones are slender, and are similar in general morphology to other bears. The medial process of MC I for the M. abductor pollicis longus is relatively weak; the medial scar for the M. extensor carpi radialis longus does not form a process; the rugous scar for the carpal ligament of MC IV is well developed, extending more than halfway along the bone shaft on its ventral surface; and MC/MT IV are longer than MC/MT V.
FIG. 6.
Upper P4–M2 of Huracan and related taxa. A. Huracan qiui, HMV 2005, Wangjiashan. B. Huracan coffeyi, B1. AMNH F:AM 76006 (reversed), Quiburis Formation; B2. AMNH F:AM 145923 (reversed), Quiburis Formation; B3. AMNH F:AM 49372, Big Sandy Formation. C. Huracan roblesi (IVPP uncatalogued cast), Venta del Moro. D. Agriotherium africanum, Langebaanweg, D1. AMNH FM 105141 (cast of PQ-L 47137, reversed); D2. AMNH FM 105141 (cast of PQ-L 41404, reversed). E. Huracan schneideri, E1. IGM 6668 (cast), Rancho Viejo Beds; E2. UF 133944 (reversed); E3. UF 21227 (reversed); E4. UF 203158 (reversed); E5. UF 206887 (reversed), not all associated. F. Indarctos atticus, F1. IVPP FV 2107 (cast), Samos; F2. IVPP V 6893.10–12, Lufeng. G. Plithocyon teilhardi, AMNH FM 26594, Tunggur Formation. Each letter represents one species and each number represents one individual.
FIG. 7.
Lower p4–m2 or m3 of Huracan and related taxa. A. Huracan schneideri: A1. IGM 6413 (cast), Rancho Viejo Beds; A2. UF 47482, Upper Bone Valley; A3. UF 53977, Upper Bone Valley; A4. UF 21226 (reversed), Upper Bone Valley, not all associated. B. Huracan coffeyi: B1. AMNH F:AM 146483; B2. AMNH F:AM 76013; B3. AMNH F:AM 146493; B4. AMNH F:AM 76023; B5. AMNH F:AM 146571; B6. AMNH F:AM 146488 (reversed); B7. AMNH F:AM 76021 (reversed); B8. AMNH F:AM 146481 (reversed); B9. AMNH F:AM 146490; B10. AMNH F:AM 76033 (reversed); B11. AMNH F:AM 146500; B12. AMNH F:AM 76031, not all associated. C. Indarctos cf. I. atticus: AMNH F:AM 22332, Baode. D. Agriotherium hendeyi: AMNH F:AM 76000, Quiburis Formation. E. Agriotherium africanum, Langebaanweg: E1. AMNH FM 105140 (cast of PQ-L45114, reversed); E2. AMNH FM 105148 (cast of PQ-L 50006); E3. AMNH FM 105149 (cast of PQ-L50007, reversed); E4. PQ-L 50446 (reversed); E5. PQ-L 50004, all not associated. F. Agriotherium palaeindicum: IVPP V 18411 (reversed), Xiaoxian. G. Plithocyon teilhardi: AMNH FM 26594, Tunggur Formation. Each letter represents one species and each number represents one individual.
TABLE 1.
Summary of most important generic differences of Agriotheriini.
TABLE 2.
Measurements and ratios of C and P4 in Agriotheriini. See text for abbreviations.
TABLE 3.
Measurements and ratios of M1 and relative size of C, P4, M2 in Agriotheriini. See text for abbreviations.
TABLE 4.
Measurements and ratios of M2 in Agriotheriini. See text for abbreviations.
FIG. 9.
Skulls (crushed) of Huracan coffeyi. A1–3. AMNH F:AM 76006; B. AMNH F:AM 76015; C. AMNH F:AM 76005; D. AMNH F:AM 30399.
FIG. 10.
Mandibles of Huracan coffeyi (lateral and dorsal views). A1, 2. AMNH F:AM 76010; B1, 2. AMNH F:AM 144551; C. AMNH F:AM 76005; D1, 2. AMNH F:AM 50032. Note variation of the premasseteric fossa.
TABLE 5.
Measurements and ratios of the cranium in Agriotheriini. See text for abbreviations.
continued
continued
FIG. 11.
Comparison of incisors among Huracan, Indarctos, and Agriotherium. A. Huracan coffeyi: A1. left I1/2, AMNH F:AM 76038; A2. left I1/2, AMNH F:AM 146477; A3. right I1/2, AMNH F:AM 146473; A4. right I1/2, AMNH F:AM 146472; A5. right i1/2, AMNH F:AM 146474; A6. right i3, AMNH F:AM 76028A; A7. left I3, AMNH F:AM 146475; A8. left I3, AMNH F:AM 146475. B. Huracan qiui: HMV 2005, left I1–3. C. Indarctos lagrelii: cast of type, left I1–3. D. Agriotherium palaeindicum: Xiaoxian, Anhui Province, China, left I1–2. Note the specialised characters on I1–3, emphasized in the figures of Huracan, which are absent in both Indarctos and Agriotherium.
FIG. 12.
Humerus of Huracan coffeyi: A. UCMP 31818, Coffee Ranch. B. AMNH F:AM 76070, Quiburis Formation. C. AMNH F:AM 76072, Quiburis Formation.
FIG. 13.
Radius of Huracan coffeyi: A. UNSM 76013, Sherman Co. Locality SM-101, A1. anterior view; A2. posterior view; A3. distal view; A4. proximal view. B. AMNH F:AM 76077, Quiburis Formation. C. AMNH F:AM 76078, Quiburis Formation.
FIG. 14.
Ulna of Huracan coffeyi: A. AMNH F:AM 50042, Guymon, A1. lateral view; A2. medial view; A3. anterior view. B. AMNH F:AM 76008, Quiburis Formation. C. AMNH F:AM 76082, Quiburis Formation.
FIG. 15.
Femur of Huracan coffeyi: A. AMNH F:AM 50086, Quiburis Formation. B. AMNH F:AM 76105, Quiburis Formation. C. AMNH F:AM 146503, Quiburis Formation.
Huracan qiui, sp. nov.
Etymology: Species name in honor of Zhanxiang Qiu, for his outstanding contributions to Chinese Cenozoic paleontology and stratigraphy, especially in the Linxia Basin.
Type specimen: HMV 2005, a nearly complete skull with little deformation (fig. 17), collected by local villagers from the Wangjiashan fossil locality, which yields the Qingbushan Fauna.
Diagnosis: Huge ailuropodine bear with weakly hypercarnivorous features; P2–3 double rooted; P4 slightly shorter than M1, with subdivided hypocone; M1 elongate with large metastyle; and M2 slightly elongate.
Differential diagnosis: Differs from both Huracan coffeyi and H. schneideri in having a narrower cranium; more developed anterior premolars; longer M1 and M2, with proportionally longer M2 talon.
Type locality: Wangjiashan, Linxia Basin, Gansu Province, China; Baodean ALMA (late Miocene).
Occurrence: Known only from the type locality.
Description: HMV 2005 represents a nearly complete cranium, lacking the zygomatic arches and the posterior part of the sagittal crest, and with a broken occipital condyle. The braincase was slightly dorsoventrally depressed postmortem. The preserved part is ∼475.7 mm in length, and the total skull length is estimated at ∼485–495 mm. Since HMV 2005 is the only specimen known for the species, its sex is unknown, which is important for comparisons among sexually dimorphic species (most ursids). Its large size could indicate that the individual is a male, whereas the low cranium, relatively small canine (CL/M1L = 0.87), weak expansion of the rostrum at the level of canines, and poorly developed sagittal crest instead could indicate a female of a very large-bodied species.
Dorsal view: The cranium is elongate. Judging from the preserved part of the anterior root of the zygomatic arch, the cranium is not wide. The anterior part of the rostrum is constant in width, but the rostrum widens posteriorly at the level of the P4. The anterior border of the nasal is broken, and the posterior border nearly reaches the level of the postorbital process of the frontal. The length of the nasal is ∼120 mm, and it gradually narrows and tapers in width posteriorly. The surface of the nasal is slightly dorsally domed. The suture between the maxilla and frontal reaches the level of the midpoint of the orbit. The nasal is separated from the maxilla by a wide bridge (11.82 mm at the middle part) composed of the posterior process of the premaxilla and the anterior process of the frontal. The premaxilla and frontal meet at the level of the P4 paracone. The forehead is wide, slightly exceeding the width of the rostrum across the canine. A postorbital process of the frontal is present but not distinct. Temporal ridges are moderately developed; the ridges from both sides unite into the sagittal crest at the anterior part of the braincase. The postorbital constriction is indistinct. The braincase is widest in its middle part, and is significantly wider than the postorbital constriction.
Lateral view: The cranium is not high. The dorsal contour of the cranium forms a gentle curve, with the highest point located in the middle part of the cranium, at the level of the postorbital constriction. The anterior contour of the naris opening is very slightly concave posteriorly, and forms a very small angle with the dorsal contour of the nasal bone. The anterior border of the nasal lies at the level of the P1. The upper toothrow is distinctly concave ventrally from P3 to M2. The lateral surface of the lower side of the maxilla undulates along the tooth roots. The infraorbital foramen is vertically oriented, and located at the level of the P4 metacone; its dorsoventral diameter is 15.42 mm on the left side and 12.22 mm on the right. The anterior border of the orbit is located between the level of M1 and M2, and its anteriormost point is distinctly higher than the infraorbital foramen. The area anterior to the orbit is slightly concave. The postorbital process of the frontal is slightly lower than the cranial roof. The anterior root of the zygomatic arch is robust and high. The ventral border of the zygomatic arch forms a distinct process at the anterior part, and the suture between the maxilla and jugal begins at the apex of this process, running anterodorsally and stopping at the anteriormost point of the orbit. The orbital region is not well preserved. The posterior opening of the infraorbital canal lies at the front of the orbit. Dorsal to that opening, the lacrimal fossa is not clearly demarcated, due to poor preservation. A large fossa for the confluent caudal palatine foramen and sphenopalatine foramen is located posteroventral to the posterior opening of the infraorbital canal, located at the level of the posterior end of M2 and relatively high above the toothrow. The inferior orbital crest is well developed and wide. It begins at the boundary between the lacrimal fossa and the posterior opening of the infraorbital canal and ends at the orbital fissure. The middle part of this crest is wide and is distinguished from the inferior temporal fossa (lateral wall of the medial pterygoid fossa) by a wide and shallow groove. The superior orbital crest is not distinct. The positions of the ethmoid foramen and optic foramen are not clear. The orbital fissure and foramen rotundum are close to each other but are separated by a thin bony partition. There is no alisphenoid canal. The anterior border of the braincase (dorsal border of the superior orbital crest) is nearly perpendicular to the orbital region, and bears several ridges for attachment of the temporal muscle. The suture between the frontal and parietal bones on its dorsal side is located at the level of the anterior part of the braincase. Its dorsal tip joins the point where the temporal ridges unite into the sagittal crest. The suture between the parietal and the squamosal part of the temporal bone cannot be seen, but a horizontal ridge is present in the expected location of this suture based on comparisons to other taxa. The postglenoid fossa faces anteriorly, and is situated distinctly higher than the toothrow. The mastoid process is broken. The ventral contour between the postglenoid process and the mastoid process is strongly concave dorsally. The sagittal crest is moderately developed. At the surface of the posterodorsal part of the braincase, a large foramen (probably for a vein) is present.
Ventral view: The rostrum is slightly elongate compared with that of other species of Huracan, slightly constricted posterior to the canine, and widest at the level of the M1. The palate is slightly vaulted. The palatine fissure is elongate, surrounded by the premaxilla, and its posterior border reaches the posterior part of the upper canine. The posterior process of the premaxilla reaches the level of the P1 along the midline. The suture between the maxilla and palatine curves and tapers anteriorly, and its anterior border reaches the level of the P4 paracone. On the left side, the major and minor posterior palatine foramina are united by a small fissure. The major posterior palatine foramen is located at the level of the P4 paracone. On the right side, the major posterior palatine foramen is situated slightly more anteriorly, between the level of the P4 and M1. Behind M2, the maxilla forms a small spine at the suture of the maxilla and palatine. Medial to this spine, a distinct lateral palatine notch is present. The anteroventral border of the medial pterygoid fossa (bony choana) lies at the level of the M2 talon, and includes a distinct medial spine near the center of this border. The posterior border of the zygomatic arch is located at the M2 metacone. Judging from the preserved part, the zygomatic arch is not very wide. The ventral borders of the medial pterygoid fossa are parallel, whereas the lateral walls of this fossa are distinctly convex laterally. Anteriorly, the ventral border of the medial pterygoid fossa forms a distinct process at the level of the M2 metaconule. Posteriorly, at the suture between the palatine and the basisphenoid, the lateral wall of the pterygoid fossa forms a spine. Both the anterior and posterior glenoid processes are well developed medially. The distance between the posterior borders of the M2 and the glenoid fossa is ∼113 mm, exceeding the P4–M2 length (101.18/100.00 mm on the left and right sides, respectively). The basicranium is not well preserved. The oval foramen is located medial to the glenoid fossa, and faces anteriorly. A bony crest is present lateral to the oval foramen. The postglenoid foramen is located at the posteromedial border of the postglenoid process. The basioccipital and basisphenoid bones are completely fused on the ventral side, so that no distinct boundary between them can be discerned. The paired attachment fossae (anteriorly) for the M. rectus capitis ventralis major (or M. longus capitis) and those for the M. rectus capitis ventralis minor, situated posterior to the former, are not distinguishable. The auditory bullae are largely broken, preserving only the anteromedial part. Details of the foramina surrounding the bulla cannot be seen. The mastoid and paroccipital processes also are broken. The hypoglossal foramen is located close to the bulla, but is still separated from the posterior lacerate foramen by a wide bony bridge. The occipital condyle is broken. The condyle neck is 76.62 mm, and the condyle would be ∼90–95 mm wide.
Occipital view: The occipital surface is poorly preserved, with most borders broken. On the left side, the lambdoidal crest and the lateral border of the mastoid form an obtuse angle. The occipital condyles are widely separated. The foramen magnum is ovoid, being wider than high. The external occipital crest is well developed in its dorsal part. Latero-dorsal to the occipital condyle, there is a concave area above the paroccipital process.
Dentition: The dentition is completely preserved. The incisor row protrudes anteriorly, forming a convex arch. Upper incisor sizes increase from I1 to I3. I2 is slightly larger than I1, whereas I3 is distinctly larger than I2. I1 and I2 are very similar in morphology, with one large main cusp and two posterior accessory cusps. The posteromedial accessory cusp is enlarged and occupies 3/4 (I1) and 2/3 (I2) of the area of the posterior part of the tooth. The posterolateral accessory cusp is small in size (more distinct in I1). I3 is large, robust, and caniniform. I3 has a long and distinct lateral ridge and a much shorter medial ridge. At the medial base of I3, a cingulumlike ridge is present along the tooth edge, subdivided into two parallel ridges in the middle with the lateral one less distinct.
All three anterior upper premolars are present, and all are small and buttonlike. All the anterior premolars show similar morphology, with one main cusp, anterior and posterior ridges, and a well-developed lingual cingulum. The posterior ridge of the P2 turns lingually at its posterior part, whereas the anterior ridges on the other two anterior premolars do not. P3 is slightly longer than P1 and P2. Both P2 and P3 are double-rooted, whereas P1 is single rooted.
P4 is large, robust, and triangular in shape. A distinct buccal concavity is present at the paracone apex. The P4 inner lobe is moderate in size, and its anterior border forms an acute angle with the lingual border of the parastyle. A buccal cingulum is only weakly developed on the buccal side of the parastyle and the anterobuccal side of the metacone. Three buccal cusps are present: parastyle, paracone, and metacone. The parastyle is large, slightly lingually shifted, and with weak buccal and lingual ridges at the anterior tip. The metacone is slightly shorter than the paracone. The hypocone dominates the inner lobe, and is weakly subdivided into two cusps, with the anterior one being slightly smaller. A tiny cusp is present at the anterobuccal corner of the tooth, representing a rudimentary protocone. A well-developed lingual cingulum flanks the side of the metacone. M1 is square in shape, with a relatively straight border, no anterior concavity, and an extremely weak buccal concavity. Three cusps are present on the buccal side of M1. The M1 metacone is smaller than the paracone, due to the large area occupied by a huge metastyle, which is approximately 2/3 the length of the metacone. A small bulge is present at the anteriormost point of the paracone, but it does not form a distinct cusp. Lingually, any cusps present are too worn to distinguish and identify. The lingual cingulum merges with the lingual slope of the protocone-metacone. M2 is similar in length to M1, but is trapezoidal rather than square in shape. The buccal concavity between the paracone and the metacone is distinct on M2. The M2 talon is relatively short. M2 generally is less worn than M1. The M2 paracone is larger than the metacone. A ridge on the posterior border of the metacone elevates slightly to form a tiny metastyle. The lingual ridge of the paracone is indistinct, whereas that of the metacone is strong. The protocone is a long ridge without distinct subdivision, whereas the metaconule is a shorter and conical cusp. A small ridge is present between the M2 metacone and the edge of the protocone and metacone cusps.
COMPARISON AND DISCUSSION
The three species from northern China and North America together represent a new taxon, herein named and diagnosed as Huracan, characterized by a unique combination of Indarctoslike and Agriotherium-like characters, as well as several newly recognized autapomorphies.
Comparison of Huracan to Indarctos and Agriotherium
The type species of Huracan was originally viewed as assignable to Agriotherium but with some Indarctos-like characters (Hunt, 1998). Indeed, the overall morphologies of species of Huracan are most similar to Agriotherium. Both genera share several characters: short rostrum, small anterior premolars, large P4 parastyle, short M2 talon, presence of a premasseteric fossa, and absence of an entepicondylar foramen of the humerus.
Nevertheless, Huracan also exhibits several characters that are never observed in or are extremely rare in typical Agriotherium. The most significant distinguishing character is the Indarctos-like m1 entoconid and metaconid complex in Huracan. In Huracan (as in Indarctos), there are always three cusps in this region, representing the metaconid and a subdivided entoconid. Nearly all Agriotherium, except one specimen of A. africanum (PQL 50446), possess no metaconid (most cases) or only an extremely weak metaconid (A. africanum: PQL 50006; A. hendeyi: AMNH F:AM 63098) (Jiangzuo and Flynn, 2019). Beyond this character, Huracan also differs from Agriotherium in: P4 with more frequently subdivided hypocone; proportionally longer M1 and M2, and larger M2 talon (fig. 18); p4 mostly lacking the medial ridge; and presence of a posterocentral accessory cusp, well developed and distinguished from the medial ridge of the I1/I2. Among these characters, M1 and M2 shapes are intermediate between the typical Indarctos and typical Agriotherium conditions. In terms of the M2/M1 length ratio, the M2 of H. coffeyi is shorter than any species of Indarctos, without overlap (fig. 18, table 3), but some specimens overlap with A. africanum in talon size. In terms of talon size, using the ratio of M2 TL/L, H. coffeyi overlaps with species of both Indarctos and Agriotherium. The standard error for this ratio in H. coffeyi is 9.4 × 10-3, larger than that represented across the entire genus Indarctos (7.1 × 10-3), suggesting extensive intraspecific variability of talon size in H. coffeyi. The second character (p4 without medial ridge) is shared between most individuals of Huracan and Indarctos, although a few Hh3-aged H. coffeyi specimens and several Hh4-aged H. schneideri specimens have this ridge. The last character (presence of I1/I2 posterocentral accessory cusp) is an autapomorphy, as it is absent in both Indarctos and Agriotherium. Huracan differs from Indarctos in the peculiar I1/I2 morphology of the new taxon; more reduced anterior premolars, larger P4 parastyle, proportionally shorter M1 and M2 with a shorter talon, presence of a distinct anterior accessory cusp on p4, more elongate m1, and presence of a premasseteric fossa on the mandible. Most of these characters are shared with Agriotherium, but do not occur in Indarctos and Ursavus (outgroup), representing synapomorphies indicating a closer relationship of Huracan to Agriotherium than to Indarctos. Some characters show variability, e.g., the I3 medial cingulum often bears a distinct central groove in Huracan, which also is the case in Indarctos (at least in all material from Lufeng and unpublished material from Ningxia, North China), but the groove seems to be absent in all material of Agriotherium (based on published specimens and unpublished material from Yushe).
FIG. 18.
Dental and mandible plots of Huracan and related taxa, with emphasis on dental length and its with/length ratio, ratio of length between different teeth, and premolar/molar row length.
In terms of cranial characters, Agriotherium has the shortest and most robust rostrum relative to Huracan and Indarctos (fig. 19), with Huracan intermediate between Agriotherium and Indarctos. Two ratios can be used to measure relative rostrum length: relative diastema length between C/c and P4/p4 and cheek tooth (including molars and P4/p4) row length, i.e., L21/L9 of the cranium and L5/L4 of the mandible (see the definitions of L4, L5, L9 and L21 in Materials and Methods). Both ratios in H. coffeyi (means of 0.36, and 0.33, respectively; see the ratios and variation of each value in tables 6, 11) are smaller than those of Indarctos (means of 0.42 and 0.38, respectively; see the ratios and variation of each value in tables 6, 11), indicating a generally shorter rostrum, although there is considerable overlap in their ranges.
Huracan, Agriotherium, and Indarctos share many derived traits relative to ursid outgroups, e.g., large body size, reduced anterior premolars, enlarged carnassial, m1 with expanded talonid, etc. Within this clade of Huracan, Agriotherium, and Indarctos, Huracan shares only retained primitive characters with Indarctos but possesses some derived characters shared with Agriotherium. Huracan thus represents a morphologically intermediate taxon between the two genera, providing an important link in understanding transformations between the morphology of the previously well-known taxa Indarctos and Agriotherium, although the presence of autapomorphies clearly indicate that it was not directly ancestral to Agriotherium (neither morphologically nor temporally, as Agriotherium also had already appeared in both Asia and North America prior to the first occurrence of Huracan), but rather represents the closest sister group to Agriotherium.
Another hypothesis about the origins of Agriotherium was that it evolved from some kind of hemicyonid bear (Qiu et al., 1991; Morales et al., 2005; Abella et al., 2014). However, the dentitions of hemicyonids differ markedly from that of Agriotherium. In comparison to Agriotherium, P4s of hemicyonids lack a parastyle (most taxa) or only possess a cingulumlike structure in the parastylar region (a very few specimens of Plithocyon, especially in P. armagnacensis). In hemicyonids, the M1 and M2 are more transversely widened and have a well-developed lingual cingulum, M2 never develops a talon and is always smaller than M1, lower molars are overall narrower, the m1 talonid is not especially wide, and the m1 buccal contour only slightly undulates or is straight (Frick, 1926; Ginsburg and Morales, 1998; Hunt 1998). Concerning cranial morphology, hemicyonid skulls are less stout, possess an alisphenoid canal, and the mastoid process is never exaggeratedly developed (Frick, 1926). In all of these aspects, Agriotherium is much more similar to Indarctos and Huracan than to hemicyonids. The limited similarities between Agriotherium and hemicyonids include a P4 inner lobe with single cusp and M2 lacking a talon. But as discussed above, these traits also show a continuous range of variation from Indarctos-type through Huracan-type to Agriotherium-type, and the latter type is convergent with hemicyonids, presumably reflecting common adaptations in these two lineages toward a more carnivorous diet. Postcranial morphology further suggests that hemicyonids were better adapted to a cursorial locomotor pattern than any other bear (Frick, 1926), even relative to the quite cursorial morphology exhibited independently by Huracan and Agriotherium.
PHYLOGENETIC ANALYSES
We performed phylogenetic analyses of 33 species, based on a large matrix containing 164 characters, including 141 dental traits, 22 craniomandibular traits, and one postcranial trait (see matrix and tables S1 and S2 in the online supplement: https://doi.org/10.5531/sd.sp.56). This matrix is partially based on the matrix of Abella et al. (2012) but expanded by our observations and analyses (Jiangzuo et al., 2019). The matrix is designed for the whole Ursidae, but only Agriotheriini and selected outgroups are included here for analyses of interrelationships within this subclade. Seven continuous traits were subdivided into different categories and are set as either additive or nonadditive.
Our phylogenetic analyses using equal weighting and implied weighting (both implemented in TNT1.6) give similar results (fig. 20), differing only in the resolution of Indarctos anthracitis. Both analyses support the monophyly of H. qiui and two North American species H. coffeyi and H. schneideri, with H. qiui as basal to a clade of the other two species. ?“H.” punjabiensis is weakly supported as the nearest relative of the clade of Huracan and Agriotherium; while it also shares some features with species of Huracan, its taxonomic assignment is thus uncertain. Two Vallesian Indarctos are here supposed to be a near relative for both Turolian and equivalent aged Indarctos and Huracan + Agriotherium, which may merit a new genus, pending further research.
TABLE 6.
Ratios of the cranium in Agriotheriini. See text for abbreviations.
EVOLUTION OF HURACAN
So far, three species can be definitely assigned to the new taxon Huracan, two from North America (the genotype H. schneideri and H. coffeyi) and another from eastern Asia (China) (H. qiui). This clade appears to include a fourth species, from Europe (Spain) (H. [“Agriotherium”] roblesi), and possibly a fifth, from southern Asia (?“H.” punjabiensis).
The two North American species may represent chronospecies, as they are temporally sequential and nonoverlapping, share diagnostic apomorphies distinguishing them from all other species, and the older species is differentiated from the younger only by lacking diagnostic apomorphies of the younger species and thus may represent a metaspecies. The Hh3-aged species H. coffeyi previously was viewed as synonymous with H. schneideri (Miller and Carranza-Castaóeda, 1996; Hunt, 1998; Jiangzuo and Flynn, 2019), whose type is from the Upper Bone Valley, Hh4. Our observations on new specimens from the type locality of H. schneideri, and our review of all Hh4-age material of this group, suggest that H. schneideri shows stable differences from Hh3-age specimens, e.g., shorter M1 and M2 with smaller M2 talon, m1 with overall smaller size, and smaller posterior entoconid and weak or no subdivision of the m2 entoconid. Other differences in H. schneideri include smaller, more reduced premolars, and weaker anterior accessory cusps and the presence of medial ridge in p4, but these traits are known only from the type mandible of H. schneideri and the variability of these two characters within this species is unknown. Hh4-aged H. schneideri varies widely in size, e.g., the type specimen has a smaller dental size than that of any Hh3 Huracan, but its mandible size falls within the variation of that of Hh3-aged Huracan (fig. 18). But whether this represents a smaller average dental (toothrow) size or stronger sexual dimorphism (mandible size) is still unclear. Among the potential names for Hh3-aged Huracan, H. coffeyi (Dalquest, 1986) from Coffee Ranch should be used. The holotype is a mandible with well-preserved teeth, similar to other Hh3-aged Huracan specimens we observed, and therefore H. coffeyi is the valid name for Hh3-aged Huracan.
FIG. 20.
Phylogenetic trees of Agriotheriini: A. strict consensus tree, parsimony with equal weighting, tree length 473, CI = 0.548, RI = 0.834; B. single most parsimonious tree, parsimony with implied weighting (k = 12), tree length 16.16, CI = 0.549, RI = 0.835. Number in the node represents bootstrap value (1000 times).
The Asian species H. qiui represents a more anatomically primitive form in having a narrower rostrum, double-rooted P2, and more elongated M1 and M2 (therefore is more Indarctos-like than H. coffeyi). Morphologically and temporally, it is probably close to the ancestral lineage of the North American and European Huracan species.
Another species that can be assigned to Huracan is “Agriotherium” roblesi from Venta del Moro of Spain, MN 13 (Morales and Aguirre, 1976). This species was initially known only from M1 and M2. Both teeth are very close in morphology to those of H. coffeyi, especially in their large size and distinct but short M2 talon. Recently, a P4 and several new mandibles have been discovered at the type locality (Montoya et al., 2002; 2006; Morales et al., 2011), although a detailed description of that material is not yet published. Based on plate 2 of Montoya et al. (2006) and figure 2 of Morales et al. (2011), we can see that this species indeed has an Indarctos-like lower m1 (metaconid and double entoconid) and a premasseteric fossa, in accord with diagnostic features of Huracan. Based on the M1 and M2, this species shares diagnostic features of Huracan in having short M1 and M2, with small but distinct talon. The shape of these two teeth in H. roblesi is closer to those of Huracan coffeyi than to those of the Asian species H. qiui, so the specieswould be a member of the Huracan clade as H. roblesi. Since only the M1 and M2 were formally published, this species is not included in the phylogenetic analysis.
The southern Asian species ?“H.” punjabiensis has an M2 with a relatively short talon, also in accord with diagnostic features of Huracan (Jiangzuo and Flynn, 2019). Its M1 also is anteroposteriorly shortened. However, the P4 of this species is different in having a small parastyle, and therefore is more Indarctos-like. A large P4 parastyle is a synapomorphy of Huracan and Agriotherium, so ?“H.” punjabiensis appears to be more closely related to these two taxa than to Indarctos based on its upper molar morphology, but is probably slightly more primitive than the ancestral anatomical form of the P4 for the Huracan/Agriotherium clade. The phylogeny (fig. 20) supports it as the nearest outgroup of Huracan and Agriotherium, so it probably should be assigned to a distinct genus. Due to the paucity of material of this species, however, erecting a new genus would be premature and could create unnecessary confusion and instability to the taxonomy for this group. We thus cannot definitively assign it to any genus, but given the overall similarity of the available material of this taxon to Huracan, we provisionally treat it as ?“H.”punjabiensis herein. Previously, Baryshnikov (2002) proposed conspecificity of this species and the type species of Indarctos, i.e., I. atticus, therefore referring all European I. atticus to I. punjabiensis. This view was followed by some authors (Abella et al., 2019). We do not regard these two species as conspecific, and instead regard the European material previously assigned to I. punjabiensis as I. atticus.
TABLE 7.
Measurements and ratios of c and p4 in Agriotheriini. See text for abbreviations.
TABLE 8.
Measurements and ratios of m1 in Agriotheriini. See text for abbreviations.
TABLE 9.
Measurements and ratios of m2 in Agriotheriini. See text for abbreviations.
TABLE 10.
Measurements and ratios of m3, and relative sizes of lower teeth compared to m1, in Agriotheriini. See text for abbreviations.
CURSORIAL ADAPTATION OF HURACAN
Postcranial material of Huracan coffeyi is the most abundant and complete for any species of Agriotheriini (Ailuropodinae, Ursidae). Even though H. coffeyi has a huge skull and very large teeth (probably the largest among all known bears, living and extinct, together with Ursus ingressus and Arctotherium angustidens), its postcranial bones are only slightly larger than those of Agriotherium africanum, are smaller than Indarctos oregonensis, and are distinctly smaller than those of the giant short-faced bear Arctodus.
Even though largely crushed, and thus many detailed traits cannot be observed, it is still clear that H. coffeyi is likely one of the most cursorial members of the crown ursid clade. Among postcranial bones, the humerus is the most characteristic bone for inferring locomotor behavior (Andersson, 2004; Day and Jayne, 2007; de Oliveira and Santos, 2018; Meloro and de Oliveira, 2019). As mentioned in the description above, the epicondyle of the humerus in H. coffeyi is very weak for a bear, and its lateral expansion is indistinct (fig. 21). The ectepicondyle is proximally extended and overlaps the deltoid tuberosity. Although most known humeri are somewhat deformed, two ratios can still be used to reliably infer the locomotor behavior of H. coffeyi. The ratio of distal width to humerus length reflects expansion of the epicondyle, where a large value often relates to scansorial or fossorial locomotor capabilities and a small one is often related to terrestrial cursoriality. In living Ursus species that have scansorial and/or fossorial adaptations, this ratio ranges from 0.25 to 0.35 with most of those species near 0.30. The living giant panda (Ailuropoda) has both a similar ratio and similar locomotor behavior to extant species of Ursus, although Ailuropoda is less cursorial than Ursus. The extinct giant short-faced bear Arctodus simus is famous for its huge size and elongated limbs, which has been interpreted as reflecting greater cursoriality than in living bears (Kurtén, 1967; Christiansen, 1999; Matheus, 2003). The ratio of distal width and length of the humerus in Arctodus is indeed smaller than that of Ursus, but the epicondyles are still well developed. In contrast, H. coffeyi not only has a small width/length ratio as a result of relative elongation of the humerus, but the epicondyles also are distinctly weaker than those of Ursus, Arctodus, and Ailuropoda. The width/length ratio is distinctly smaller in H. coffeyi than in Ursus, Arctodus and Ailuropoda, but comparable to that of extant Panthera (Felidae) species, a large and relatively cursorial carnivore. The medial epicondyle provides the attachment for the M. pronator teres and several palm/digit flexion muscles, and the lateral epicondyle provides the attachment for the M. anconeus and several palm/digit extension muscles, and in its dorsal part for the M. brachioradialis and M. brachialis (Davis, 1964; Evans and de Lahunta, 2013). Weak expansion of the epicondyle of the humerus in H. coffeyi suggests it had weak musculature for rotation, extension, and flexion of the distal forelimb and manus. However, the lateral epicondyle crest is distinctly dorsally extended, as in Ailuropoda, suggesting relatively long fibers for the M. brachioradialis and M. brachialis. In distal view, the medial epicondyle is nearly perpendicular to the condyle and is well developed, making the olecranon fossa deep and narrow. Brown bears (Ursus arctos) have a similar morphology in this aspect, although the olecranon fossa is wider, whereas Arctodus, Ailuropoda, and Panthera all have a shallower olecranon fossa and the medial epicondyle forms an angle with the condyle. The olecranon fossa accommodates the olecranon process of the ulna, and a deep and narrow olecranon fossa suggests restricted mediolateral rotation ability of the ulna. Together, these traits suggest that H. coffeyi was a cursorial animal, more so than any living bear, the extinct giant short-faced bear, and probably extant lions, with a restricted ability to control prey because of forelimb adaptations to increased cursoriality. Indarctos has a variable humerus morphology, although all Indarctos species are clearly distinguished from H. coffeyi in having an entepicondylar foramen. The type specimen of I. oregonensis is very robust, with expanded humerus epicondyles like those of Ursus (Merriam et al., 1916), whereas I. atticus from Pikermi is distinctly more slender, similar to the morphology of H. coffeyi (Roussiakis, 2001). The humerus of Indarctos from Withlacoochee 4A, Florida also is slender, with weak epicondyles (Jiangzuo and Hulbert, 2021). Agriotherium africanum from southern Africa also lacks the entepicondylar foramen of the humerus, and the general postcranial morphology is similar to that of H. coffeyi, also suggesting a cursorial adaptation in this species.
FIG. 21.
Comparison of humerus morphology and ratios of Huracan coffeyi and related taxa. A. Ursus arctos, Alaska, AMNH M 135504. B. Arctodus simus, Alaska, AMNH F:AM 95656. C. Huracan coffeyi, Coffee Ranch, UCMP 31818. D. Agriotherium africanum, Langebaanweg, PQ-L 45063 (reversed; photos courtesy of A. Valenciano). E. Indarctos cf. I. oregonensis, Withlacoochee River 4A, Florida, UF 13799. F. Ailuropoda melanoleuca, AMNH M 147746. G. Panthera leo, Tanzania, AMNH M 85143. H, I. Ratios of humerus measurements in these taxa.
TABLE 11.
Measurements and ratios of mandibles in Agriotheriini. See text for abbreviations.
continued
continued
Other postcranial bones of Huracan coffeyi are less well preserved and provide little information relating to its locomotor behavior, but the relatively long neck of the astragalus and a moderately elongate cuboid are in accord with cursorial locomotor inferences based on the humerus. Among the few metapodials that can be used to calculate ip (index of plumpness) values (distal width/length × 100), a useful index with locomotor implications (Withalm, 2001), these values are distinctly smaller in H. coffeyi than in Indarctos oregonensis (tables 13, 14, 17, 18), and much smaller than in Ursus, Arctodus or Tremarctos (Kurtén, 1967; Lundelius, 1967; Withalm, 2001).
Together, these postcranial traits all point to a cursorial adaptation in Huracan coffeyi. Very few articulated limb bones of this species are known, but based on limited information (tables 14, 18), the radius/humerus length ratio and tibia/femur length ratio are 0.89 and 0.73, respectively. The ratios calculated from larger samples of pooled data for disarticulated elements (ratios of mean lengths of these bones) are similar (0.89 and 0.75, respectively), suggesting that these measures are reliable for the species. These two ratios suggest that the distal forelimb is slightly more elongate than in Ursus (mean of Ursus 0.86, n = 45), but almost identical in proportions in the distal hind limb (mean of Ursus 0.74, n = 58). In contrast, Ailuropoda has very short distal forelimb bones but a slightly elongated distal hind limb (0.78, n = 22 and 0.77, n = 19, respectively).
Arctodus simus lacks evidence of cursorial adaptations except for the overall elongation of its limb bones. The limb bones of Arctodus simus are proportionally longer than those of other bears, leading to a “gracile” appearance, however, it still overlaps with Ursus and the limb bones are stouter than in the large-bodied felids (Panthera). Well-developed humerus epicondyles and a wide range of ulna rotation support an inference that Arctodus was powerful and could subdue large prey, unlike the suggestion of Sorkin (2006) that this genus has a weak medial epicondyle and reduced development of the pronator teres muscle. We agree with Matheus (2003) that elongation of the limbs of Arctodus do not reflect high-speed short-distance locomotion, but instead evolved for increased locomotor efficiency during prolonged travel, and Arctodus should not be regarded as a cursorial bear as it often has been in the past. In contrast, Huracan (and Agriotherium) are cursorial bears, with elongated and gracile limb bones, and a humerus with weakly developed epicondyles. The late Hemphillian rhino Aphelops, and the proboscideans Gomphotherium, Rhynchotherium, and Mammut, all are probably too large and stout to have represented prey species for North American Huracan and, instead, the smaller rhino Teleoceras, the moderately cursorial large camelid Megatylopus, and the dromomerycid Pediomeryx probably were the preferred prey of Huracan.
EVOLUTION, DISPERSAL, AND PALEOECOLOGY OF AGRIOTHERIINI
Based on the new material and taxonomic revisions above, a close relationship of Indarctos and Agriotherium is further supported by analyses of the new taxon Huracan. A comprehensive phylogeny of Ailuropodinae is under preparation and will be published elsewhere when completed, but based on currently available evidence and a phylogenetic analysis of the Agriotheriini (fig. 20), all three genera can be assigned to Agriotheriini (as has been detailed above) within Ailuropodinae (based on diagnostic features like large cheek teeth, P4 with parastyle and enlarged inner lobe, wide M1 and M2, and high mandible as compared to the lower toothrow). As specialized carnivorous members of Ailuropodinae, Agriotheriini differ from the living panda Ailuropoda melanoleuca in many aspects, although both Agriotheriini and Ailuropoda share a large P4 with well-developed parastyle, large P4 inner lobe, square M1, large p4 with a frequently developed accessory cuspid, large m1, and proximally extended humerus lateral epicondylar crest. Since early fossil members of Ailuropodini are relatively smaller bodied, large size in later-diverging members of those clades is convergent.
Miomaci is established based on a fragmentary isolated jaw from Rudabánya, Hungary, and has been proposed as closely related to Indarctos (de Bonis et al., 2017). It is definitely an ailuropodine, as it has parastyle and the anterior border of the coronoid process is nearly perpendicular to the horizontal corpus; both of these features are quite common in Ailuropodinae, but absent or very rare in Ursinae or other stem ursids. It also possesses unique traits, e.g., short M2 talon (which is elongated in Indarctos vireti) and distinct p4 posterior accessory cuspid (absent in Indarctos, Agriotherium, and Huracan), which seems to suggest a separate evolutionary tendency. Its direct link with Indarctos, rather than, e.g., Ailuropodini, is not sufficiently supported based on available evidence. We suggest that this genus should not yet be placed taxonomically in any ailuropodid tribe, pending recovery or description of better material or more comprehensive analyses.
TABLE 12.
Measurements and ratios of the humerus in Agriotheriini. See text for abbreviations.
TABLE 13.
Measurements and ratios of the radius, ulna and the first metacarpal (MC1) in Agriotheriini. See text for abbreviations.
TABLE 14.
Measurements and ratios of metacarpals (MC2, 4, 5) and radius/humerus ratios, in Agriotheriini. See text for abbreviations.
TABLE 15.
Measurements and ratios of the femur and tibia in Agriotheriini. See text for abbreviations.
The morphology of Indarctos is closest to that of the earlier diverging Ursavus clade (Jiangzuo and Flynn, 2019), and Indarctos is the earliest appearing taxon among the three large Agriotheriini. The earliest species, Indarctos vireti, appeared in MN9 in Europe, and this small bear has a relatively long rostrum, slender mandible, elongated M2 talon, and no P4 parastyle (Crusafont Pairó and Kurtén, 1976; Jiangzuo et al., 2021). A slightly more anatomically derived species, Indarctos arctoides, has a small P4 parastyle and somewhat shorter M2 talon, although the skull and mandible also seem to be slender as in I. vireti (Montoya et al., 2001; Abella and Valenciano, 2017). The younger and more anatomically derived Turolian or equivalent-aged species of Indarctos are represented by I. lagrelii, I. atticus, I. zdanskyi, and I. oregonensis (Merriam et al., 1916; Zdansky, 1924; Montoya et al., 2001; Qiu and Tedford, 2003). Beginning at that time, Indarctos greatly expanded its geographic distribution to what is now termed the Holarctic + Oriental realms. These four Indarctos species share several anatomically derived traits relative to older species of Indarctos, including smaller anterior premolars, larger P4 parastyle, shorter M2 talon, and the skull and mandible become stouter. Therefore, evolution within this lineage through time, from basal Agriotheriini (“I.” vireti and “I.” arctoides) to Turolian and equivalent-aged Indarctos, is characterized by enlargement of the carnassials and reduction of the posterior molars, exemplifying increasing specialization for carnivory. Isotopic analysis of I. arctoides from Batallones 3 suggested that this species was largely carnivorous (Domingo et al., 2016), and therefore Turolian and equivalent-aged Indarctos species with similar dental morphologies very likely were carnivorous.
Agriotherium first appeared in Asia and North America during the late Miocene. The earliest records are Agriotherium palaeindicum from the Dhok Pathan Zone of the Siwaliks, southern Asia (Lydekker, 1884; Matthew, 1929; Colbert, 1935); A. inexpetans from Jiegou, in the Yangjiashan Fauna of the “Liushu” Formation, Linxia Basin, China, of eastern Asia (Qiu et al., 1991; Deng et al., 2013); and A. hendeyi from the Quiburis Formation, western North America (Jiangzuo and Flynn, 2019). These early species of Agriotherium were more or less contemporary with each other, and all are relatively small in size. Across the Mio-Pliocene boundary, Agriotherium subsequently arrived in Europe (A. insigne) and Africa (A. africanum) (Stehlin, 1907; Hendey, 1972, 1980). Agriotherium species quickly disappeared after their first occurrences in North America and Africa, but survived into the earliest Pleistocene in Eurasia (Flower, 1877; Lydekker, 1884; Nanda, 2008).
Huracan appeared at approximately the same time as Agriotherium. In the Linxia Basin, Huracan qiui occurs in the Qingbushan Fauna (Baodean, ∼7 Ma), which is slightly younger than the Yangjiashan Fauna (Latest Bahean, ∼8 Ma) that yields Agriotherium inexpetans. Slightly younger than H. qiui, the anatomically more derived species H. coffeyi occurred in North America, and H. roblesi appeared in Europe, suggesting the spread of this genus across the northern hemisphere at this time. Huracan disappeared in Europe soon thereafter, whereas it survived in North America until the middle Pliocene (Samuels et al., 2009). The full temporal distribution of Huracan in Asia remains unclear, but unpublished material from the Yushe Basin suggests that it survived into the earliest Pleistocene.
TABLE 16.
Measurements and ratios of the calcaneum and astragalus in Agriotheriini. See text for abbreviations.
Based on morphological and chronological evidence, it is clear that Indarctos is the oldest and earliest diverging taxon within the tribe Agriotheriini. Early Indarctos species, like Indarctos vireti and I. arctoides, are probably sister taxa to all younger Agriotheriini (fig. 20), thus likely rendering Indarctos paraphyletic when a more comprehensive phylogenetic analysis of the Agriotheriini and ailuropodid relatives is undertaken. During the Turolian (or possible late Vallesian), Indarctos dispersed to eastern Asia, and this paraphyletic lineage of early Agriotheriini species probably was ancestral to the clade containing Agriotherium and Huracan, as the earliest and anatomically most anatomically primitive species of both genera also were found in this region of Asia. In eastern Asia, Indarctos, Agriotherium, and Huracan coexisted in the Hezheng region during the late Late Miocene, although Indarctos seems to have outcompeted the other two genera in terms of abundance of individuals in any given fauna. Interestingly, at Baode, Shanxi Province, northeast of Hezheng, China, which also produced an extremely abundant assemblage of fossil Carnivora, only Indarctos has been found (Zdansky, 1924; Qiu and Tedford, 2003). In Lufeng and Yuanmou, southern China (late Bahean and Baodean ALMAs), only Indarctos existed (Qi, 2006). In southern Asia, Indarctos, Agriotherium, and possibly Huracan occur within the Dhok Pathan Zone (∼6-8 Ma) of the Siwaliks (Lydekker, 1884; Colbert, 1935; Jiangzuo et al., 2021), but the detailed chronology of occurrence of material of each of these taxa is unclear, leading to uncertainty as to whether all three genera coexisted here at any given time within the Dhok Pathan Zone. In Europe, Indarctos disappeared in MN13 (6.6–8 Ma) and Huracan (H. roblesi) appeared in MN13 (4.9–6.6 Ma) (Abella et al., 2019), but the two taxa never coexisted in the same fauna. Agriotherium arrived in Europe in MN14 (4.2–4.9 Ma) (Stehlin, 1907), thus these three genera (Indarctos, Huracan, and Agriotherium) do not overlap temporally in Europe. In North America, Indarctos appeared in the late Early Hemphillian (Hh2, 6.5–7.5 Ma), and went extinct by the end of Hh2, and Huracan and Agriotherium first occurred and coexisted in Hh3 (5.5–6.5 Ma). After Hh4 (4.6–5.5 Ma), only Huracan survived in North America, persisting until the middle Blancan NALMA (∼3.6 Ma). A summary of distributional changes through time for these Agriotheriini taxa in the North Hemisphere is shown in figures 22 and 23.
TABLE 17.
Measurements and ratios of metatarsals (MT1, 2) in Agriotheriini. See text for abbreviations.
TABLE 18.
Measurements and ratios of metatarsals (MT3, 4, 5), and relative tibia and MT3 lengths, in Agriotheriini. See text for abbreviations.
FIG. 23.
Geographic distribution pattern of Agriotheriini (Ailuropodinae, Ursidae) in the Northern Hemisphere through time (early Late Miocene to late Pleistocene). Artwork by Yu Chen.
FIG. 24.
Reconstructions of Huracan. Left Huracan qiui, and right Huracan coffeyi. Artwork by Qigao Jiangzuo.
Ecologically, Huracan and Agriotherium have more specialized dentitions, adapted to hypercarnivory, and the postcranial bones reflect greater cursoriality in locomotion relative to those of Indarctos species. Agriotherium advanced further than Huracan in terms of modifications to the dentition and cranial morphology. The temporal replacement of Indarctos species by Huracan and Agriotherium worldwide during the Mio-Pliocene is correlated with times of global cooling and aridification (Zachos et al., 2001; Westerhold et al., 2020), and the appearance of more open environments associated with those climatic conditions seems to have favored the evolution of more cursorial species. In North America, where faunal records are the most complete, the temporal replacement of Indarctos by Huracan and Agriotherium occurred during the transition from Early to Late Hemphillian (∼6.5–7 Ma) (Tedford et al., 2004). Marked faunal turnover occurred at this transition, as most typical “Miocene” components across many major clades, such as Hystricops, Leptodontomys, Leptarctus, Sthenictis, and Barbourofelis went extinct (Tedford et al., 2004). This coincides with C4 grassland expansion in North America, especially in its southern part (Cerling et al., 1997). Similar to the scenario of replacement of Indarctos species by Huracan and Agriotherium, Nimravides catocopis was fully replaced temporally by the more cursorial Amphimachairodus coloradensis at this turnover (Martin, 1998; Tedford et al., 2004; Antón et al., 2013; Jiangzuo and Hulbert, 2021).
Huracan and Agriotherium seem to overlap substantially with each other in ecomorphology, as both genera were likely hypercarnivorous and cursorial. In North America, Huracan survived longer than Agriotherium, whereas in Europe, Agriotherium survived longer than Huracan. Based on unpublished material from the Yushe basin and several other localities in China, both Huracan and Agriotherium survived into the earliest Pleistocene in eastern Asia.
CONCLUSIONS
Based on a large assemblage of new and newly studied fossil material from North America and north China, and extensive comparisons across species of agriotheriine bears, we erect a new genus of Agriotheriini (Ailuropodinae), Huracan, as reconstructed in figure 24. Four species are recognized in Huracan, including H. coffeyi (new combination) and H. schneideri (new combination, genotype species) in North America, H. qiui (n. sp.) in Asia, and H. roblesi (new combination) in Europe. The morphology of Huracan is intermediate between Indarctos species and Agriotherium in many features, such as an Indarctos-like m1 structure and P4 hypocone; Agriotherium-like P4 parastyle; and anatomically intermediate M2 traits (between those of Indarctos and Agriotherium); as well as autapomorphic traits of the upper incisors and lower p4, and was supported as a valid monophyletic taxon by our phylogenetic analyses of the Agriotheriini. Postcranial anatomy suggests that Huracan is one of the most cursorial crown ursids, together with its sister taxon Agriotherium. Huracan and Agriotherium appear to have originated in eastern Asia from an Indarctos-like ancestor, and dispersed across the northern hemisphere and into Africa afterward. Huracan and Agriotherium likely had similar paleoecologies and exhibited distinct spatiotemporal distribution patterns on different continents—Huracan survived longer in North America, Agriotherium survived longer in Europe, and both persisted contemporaneously late in their histories in eastern Asia.
ACKNOWLEDGMENTS
We thank Z. Qiu for allowing us to study the Chinese specimens and encouraging the study of systematics of North American Agriotheriini bear. We thank A. Valenciano from the Iziko South African Museum (Cape Town, South Africa) for providing photos of Agriotherium africanum, and P. Brewer and H. Taylor from the Natural History Museum (London) for providing the photos of Agriotherium sivalense. We thank Y. Chen for excellent artwork for Huracan, Agriotherium, and Indarctos. For their help in accessing collections, we also thank J. Meng, R. O'Leary, and J. Galkin (AMNH fossil mammal collections); M. Surovy, E. Hoeger, S. Ketelsen, N. Duncan, and N. Simmons (AMNH modern mammal collections); X. Wang and S. McLeod (LACM fossil collections); G. Takeuchi (LACMHC fossil collection of Rancho La Brea); P. Holroyd (UCMP fossil collections); R. Secord, R.M. Hunt, Jr., and G. Corner (UNSM fossil collections); B. MacFadden and R. Hulbert (UF fossil collections); A. Millhouse, D. Lunde and J.J. Ososky (USNM fossil and modern mammal collections); M. Omura (MCZ modern mammal collections); and Z. Qiu and J. Chen, R. Yang, W. He, S. Chen, and L. Zhang (fossil collections of the IVPP and HMV). Thanks for the valuable suggestions from J. Wagner and an anonymous reviewer, which greatly improved the quality of the paper. This work was supported by the Strategic Priority Research Program of Chinese Academy of Sciences (grant no. XDB26000000 and XDA20070203), Key Frontier Science Research Program of the Chinese Academy of Sciences (grant nos. QYZDYSSW-DQC-22 and GJHZ1885), National Natural Science Foundation of China (grant nos. 41430102, 41772018 41872001, 41872005, and 42102001), China Scholarship Council, and the Frick Fund, Department of Vertebrate Paleontology, Division of Paleontology, AMNH.
Copyright © American Museum of Natural History 2023
