Pygmy Sperm Whales (Odontoceti, Kogiidae) from the Pliocene of Florida and North Carolina

ABSTRACT 
 We describe fossil kogiid periotics from the Lower Pliocene upper Bone Valley Formation in central Florida and the Lower to Upper Pliocene Yorktown Formation at Lee Creek Mine, North Carolina. The fossils show diagnostic characters that identify them as belonging to Kogiidae, such as three spines in the anterior process, presence of an incudal process, and a posterior process oriented along the long axis of the bone. Morphological comparisons and morphometric and statistical analyses of periotic proportions confirm the presence of a large and a small morphotype within the sample. The large morphotype (mean length = 39.76 mm) belongs to an unknown kogiid that occurs in both formations, whereas the small morphotype (mean length = 28.64 mm), referred to aff. Kogia sp., occurs only in the Yorktown Formation. The cooccurrence of two taxa in North Carolina may represent one of the earliest evidences of sympatry in kogiids and may demonstrate that this ecological behavior has been part of the natural history of this group at least since the deposition of the Yorktown Formation at Lee Creek (∼4.8–3.1 Ma). In addition, the occurrence of the large morphotype in the upper Bone Valley Formation, herein reported for the first time, shows that we are still far from understanding the diversity of marine mammals of that formation and that revision of newly acquired material is necessary.


INTRODUCTION
The Palmetto Fauna of the upper Bone Valley Formation in Florida and the Yorktown Formation at Lee Creek Mine in North Carolina are considered two of the most diverse Pliocene marine mammal deposits (Morgan, 1994;Tedford et al., 2004;Ward and Bohaska, 2008). Identification of material from these formations is usually based on diagnostic cranial material, primarily skulls, mandibles, and ear bones (e.g., periotics), and, to a lesser extent, postcranial elements (e.g., Kaz ar and Bohaska, 2008). The marine mammal fauna of the upper Bone Valley Formation was reviewed by Morgan (1994), who listed a total of 17 taxa that included cetaceans, sirenians, and pinnipeds. In addition, Whitmore and Kaltenbach (2008), Kohno and Ray (2008), and Koretsky and Ray (2008) revised the cetaceans and pinnipeds of the Yorktown Formation at Lee Creek where they listed a total of 34 taxa.
The Bone Valley and Yorktown formations share taxa at a higher taxonomic level; however, they also present many unique occurrences (Appendix 1). Because of the inherent uncertainties of the fossil record, these unique occurrences could be either true absences or the result of sampling bias. New discoveries and additions to these taxonomic lists will enhance our knowledge of early Pliocene marine mammal communities, ultimately allowing us to better understand changes in mammal diversity in deeptime and their biogeographic implications.
During a recent revision of the fossil material from the upper Bone Valley Formation at the Florida Museum of Natural History, the senior author of this work (JVJ) noted the presence of kogiid periotics, which have not been previously reported from this formation. Though kogiid periotics are known from the Yorktown Formation at Lee Creek (Luo and Marsh, 1996;Whitmore and Kaltenbach, 2008), they are in need of detailed revision. Here we describe kogiid periotics from both the Palmetto Fauna (5-4.7 Ma: Morgan, 1994;Webb et al., 2008) of the Lower Pliocene upper Bone Valley Formation in central Florida and from the slightly younger Lower to Upper Pliocene (»4.8-3.1 Ma: Snyder et al., 1983) Yorktown Formation at Lee Creek Mine, North Carolina. Furthermore, by using traditional morphological comparisons, morphometric observations, and statistical analyses, we identify two different cooccurring morphotypes at Lee Creek Mine. The presence and cooccurrence of these morphotypes potentially provide evidence of earliest record of sympatry in kogiids, showing that this type of biogeographic distribution has been part of the natural history of this group at least since the Pliocene.
The lateral (ventrolateral) tuberosity is located lateral to the mallear fossa; it is low (»2 mm), blunt, and subtriangular to oval in outline; its smooth surface indicates that the lateral tuberosity did not contact the sigmoid process of the tympanic bulla as it does in Physeter and Zarhachis (Luo and Marsh, 1996). The hiatus epitympanicus is short and shallow, with some specimens showing little to no break between that surface and the posterior bullar facet. The articulation with the incus is marked by a raised circular area, the incudal process ( Fig. 1E-H), located toward the medial extent of the hiatus epitympanicus; this raised area contrasts with the fossa observed in other odontocetes but is typical of kogiids (Luo and Marsh, 1996). A deep (»2 mm), round (»4 mm in diameter) mallear fossa is located posterior to the accessory ossicle and medial to the lateral tuberosity. A shallow groove extends along the posterior and posteromedial borders of the mallear fossa in some specimens (e.g., UF 203525). The pars cochlearis has a rounded outline in ventral view. The fenestra rotunda is located on the posteroventral surface of the pars cochlearis; it is large (»5 mm high by 3 mm wide) and separated by 4 mm from the small (»2 mm in diameter), posterodorsally located perilymphatic foramen (D cochlear aqueduct).
The posterior process is oriented posteriorly along the long axis as the rest of the bone; it is large relative to the total size of the periotic, with a posterior bullar facet that is, on average, about 40% of the total length of the periotic and has a total width that equals about 70% of the maximum width of the periotic (Table 1). The posterior process has a ventromedially oriented convex posterior bullar facet, with its medial edge upturned, thus differing from other physeteroids displaying a more posteroventrally oriented posterior bullar facet; fine grooves are present posteriorly on this otherwise smooth surface. In some specimens the length and width of the posterior bullar facet are subequal, giving it a plate-like shape ( Fig. 1), similar to the periotics referred to Scaphokogia cochlearis (Muizon, 1988:fig. 37; Fig. 4A, C, E) and some of the periotics described from the Yorktown Formation (Whitmore and Kaltenbach, 2008:fig. 80). These rounded, plate-like posterior bullar facets are unlike the shorter, more rectangular surfaces seen in Kogia and the Yorktown periotic described by Luo and Marsh (1996: fig. 1). The dorsolateral and dorsal surfaces of the posterior process are relatively smooth and flat, in contrast to the more recurved and rugose surface in physeterids (e.g., Aulophyseter morricei Kellogg, 1927). The rim of the internal acoustic meatus is raised and the meatus is somewhat funnel shaped. The height of the transverse septum within the internal acoustic meatus is variable, in some specimens being lower (e.g., UF 135933) or nearly as high (e.g., UF 203525) as the rim of the meatus; nonetheless, it clearly separates the facial canal from the cochlear foramina ( Fig. 1A-D). The ventral foramen of the facial canal is round (<3 mm in diameter) and continuous, with a groove whose medial edge forms a ridge that separates these from the stapedial fossa. The fenestra ovalis is located medial to the ventral foramen of the facial canal, has smooth rounded edges, and its long axis is oriented anteromedial to posterolateral. The stapedial fossa is rounded in outline (»5 mm in diameter) and deep, with Abbreviations: ads, anterodorsal spine; ao, accessory ossicle; ap, anterior process; avs, anteroventral spine; ds, dorsal spine; dt, dorsal tuberosity; ef, endolymphatic foramen; fc, facial canal; fo, fenestra ovalis; fr, fenestra rotunda; fs, foramen singulare; iaf, inferior acoustic foramen; iam, internal acoustic meatus; ip, incudal process; mf, mallear fossa; pc, pars cochlearis; pf, perilymphatic foramen; pp, posterior process; sf, stapedial muscle fossa; smf, suprameatal fossa; ts, transverse septum; vlt, ventrolateral tuberosity. well-demarcated edges, and is located posterior to the fenestra ovalis and posterolateral to the fenestra rotunda, separated from the latter by a low caudal tympanic process.
Both the perilymphatic and endolymphatic foramina are close to the internal acoustic meatus, being separated from each other by an approximately 1-mm bony septum. The endolymphatic foramen has an oval outline (»4 mm wide by »2 mm long) and is located between the perilymphatic foramen and the internal acoustic meatus. The dorsal tuberosity is prominent (being the highest point on the dorsal surface), blunt, and located posterolateral to the internal acoustic meatus and endolymphatic foramen (Figs. 1A-D, 2). A concave area just lateral to the internal acoustic meatus is likely the suprameatal fossa (Luo and Marsh, 1996).
Remarks-Whitmore and Kaltenbach (2008) described, in part, the large morph from the Yorktown Formation at Lee Creek Mine. The overall morphology of the Yorktown periotics does not differ markedly from those of the upper Bone Valley. However, we do note that mean and median values (Table 1) and the results of our morphometric analyses (see below) indicate slight differences between periotics from the respective formations, potentially representing distinct but closely related (congeneric) species. KOGIA Gray, 1846 aff. KOGIA sp. indet. (Fig. 3A, B, D, E, G, H, J-K; Tables 1, S4) Kogiinae genus indeterminate, Luo and Marsh, 1996:331. Kogiinae incertae sedis, Whitmore and Kaltenbach, 2008:235 (in part).
Locality-The material was collected from spoil piles at Lee Creek Mine, located on the south side of Pamlico River, Aurora, Beaufort County, North Carolina (see location of mine in Ward and Blackwelder, 1987: fig. 1).
Remarks- Luo and Marsh (1996) provided a very thorough description of this morphotype. The specimens differ from the other kogiid periotics from the Yorktown and Bone Valley by their smaller size (Table 1, Fig. 5), being on average about 10 mm shorter, and the more rectangular and mediolaterally narrower outline of the posterior bullar facet. Most of the characteristics that make this morphotype unique are shared with Kogia breviceps and K. sima and are why we tentatively refer it to aff. Kogia sp.

Morphological Comparison
Both periotic morphotypes studied show several characteristics diagnostic of kogiids, such as the presence of three spines, dorsal, anterodorsal, and anteroventral, on the anterior process; a raised incudal process; and a posterior process that is oriented along the long axis of the bone and is not recurved ventrally (Muizon, 1988;Luo and Marsh, 1996;Lambert, 2008). The large morphotype from Bone Valley and Yorktown compares best with the referred periotic of Scaphokogia cochlearis in size and overall morphology (Muizon, 1988; Table 1, Fig. 4A, C, E). Though differing from Yorktown aff. Kogia sp., K. sima, and K. breviceps in its larger size (Figs. 1-4; Tables 1, S1; Kasuya, 1973:99), the large morphotype further differs by exhibiting a disproportionately larger posterior process with an oval or rounded outline of the posterior bullar facet and, with respect to Kogia sima and K. breviceps, has a less triangular outline of the cochlear portion in ventral view (Figs. 1E-H, 3F). A recently described kogiid periotic from the late Tortonian of Malta (Bianucci et al., 2011) differs from the Bone Valley material in the unique morphology of the posterior process, which is shallowly concave and posterolaterally directed relative to the long axis of the rest of the periotic.
The second periotic morphotype, found only in the Yorktown Formation, resembles those of Kogia sima and K. breviceps in its small size and overall morphology ( Fig. 3; Luo and Marsh, 1996). However, the Yorktown aff. Kogia differs by having a less recurved and dorsoventrally thinner anterior process, more inflated body of the periotic, and more rounded pars cochlearis ( D more inflated pars cochlearis sensu Luo and Marsh, 1996; Fig. 3).
There are other records of Kogia from Pliocene deposits. Pilleri (1986,1987) described several periotics from the Pliocene of Italy and referred them to extant taxa. Some of these periotics and a tympanic bulla were later restudied by Bianucci (1996), who referred to them only as Kogia sp. Based on their larger size, some of these Italian periotics are more similar to the large morph from Bone Valley and Yorktown but do share morphological similarities with Kogia as reported by Bianucci (1996: fig . 2). In addition to the periotics, there is a skull from Piacenzian-age deposits that represents an extinct species, Kogia pusilla (Pilleri, 1987); it is the earliest record of the genus known from cranial material (Bianucci and Landini, 1999). Unfortunately, no periotics were associated with the skull.  The third Pliocene record of Kogia comes from the late Zanclean Tirabuzon Formation in Baja California Sur. Barnes (1998) mentioned the presence of periotics representing aff. Kogia and cf. Scaphokogia sp. as part of his Santa Rosalia Local Fauna. Upon reexamination of the material by one of us (JVJ), it is evident that all of the Tirabuzon material represents a single large morph (e.g., LACM 143481: Fig. 4B, D, F). In general, the morphology and size of the Tirabuzon kogiid periotics cannot be readily differentiated Abbreviations: ads, anterodorsal spine; ao, accessory ossicle; ap, anterior process; avs, anteroventral spine; ds, dorsal spine; dt, dorsal tuberosity; ef, endolymphatic foramen; fc, facial canal; fo, fenestra ovalis; fr, fenestra rotunda; fs, foramen singulare; iaf, inferior acoustic foramen; iam, internal acoustic meatus; ip, incudal process; mf, mallear fossa; pc, pars cochlearis; pf, perilymphatic foramen; pp, posterior process; sf, stapedial fossa; smf, suprameatal fossa; tc, transverse crest; vlt, ventrolateral tuberosity. from the referred periotic of S. cochlearis or from any of the large morphs from the Bone Valley or Yorktown formations examined in this study and is better referred to Kogiidae gen. et sp. indet.
Other fossil kogiids from the Bone Valley and Yorktown localities include Kogiopsis floridana Kellogg, 1929 and Aprixokogia kelloggi Whitmore and Kaltenbach, 2008. Although fragmentary, the type material of K. floridana (cast LACM 123975) represents a taxon of a size larger than any known kogiid and more consistent with physeterids. Therefore, we agree with other workers (e.g., Barnes, 1973;de Muizon, 1988) who consider K. floridana a physeterid instead. The stem kogiid Aprixokogia kelloggi was described based on a nearly complete skull, with no periotics associated with it, which limits our comparison with this species. Additional fossil kogiids, such as Thalassocetus antwerpiensis Abel, 1905, Praekogia cedrosensis Barnes, 1973, andNanokogia isthmia Velez-Juarbe et al., 2015, are known only from cranial material lacking associated periotics.

Morphometric Analysis
Qualitative comparisons of the Bone Valley and Yorktown kogiid periotics (n D 178) and assessment of the measurements presented in Table 1 and complete measurements of each periotic reported in the Supplementary Material (Tables S2-S4, Fig. S1) conservatively indicate the presence of two morphotypes in the sample. We conducted morphometric analyses, using the statistical software R (R Development Core Team, 2012) to further test whether the fossil periotics actually reflect the presence of two distinct taxa. First, we performed principal component analysis on linear measurements (centered and geometrically scaled) from the 172 complete fossil periotics and 16 periotics from extant taxa to detect groupings along statistically significant principal component (PC) axes. Results from pair-wise Anderson tests (Zelditch et al., 2012) show that the eigenvalues associated with the first two PC axes are significantly greater than subsequent eigenvalues (x 2 D 158.06, df D 2, P < 0.001 and x 2 D 27.17, df D 2, P < 0.001), indicating that PC1 and PC2 are aligned along biologically significant dimensions of variance.
Along PC1, the periotics separate into a group with highly negative values (i.e., small morphotype and Kogia spp.) and a group of more positive values (large morphotype; Fig. 5A). The highest PC1 loadings correspond to the width and length of the posterior process (0.64 and 0.47, respectively), indicating that the large morphotype differs significantly from the small morphotype by exhibiting a disproportionately larger posterior process. The grouping of taxa along the PC2 axis is based primarily on the proximal thickness of the anterior process (proximal thickness of the periotic anterior process loading D 0.93). The noticeable separation of extant Kogia spp. along PC2 suggests that the morphological variation represented by this axis is restricted to the species level.
We also used a Welch's two-sample t-test (a D 0.05) to compare ratios and measurements between the distinct morphotypes  Fig. 5B). Similarly, the Mahalanobis distance between the multivariate means of the two fossil morphotypes is greater than that of the extant species (3.51 and 1.76, respectively). Consequently, we hypothesize that the differences between the fossil morphotypes large and small from the Yorktown Formation do not reflect an ontogenetic series or sexual dimorphism within a single species. Instead, these morphotypes potentially indicate the presence of two sympatric species of kogiids during the early Pliocene.

DISCUSSION
Previous work had pointed out the presence of at least two kogiids in the Yorktown Formation at Lee Creek, based on periotics (Luo and Marsh, 1996;Whitmore and Kaltenbach, 2008) and postcranial elements (Kaz ar and Bohaska, 2008). Here, we have confirmed the presence of these two morphotypes in Lee Creek, as well as identified one of them (the large morph) in the Palmetto Fauna of the upper Bone Valley Formation in Florida. The Yorktown periotics represent two different taxa, potentially different genera (one referred to aff. Kogia), and is interpreted as one of the earliest evidences of sympatry in kogiids. However, because the fossils from Lee Creek Mine come from spoil piles and therefore lack a well-resolved stratigraphic context, we consider sympatry as a tentative scenario. The newly noted presence of kogiids in the upper Bone Valley shows that we are still far from understanding the diversity of marine mammals of the Palmetto Fauna and that a revision of material acquired in more recent years is needed. Unfortunately, until diagnostic cranial material associated with any of these morphotypes is collected, we cannot determine their true affinities and know whether they belong to any of the Pliocene kogiids known from more diagnostic cranial material, such as Aprixokogia kelloggi from the Yorktown Formation. Alternatively, it is possible that the large morph reported here actually represents A. kelloggi, because it is the largest of the two better known Pliocene-age kogiids from the Atlantic-Mediterranean regions, whereas the small morph  represents Kogia pusilla. Nonetheless, our work highlights the importance of understanding morphological variation in less diagnostic and/or more fragmentary elements, such as periotics and postcranial elements, because they can uncover a greater degree of diversity than that represented by well-preserved cranial material (e.g., monodontids in the Yorktown Formation; Kaz ar and Bohaska, 2008; V elez-Juarbe and Pyenson, 2012). Our methodology should provide a basis to study other taxa represented in collections by large numbers of periotics, such as other material from the Yorktown Formation or from the middle Miocene Sharktooth Hill Bonebed (Pyenson et al., 2009).

CONCLUSIONS
Our description and quantitative comparisons of the kogiid periotics from the early Pliocene Palmetto Fauna of the upper Bone Valley Formation, Florida, and Yorktown Formation, North Carolina, identify significant differences in morphology, size, and shape amongst the specimens and confirm the presence of two morphotypes. The large morphotype, which we designate as an indeterminate taxon of Kogiidae, is found in both Bone Valley and Yorktown, whereas the second, small morphotype is found only in the latter and represents an unknown species with similarities to Kogia. In addition, our reexamination of other published Pliocene kogiid periotics from North America revealed that previous reports of aff. Kogia and cf. Scaphokogia from the Tirabuzon Formation in Baja California Sur represent a single taxon, best considered as Kogiidae gen. et sp. indet.
The occurrence of kogiids in the Palmetto Fauna of the upper Bone Valley Formation represents a new record, expanding the diversity of marine mammals in the early Pliocene of Florida. The presence of two species of kogiids in the Yorktown Formation potentially represents the earliest evidence of sympatry in kogiids and indicates that this type of biogeographic distribution was likely present at least since the time of deposition of the Yorktown Formation at Lee Creek (»4.8-3.1 Ma) and thus is not restricted to extant taxa.