Oligocene plotopterid skulls from western North America and their bearing on the phylogenetic affinities of these penguin-like seabirds

ABSTRACT Plotopterids are penguin-like, wing-propelled birds with controversial phylogenetic affinities. They are usually regarded as closely related to Suloidea (gannets, cormorants, and allies), with the penguin-like features considered to be of convergent origin. However, it has also been proposed that the similarities shared by plotopterids and penguins are due to common ancestry. An in-depth assessment of plotopterid affinities has been hampered by the fact that very little data about the skull of these birds were available. New fossils of Tonsala from the Oligocene Pysht Formation in Washington State (U.S.A.) include the first well-preserved cranial remains of this taxon. They show that although plotopterids share derived cranial features with members of Suloidea that are absent in species of Sphenisciformes (penguins), they lack diagnostic derived features of the representatives of crown group Suloidea. To assess the affinities of plotopterids, we performed a phylogenetic analysis that included, for the first time, early stem group representatives of Sphenisciformes, resulting in a sister-group relationship between Plotopteridae and Suloidea. Intriguingly, however, our reanalysis of the emended data of a more comprehensive recent analysis that supported a position of Plotopteridae within Suloidea recovered a sister-group relationship between Plotopteridae and Sphenisciformes. Although cranial morphology challenges the hypothesis of close affinities between plotopterids and penguins, more data on early stem lineage representatives of penguins are needed for a robust placement of Plotopteridae relative to Sphenisciformes. SUPPLEMENTAL DATA—Supplemental materials are available for this article for free at www.tandfonline.com/UJVP


INTRODUCTION
The extinct Plotopteridae were wing-propelled, flightless seabirds; their fossils have been found in late Eocene to early middle Miocene rocks of Japan and western North America (e.g., Howard, 1969;Olson and Hasegawa, 1979, 1985, 1996Olson, 1980;Goedert, 1988;Goedert and Cornish, 2002;Sakurai et al., 2008;Dyke et al., 2011). These birds appear to have been fairly diverse in the late Eocene/Oligocene Makah and Pysht formations of the Olympic Peninsula in Washington State, U.S.A. (Goedert and Cornish, 2002); however, only two species have been described from these localities, the Oligocene Tonsala hildegardae Olson, 1980, and the coeval but larger T. buchanani Dyke et al., 2011. Plotopterids possessed flipper-like wings that would closely resemble those of extant penguins (Sphenisciformes), a similarity that was attributed to convergence due to the functional constraints imposed on wing-propelled diving Hasegawa, 1979, 1996;Olson, 1980). Olson (1980) classified plotopterids into the taxon Suloidea, which includes Sulidae (gannets and boobies), Anhingidae (anhingas), and Phalacrocoracidae (cormorants). Mayr (2005Mayr ( , 2009, by contrast, hypothesized that Plotopteridae are the sister taxon of Sphenisciformes, and a phylogenetic analysis resulted in a sister-group relationship between the clade (Plotopteridae C Sphenisciformes) and Suloidea (Mayr, 2005). In order to account for the results of molecular analyses (see below), Mayr (2009) emended this hypothesis to a sister-group relationship between the clades (Plotopteridae C Sphenisciformes) and (Fregatidae C Suloidea).
The hypothesis of close affinities between Plotopteridae and Sphenisciformes was met with skepticism (Sakurai et al., 2008;Kawabe et al., 2014), and was not supported by a comprehensive analysis of 464 morphological characters (Smith, 2010) that resulted in a sister-group relationship between Plotopteridae and Phalacrocoracidae and Anhingidae. Virtual brain endocasts of plotopterids, by contrast, resemble those of penguins in overall shape and some derived features, including a very large flocculus (Kawabe et al., 2014).
At the time the hypothesis of sphenisciform affinities of plotopterids was presented (Mayr, 2005), the early history of penguins was little known. Since then, several early Paleogene stem group Sphenisciformes have been described, of which the two species of the taxon Waimanu from the early Paleocene of New Zealand are the oldest and earliest diverging representatives (Slack et al., 2006;Ksepka and Ando, 2011).
In the past few years, new molecular analyses have further provided a framework for the interrelationships of modern birds, although the affinities of penguins remain controversial (Fig. 1). Sphenisciformes have often been considered most closely related to Procellariiformes (tubenoses and allies) based on morphological analyses (for example, Livezey and Zusi, 2007;Ksepka and Ando, 2011). A clade including these two taxa has been supported by analyses of nuclear gene sequences (Hackett et al., 2008;McCormack et al., 2013;Yuri et al., 2013) and complete mitochondrial genomes (Gibb et al., 2013). Another analysis of complete mitochondrial genomes, however, supported a sister-group relationship between Sphenisciformes and Ciconiidae (Pacheco et al., 2011), but in that analysis only a limited sample of taxa traditionally considered members of 'Pelecaniformes' was included. Earlier analyses of molecular data resulted in a sister-group relationship between Sphenisciformes and a clade including Fregatidae and Suloidea (nuclear beta-fibrinogen sequences; Fain and Houde, 2004), or in a sister-group relationship between the clades (Phaethontidae C Sphenisciformes) and (Fregatidae C Suloidea) (Brown et al., 2008).
Aside from these uncertainties surrounding the early evolution and phylogenetic position of penguins, a well-founded assessment of the affinities of plotopterids has been impeded by the limited amount of information available on the cranial anatomy of these birds. Although plotopterid skulls from the late Oligocene of Japan were figured by Hasegawa et al. (1979) and Kawabe et al. (2014), these specimens have not yet been described, and the published figures do not allow the recognition of critical osteological details. The only previously described cranial material of North American plotopterids are mandible fragments of Tonsala (Goedert and Cornish, 2002;Dyke et al., 2011). Here we describe well-preserved partial skulls of Tonsala from Oligocene rocks of Washington State and revise the phylogenetic affinities of plotopterids. Furthermore, Waimanu and plotopterids are for the first time considered together in a phylogenetic analysis, and new scorings for plotopterids are added to the revised character matrix of Smith (2010) to evaluate their influence on the resulting tree topology.
We further added scorings for Tonsala to the analysis of Smith (2010), and corrected some erroneous scorings for other taxa included in that analysis (see Appendix S3). The new phylogenetic analyses were performed with the heuristic search modus of NONA 2.0 (Goloboff, 1993) through the WINCLADA 1.00.08 interface (Nixon, 2002), using the commands hold 10000, mult*1000, hold/10, and max*. The consistency index (CI) and retention index (RI) were calculated, as well as bootstrap support values with 1000 replicates, 10 searches holding 10 trees per replicate, and tree bisection and reconnection (TBR) branch swapping without max*.
Institutional Abbreviations-SMF, Senckenberg Research Institute and Natural History Museum Frankfurt, Germany; UWBM, Burke Museum of Natural History and Culture, University of Washington, Seattle, Washington, U.S.A. SYSTEMATIC PALAEONTOLOGY AVES Linnaeus, 1758PLOTOPTERIDAE Howard, 1969TONSALA Olson, 1980TONSALA HILDEGARDAE Olson, 1980 Referred Specimens-SMF Av 599 (Figs. 2-4; partial skull including caudal part of upper beak and mandibles, both ossa palatina, and ventral portion of right quadrate), collected in 1989 by J.L.G. SMF Av 600 (Figs. 5, 6; poorly preserved partial skull including braincase fragments, caudal portion of upper beak, and caudal portions of both mandibles [the right ramus mandibulae is broken and its caudal end is displaced]), collected in 1985 by J.L.G. SMF Av 601 ( Fig. 7; largely complete cervical vertebra and craniodorsal fragment of another cervical vertebra, left ulna lacking the distal end, damaged right femur, both tibiotarsi, and unidentified bones [?rib fragment and ?synsacrum fragments]), collected in 2008 by J.L.G.
Locality and Horizon-All specimens are float concretions that eroded from the beach terrace and low banks west of the mouth of Murdock Creek, south shore of Strait of Juan de Fuca, Clallam County, Washington, U.S.A. (GPS data: SMF Av 599: 48.1567 N, 123.8696 W;SMF Av 600: 48.1545 N, 123.8655 W;SMF Av 601: 48.1558 N, 123.8685 W); lower part of Pysht Formation, latest early Oligocene/early late Oligocene Prothero et al., 2001), or late Oligocene (Nesbitt et al., 2010).
Measurements ( Remarks on Taxonomy and Taphonomy-Our identification of the two skulls as belonging to Plotopteridae is supported by the great similarity of the specimens to figured but undescribed plotopterid skulls from Japan (Kawabe et al., 2014). No large birds other than plotopterids are known from the Pysht Formation, and the only plotopterid species of similar size belong to the taxon Tonsala. The two named plotopterids from the Pysht Formation differ in size, with Tonsala buchanani being about 1.3 times larger than T. hildegardae. With the exception of the T. buchanani holotype, all of the described plotopterid specimens from the Pysht Formation are from T. hildegardae, and based on size and morphology of the bones, SMF Av 601 can also be referred to this latter species (the length of the femur is 109.4, versus 106.5 in T. hildegardae and 134.2 in T. buchanani; Dyke et al., 2011). Assuming that T. hildegardae had similar proportions to the King Penguin (Aptenodytes patagonicus), the two skulls fit the bones of SMF Av 601 in size and are also assigned to T. hildegardae rather than to the much rarer and larger T. buchanani.
A revision of the plotopterid material from the Makah and Pysht Formations described by Dyke et al. (2011) is beyond the scope of the present study. We note, however, that of the specimens assigned to T. buchanani by Dyke et al. (2011), UWBM 86869 (Whiskey Creek specimen) and UWBM 86875 (holotype) are demonstrably as much as 5 million years different in age (Goedert and Cornish, 2002:fig. 2), with the Whiskey Creek specimen being possibly the oldest known plotopterid from North America (Goedert and Cornish, 2002). The bone labeled as a pterygoid of UWBM 86869 by Dyke et al. (2011: fig. 2A) is actually a vertebra, although a pterygoid is present and was listed by Goedert and Cornish (2002). Morphological differences between the distal ends of the humeri of UWBM 86869 and FIGURE 2. Partial skull with mandibles of Tonsala hildegardae from the Oligocene Pysht Formation of Washington State, U.S.A. (SMF Av 599) in A, B, dorsal, C, D, distal, E, ventral, F, right lateral, and G, left lateral views. Specimen in B coated with ammonium chloride to enhance contrast. In D, the matrix was digitally removed and the mandibles were brought in their correct anatomical positions. Abbreviations: ccv, caudodorsal portion of cranial cavity; cor, processus coronoideus; cvx, convex bulge on ventral surface of upper beak; gap, midline gap between ossa maxillaria; lmd, left ramus mandibulae; max, os maxillare; nas, nasal bar; nfh, nasofrontal hinge; nas, nasal bar; nos, nostril; nvs, longitudinal neurovascular sulci (along lateral surfaces of upper beak and mandible, respectively); pal, os palatinum; qdr, ventral portion of right quadrate; rmd, right ramus mandibulae; smd, marked fossa along ventral section of medial mandibular surface; sul, sulcus along tomium maxillare; tmd, tomium mandibulare; tmx, tomium maxillare; uib, unidentified rod-like bone. Scale bars equal 50 mm, same scale for A, B, and E-G.
another specimen that was referred to T. buchanani by Dyke et al. (2011), UWBM 86871 (Makah Formation, Jansen Creek Member, early? Oligocene), also indicate that these are likely different taxa (Goedert and Cornish, 2002). Dyke et al. (2011) acknowledged these differences, but nevertheless included UWBM 86869 and UWBM 86871 in T. buchanani. Morphological disparity of the distal humerus of the T. hildegardae holotype (Olson, 1980) and UWBM 86871 further seems to indicate a different taxon at the genus level. The ulna and scapula are preserved in the partial skeleton UWBM 86871 (Goedert and Cornish, 2002) and provide important additional data, but even though the proximal ulna was figured by Dyke et al. (2011: fig . 2M), it was identified as the proximal end of a radius, and the scapula was not mentioned by Dyke et al. (2011). Therefore, pending a thorough restudy of these fossils along with newly available specimens undergoing preparation, we currently recognize only the holotype specimen (UWBM 86875) as T. buchanani; the others that were referred to T. buchanani (UWBM 86869, 86870, and 87871) are herein regarded as Plotopteridae incertae sedis.
Like other bones from the same part of the Pysht Formation (Goedert et al., 1995;Kiel et al., 2010;Goedert and Cornish, 2002), the present fossils were heavily damaged prior to fossilization. Boreholes of the marine annelid Osedax, like those reported on other plotopterid bones from the Pysht Formation (Kiel et al., 2011), are visible on most bones of SMF Av 601. Notably, some of the bones of this specimen also exhibit distinct parallel scratches on the bone surface (Fig. 7R); these are not preparation artifacts and probably represent traces of marine scavengers. Similar scratches were noted on Osedax-damaged fossil whale teeth (Kiel et al., 2013) from this same part of the Pysht Formation. The nodule that held SMF Av 601 also contained a tooth of a cow shark (Hexanchidae), and we consider it possible that some of the breakage of the bones-albeit not the scratches on the bone surfaces-is due to the fact that sharks were feeding on the carcasses. In specimen SMF Av 600, fecal pellets of an invertebrate are preserved within the right antorbital fenestra, and these have also been found in association with other plotopterid bones from the Pysht Formation (Goedert and Cornish, 2002).
Description and Comparisons-The dimensions of the skulls correspond to those of a large albatross (e.g., Diomedea epomophora), thus surpassing the skull size of all extant suloideans. The beak is not completely preserved in the new specimens, but an undescribed fossil from Japan shows that it was proportionally longer in plotopterids than in most extant suloideans (Kawabe et al., 2014: fig. 1). The upper beak of Tonsala is dorsoventrally very low, as in frigatebirds and pelicans, with a wide nasal bar as in species of Pelecanidae, Fregatidae, and Suloidea, but unlike in sphenisciforms, in which the nasal bar is narrow (Figs. 2, 3). One of the most notable features is the presence of very long, slit-like nostrils, as in the early Eocene taxa Limnofregata (Fregatidae) and Prophaethon (Phaethontidae; Harrison and Walker, 1976). In extant species of Fregatidae and Suloidea, by contrast, the nostrils are reduced to very small caudal foramina or are completely absent (Sulidae), but a longitudinal furrow still denotes the fused narial openings. Unlike in extant suloideans and all other 'pelecaniform' birds, the caudal sections of the paired ossa maxillaria of Tonsala are not fused together medially, so that the ventral surface of the beak is less heavily ossified and a midline gap is visible (Fig. 2D), which is, however, relatively narrower than in sphenisciforms. The ventral surface of the upper beak is further not flat but has sulci along the tomia (as in frigatebirds and sulids, but unlike in phalacrocoracids and anhingas), and the medially adjacent part forms a markedly convex bulge (Fig. 2C, D). On the lateral side of the upper beak, there is a narrow neurovascular sulcus just above the tomium (Fig. 2F).
The well-preserved ossa palatina of SMF Av 599 (Figs. 2, 4) do not exhibit the unique and diagnostic derived shape of the palatines of extant suloideans, which form an essentially rectangular, dorsally flat platform (Fig. 3B). In overall shape, they more closely resemble the palatine bones of sphenisciforms than those of suloideans (Fig. 3), but there are some differences in detail. Whereas the palatines of suloideans are of equal width over their entire length, the ossa palatina of Tonsala widen caudally, with  . Fossil specimens were coated with ammonium chloride. Abbreviations: cdc, condylus caudalis, cdl, condylus lateralis; cdm, condylus medialis; cer, fossa of cerebral hemisphere; cho, choana; den, dorsal lamina of os dentale; fap, hooklike facies articularis pterygopalatini; fos, fossa on ventral surface of processus mandibularis; iom, internal ossification of Meckel's cartilage; jug, os jugale; nvs, neurovascular furrow along lateral surface of mandible; oss, ossicle on rostral end of os jugale; plt, plate-like bone ventromedially adjacent to jugal bar; pmp, processus maxillopalatinus; scl, scleral ossicle; soc, sulcus for occipital sinus; uib, unidentified rod-like bone; vpn, ventral process of os nasale. Scale bars equal 10 mm. the pars lateralis being wider than in suloideans and the angulus caudolateralis more pronounced (Fig. 3). Furthermore, the lamella choanalis (lamella dorsalis sensu Zusi and Livezey, 2006) extends over more than half of the length of the palate (exclusive of the pars maxillaris), whereas it is restricted to its rostral portion in suloideans. As in suloideans, the lamellae choanales are rostrally fused and do not form a processus choanalis or a cornu nasale palatini (terminology after Zusi and Livezey, 2006); a processus vomeralis is absent. Both ossa palatina are closely conjoined, but unlike in sulids and anhingas, they do not appear to have been fused in their caudal section, where a suture between the lamellae dorsales is visible. The dorsal surface of the palatines is sculptured; the ventral surface is slightly convex with the lateral portions being dorsally bent. As in sulids but unlike in anhingas and phalacrocoracids, the trough for articulation with the rostrum parasphenoidale is restricted to the caudal portion of the palatines. The processus pterygoideus is short, and the dorsoventrally wide and convex facies articularis pterygopalatini forms a distinct dorsal hook as in the species of Podicipediformes (Fig. 4A, B). Only the caudalmost portion of the crista ventralis (lamina ventralis choanalis sensu Zusi and Livezey, 2006) is visible in SMF Av 599. As in suloideans but unlike in frigatebirds and sphenisciforms, a vomer is absent (SMF Av 599).
In SMF Av 599, sheet-like processus maxillopalatini (processus palati maxillares sensu Zusi and Livezey, 2006) are preserved (Fig. 4D). The nasofrontal hinge (Figs. 2B, 6A) is a rostrocaudally extensive bending zone as in phalacrocoracids rather than a sharply delimited hinge as in sulids and anhingas. An os lacrimale is not preserved, and in SMF Av 600, the corresponding area of the skull does not show articulation facets for this bone (overall, however, this region of the skull is rather poorly preserved). In SMF Av 599, there is a small dorsal tuberosity on the rostral end of the right os jugale and an ossicle on the left one (Fig. 4C); whether this represents a homologue to the os suprajugale of suloideans is uncertain. A plate-like, elongated bone ventromedially adjacent to the jugal bar (Fig. 4D) is identified as a caudal process of the tomium of the upper beak, which is absent in extant suloideans. In caudal view into the beak, two ventral processes of the nasal bar are visible (Fig. 4D).
In SMF Av 599, only a fragment of the braincase is preserved, which is here tentatively identified as the caudodorsal portion of the cranial cavity. The fragment exposes the interior surface and two fossae can be discerned, which are separated by a marked sulcus (Fig. 4E). These structures are interpreted as the fossae for the two cerebellar hemispheres and the sulcus for the occipital sinus, which is very prominent in plotopterids (Kawabe et al., 2014). The braincase of SMF Av 600 is more complete, but the caudal portion is crushed and the poor preservation of the fossil does not allow the recognition of many osteological details (Figs. 5, 6). In its proportions, the braincase of Tonsala closely matches those of sulids, and as in the latter but unlike in sphenisciforms, the interorbital section is wide and there are no supraorbital fossae for glandulae nasales (Fig. 6). The processus postorbitales appear to have been short in SMF Av 600, but these processes may be broken because they are better developed in a plotopterid skull figured by Kawabe et al. (2014: fig. 1). In SMF Av 600, the cranial margins of the fossae temporales are preserved, and these are sigmoidally curved as in sulids (Fig. 6A); as shown by the undescribed skull figured by Kawabe et al. (2014: fig. 1), the temporal fossae of plotopterids were well developed. The crushed caudal portion of the cranium of SMF Av 600 is difficult to interpret owing to its poor preservation, and it is to be hoped that future radiograph investigations will shed more light on the identity of the bones. Clearly visible is the caudoventrally projected left processus paroccipitalis, as well as the dorsal rim of the left otic cavity. A circular foramen in the left caudolateral portion of the cranium (Fig. 5D) may represent the dorsal opening of the recessus tympanicus dorsalis. SMF Av 599 includes the ventral portion of the right quadrate. As in sulids but unlike in sphenisciforms, there is a deep fossa on the ventral surface of the processus mandibularis ( Fig. 4H-J). The condylus caudalis is mediolaterally broad as in sulids, but the condylus lateralis and the condylus medialis are less splayed. The medial surface of the condylus medialis is concave; unlike in sulids and members of crown group Sphenisciformes, however, there is no distinctly concave articulation facet lateral of this condyle.
The preserved portions of the mandibles agree with the mandible of sulids in proportions. The mandibular ramus deepens in the midsection, and the dorsal portion of its lateral surface exhibits a distinct longitudinal neurovascular sulcus, perpendicular to which dendrite-like smaller furrows diverge, as in phalacrocoracids and sulids (SMF Av 599; Fig. 4F). The medial mandibular surface bears a marked fossa along its ventral section (SMF Av 599; Fig. 2D), with such a fossa being also present in species of Fregatidae, Sulidae, Phalacrocoracidae, Gaviidae, and the late Eocene sphenisciform Perudyptes (Ksepka and Clarke, 2010). The dorsal surface of the tomium is ridge-like and not planar as in sulids (Fig. 2D). There are no fenestrae mandibulae. The os dentale is deeply forked caudally, with a well-developed dorsal lamina (sensu Zusi and Warheit, 1992: fig. 1). There also appears to have been an internal ossification of Meckel's cartilage as in extant frigatebirds and suloideans, in which this ossification is associated with an intraramal joint ( Fig. 4F; Zusi and Warheit, 1992). The processus coronoideus is well defined and marked (Fig. 2F). The articular end of the mandible is best visible in SMF Av 600 and closely resembles that of extant sulids. As in the latter, it is caudally truncate, with a concave caudal surface, and unlike in sphenisciforms, there is no caudally prominent processus retroarticularis and the processus medialis is developed as a dorsoventrally deep crest (this process is narrow and pointed in sphenisciforms; Fig. 64D-F). Also as in sulids, the cotyla lateralis is markedly concave and forms a distinct notch (Figs. 5F, 6H). On the ventral surface, there is a sharp ridge, which is medially bordered by a fossa (Fig. 6D).
In SMF Av 599, two rod-like bones are preserved (Figs. 2E, 4B), and even though these are situated in the appropriate anatomical position, they are too large and straight to be ossa ceratobranchialia of the hyoid apparatus. Likewise, the morphology of these bones does not suggest them to be jugal bars, and they are too long for sternal ribs.
The nodule containing the bones of SMF Av 601 also included a caudal cervical vertebra, which is probably vertebra 10, 11, or 12 (if compared with suloideans) or 7, 8, or 9 (if compared with sphenisciforms), as well as a fragment of the craniodorsal portion of another cervical vertebra (Fig. 7A-F). The more complete vertebra (Fig. 7A-D) lacks the zygapophyses craniales, ansae costotransversariae, and processus carotici, and some edges of the bone are broken or damaged. Because the bone was completely encased in matrix, all of the damage to this and the other bones in the nodule occurred prior to embedding in the sediment and may indicate action of a predator or scavenger (see above). The vertebrae are only slightly larger than the caudal thoracic vertebrae of Sula bassana and Aptenodytes patagonicus (Fig. 7G, H). They exhibit a rather generalized morphology and differ from the caudal cervical vertebrae of both sphenisciforms and suloideans. In the fragmentary vertebra, the articulation facet of the left zygapophysis cranialis is preserved and matches that of sphenisciforms (in suloideans this facet is less horizontally oriented). As in sphenisciforms but unlike in suloideans, the cranial rim of the arcus vertebrae is deeply concave. The processus spinosus forms a very low ridge. On the lateral surface of the corpus, there is a shallow, elongate fossa. The zygapophysis caudales are short, the lacuna interzygapophysialis is broadly rounded; the torus dorsalis is low, and there is no crista transverso-obliqua. A processus ventralis is absent. The ventral  surface of the corpus vertebrae is wide as in suloideans and not narrow as in sphenisciforms. The facies articulares cranialis et caudalis are heterocoelous, and the facies articularis cranialis is not as deeply concave as in sphenisciforms.
The ulna of Tonsala hildegardae was described by Olson (1980), and in its proportions and overall shape, the bone is remarkably similar to the ulna of Waimanu. The new specimen (SMF Av 601; Fig. 7I-K) lacks the distal end and the processus condylaris dorsalis. It matches the description of the T. hildegardae holotype except that the row of eleven pits for the attachment of feather quills is less strongly marked. The cotyla dorsalis has a slightly convex plane, as noted by Olson (1980), and is more elevated than the cotyla ventralis. The low and caudally situated olecranon forms a rounded knob; on its ventral surface, there is a marked fossa. The cranial surface of the shaft is rounded, whereas the caudal surface forms a ridge.
The femur (SMF Av 601; Fig. 7L, M) agrees well in size and morphology with a femur of T. hildegardae described by Goedert and Cornish (2002). The specimen is badly damaged, with the caput femoris and the medial portion of the proximal half of the shaft being broken. As already detailed by Goedert and Cornish (2002), the bone is much more elongated and slender than the femora of the plotopterid taxa Copepteryx and Hokkaidornis. The distal section of the shaft has a flat medial surface, which is oriented perpendicular to the cranial and caudal planes of the bone, so that the cross-section of the distal femur shaft is nearly rectangular. The sulcus patellaris is wide and shallow. Unlike in suloideans but as in sphenisciforms, the fossa poplitea is deeply excavated (Fig. 7N, O). The crista supracondylaris medialis is distinct and sharply defined.
Only fragmentary proximal ends of the tibiotarsi of Tonsala were previously known (Goedert and Cornish, 2002;Dyke et al., 2011). In SMF Av 601, both tibiotarsi are preserved, with the right one being nearly complete, except for some breakage on the proximal end ( Fig. 7P-V). The left tibiotarsus, however, is severely damaged, with extensive areas of the shaft being broken; the caudal surface of the shaft exhibits numerous parallel scratches ( Fig. 7R; see above). The marked length difference between the left and right tibiotarsi is due to the fact that the left bone is broken, with the fragments of the distal fourth being misaligned (as a result of preservation, not preparation). The tibiotarsus of Tonsala is proportionally more slender than that of Copepteryx, and is also much more elongated than the tibiotarsi of sulids. Most of the crista cnemialis cranialis is broken, but its preserved distal section is unusually well developed and reaches far down the tibiotarsus shaft, almost to the distal end of the crista fibularis. As in anhingas and phalacrocoracids, there is an elongate fossa on the lateral surface of the crista cnemialis lateralis. The proximal section of the shaft is very wide, with a flat to slightly concave cranial surface and a convex caudal surface; the medial part of the bone, that is, the distal section of the crista cnemialis cranialis, is markedly cranially slanted. The fossa flexoria, on the caudal surface of the proximal end, is very shallow. The distal end of the shaft exhibits a marked medial constriction, just above the condylus medialis (Fig. 7V). The distal end of the bone is medially inflected. There is a marked, cranially protruding embossment lateral of the pons supratendineus, which also occurs in some sphenisciforms but is absent in suloideans (Fig. 7V-X). The sulcus extensorius is wide, and as in suloideans the pons supratendineus is oriented obliquely to the longitudinal axis of the tibiotarsus; in sphenisciforms, this osseous bridge is wider and extends perpendicular to the axis of the bone. The proximodistally deep trochlea cartilaginis tibialis forms a prominent medial ridge. With regard to the craniocaudally flattened and wide proximal end, the tibiotarsus of Tonsala resembles that of suloideans, whereas it corresponds with the tibiotarsus of sphenisciform birds in the long crista cnemialis cranialis and the embossment lateral to the pons supratendineus.
Analysis of the revised and emended data matrix of Smith (2010) did not support the tree topology obtained by that author, that is, a position of plotopterids within Suloidea as sister taxon of the clade (Phalacrocoracidae C Anhingidae). Instead, the analysis resulted in 12 most parsimonious trees, with Plotopteridae being the sister taxon of Sphenisciformes (Fig. 8B).
We added 38 new scorings to the matrix of Smith (2010) and corrected 35 erroneous scorings of altogether 23 characters (see Appendix S3). A (Plotopteridae C Sphenisciformes) clade is, however, already obtained in half of the 24 trees resulting from an analysis in which only four character scorings are modified, with characters 4, 8, and 22 of Smith (2010) scored as absent for Tonsala, and character 198 scored as present for Eudyptula. Three of these characters (4,8,22) are derived features of suloideans and are absent in Tonsala, and character 198 is a derived feature shared by plotopterids and sphenisciforms, which, together with a few others (see Appendix S3), was erroneously scored variable for sphenisciforms by Smith (2010).
Smith's (2010) data set is more comprehensive than ours, but only includes crown group representatives of Sphenisciformes. In the resulting trees of the revised character matrix of Smith (2010), the clade including Plotopteridae and Sphenisciformes is the sister taxon of a clade including Gaviiformes and Podicipediformes. A clade including these three extant taxa is not supported by any current analyses of molecular data.
As noted by all previous authors, the osteology of plotopterids indicates close affinities to Suloidea Hasegawa, 1979, 1996;Olson, 1980;Mayr, 2005;Sakurai et al., 2008). In addition to some postcranial characters (Olson, 1980;Mayr, 2005, and above), plotopterids and suloideans share derived skull features, including a nasofrontal hinge, caudally projecting processus paroccipitales (Kawabe et al., 2014: fig. 1), loss of the vomer, and the presence of an internal ossification of Meckel's cartilage, which is also found in frigatebirds and tropicbirds (Zusi and Warheit, 1992). On the other hand, plotopterids lack diagnostic apomorphies of crown group suloideans. In particular, the palatine bones are not fused in their caudal sections and do not form a dorsally flat and subrectangular platform, the presence of which is a unique apomorphy of suloideans ( Fig. 3; Mayr, 2003). Furthermore, the nostrils are not reduced to tiny openings, the caudal sections of the maxillary bones are not ventrally fused (Fig. 3), and unlike in suloideans (including the extinct giant anhingas; Noriega, 1992;Alvarenga, 1995), the cristae iliacae dorsales of plotopterids are not fused with the crista spinosa of the synsacrum (Goedert and Cornish, 2002: fig. 7). Based on the above character evidence, we thus consider a position of Plotopteridae outside crown group Suloidea to be well supported, and the derived similarities shared with Phalacrocoracidae and Anhingidae, such as a very large patella, opisthocoelous vertebrae, and laterally expanded alae praeacetabulares iliorum, are here regarded to be of convergent origin.
The new data challenge the hypothesis of a sister-group relationship between Plotopteridae and Sphenisciformes (Mayr, 2005), because penguins lack all of the above-listed derived features shared by plotopterids and suloideans. We note, however, that there exists a possibility that character states were reversed into the primitive condition in penguins due to paedomorphosis (see Mayr, 2005), and in some osteological aspects, such as the shape of the palatines, plotopterids more closely resemble penguins than the taxa of the Suloidea.
An ultimate appraisal of the affinities between Plotopteridae and Sphenisciformes depends on a more detailed study of the osteology of Waimanu and other early representatives of Sphenisciformes as well as on a robust higher-level phylogenetic placement of extant penguins based on molecular data. Overall, the osteology of Waimanu is more similar to that of plotopterids than to that of Gaviiformes or Procellariiformes, but there are some distinct differences in detail. The extremitas omalis of the furcula of Waimanu, for example, exhibits the derived shape of sphenisciforms and lacks a well-developed facies articularis acrocoracoidea, which is found in plotopterids and suloideans. Unlike plotopterids and crown group sphenisciforms, the corpus of the scapula of Waimanu is not greatly widened (Slack et al., 2006: fig. 1A [j, k]), and the distal end of the tarsometatarsus of Waimanu is distinguished from that of plotopterids in that the trochlea metatarsi II is not markedly longer than the trochlea metatarsi IV, with the latter furthermore not being asymmetric in dorsoplantar view. The shape of the Waimanu coracoid is different from that of the elongate coracoid of plotopterids. The upper beak of the late Eocene stem group sphenisciform Icadyptes resembles that of plotopterids in the very long and slitlike nostrils and the presence of distinct vascular impressions on the lateral surfaces (Ksepka et al., 2008), but because the bill morphology of Waimanu and that of other Paleogene sphenisciforms is poorly known, the significance of these features is difficult to assess.
If sphenisciform affinities of plotopterids can be upheld, these birds most likely are the sister taxon of a clade including Waimanu and other sphenisciforms, which share well-developed dorsal fossae for glandulae nasales, a derived shape of the extremitas omalis of the furcula, and, probably correlated therewith, a narrow, hook-like extremitas omalis of the coracoid. Because the scapula of Waimanu is, however, not greatly widened as in other sphenisciforms and plotopterids, widening of the scapula must then have occurred convergently in members of Plotopteridae and Sphenisciformes other than Waimanu.
If, on the other hand, the diving adaptations of plotopterids and penguins are of convergent origin, they would not only testify to a remarkable adaptive radiation within Aequornithes, the 'waterbird clade,' to which penguins and Suloidea belong (Mayr, 2011), but also very divergent adaptations would have evolved in closely related taxa, that is, wing-propelled diving in Plotopteridae and Sphenisciformes and long-distance soaring in Fregatidae and the procellariiform Diomedeidae.
In any case, recognition of a sister-group relationship between Plotopteridae and Suloidea allows a reconstruction of the ancestral state of some morphological features of suloideans. In particular, long open nostrils are likely to be plesiomorphic for a clade including Fregatidae, Plotopteridae, and Suloidea and were closed at least twice independently in Fregatidae and in Suloidea (if Phaethontidae are indeed the sister taxon of Fregatidae and Suloidea, a third reduction of the nostrils occurred in tropicbirds). Limnofregata lived in a lacustrine palaeoenvironment (Olson, 1977), and closure of the nostrils in the stem lineage of Fregatidae may be related to the transition into a marine environment. Plotopterids, however, were highly specialized seabirds, and why open nostrils are present in these birds but were reduced in members of crown group Suloidea remains an open question.