Introduction

Although extant Anseriformes appear to occupy a relatively narrow range of dominantly aquatic ecologies, both extinct and extant taxa show a broad range of locomotor and feeding modes. Extinct ecotypes exhibited in the comparatively rich Cretaceous-Paleogene fossil record of this clade include the terrestrial, giant flightless birds within Gastornis [1,2, vide 3–5] with dorsoventrally broad beaks and debated diets [6–8]; the marine albatross-like Pelagornithidae with pseudotooth projections for piscivorous diets [9–11]; and the more aquatic, duck-like species such as the wide-billed, but long-legged wader Presbyornis [12–14]. Extant Anseriformes include the largely terrestrial Anhimidae that are primarily herbivorous, semiaquatic foragers; Anseranas, an aquatic surface swimmer and grazer that is primarily herbivorous; and a variety of ecotypes within Anatidae [15,16]. Within Anatidae, taxa possess both herbivorous and omnivorous diets with various degrees of specialization; different filter feeding modes comprising grazing, mixed feeders, and diving graspers; and a variety of swimming ecologies that comprise a terrestrial ecology, foot propelled swimming, wing propelled swimming, surface swimming, plunging, and foot and wing propelled swimming [17–19]. Diving, as either an escape behavior or in feeding, occurs throughout Anatidae but not in Anhimidae and possibly not in Anseranas [16,20,21].

Despite the ecological breadth of anseriform ecologies and an abundance of recovered extinct taxa, important questions remain regarding the phylogeny and the ecological and behavioral evolution of Anseriformes. It is debated when within stem or crown Anseriformes they evolve more aquatic ecologies, especially as vestigial rhamphothecal lamellae and semipedal webbing were proposed to be present in extant Anhimidae [22]. It is similarly uncertain how herbivory and beak shape evolved within this group. Simulated trait evolution has supported two main patterns of diversification of beak shape and related diet in crown Anseriformes: either a single evolutionary trajectory or several independent and parallel transitions to a narrower, more herbivorous “goose-like” beak are estimated [19]. Phylogenetic placement of extinct anseriform taxa is necessary to inform evolution of these traits in both the stem and the crown.

Morphological analyses containing extinct taxa have resulted in drastically differing topologies [11,13,14,23–28]. Additionally, specimens that represent or are referrable to several extinct taxa have not been included or fully captured by scorings in previous analyses, resulting in a higher level of missing data in these analyses. At the same time, fine-grained molecular analyses of extant taxa have produced conflicting results within Anatidae [29–31]. Robust placement of extinct taxa is critical as ancestral state reconstructions are optimized differently depending on variations in extinct taxon placement. Better understanding the evolutionary relationships and ecological evolution of Anseriformes and their stem lineages is also necessary for better understanding early avian evolution and biogeography. New Paleogene fossils and robust placement of these and previously described taxa within a phylogenetic context is needed to better understand the answers to these questions.

The predominantly lacustrine Green River Formation of North America preserves an exceptional snapshot of early Eocene diversity in North America [32], but has only produced two aquatic taxa to date: Presbyornis, an anseriform, and Limnofregata Olson 1977 [33–35], likely an extinct relative of living frigatebirds [33,34]. Two semiaquatic taxa have also been recovered: an ibis-like taxon [36] and Messelornis nearctica, a stem ralloid [37,38]. Here we describe a new aquatic avian taxon from the early Eocene Fossil Butte Member (FBM, 51.97 ± 0.16 Ma [39]) of the Green River Formation. The taxon was originally illustrated and suggested to be anseriform in [32]. We recover the taxon as a member of Anseriformes. Most known Paleogene Anseriformes have a wide, duck-like beak (i.e. Presbyornis, Nettapterornis [40]), narrow pseudotoothed or dorsoventrally broad beaks (eg. Pelagornis [41] and Gastornis, respectively), or largely unknown beak morphology (i.e. Conflicto antarcticus [28]; Vegavis iaai [23,42,43]). The only other known exceptions to this are the latest Paleocene Anachronornis anhimops Houde et al. [44], the early Eocene Danielsavis nazensis Houde et al. [44], and the Middle Eocene Perplexicervix microcephalon Mayr [45]. Like the latter taxa, the new fossil presents a narrow bill that is most similar to those of the Anhimidae but differs from all extant Anseriformes. We identify this fossil as the holotype specimen of a new species. New x-ray computed tomography (CT) images allow recovery of previously hidden morphologies, revealing it to represent a new taxon and ecology for the Green River Formation. The specimen was found at a near-shore locality of the Fossil Butte Member of the Formation, locality H [32,46], where several lithornithid and neoavian fossils have been previously described [32,47].

We additionally present a new morphological dataset for extinct and extant Galloanseres that includes more exemplars of extinct and extant Anseriformes than any previous matrix. Creation of this dataset included the examination and scoring of specimens of Presbyornis and Telmabates antiquus [48] that have not been included in previous studies. This included assessment of all Presbyornis material housed in the Vertebrate Paleontology Collection of the Smithsonian National Museum of Natural History and all Telmabates material housed in the Vertebrate Paleontology Collection of the American Museum of Natural History. We perform parsimony analysis on the new morphological matrix, Bayesian analysis on a matrix that combines the new morphological matrix with molecular data, and ancestral state reconstruction of ecological and behavioral traits. Results provide new insights into outstanding issues of avian phylogeny, evolution and diversification.

Institutional abbreviations

AMNH, American Museum of Natural History, New York, NY, U.S.A.; FMNH, Field Museum of Natural History, Chicago, IL, U.S.A.; TMM, the Texas Memorial Museum, Austin, Texas, U.S.A.; USNM, National Museum of Natural History, Smithsonian Institution, Washington, D.C., U.S.A. Specimen numbers are presented in Table 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Specimen numbers of skeletal specimens used for comparison during fossil description and phylogenetic analyses.

https://doi.org/10.1371/journal.pone.0278737.t001

Systematic paleontology

AVES Linnaeus, 1758 sensu [61]

NEOGNATHAE Pycraft, 1900 sensu [61]

ANSERIFORMES Wagler, 1831 [62]

Paakniwatavis grandei, gen. et sp. nov.

Holotype specimen

FMNH PA725, a partial skeleton and tracheal rings preserved in a kerogen-poor laminated micrite slab (Figs 1 and 2). Measurements are provided in Table 2. Most of the vertebrae are absent or obscured where present. The shoulder girdle, thoracic vertebrae, ribs, pelvis, femora, synsacrum, caudal vertebrae and pygostyle have been eroded due to taphonomic processes. It appears that bacteria-induced or some other organic erosion of the bone has occurred. This type and extent of organic erosion is unique within recovered avian fossils from FBM. Similar taphonomy has been reported in an early Cretaceous Enantiornithine [63], the early Cretaceous Microraptor gui [64], and several other Jurassic and Cretaceous avialan theropods [65,66], although much less erosion and deformation of the bone has occurred in these specimens compared to that of the holotype specimen of P. grandei. It has been suggested that this taphonomic phenomenon is due to changes in matrix chemistry caused by water being trapped between the feathers and the body [64]. Scanning electron microscopy and additional analysis of the holotype specimen of P. grandei is necessary to determine the cause of this rare taphonomy.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Photograph (A) and line drawing (B) of the holotype specimen of Paakniwatavis grandei (FMNH PA725).

Extremely crushed bone and bone margin is delimited with dashed margins. Anatomical abbreviations: prx, premaxilla; orb, orbital margin; mnd, mandible; cvt, cervical vertebrae; tvt, thoracic vertebrae; syn, synsacrum; pyg, pygostyle; cor, coracoid; scp, scapula; fur, furcula; str, sternum; rbs, ribs; hum, humerus; uln, ulna; rad, radius; rde, radiale; cmc, carpometacarpus; mII:1, phalanx 1 of manual digit II; mtII:2, phalanx 2 of manual digit II; ili, ilium; fem, femur; tbt, tibiotarsus; tmt, tarsometatarsus; mtI, metatarsal I; I:1, phalanx 1 of pedal digit I; II:1, phalanx 1 of pedal digit II; III:1, phalanx 1 of pedal digit III; IV:1, phalanx 1 of pedal digit IV.

https://doi.org/10.1371/journal.pone.0278737.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Photograph (A) and line drawing (B) of the holotype specimen of Paakniwatavis grandei (FMNH PA725).

Bone is unfilled. Extremely crushed bone and bone margin is delimited with dashed margins. Anatomical abbreviations: prx, premaxilla; max, maxilla; jug, jugal; orb, orbital margin; rmf, rostral mandibular fenestra; scl, scleral ossicles; mnd, mandible; rde, radiale; cmc, carpometacarpus.

https://doi.org/10.1371/journal.pone.0278737.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Selected measurements of Paakniwatavis grandei in millimeters (mm), taken from surface of the holotype specimen slab (left/right) compared with taken and previously published measurements of Presbyornis [22,52], Telmabates antiquus [48], and Nettapterornis oxfordi [54].

Pedal phalanges are described using the format (digit:phalanx). Measurements are given for holotype specimens only.

https://doi.org/10.1371/journal.pone.0278737.t002

The holotype specimen was scanned using dual tube x-ray computed tomography at the PaleoCT Lab at the University of Chicago, which can scan specimens with a resolution of up to 0.4 mm. As the specimen slab was large, it was scanned using a two-part multiscan that was combined to form one image sequence. The voxel size of the combined scan is 103.8710. The specimen is housed in the Department of Geology of FMNH. CT data generated during the current study are available in the Supplementary Data via DRYAD under DOI 10.5061/dryad.v15dv4208, and the rest of the Supplementary Data is available via Morphobank [67] under Project 4001 (http://morphobank.org/permalink/?P4001).

Diagnosis

Paakniwatavis grandei is diagnosed by a proposed unique combination of characters comprising (1) a mediolaterally narrow rostrum (character 3:state 1; Figs 1 and 3), (2) an elongate retroarticular process (Fig 3; 241:2, 254:2), (3) a dorsoventrally thick furcula (Figs 1 and 3; 424:2), (4) thoracic vertebrae that are not solely heterocoelous (286:2), (5) presence of a supracoracoid nerve foramen (Fig 4; 391:1), (6) lack of a spur on the carpometacarpus (509:1), (7) femora that are half the length of the tibiotarsi (Fig 1; 597:1), (8) presence of a prominent tubercle laterodistal to the pons supratendineus of the tibiotarsus (643:1), (9) tarsometatarsi that are just over half the length of the tibiotarsi (Fig 1; 656:1), (10) a medial hypotarsal crest that is projected farther plantar than the lateral crest (Fig 4; 668:1), and (11) a deep sulcus extensorius of the tarsometatarsus (Fig 1; 686:2). Diagnosis for the genus as per the species.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Comparison of anseriform traits in Paakniwatavis grandei (FMNH PA725) to those of the extinct Presbyornis and extant Anseriformes (Chauna 21orquate, Anseranas semipalmata, and Chloephaga melanoptera).

Photograph (A) and CT scan slice (B) of the holotype specimen of P. grandei (FMNH PA725), with close-ups (B1 and B2) of CT scan slices of the caudal mandible. Photographs of the skull and mandible of (C) C. 21orquate, (D) A. semipalmata, I C. melanoptera, and (F) Presbyornis sp. (USNM 299846). Photograph (G) and CT scan slice (G1) of the left tarsometatarsus and pedal phalanges of P. grandei. Photographs of (H) the left tarsometatarsus and pedal phalanges of Presbyornis sp. (USNM ACC 335940) and (I) the right tarsometatarsus and pedal phalanges of 21orquateata. Photographs of the furcula of (J) 21orquateata, (K) A. semipalmata, (L) Anas platyrhynchos, and (M) Presbyornis sp. (USNM ACC335940). Photograph (N) and CT scan slice (O) of the furcula of P. grandei from the holotype specimen. Anatomical abbreviations: prx, premaxilla; orb, orbital margin; mnd, mandible; rta, retroarticular process; rde, radiale; cmc, carpometacarpus; sym, symphysis; tmt, tarsometatarsus; I:1, phalanx 1 of pedal digit I; II:1, phalanx 1 of pedal digit II; III:1, phalanx 1 of pedal digit III; III:2, phalanx 2 of pedal digit III; IV:1, phalanx 1 of pedal digit IV.

https://doi.org/10.1371/journal.pone.0278737.g003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Comparison of the coracoid and hypotarsus of the tarsometatarsus of Paakniwatavis grandei (FMNH PA725) to that of the extinct Presbyornis, Telmabates antiquus, Chaunoides antiquus and Nettapterornis oxfordi and extant Anseriformes.

Line drawings of the coracoids and hypotarsi of (A) Anhima cornuta, (B) C. antiquus, (C) Anseranas semipalmata, (D) Dendrocygna guttata, (E) Presbyornis sp., (F) P. grandei, (G) T. antiquus and (H) A. oxfordi. Coracoids are depicted in dorsal aspect, hypotarsi are depicted in proximal and plantar aspects. L and M in panel (A) denote the lateral and medial sides of each element. (F1 and F2) are segmented hypotarsi of the left tarsometatarsus of P. grandei in lateroplantar aspect. (F3 and F4) are CT scan slices of the same tarsometatarsus of P. grandei. Anatomical abbreviations: acr, processus acrocoracoideus; hmf, humeral facet; sct, scapular facet; prc, procoracoid process; fns, supracoracoid nerve foramen; pno, pneumatic opening; lpr, lateral process; mpr, medial process; mhc, medial hypotarsal crest; 1, sulcus for tendon of musculus flexor hallucis longus (fhl); 2, sulcus or canal for tendon of musculus flexor digitorum longus (fdl); 3, sulcus for tendon of musculus flexor perforatus digiti 2 (fp2); 4, sulcus for tendon of musculus flexor perforans et perforatus digiti 2 (fpp2); 5, sulcus for muscularus fibularis longus (fbl).

https://doi.org/10.1371/journal.pone.0278737.g004

Differential diagnosis

The rostrum shape exhibited by this taxon is unlike most other previously recovered Paleogene Anseriformes. Due to this Paakniwatavis grandei is easily distinguished from Nettapterornis, which has a mediolaterally wide, duck-like bill that is mediolaterally wider than the width measured from the left paroccipital process to the right paroccipital process. Paakniwatavis in contrast possesses a mediolaterally narrow bill that is narrower than the width between its parocciptal processes and, in this feature, is more similar to extant Anhimidae (Figs 1 and 3). The tomial margin of the bill of Nettapterornis is similarly dorsoventrally thick and recurved, unlike the straight and dorsoventrally narrow facial margin of the bill in Paakniwatavis (Fig 3). The nares of Paakniwatavis are over half the length of the rostrum, whereas that of Nettapterornis is less than half of their rostral length. The coracoid in Paakniwatavis is more elongate relative to the width of the sternal facet than that of Nettapterornis (Fig 4). The rami of the furcula are extremely thick in Paakniwatavis (Fig 1) compared to the thin furcula of Nettapterornis (Fig 3). The sulcus ligamentosus transversus is more truncate in Paakniwatavis than in Nettapterornis. The fossa olecrani of Paakniwatavis is more shallow than that of Nettapterornis. Paakniwatavis has a deeper fossa infratrochlearis of the carpometacarpus than Nettapterornis. The craniocaudal lengths of manual digits II and III at the synostosis are subequal in Paakniwatavis, whereas II extends further distally than III in Nettapterornis.

As with Nettapterornis, Paakniwatavis can be differentiated from Presbyornis due to Presbyornis possessing a mediolaterally wide, duck-like bill that is mediolaterally wider than the width measured from the left paroccipital process to the right paroccipital process. The tomial margin of the bill of Presbyornis is again like that of Nettapterornis and dorsoventrally thick and recurved, unlike the straight and dorsoventrally narrow facial margin of the bill in Paakniwatavis (Fig 3). Similarly, the nares of Paakniwatavis are over half the length of the rostrum, whereas those of Presbyornis are less than half of their rostral length. The crista tympanica of the quadrate terminates within the ventral half of the quadrate in Paakniwatavis, whereas it terminates within the dorsal half in Presbyornis. The tuberculum subcapitulare is separated from the squamosal capitulum in Paakniwatavis, but is contiguous with the capitulum in Presbyornis. The relative heights of the rostral and caudal apices of the coronoid process of the mandible are subequal in Paakniwatavis, but the rostral apex is higher in Presbyornis. The mandibular ramus caudal to the rostral fenestra mandibulae is deep and concave along the medial face in Paakniwatavis, and is relatively shallow in Presbyornis. A mandibular ventral angle is prominent in Presbyornis, but not in Paakniwatavis. While the fenestra rostralis mandibulae is slit-like in Paakniwatavis, it is transverse and largely perforate in Presbyornis (Fig 3). The synsacral count of Paakniwatavis is within 14–19 vertebrae, whereas it is within 10–13 for Presbyornis. In the coracoid, a small pneumatic foramen directly below the scapular cotyla is present in Presbyornis but is absent in Paakniwatavis and Nettapterornis. The rami of the furcula are again extremely thick in Paakniwatavis compared to the thin furcula of Presbyornis. The acromion process of the scapula is truncate unlike the cranially elongate processes of Presbyornis. The dorsal angle of the scapula is caudal to the midpoint of the shaft in Presbyornis, but is at the midpoint in Paakniwatavis. The fossa pneumotricipitalis of the humerus is pneumatic in Paakniwatavis and Nettapterornis but apneumatic in Presbyornis. Paakniwatavis has a narrower crista deltopectoralis of the humerus than Presbyornis. Paakniwatavis has a deeper impressio coracobrachialis cranialis than Presbyornis, The sulcus ligamentosus transversus is more truncate in Paakniwatavis than in Presbyornis. The fossa m. brachialis is located more medially in Paakniwatavis. The fossa olecrani of Paakniwatavis is again more shallow than that of Presbyornis. The depression radialis of the ulna is deeper in Paakniwatavis. The labrum dorsalis of the carpometacarpus is more sharply angled in Paakniwatavis. The epicondylus medialis of the tibiotarsus is more pronounced in Paakniwatavis than in Presbyornis. The epicondylus medialis is less pronounced than those of Presbyornis or Chaunoides antiquus. The sulcus extensorius tibiotarsus opens under the pons supratendineus along the midline in Paakniwatavis rather than medially. The tarsometatarsus is approximately half the length of the tibiotarsus or less in Paakniwatavis, whereas the length of these elements is subequal in Presbyornis. The lateral cotyle of the tarsometatarsus is more shallow in Paakniwatavis. In Paakniwatavis, pedal digit IV: phalanx IV is longer than IV: III, whereas the opposite is true in Presbyornis.

When compared to extinct taxa, the narrow bill of Paakniwatavis is most like those of Anachronornis anhimops and Danielsavis nazensis. The rostrum of Paakniwatavis can be distinguished from that of Danielsavis as the nares of Paakniwatavis do not extend as far anteriorly into the rostrum as those of Danielsavis do. The coronoid processes of the mandible are tuberculate and subtle compared to those of Danielsavis (Fig 2, Mayr et al. [60]). The rami of the furcula are much thicker in Paakniwatavis compared to the thin furcula of of Danielsavis. The acromion process of the scapula is truncate and unlike the cranially elongate processof Danielsavis. The acrocoracoid process of the coracoid is more sharply hooked than that of Danielsavis (Fig 5, Mayr et al. [60]). The shaft of the coracoid of Paakniwatavis is wider and more robust compared to that of Danielsavis. The impressio m. sternocorocoidei is much deeper in Paakniwatavis compared to the shallow impression in Danielsavis (Fig 4). A robust additional medial projection above the medial angle that is part of the origin of the sternocoracoclavicular ligament is present in Danielsavis but absent in Paakniwatavis. The length of the ulna is longer than that of the humerus in Danielsavis, whereas in Paakniwatavis the ulna is slightly more truncate than the humerus. Danielsavis has a robust index process that extends beyond the articular facet of manual digit II:1 which is not present in Paakniwatavis.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Resulting consensus tree from Bayesian analysis of 719 morphological characters and 158,368 base pairs.

Clade credibility values greater than 50% are annotated above branches. Extinct taxa are delimited with daggers. A line drawing of the holotype specimen of Paakniwatavis grandei is overlaid on the tree and a reconstruction of this species is shown to the right of the tree. Icons represent definitive ancestral state reconstruction of the earliest transitions. Only the first major transitions for Anseriformes are shown, with subsequent transitions excluded. The insert to the right of the tree displays how results changed when Wilaru was included in analysis.

https://doi.org/10.1371/journal.pone.0278737.g005

The angulus mandibulae of the mandible is more truncate in height compared to that of Anachronornis (Fig 2, Mayr et al. [60]). The rami of the furcula are much thicker in Paakniwatavis compared to the thin furcula of Anachronornis. The shaft of the coracoid of Paakniwatavis is again wider and more robust compared to that of Anachronornis. The impressio m. sternocorocoidei is much further excavated in Paakniwatavis compared to the shallow impression in Anachronornis (Fig 4). A robust additional medial projection above the medial angle is again present in Anachronornis but absent in Paakniwatavis. Paakniwatavis has a deeper impressio coracobrachialis cranialis than Anachronornis. The sulcus ligamentosus transversus of the humerus is deeper in Paakniwatavis than in Anachronornis. The head of the humerus in Anachronornis is triangular unlike the rounded head in Paakniwatavis and is projected much further proximally than that of Paakniwatavis.

As in Anachronornis and Danielsavis, Paakniwatavis can be differentiated from Perplexicervix due to the shaft of the coracoid of Paakniwatavis being wider and more robust in Paakniwatavis, and again Paakniwatavis possesses a more deeply excavated impressio m. sternocorocoidei (Fig 4). Additionally, the supracoracoid nerve foramen of Perplexicervix is smaller and located along the midline of the shaft, whereas it is larger and located further medially in Paakniwatavis.Paakniwatavis can be distinguished from Telmabates due to the number of synsacral vertebrae present. The synsacral count of Paakniwatavis is within 14–19 vertebrae, whereas it is within 10–13 for Telmabates as it is in Presbyornis. Telmabates has a small pneumatic foramen directly below the scapular coracoid like Presbyornis, and again this is absent in Paakniwatavis. The acromion process of the scapula is again truncate and unlike the cranially elongate processes of Telmabates. The dorsal angle of the scapula is caudal to the midpoint of the shaft in Telmabates, but is at the midpoint in Paakniwatavis. The incisura capitis of the humerus is deep in Paakniwatavis but shallow in Telmabates. The fossa pneumotricipitalis of the humerus is pneumatic in Paakniwatavis but apneumatic in Telmabates. Paakniwatavis has a deeper impressio coracobrachialis cranialis than Telmabates, The sulcus ligamentosus transversus of the humerus is more truncate in Paakniwatavis than in Telmabates. The epicondylus medialis of the tibiotarsus is more pronounced in Paakniwatavis than in Telmabates. Paakniwatavis has a dome-like crista that overhangs the proximal margin of the fossa pneumotricipitalis of the humerus, whereas this crista is typical in Telmabates. The impressio m. pectoralis is deeper in Paakniwatavis. The trochlea carpalis of the carpometacarpus is deeper in Paakniwatavis. The epicondylaris medialis depression is more shallow in Paakniwatavis than in Telmabates.

Paakniwatavis can be differentiated from Chaunoides Alvarenga 1999 due to differences in the coracoid and tarsometatarsus. Within the coracoid, the primary axis of the scapular cotyla is skewed laterally in Paakniwatavis, Presbyornis, Telmabates and Nettapterornis but is centralized in Chaunoides (Fig 4). The sternal facet is skewed ventrally in Chaunoides, whereas it projects caudally in Paakniwatavis. In the tarsometatarsus, the major hypotarsal ridge in Paakniwatavis is hooked distally. The epicondylus medialis of the tibiotarsus is more pronounced in Paakniwatavis than in Chaunoides.

Materials and methods

Results

Morphological results denote the results that were acquired from parsimony analysis of the morphological data, whereas the combined results comprise results of Bayesian analysis of the combined morphological and DNA data. Paakniwatavis grandei Musser and Clarke sp. nov. urn:lsid:zoobank.org:act:05722EDF-EDF1-4B2C-87A0-493D97DC7B3D is recovered across all analyses as the sister taxon of Anseranatidae+Anatidae (~50% bootstrap value in both morphological results, 85% clade credibility in both combined results (Figs 5, S1 and S2 in Supplementary Data via 10.5061/dryad.v15dv4208).

Across both the parsimony analysis of morphological data and the Bayesian analysis of combined data, addition of Wilaru results in differing topologies and causes nonmonophyly of Galloanseres. Due to this, the monophyly of Galloanseres was minimally constrained using a molecular backbone constraint when Wilaru was included in morphological analysis, and the following results discussed with Wilaru included comprise a monophyletic Galloanseres (S1 and S2 Figs, Supplementary Data via 10.5061/dryad.v15dv4208).. In the morphological results, Wilaru is placed as the sister taxon of Gastornis within a stem anseriform clade that also contains the Pelagornithidae (<50% bootstrap support; S2 Fig, Supplementary Data via 10.5061/dryad.v15dv4208). Addition of Wilaru also changes the relationships within the clade containing Vegavis iaai, Telmabates, Presbyornis, Conflicto, and Nettapterornis 32xford that is sister to Anatidae recovered across all analyses. With Wilaru excluded, Telmabates and Presbyornis are placed as sister taxa and form the sister clade to a group comprising Vegavis(Conflicto+Nettapterornis), whereas inclusion of Wilaru results in Vegavis being placed as the sister taxon of a clade containing Telmabates(Presbyornis(Conflicto+Nettapterornis). When Wilaru is included, bootstrap support scores rise to 99% per node within the group Presbyornis(Nettapterornis+Conflicto). Additional topology changes when Wilaru is included in the morphological analyses include Dendrocygna guttata being placed as the sister taxon of Thalassornis leuconotus and differences in the placements of several additional crown anatids (Mergus serrator, Netta rufina, Cygnus atratus and Coscoroba coscoroba; S1 and S2 Figs, Supplementary Data via 10.5061/dryad.v15dv4208).

When a Galloanserine molecular backbone constraint was not enforced, Danielsavis nazensis was recovered as the sister taxon of Anachronornis anhimops (77% bootstrap support) and placed within a group comprising Pelagornithidae(Gastornis(Anachronornis+Danielsavis) (<50% bootstrap support; S4 Fig, Supplementary Data via 10.5061/dryad.v15dv4208). This clade was placed within a polytomy containing Ichthyornis dispar and the rest of included Aves. Perplexicervix microcephalon was recovered as the sister taxon to Pan-Anseriformes (<50% bootstrap support). When Perplexicervix was excluded, Danielsavis was again placed as the sister taxon of Anachronornis (83% bootstrap support); however, the clade including these taxa, the Pelagornithidae and Gastornis collapsed into a polytomy containing these taxa (S5 Fig, Supplementary Data via 10.5061/dryad.v15dv4208). When Galloanserine molecular backbone constraints were enforced, Danielsavis, Anachronornis and Perplexicervix were recovered in a polytomy with Ichthyornis and the rest of included Aves (S6 Fig, Supplementary Data via 10.5061/dryad.v15dv4208). Exclusion of Perplexicervix did not change these results except that Anachronornis and Danielsavis were recovered as sister taxa (77% bootstrap support; S7 Fig, Supplementary Data via 10.5061/dryad.v15dv4208).

In the combined results, addition of Wilaru further defines the base of included Galliformes recovered in the morphological analyses (Gallinuloides wyomingensis(Macrocephalon maleo+Megapodius freycinet)) and collapses the base of a clade containing Telmabates, Presbyornis, Nettapterornis, Conflicto, Vegavis and Wilaru that is sister to Anatidae into a polytomy (94% clade credibility; S3 Fig, Supplementary Data via 10.5061/dryad.v15dv4208. Both combined results recovered a Presbyornis(Nettapterornis+Conflicto) clade, with posterior probability values for this clade rising when Wilaru is included. When Wilaru is excluded, Vegavis is placed as the sister taxon to a Telmabates(Presbyornis(Nettapterornis+Conflicto)) group, as in the morphological results with Wilaru included, although the clade credibility of this recovered group is low (85% or less, see Figs 5 and S3, Supplementary Data via 10.5061/dryad.v15dv4208). These drastically different placements across results from both the combined and morphological datasets are likely due to the lack of a skull known for Wilaru, although it does suggest that Wilaru may not be a presbyornithid as previously hypothesized in De Pietri et al. [58]. Due to this, we will largely focus on discussing analyses that exclude Wilaru.

Results from Bayesian analysis of the combined data matrix with Wilaru excluded are presented in Fig 5. Placement of extant taxa in the combined results are consistent with the resulting phylogram of Sun et al. [31] based on mitochondrial genomes; however, this tree included relatively few anseriform taxa used in our analysis. Combined results remain largely consistent with those of the phylogram of Sun et al. [31] based on two mitochondrial genes, which contains almost all extant anseriform taxa used in our analysis. Differences arise in the placement of several non-goose taxa within crown Anatidae. We recover a clade containing Oxyura+Stictonetta (Anas+Amazonetta(Netta(Mergus(Tadorna+Chloephaga)))) as being sister to the goose clade Cygnus+Coscoroba(Branta(Anser+Anser)), whereas Sun et al. [31] recovers a Mergus(Tadorna+Chloephaga(Netta(Anas+Amazonetta))) group that is sister to Oxyura+the goose clade. Differences between these analyses are likely due to our inclusion of morphological data, mitochondrial genomes for more taxa and nuclear genes. Additionally, the Sun et al. [31] analysis did not include Stictonetta, Coscoroba, or Thalassornis, and our analysis did not include several genera included in Sun et al. [31] (eg. Neochen, Melanitta, Aythya and more).

Within the combined results (Fig 5), clade credibility values of all nodes were 94% or higher with several exceptions. Exceptions comprise placement of Asteriornis maastrichtensis as the sister taxon of a Tinamidae+Lithornithidae clade (54%), placement of Crax as the sister taxon of a Lophura+Gallus gallus sister group (69%), placement of Paakniwatavis grandei as the sister taxon to Anseranas semipalmata+(a fossil clade+crown Anatidae) (85%), placement of Anseranas as the sister taxon to a a clade that contains extinct taxa and crown Anatidae (89%), placement of Telmabates as the sister taxon to Prebyornis(Conflicto+Nettapterornis) (59%), placement of Presbyornis (55%), placement of Nettapterornis as the sister taxon of Conflicto (88%), and placement of Netta rufina as the sister taxon to a clade containing Mergus(Tadorna+Chloephaga) (85%).

The strict consensus tree from analysis of morphological data is broadly congruent with combined results but presents alternative hypotheses for the placement of several extinct and extant taxa within Pan-Anatidae, the name we are giving to the clade that includes crown Anatidae and the clade of extinct taxa that was placed as the sister taxon to crown Anatidae across all analyses (S1 Fig, Supplementary Data via 10.5061/dryad.v15dv4208). Unconstrained parsimony analysis of morphological data resulted in one most parsimonious tree (MPT) of 2,821 steps (CI = 0.291, RI = 0.580, RC = 0.169, HI = 0.709). Morphological analysis recovers a clade containing Telmabates+Presbyornis(Vegavis(Conflicto+Nettapterornis)) as the sister taxon to Anatidae. This clade containing the same taxa is structured as follows within the combined data results: Vegavis(Telmabates(Presbyornis(Conflicto+Nettapterornis))). The positions of several extant anatids differ across the analyses as well. In the morphological results, Dendrocygna is sister to Thalassiornis+crown Anatidae, whereas Dendrocygna and Thalassiornis are sister taxa in the combined data results. Additional differences within the morphological results comprise the placements of Netta, Oxyura, Stictonetta, the Chloephaga+Tadorna sister group, and Mergus. Cygnus and Coscoroba are additionally no longer sister taxa in the morphological results.

Analysis of the combined dataset and morphology only dataset resultin Gastornis (Pelagornis+Protodontopteryx) as a stem anseriform clade (<50% bootstrap support, 99% clade credibility;(Figs 5 and S3, Supplementary Data via 10.5061/dryad.v15dv4208). Both datasets also recover a clade containing Telmabates, Presbyornis, Vegavis, Conflicto, and Nettapterornis (<50% bootstrap support, 95% clade credibility) as the sister taxon of Anatidae. Recovery of this clade containing solely extinct taxa may be real or discovered to be a paraphyletic assemblage; all synapomorphies for this clade exhibited CI≤0.5 in the morphological results and CI≤0.5 in the combined data results (see Supplementary Data via 10.5061/dryad.v15dv4208).

Ancestral state reconstruction using the combined dataset indicates that P. grandei likely preferred an aquatic or semi-aquatic environment, was either not specialized for aquatic feeding modes (hereafter referred to as a “non-swimmer”) or was a surface swimmer, was primarily herbivorous, was a grazer, had some form of rhamphothecal lamellae, had a pedal digit III that was subequal to or longer than that of the tarsometatarsus, had an ossified pessulus of the syrinx, and did not have asymmetry at the tracheobronchial juncture of the syrinx. Status of pedal webbing could not be reconstructed for P. grandei; however, the anatomy and skeletal proportions of P. grandei suggest that it was an aquatic surface swimmer. This is especially likely as P. grandei has a pedal digit III that is longer than the tarsometatarsus, indicating that this taxon likely was a surface swimmer and led an aquatic or semi-aquatic lifestyle [103]. Optimizations for the Anseriformes total group including stem-Anseriformes indicate that this clade preferred either a terrestrial or aquatic habitat, were non-swimmers, were omnivorous or carnivorous, were mixed feeders or grazers, did not have pedal webbing, had a pedal digit III was shorter than the length of the tarsometatarsus, had an ossified pessulus, and had no asymmetry at the tracheobronchial juncture. The status of rhamphothecal lamellae could not be reconstructed for this clade. Optimizations for crown Anseriformes were identical to those for total group Anseriformes with the exception of crown Anseriformes preferring an aquatic or semiaquatic habitat, being primarily herbivorous, having a form of rhamphothecal lamellae, and being grazers. Optimizations for Anatidae and Pan-Anatidae indicate that they originally preferred an aquatic or semiaquatic habitat, were surface swimmers, were primarily herbivorous, had full rhamphothecal lamellae present, were grazers, possessed pedal webbing, had a pedal digit III was subequal in length or longer than the tarsometatarsus, had an ossified pessulus, and had no asymmetry at the tracheobronchial juncture. Optimizations for crown anatids were identical with the exception of optimizations suggesting that they preferred aquatic environments only.

Ancestral state reconstruction using the morphological data provided identical results for P. grandei. Status of pedal webbing again could not be reconstructed for P. grandei. Optimizations for total group Anseriformes were identical to those of the combined data analysis. Again, the status of rhamphothecal lamellae could not be reconstructed for this clade. Optimizations for crown Anseriformes remained identical. Optimizations for Anatidae and Pan-Anatidae remained identical with the exception of asymmetry at the tracheobronchial junction being unable to be reconstructed. Optimizations for crown Anatidae were identical to those recovered using the combined data results, with the exception of asymmetry at the tracheobronchial junction being unable to be reconstructed.

Discussion

We recover Paakniwatavis grandei as the sister taxon to Anseranatidae across all analyses, regardless of taxon sampling. This placement is consistent with the unique combination of anhimid-like and anseranatid-like morphologies displayed by the taxon as well as its aquatic morphologies. A Gastornis+Pelagornithidae group was also recovered across all analyses as the sister taxon of crown Anseriformes, and a clade containing Vegavis iaai, Presbyornis, Telmabates, Conflicto, and Nettapterornis was recovered as the sister taxon of crown Anatidae across all analyses (called Pan-Anatidae here). Asteriornis maastrichtensis Field et al. 2020 is recovered as a stem Galloanserine in the morphology only topology (<50% bootstrap support), but is the sister taxon to a Tinamidae+Lithornithidae clade within the combined results (54% clade credibility).

Our results for Gastornis and the Pelagornithidae are consistent with those of Bourdon [24], in which Pelagornithidae are the sister taxon of Anseriformes; however, this study did not recover a monophyletic Galloanseres. Field et al. [11] recovered the Pelagornithidae either as the sister taxon to an Anseriformes(Conflicto+Nettapterornis) group or within a polytomy containing Neoaves and Galloanseres. Mayr [25] recovered Pelagornithidae as the sister taxon to a Sylviornithidae(Dromornithidae(Galloanseres))) group, and Mayr et al. [51] recovered a polytomy comprising Pelagornithidae, Galloanseres and Neoaves. Using the dataset of Field et al. [11], Houde et al. [44] recovered Pelagornithidae as the sister taxon to a small sample of four neoavian species using Bayesian analysis; however, when parsimony analysis was used on the same dataset Galloanseres is nonmonophyletic (see Supplement of [44]), so much so that Galliformes are nested within Palaeognathae and the Pelagornithidae are sister to a polytomy containing the four included neoavian species and Anseriformes (including Pan-Anatidae). This level of non-monophyly calls the legitimacy of the Field et al. [11] dataset into question and indicates that character revision and/or revision of the scorings of this dataset are necessary before use in future studies.

Placement of Presbyornis, Nettapterornis, and Vegavis within a clade that is sister to Anatidae is consistent with the results of Ericson [14], Livezey [13], and Elzanowski and Stidham [52] for Presbyornis and both Vegavis and Presbyornis in the Livezey [13] matrix (see also Clarke et al. [23]). Worthy et al. [27] recovered Presbyornis and Wilaru as either the sister group of Anseranas+Anatidae or Anseranas, and recover Vegavis as either the sister taxon of the anseriform-like Gastornithiformes (eg. Gastornis) or the sister taxon to Anseriformes. Tambussi et al. [28], using the Worthy et al. [27] data matrix, similarly recover Presbyornis, Wilaru, Nettapterornis and Conflicto as stem Anseriformes and place Vegavis as the sister taxon of Gastornithiformes within stem Anseriformes. Field et al. [11] recover Wilaru, Presbyornis, Conflicto, and Nettapterornis as stem Anseriformes outside Anhimidae+Anatidae or Wilaru and Presbyornis as the sister group of Anseranas. Field et al [11] also recovered Vegavis as a stem Galloanserine or Neoavian taxon (it remains in an unresolved polytomy) or as a stem Neornithine, and place Asteriornis as a stem Galloanserine or stem galliform. Torres et al. [104] places Conflicto and Vegavis within a polytomy containing Anatidae and recovers Asteriornis as the sister taxon of a Lithornis+tinamou clade.

Anachronornis anhimops and Danielsavis nazensis were placed as sister taxa across almost all morphology only analyses (S4, S5 and S7 Figs, Supplementary Data via 10.5061/dryad.v15dv4208), and were placed within a clade containing Gastornis gigantea and the Pelagornithidae in results where Wilaru was excluded and Galloanserine molecular backbone constraints not enforced (S4 and S5 Figs, Supplementary Data via 10.5061/dryad.v15dv4208). Perplexicervix microcephalon was recovered as the sister taxon to Pan-Anseriformes in results where Wilaru was excluded and Galloanserine molecular backbone constraints not enforced (S4 Fig, Supplementary Data via 10.5061/dryad.v15dv4208). Houde et al [44] recover Anachronornis and Danielsavis as stem Anseriformes using both parsimony analysis and Bayesian analysis with low support. When excluding Danielsavis from the employed datasets, Houde et al. [44] place Anachronornis as either a stem anseriform, stem anhimid, stem anseranatid, or within a polytomy containing other Anseriformes and/or stem Anseriformes depending on the dataset and analysis method used. Thus the phylogenetic placement of both extinct taxa remains tenuous at best and will require further study.

Perplexicervix has not been included in any previously published phylogenetic analysis. Perplexicervix was originally thought to be anseriform [45]; however, Mayr et al. [60] remarked on a few similarities between Perplexicervix and Otididae but did not perform phylogenetic analysis that included this taxon or identify unambiguous character evidence to support this hypothesis. Like Galloanserines, Perplexicervix exhibits basipterygoid processes, which are absent in most of Neoaves. Combined data results yielded 7 unambiguous and 3 ambiguous optimized synapomorphies of the quadrate, coracoid, scapula, humerus, femur, tibiotarsus, tarsometatarsus, and pedal phalanges (two with CI = 1) that support placement of P. grandei within Anseriformes. Within the quadrate, the ventral apex of the crista tympanica terminates within the ventral half of the quadrate (character 172: state 2, ambiguous). The labrum externa along the lateral angle of the sternal coracoid is ventrocranially angled (416:2, unambiguous). Within the scapula, the acromion process is truncate and does not reach cranially beyond the articular faces of the head (437:1, ambiguous). The tuberculum ventral and the crista along the proximal margin of the fossa pneumotricipitalis are domed and distally prominent, overhanging the fossa pneumotricipitalis (456:2, unambiguous, CI = 1). The length of the femur is approximately one half the length of the tibiotarsus (597:1, unambiguous). The distal opening of the pons supratendineus of the tibiotarsus is centered along the midline (653:3, unambiguous). Within the tarsometatarsus the medial margin of the medial cotyle is exceptionally projected proximally and crista-like (658:2, unambiguous, CI = 1), the lateral cotyle is flattened or only slightly concave (662:2, ambiguous), and the hypotarsal eminence is proximally prominent (664:2, unambiguous). Within the pedal phalanges, pedal phalanx II: digit 2 is slightly more elongate than III:2 (706:2, unambiguous).

Seven unambiguous and three ambiguous optimized synapomorphies of the quadrate, mandible, furcula, scapula, carpometacarpus, tarsometatarsus and pedal phalanges (one with CI = 1) support placement of P. grandei as a sister taxon to the clade (Anseranatidae+Anatidae). Within the quadrate, the crista tympanica is present and exhibits an extremely prominent crista (171:3, unambiguous), and the caudal face of the otic process is deeply concave (176:2, unambiguous). Within the mandible, the medial portion of the ramus caudal to the coronoid process (or homologous site) is extremely deep and concave (244:2, unambiguous), and a true retroarticular process is present and exceptionally tapered throughout (256:2, ambiguous). The width of the lateral diameter of the furcular ramus is larger than that at the symphysis (436:3, ambiguous), and the scapula is shorter than the humerus in length (449:1, unambiguous). Within the carpometacarpus, the proximal terminus of the dorsal rim of the trochlea carpalis is strongly angular and proximally elongated (514:2, unambiguous). Within the tarsometatarsus, the distal-most terminus(i) of medial crest(s) are much more distally extensive and the lateral crista(e) are proximodistally truncate, about 1/2-2/3 proximodistal length of the medial crista (669:1, unambiguous, CI = 1). The proximal portion of the sulcus extensorius medial to the dorsal foramina vascularia proximalia is present and deeply excavated (680:2, ambiguous). Within the pedal phalanges, phalanx III is longer than the tarsometatarsus (719:1, unambiguous).

Our phylogenetic placements of the abovementioned extinct taxa suggests that extinct anseriform diversity is likely much more vast than previously known and remains poorly understood. Likely plesiomorphic traits shared by these taxa are additionally necessary to further explore and understand to better resolve the relationships and evolutionary histories of these taxa, especially as new relevant fossils are discovered. Analysis of the Cretaceous ornithurine Janavis finalidens [105] and recent evidence on the ichthyornithine palatine [104] similarly indicate that reevaluation of purported galloanseran affinities of early Cenozoic groups is necessary, especially within the Pelagornithidae. Recovery of additional, more complete and better preserved anseriform-like fossils is ultimately necessary to more robustly resolve the phylogenetic placement of these important taxa. Adding Neoaves to the dataset is also important in future iterations of the Musser and Cracraft [79] dataset. Despite this, consistent results across several analyses using different data types and methods suggest that most placements of the included taxa are fairly robust. P. grandei represents a unique ecology for the Green River Formation, and this new fossil along with this new dataset and re-evaluation of Presbyornis material thus begins to elucidate several critical issues in anseriform evolution.

Ancestral state reconstruction across all analyses suggests the evolution of a combination of terrestrial and semi-aquatic traits at the base of Anseriformes, with crown Anseriformes exhibiting a shift toward more aquatic traits such as preferring an aquatic or semiaquatic habitat, being primarily herbivorous, having a form of rhamphothecal lamellae, and being grazers (Fig 5). This is inconsistent with the assertion that Anseriformes were ancestrally terrestrial as suggested by Olson and Feduccia [22], Ericson [14], and Livezey [13]. The placement of P. grandei and its influence on the optimization of these reconstructions suggests aquatic or semi-aquatic ancestry of Anseriformes, especially combined with a reduced form of rhamphothecal lamellae present in extant Anhimidae [22].

The question of how anseriform beak morphology and filter feeding evolved remains an open one. Either a narrower, “goose-like” beak was ancestral for Anseriformes, or this narrower beak evolved several times within Anseriformes [19]. The “goose-like” beak is associated with increased leaf consumption, decreased invertebrate consumption, and an increase in the mechanical advantage of the beak that allows for more effective cropping of plants [19], whereas “duck-like” beaks are associated with increased filter feeding and consumption of invertebrates. We take this classification of a “goose-like” beak a step further in the broader context of both stem and crown Anseriformes: We first identify whether the rostrum is mediolaterally wider than the width of the paroccipital processes (indicating an anseranatid-like beak; character 3) and, if so, whether it is anteriorly tapered and narrowed further (a “goose-like” beak; character 4: state 1), or whether it remains subequal in width (a “duck-like” beak; 4:2). P. grandei is the only Paleogene anseriform currently known to present a narrow, anhimid-like beak other than the Pelagornithidae, Danielsavis and Anachronornis. Its beak is mediolaterally narrower than the width of its paroccipital processes, as in extant Anhimidae. Chaunoides and Telmabates have no known preserved skull and Vegavis has no known preserved rostrum or braincase, while other Paleogene fossils with a beak preserved such as Presbyornis, Nettapterornis and Conflicto all present an anseranatid-like beak based on currently known remains. All of our analyses posit the first appearance of a beak that is mediolaterally wider than the width of the paroccipital processes as ancestral to the node containing Anseranas semipalmata. It would have been an anteriorly tapered, “goose-like” beak like that of the extant Anhmidae or Anseranas (Fig 5). Across all analyses, ancestral state reconstruction indicates that this “goose-like” beak is ancestral to Anseranatidae+Anatidae. All analyses indicate that anteriorly wider “duck-like” beaks evolve at least twice: once after the divergence of Anseranas in taxa closer to Anatidae (present in Presbyornis and Nettapterornis), and once within crown Anatidae (see Fig 5 and Supplementary Data via 10.5061/dryad.v15dv4208). These results contradict those of Olsen [19], who found that a duck-like beak was ancestral for most Anatidae, followed by multiple transitions toward a goose-like beak; however, this study performed ancestral state reconstruction using a phylomorphospace of beak curvature measurements, beak function metrics, and quantified diet data for a smaller sample of anseriform taxa that included only two extinct taxa, Presbyornis and the recently extinct moa-nalo Thambetochen chauliodous Olson and Wetmore [106] (see also Olson and James [107]). More robust placement of Anachronornis and Danielsavis in future studies will be critical to better understanding anseriform ancestral state reconstructions and the evolution of beak morphology.

At the same time, our results are consistent with the results of Olsen [19] in that we find rhamphothecal lamellae to have been present ancestrally for crown Anseriformes and Anatidae (including within both the stem and crown lineages of these groups), indicating that herbivory and/or filter feeding was ancestral for these clades. This again contradicts the hypothesis that Anseriformes were ancestrally terrestrial and would explain the presence of reduced rhamphothecal lamellae in extant Anhimidae [22]. Anhimidae represent the only known example of rhamphothecal lamellae being present without pedal webbing in extant birds; however, similar lamellae-like ridges have been found in Ornithomimus [108,109], Gallimimus, chelonians, hadrosaurs [109] and an edentulous ceratosaur [110]. A partial correlation between the presence of these ridges and exclusively herbivorous diet among terrestrial chelonians has been found [111,112], suggesting that more prominent ridges were present when more coarse vegetation was eaten. Studies on the jaw mechanics, locomotion and gut contents of hadrosaurs have similarly demonstrated that they were obligate terrestrial herbivores that used their beak for cropping through vegetation [113,114]. If our ancestral state reconstructions for crown Anseriformes are correct, lamellae coevolved with a shift toward herbivory and grazing along with the preference for a more aquatic habitat despite a lack of pedal webbing. Based on the available evidence and our results, some form of rhamphothecal lamellae was ancestral to crown Anseriformes and could have developed due to aquatic grazing, then remained (or became reduced) within extant Anhimidae while developing further within more derived crown anseriform taxa. Webbing then was maximally ancestral to P. grandei, and minimally was ancestral to Anseranas. Our results thus suggest that crown Anseriformes were maximally ancestrally aquatic or semi-aquatic, and fully aquatic ancestral to crown Anseranatidae. In general our results suggest a trend within Anseriformes toward aquatic grazing and the anseranatid “goose-like” beak to obtain an herbivorous (increased leaf and root consumption) diet, with at least two evolutions of a “duck-like” beak associated with increased filter feeding and invertebrates obtained by this feeding mode at least once within Pan-Anatidae and once within Anatidae [115,116]. If these placements and ancestral state reconstructions are correct, this would add to mounting support that feeding ecology has acted as the primary selective force in waterfowl beak shape diversification [19].

Further elucidation of anseriform behavioral evolution is indicated in ancestral state reconstruction of syringeal characters. Ancestral state reconstruction within the combined results indicates that P. grandei, Anseriformes (inclusive and exclusive of stem Anseriformes) and anatids (inclusive and exclusive of the clade containing extinct taxa that was placed as the sister to Anatidae) had an ossified pessulus of the syrinx, a derived neognath bird feature that has been proposed to anchor enlarged vocal folds or labia [117], consistent with the results of Clarke et al. [42]. These taxa also were indicated to ancestrally not possess asymmetry at the tracheobronchial juncture of the syrinx. Ancestral reconstruction within the morphological results is identical with the exception of ambiguity within Pan- and crown Anatidae regarding asymmetry at the tracheobronchial juncture. Both results indicate that asymmetry at the tracheobronchial juncture must have evolved at least once within Anseranatidae+Anatidae. This is somewhat consistent with Clarke et al. [42]; however, while Clarke et al. [42] considered a single origin in Anatidae likely, our results may suggest more gains and losses within this Anseranatidae+Anatidae clade. Further study and coding of extant syrinx asymmetry is necessary as previous descriptions and figures of extant syrinxes largely focus on pronounced asymmetrical bullae in some male anseriform taxa rather than the more subtle asymmetry of the rings found in females, and the large range of variation across differing taxa and sexes within Anseriformes is not well understood [20,117]. Better understanding asymmetry in extant taxa has important implications for extinct taxa as well; for example, the subtle asymmetry present in Vegavis may suggest that this taxon exhibited sexual dimorphism within the syrinx, as in some extant Anatidae. Asymmetry is an important trait to further study as it is correlated with the presence of a dual sound source and the presence of labia [42,117]. Recovery of further fossils that include syrinxes and further study of extant syrinx anatomy and function in extant birds is thus necessary to understand the evolution of this organ.

All analyses recover the K-Pg taxa Vegavis, Nettapterornis, Conflicto, Presbyornis and Telmabates within a clade that is the sister taxon to Anatidae. Within the combined data results, six unambiguous and 10 ambiguous synapomorphies (CI < 1) were recovered for this clade that united three or more of these taxa within the axial skeleton and hindlimbs (see Supplementary Data via 10.5061/dryad.v15dv4208). Although several appear to be plesiomorphic, these characters may represent evolutionary and biological/ecological significance pending recovery of key elements from taxa such as Vegavis, Telmabates and Conflicto.

P. grandei exhibits non-heterocoelous vertebrae. Although this character was not optimized as a synapomorphy for this clade, Nettapterornis, Telmabates, Presbyornis, and the Neogene Dromornithidae have amphicoelous thoracic vertebrae [14,56,118], whereas Vegavis, Pelagornithidae and gastornithids have heterocoelous thoracic vertebrae [23,42]. Although it cannot be discerned which non-heterocoelous form the vertebrae of P. grandei possess due to taphonomic distortion, it is likely that P. grandei possessed amphicoelous vertebrae as well. Amphicoelous thoracic vertebrae are plesiomorphic within Avialae but also present in well-nested neoavian clade Charadriiformes [14,56,88,118]. Opisthocoelous thoracic vertebrae are only known within Neoaves [79,88,89], and amphiplatyan and procoelous vertebrae are only known in non-avian dinosaurs [89].

Further study on the function of amphicoelous vertebrae in the context of avian evolution and ecology is needed, especially since birds possess a unique dorsal intervertebral joint [119]; however, the literature on this in fishes and crocodylomorphs suggests that amphicoelous vertebrae provide a more rigid spine [120–122] that can withstand increased stress without deformation that may be caused by powerful movements of musculature [120,121], allowing for rapid flexure of the spine. Amphicoelous vertebrae have evolved several times in sharks, dipnoans, bony ganoids and teleosts, associating their appearance with improved speed of motion [120]. Amphicoelous vertebrae are also associated with aquatic environments, as transitions from amphicoelous to platycoelous vertebrae has been hypothesized to represent a transition from aquatic to terrestrial environments [123].

Results further clarify the complex picture of avian evolution around the K-T boundary, indicating that several lineages within Anseriformes with a variety of ecologies not represented in the crown were present by the latest Cretaceous and into the early Paleogene. P. grandei represents an early Eocene lacustrine taxon that was likely aquatic, swam and used its narrow bill in aquatic grazing or mixed feeding. The Cretaceous-early Paleogene Nettapterornis and Presbyornis were aquatic taxa that likely filter fed on a more invertebrate-heavy diet within both marine and marine and lacustrine environments, respectively [19]. The early Eocene Telmabates may have shared a similar ecology to Presbyornis given its amphicoelous thoracic vertebrae and evidence that the Casamayor formation in which it was found is known to be a marine-fluvial transition zone [124]. Other Cretaceous and Paleogene material has been referred to Presbyornis from more fluvial as well as marine settings possibly suggesting a cosmopolitan, flexible habitus [125–128]. At the same time the Cretaceous Vegavis, with heterocoelous thoracic vertebrae, was present in a near shore marine environment with unknown diet, and the Cretaceous Conflicto was present within a near shore marine/transitory estuarine environment, again with unknown diet and locomotion due to missing data [28]. In addition to this array of taxa and ecologies, the specialized Cretaceous-Paleogene marine, piscivorous pseudodontorns and the giant terrestrial Paleogene gastornithids were present within the stem anseriform lineage. If Vegavis and Conflicto are also found to have had omnivorous, mixed and/or piscivorous diets, a proliferation of Cretaceous-early Paleogene non-herbivorous stem and crown Anseriformes may have arisen. Proposed significant loss of plant cover due to global cooling and the K-T impact event [129–134] could suggest a short but strong selective regime favoring mixed and non-herbivore specialists. Fossil evidence suggests that early Anseriformes were diversifying rapidly since at least the late Cretaceous and were already widespread within the same time frame, as the early Eocene Telmabates was found in Patagonia and Paleocene and Eocene Presbyornis material has been recovered from North America, Europe and Mongolia [33,126–128].

An approximate mean body mass estimate for P. grandei is 304g based on the published allometric equation using femoral length from Field et al. [81]. This body mass level is quite small; it is most comparable to the mass of many Anas (within the 300s range). Many other anatids are generally larger (which typically range from 600-over 1,000g) [135]. Its body mass is estimated to be less than half that of Presbyornis and Telmabates (882g and 1423g, respectively). This is consistent with recent evidence that correlation between herbivory and body mass is not significant when accounting for phylogeny [18].

Further analysis of these Paleogene anseriform fossils in the context of broader extinct taxon sampling, especially in the context of extinct anseriform-like taxa, is necessary to further evidence placement of P. grandei and other Paleogene anseriform-like taxa and to gain more robust insight into ancestral states; however, P. grandei represents a key taxon, a unique ecology within known Anseriformes and the Green River Formation, and a potential calibration point for anseranatids. Other included extinct taxa also represent potential calibration points that would be valuable for stem Anseriformes, Anhimidae and Pan-Anatidae. Recovery of additional fossils and further phylogenetic analyses (especially of taxa such as Chaunoides and Wilaru) are preferable to confirm these relationships, further reveal the ecological and behavioral evolution and biogeography of Anseriformes, and better elucidate our understanding of avian evolution.

Supporting information

Acknowledgments

We thank all of the staff of FMNH, especially Lance Grande, Jingmai O’Connor, William Simpson, Adrienne Stroup, Shannon Hackett, John Bates, and Ben Marks for specimen access and valuable discussion. We thank April Neander for scanning the specimen, and Matthew Colbert for aid in scan visualization and segmentation. We thank Helen James, Christopher Milensky, Brian Schmidt and Mark Florence of the Smithsonian National Museum of Natural History for access to the Ornithology and Vertebrate Paleontology collections. We thank Joel Cracraft, Paul Sweet, Mark Norell, Ruth O’Leary and Carl Mehling of AMNH for access to the Ornithology and Vertebrate Paleontology Collections. We additionally thank Kenneth Bader, Matthew Brown, and Christopher Sagebiel for access to additional TMM collections and their aid in working with AMNH loans. Finally, we thank Joel Cracraft, Zhiheng Li, Melissa Kemp, Christopher Bell, Daniel Field, Daniel Ksepka and Christopher Torres for valuable discussion.