Partial skeleton from the Paleocene of New Zealand illuminates the early evolutionary history of the Phaethontiformes (tropicbirds)

Abstract We describe a new stem group representative of the Phaethontiformes (tropicbirds) from the Paleocene Waipara Greensand in New Zealand. The fossil consists of a partial skeleton with a nearly complete skull and represents the first unambiguous record of the Phaethontiformes from the Paleocene of the Southern Hemisphere. Clymenoptilon novaezealandicum gen. et sp. nov. has a proportionally shorter pelvis than Prophaethon shrubsolei from the early Eocene London Clay and appears to have been less adapted to foraging in an aquatic environment at or below sea-level. It is furthermore distinguished from P. shrubsolei and Lithoptila abdounensis from the late Paleocene/early Eocene of Morocco in a proportionally smaller foramen magnum of the skull. Together with other plesiomorphic features, this suggests that C. novaezealandicum is the sister taxon of a clade including Lithoptila, Prophaethon and crown group Phaethontiformes, and as one of the oldest stem group phaethontiforms the new species may indicate a Southern Hemispheric centre of origin of tropicbirds. After a recently described bony-toothed bird, C. novaezealandicum is the second seabird species from the Waipara Greensand to show affinities to taxa from the early Paleogene of the Northern Hemisphere. The wide early Paleogene distribution of the Phaethontiformes stands in sharp contrast to the geographic restriction of coeval diving seabirds, and different factors appear to have limited the dispersal of aquatic and pelagic seabird taxa in the early Paleogene. Gerald Mayr [Gerald.Mayr@senckenberg.de], Senckenberg Research Institute and Natural History Museum Frankfurt, Ornithological Section, Senckenberganlage 25, D-60325 Frankfurt am Main, Germany; Vanesa L. De Pietri [vanesa.depietri@canterbury.ac.nz], University of Canterbury, School of Earth and Environment, Private Bag 4800, Christchurch 8140, New Zealand; Al Mannering [alman@slingshot.co.nz], Canterbury Museum, Rolleston Avenue, Christchurch 8050, New Zealand; Leigh Love [lvlove@xtra.co.nz], PO Box 49, Waipara 7483, New Zealand; Erica Crouch [e.crouch@gns.cri.nz], GNS Science, 1 Fairway Drive, Lower Hutt 5040, New Zealand; Catherine Reid [catherine.reid@canterbury.ac.nz], University of Canterbury, School of Earth and Environment, Private Bag 4800, Christchurch 8140, New Zealand; R. Paul Scofield [pscofield@canterburymuseum.com], Canterbury Museum, Rolleston Avenue, Christchurch 8050, New Zealand, and University of Canterbury, School of Earth and Environment, Private Bag 4800, Christchurch 8140, New Zealand.

OWING TO ongoing collecting activities of one of the authors (LL) during the past years, the marine strata of the Waipara Greensand in Canterbury, New Zealand have yielded significant numbers of early and middle Paleocene bird fossils.Most of the specimens belong to the Sphenisciformes (e.g., Mayr et al. 2017Mayr et al. , 2020)), but three non-sphenisciform species were also described.The best-represented of these is Protodontopteryx ruthae Mayr, De Pietri, Love Mannering & Scofield, 2019, which is based on a partial skeleton and is the oldest record of the Pelagornithidae (bony-toothed birds; Mayr et al. 2021).Another species, Australornis lovei Mayr & Scofield, 2014, is known from incomplete wing and pectoral girdle bones and is a presumably aquatic bird of uncertain phylogenetic affinities (Mayr & Scofield 2014).Mayr & Scofield (2016) reported fragmentary remains of a putative stem group representative of the Phaethontiformes (tropicbirds).Here we describe a wellpreserved partial skeleton of a new phaethontiform species from the Waipara Greensand at mid-Waipara River, North Canterbury, which is distinctly larger than the fossil reported by Mayr & Scofield (2016) and illuminates the early evolutionary history of the Phaethontiformes.
Tropicbirds are highly aerial, plunge-diving and predominantly piscivorous birds, which occur in tropical and subtropical seas.The three extant species are classified into the taxon Phaethon and mainly differ in size and plumage features.These pelagic birds are relics of a once much more widespread and diverse group of birds, whose fossil record goes back into the earliest Paleogene (Mayr 2022).Current sequence-based analyses support a sister group relationship between the Phaethontiformes and the terrestrial Eurypygiformes (Sunbittern and Kagu), with the clade formed by both taxa being recovered as the sister taxon of the Aequornithes, which include most ß 2023 The Author(s).Published by Informa UK Limited, trading as Taylor & Francis Group.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/),which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.The terms on which this article has been published allow the posting of the Accepted Manuscript in a repository by the author(s) or with their consent.
The first fossil species to have been assigned to the Phaethontiformes is Prophaethon shrubsolei Andrews, 1899 from the early Eocene of the British London Clay, which is known from two partial skeletons and a few isolated bones (Andrews 1899;Harrison & Walker 1976;Mayr 2015).Prophaethon shrubsolei is clearly distinguished from extant Phaethontiformes and was assigned to the family-level taxon Prophaethontidae.The differences are particularly obvious in the morphologies of the pelvic girdle and legs, which indicate a more aquatic ecology than in extant Phaethontidae, and P. shrubsolei most likely was foraging at or below sea-level and used its hind limbs for propulsion in the water (Mayr 2015).
Another well-represented Paleogene phaethontiform is Lithoptila abdounensis Bourdon, Bouya & Iaroch ene, 2005 from the late Paleocene and early Eocene of the Ouled Abdoun Basin in Morocco (Bourdon et al. 2005(Bourdon et al. , 2008a)).This species was classified into the Prophaethontidae and is represented by a large number of isolated bones.Phaethusavis pelagicus Bourdon, Amaghzaz & Bouya, 2008 from the early Eocene of Morocco, by contrast, is only known from a proximal humerus and was considered to be more closely related to crown group Phaethontiformes (Bourdon et al. 2008b).
Zhylgaia aestiflua Nessov, 1988 from the late Paleocene of Kazakhstan (Nessov 1988(Nessov , 1992) ) was referred to the Phaethontiformes by Bourdon et al. (2008a).This species was originally based on fragmentary humerus remains, but Mayr & Scofield (2016) detailed that Tshulia litorea Nessov, 1988 from the same locality, which is known from leg bones, is also likely to be a phaethontiform and to be conspecific with Z. aestiflua.Mayr & Scofield (2016) furthermore considered possible phaethontiform affinities for Novacaesareala hungerfordi Parris & Hope, 2002 from the latest Cretaceous or earliest Paleocene of New Jersey, USA (Parris & Hope 2002).Remains of putative Prophaethontidae were also reported from the late Paleocene of Maryland, USA, but only consist of a few incomplete bones (humeri and coracoids), which have not yet been figured (Olson 1994).
Whereas various Paleocene or early Eocene fossils were assigned to the Phaethontiformes, the Oligocene and Neogene fossil record of tropicbirds is much more scant, which may indicate that these birds became more pelagic in their evolutionary history (Mayr 2022).A coracoid from the middle Eocene of Belgium was assigned to the Prophaethontidae by Mayr & Smith (2002), and Mayr & Smith (2013) tentatively assigned a femur from the early Oligocene of Belgium to the Phaethontiformes.A few fragmentary bones of phaethontiforms also exist from early Neogene localities in Europe and North America (Olson 1985, Ml ıkovsk y 1997, Olson & Walker 1997).
The new fossil from the Waipara Greensand provides robust evidence for the presence of stem group tropicbirds in the Paleocene of New Zealand.It is one of the few substantial records of a non-sphenisciform bird from the earliest Paleogene and belongs to a new taxon, which illuminates the earliest evolution of the Phaethontiformes.

Material and methods
X-ray computed tomography (CT) scans were conducted at Pacific Radiology, St Georges Hospital, Christchurch using a medical Siemens Somatom Definition Scanner.One thousand and sixteen 400 lm slices were made at 200 lm intervals using an apparent power of 140 KVA.Images taken were 512 by 512 pixels and the field of view was 170 mm.Scans were rendered using Materialise Mimics Research version 25.0 (Materialise NV, Leuven, Belgium).
Osteological terminology follows Baumel and Witmer (1993), except that the correct Latin term praemaxillary, instead of premaxillary, is used (see Mayr 2020).The taxonomy of the extant species is after the IOC World Bird List (https://www.worldbirdnames.org).

Diagnosis
Clymenoptilon differs from Lithoptila Bourdon, Bouya & Iaroch ene, 2005 in foramen magnum proportionally smaller and condylus occipitalis larger (the condylus occipitalis is almost as wide as the foramen magnum in Clymenoptilon, whereas it is much narrower than this foramen in Lithoptila); processus paroccipitales more strongly developed; coracoid with tip of extremitas omalis ventrally slanting and forming a ventromedial projection, facies articularis clavicularis exhibiting a marked concavity; proximal end of humerus proportionally wider (ratio of proximal width to length of humerus 0.19 in Clymenoptilon versus 0.16 in Lithoptila).Clymenoptilon differs from Prophaethon Andrews, 1899 in beak proportionally longer (the beak is longer than the neurocranium in Clymenoptilon, whereas it is as long as the neurocranium in Prophaethon); foramen magnum proportionally smaller; nasofrontal hinge less well-defined; facies articularis clavicularis of coracoid with marked concavity in its medial portion; pelvis proportionally shorter (measuring only half the length of the skull, whereas the pelvis length is more than two thirds of the skull length in Prophaethon).Clymenoptilon differs from Phaethusavis Bourdon, Amaghzaz & Bouya, 2008b in apex (dorsalmost point) of crista deltopectoralissituated much farther distally (it reaches well beyond the distal terminus of the crista bicipitalis in Clymenoptilon, whereas it is on the same level in Phaethusavis).Clymenoptilon differs from Zhylgaia Nessov, 1988 andNovacaesareala Parris &Hope, 2002 in condylus ventralis of humerus being proportionally larger; ventral portion of distal end of humerus with processus flexorius more strongly developed.Clymenoptilon differs from Phaethon Linnaeus, 1758 in nostrils long and slit-like; nasofrontal hinge less well-defined; foramen magnum proportionally much smaller; humerus with crista deltopectoralis not notched in its proximal section; ulna not distinctly exceeding humerus in length; pelvis mediolaterally much narrower.Clymenoptilon differs from the similar-sized Australornis lovei Mayr & Scofield, 2014 from the Waipara Greensand in extremitas omalis ofcoracoid much wider in dorsoventral direction; intumescentia humeri forming a much more pronounced convexity; tuberculum dorsale of humerus smaller.

Etymology
The taxon name is derived from Kktl eg (Clymene, Gr.), who in Greek mythology is the mother of Phaethon, and psίko (ptilon, Gr.), feather.It refers to the phaethontiform affinities of the new taxon and its morphological similarity to Lithoptila and is neuter in gender; the connecting vowel "o" is inserted to improve pronounceability.

Etymology
The species epithet refers to the geographic provenance of the holotype.

Holotype
UC 22048, partial skeleton including the skull, vertebral column, right wing and pectoral girdle elements, pelvis.

Remarks
As detailed by Mayr et al. (2021), the holotype of the pelagornithid Protodontopteryx ruthae from the Waipara Greensand appears to be associated with bones of another bird belonging to a different, undetermined taxon.The humerus of this latter bird has a similar length to that of the new phaethontiform species described in the present study but is distinguished in the morphology of the distal end (including a more distally protruding processus flexorius, among other features); the ulna of the undetermined bird associated with the holotype of P. ruthae is furthermore longer than that of the new phaethontiform species.The affinities of the specimen associated with the P. ruthae holotype still have to be determined, but it probably represents a fifth non-sphenisciform avian species from the Waipara Greensand, in addition to P. ruthae, Australornis lovei, the unnamed phaethontiform reported by Mayr and Scofield (2016), and the new phaethontiform species described in the present study.
The skull (Fig. 3A, D) is nearly complete, but adhering matrix and the fact that the mandible is preserved in articulation obscure many osteological details.Diagenetic compression furthermore led to some distortion of the neurocranium.The tapering beak exceeds the neurocranium in length and is proportionally longer than in extant Phaethontiformes (Phaethon spp.), in which it is about as long as the neurocranium.As in Phaethon, its pointed tip is slightly deflected, but the culmen is sigmoidally curved rather than being convex as it is in Phaethon.As in Prophaethon and unlike in adults of extant Phaethontiformes, the nostrils of Clymenoptilon novaezealandicum are very long; they are, however, dorsoventrally narrower than in Prophaethon and reach less far rostrally (even though the tip of the upper beak is broken in the P. shrubsolei holotype).The caudal nasal bar is wider than in Prophaethon.The ventral nasal bar, formed by the praemaxillary and maxillary bones, is dorsoventrally deep.The caudal end of the nostril is slit-like but does not reach beyond the nasofrontal hinge as it does in typical schizorhinal birds (such as the Eurypygiformes).The maxillary bone forms a caudal projection ventral to the jugal bar (Fig. 3A), which was termed angulus tomialis by Livezey & Zusi (2006, p. 84).The left lacrimal is preserved in situ but is slightly displaced; it is large and rostrocaudally extensive and was not co-ossified with the frontal.The right lacrimal is detached and moved into the orbit; only a part of the bone is exposed in the specimen.The nasofrontal hinge (zona flexoria craniofacialis sensu Baumel & Raikow 1993, p. 153) is less well defined than in Prophaethon and extant Phaethontiformes, where it forms a pronounced transverse sulcus that runs across the entire breadth of the skull.The portion of the frontal bone immediately caudal to the nasofrontal hinge is not as strongly dorsally bulging as in Prophaethon (Fig. 3B, E) and Phaethon (Fig. 3C, F).The orbit of C. novaezealandicum is large as it is in other Phaethontiformes from which the skull is known.As in extant Phaethontiformes, the jugal bar is stout, but its rostral attachment to the beak is more solid in C. novaezealandicum (in Phaethon it is only a thin, strap-like sheet of bone).The neurocranium is slightly wider than in Lithoptila and more similar to that of Prophaethon in its relative width.As in other Phaethontiformes but unlike in the Procellariiformes and Sphenisciformes, there are no supraorbital fossae glandularum nasales.The processus postorbitalis is long and laterocaudally projected, resembling the process of other phaethontiforms; however, its tip is more caudally directed and less ventrally deflected than in Lithoptila.The processus zygomaticus is a laterally protruding, dorsoventrally extensive flange as it is in other Phaethontiformes; immediately ventral to the left processus zygomaticus, a piece of matrix containing a small isolated ossicle adheres to the skull and projects caudally beyond the neurocranium.Details of the otic region cannot be discerned.As in Phaethon, the fossa temporalis extends onto the caudodorsal portion of the processus postorbitalis.The temporal fossae of C. novaezealandicum reach further towards the midline of the neurocranium than in Prophaethon, but they are shallower than in Phaethon.The ventrocaudally projecting processus paroccipitales are better developed than in Lithoptila, Prophaethon and Phaethon; they border a recess on the caudal surface of the cranium, which encompasses the foramen magnum.The tuberculum basilare (processus medialis parasphenoidalis) is likewise prominent.The crista nuchalis transversa of the right side of the skull is deformed, which either represents a pathological condition or, perhaps more likely, a diagenetic artefact.Rather than being broadly convex as in Lithoptila, Prophaethon and Phaethon, the prominentia cerebellaris is mediolaterally compressed and ridge-like.On the caudal surface of the neurocranium, the paired foramina for the ramus occipitalis of the arteria ophthalmica externa are visible.The foramen magnum is proportionally much smaller than in Lithoptila, Prophaethon and Phaethon.The condylus occipitalis (Fig. 3G) is mediolaterally wider than it is dorsoventrally deep, being much wider than the condylus occipitalis of Lithoptila, Prophaethon (Fig. 3H) and Phaethon (Fig. 3I; in Clymenoptilon, the condylus occipitalis is almost as wide as the foramen magnum, whereas it is much narrower than this foramen in other phaethontiform taxa and most extant neornithine birds).
The left quadrate is preserved in situ, whereas the right one is missing.The bone appears to be of similar proportions to the mediolaterally wide quadrate of Prophaethon (see Mayr 2015), but due to overlying matrix no morphological details of the bone can be discerned in the specimen.
The caudal portion of the right mandibular ramus is missing.The caudal end of the left ramus is truncate and has a concave caudal surface.The processus medialis mandibulae is dorsoventrally deep, with a ridge-like caudal margin.
The mandibular rami are closely adjacent in the rostral half of mandible.The rostrum mandibulae is dorsoventrally deeper than the pointed rostrum mandibulae of extant Phaethontiformes; its very tip appears to be damaged so that the truncate appearance possibly is an artefact of preservation (in Phaethon, the rostrum mandibulae is pointed).The angulus mandibulae is pronounced.A lateral mandibular groove is present but less distinct and shorter than in Phaethon (the condition in Clymenoptilon may be due to the fact that matrix obscures the sulcus).Unlike in Lithoptila and extant Phaethontiformes, there is no fenestra rostralis mandibulae.The suture between the os spleniale and the os supra-angulare is well visible.
The holotype includes the caudal-most 10 presacral vertebrae, which are preserved in articulation.The penultimate thoracic vertebra exhibits large pleurocoels, that is, distinct fossae on the lateral surfaces of the corpus, which also occur in Prophaethon but are absent in extant Phaethontiformes (Mayr 2021).The caudal-most three praesacral vertebrae lack a processus ventralis, which appears to be present in the more cranial vertebrae.Because most vertebrae are partly covered by matrix, other osteological details of potential taxonomic or phylogenetic significance cannot be discerned in the specimen.The synsacrum (Fig. 4H) is exposed in ventral view and consists of 13 co-ossified vertebrae (by contrast there are about 14 in Prophaethon [the exact number is difficult to determine in the specimen] and 12 in Phaethon).Unlike in Prophaethon and extant Phaethontidae, the synsacrum of Clymenoptilon is mediolaterally narrow and the ventral surface of its cranial half is essentially flat; the caudal portion exhibits a midline ridge.
The coracoid of the holotype is complete and well preserved (Fig. 5A, B), even though parts of the bone are concealed by matrix.The extremitas omalis is mediolaterally wide.It differs from the extremitas omalis of the coracoid of Lithoptila in that, in medial view, its tip is slanting ventrally and forms a ventromedial projection; the facies articularis clavicularis exhibits a marked concavity in its medial portion which is absent in both Lithoptila and, as shown by the referred specimen NHMUK A 5359 (Fig. 5H-J), Prophaethon.The facies articularis scapularis is not visible, but because of the absence of a pronounced tuberculum coracoideum on the scapula it is likely to have been shallow, as it is in Prophaethon.The long processus procoracoideus is deflected towards the extremitas omalis.As in Lithoptila (see Bourdon et al. 2008a), the shaft of the bone is more slender than in Prophaethon.Whether a foramen nervi supracoracoidei was present cannot be ascertained (the foramen is present in all other phaethontiforms of which the coracoid is known).The processus lateralis is very long as it is in Prophaethon (see Harrison & Walker 1976;Mayr 2015); in extant Phaethontiformes (Fig. 5F, G), by contrast, the process is short.The medial margin of the extremitas sternalis forms a convex bulge.The bone differs from a coracoid from the middle Eocene of Belgium (Fig. 5C, D), which was described by Mayr & Smith (2002), in being mediolaterally more slender and in that, in medial view, the tip of the processus acrocoracoideus is more slanting.
The scapula has a fairly straight and narrow blade (Fig. 5P).The extremitas omalis does not exhibit a well-developed tuberculum coracoideum.The scapula of the smaller putative Phaethontiformes from the Waipara Greensand has not been described by Mayr & Scofield (2016), and the cranial extremity of the bone was only detected amongst associated bone fragment after the latter study was published.The acromion of this specimen (Fig. 5L) is longer than in C. novaezealandicum.
The long humerus (Fig. 5N, P) corresponds well to that of Lithoptila in its overall proportions, but has a somewhat wider proximal end.The ventral portion of the sulcus transversus is pronounced, as it is in other phaethontiforms; the dorsal section of the sulcus is obscured by matrix and also not clearly visible on the CT scans.As in Lithoptila and Zhylgaia, the cranially deflected crista deltopectoralis is long, with little dorsal extent; its caudal surface is concave.The apex (that is, the dorsal-most point) of the crista is situated distal to the distal terminus of the crista bicipitalis, whereas it is on the same level in Phaethusavis.Unlike in extant Phaethontiformes (Fig. 5O), the crista deltopectoralis of C. novaezealandicum does not form a pronounced notch in its proximal section.The tuberculum dorsale, on the caudal surface of the proximal end, is small.The crus dorsale fossae is well developed.The intumescentia humeri is markedly raised, forming a strongly convex surface.The impressio coracobrachialis, which was described as "exceedingly small" for Lithoptila (Bourdon et al. 2008a, p. 755), forms a distinct and fairly wide fossa.The crista bicipitalis is poorly developed, as it is in Lithoptila.The attachment area of musculus scapulohumeralis caudalis just distal to the tuberculum ventrale, on the caudal surface of the bone, is well delimited.Owing to diagenetic compression, the fairly straight shaft of the bone is probably more flattened craniocaudally than it actually was.Still, it appears to have been proportionally wider than the humerus shaft of Lithoptila.The distal end of the bone resembles the distal humerus of Lithoptila, with the ventral portion around the processus flexorius being better developed and more protruding than in Zhylgaia (see Bourdon et al. 2008a: fig. 4).Unlike in Prophaethon (Mayr 2015: supplemental Fig. S1), there is no well-delimited tuberculum supracondylare dorsale.The condylus ventralis is larger than in Zhylgaia and of similar size to that of Lithoptila and extant Phaethontidae.The condylus dorsalis Abbreviations: acr, acromion; cdd, condylus dorsalis; cdp, crista deltopectoralis; cdv, condylus ventralis; cor, coracoid; exo, extremitas omalis of coracoid; flx, processus flexorius; hum, humerus; ntc, notch in crista deltopectoralis; pex, processus extensorius; pii, processus internus indicis; rad, radius; sca, cranial extremity of right scapula; tbd, tuberculum dorsale; uln, ulna.Scale bars equal 10 mm. also corresponds to that of Lithoptila.The tuberculum supracondylare ventrale, which is very long in Lithoptila, is concealed by the overlying ulna.The sulcus scapulotricipitalis appears to have been well developed but is filled with sediment in the specimen; the sulcus humerotricipitalis is deeply excavated.
The ulna exceeds the humerus only slightly in length, whereas it is much longer in extant Phaethontiformes (the ulna of Lithoptila was estimated to have been much longer than the humerus by Bourdon et al. 2008a).The bone has a narrow shaft with a roundish cross-section, but because it is preserved in articulation, only few details of its ends can be discerned.As far as comparisons are possible, the proximal end resembles the proximal ulna of Lithoptila, with the olecranon being fairly long, the cotyla dorsalis dorsoventrally narrow, and the processus cotylaris dorsalis being of moderate size.The distal end also resembles the distal ulna of Lithoptila, and the rather short condylus dorsalis and ventrally prominent condylus ventralis are well visible.The tuberculum carpale is only partly exposed, but as in Lithoptila it appears to be separated from the condylus ventralis by a notch.
The radius does not allow the recognition of morphological details, and features of potential taxonomic and phylogenetic significance could not be detected.The bone is slightly sigmoidally curved and has a subovate cross-section.
The carpometacarpus (Fig. 5Q) is a long bone with a straight os metacarpale minus and a narrow spatium intermetacarpale.It has similar proportions to the carpometacarpus of Prophaethon (see Mayr 2015) and is craniocaudally narrower than that of extant Phaethontiformes (the carpometacarpi of Lithoptila and Zhylgaia are unknown).The processus extensorius is proximodistally wide and has a broadly rounded tip, being of similar shape to the processus extensorius of Prophaethon (Fig. 5R; Mayr 2015) and not as pointed as in the unnamed putative phaethontiform from the Waipara Greensand reported by Mayr & Scofield (2016).Furthermore, unlike in the latter fossil (Fig. 5M), there is no notch in the caudal rim of the trochlea carpalis.
The hand section of the wing is very long, and-depending on the length of the missing phalanx distalis digiti majoris-it approached or even exceeded the length of the humerus.The long phalanx digiti alulae is preserved in articulation.The phalanx proximalis digiti majoris is craniocaudally narrower than in extant Phaethontiformes and has similar proportions to the corresponding phalanx of Prophaethon (see Mayr 2015).The processus internus indicis, which is well developed in Prophaethon and extant Phaethontiformes, is broken in the fossil.The os carpi radiale appears to be of similar shape to that of extant Phaethontiformes, but its surfaces are too eroded for close comparisons.The os carpi ulnare cannot be discerned in the specimen and seems to be missing, as is the phalanx digiti minoris.
The pelvis (Fig. 4A-C, E-G) is quite short compared to the size of the bird and measures only about half of the length of the skull.In Prophaethon, by contrast, it is distinctly longer, and even the incomplete pelvis preserved in the holotype of P. shrubsolei (Fig. 4C, I) already reaches two thirds of the skull length (the pelvis of other fossil Phaethontiformes is unknown).The holotype includes the right os coxae, which is detached from the synsacrum and moved onto the left side of the latter bone.In contrast to Prophaethon and extant Phaethontidae (Fig. 4D, J), the ilium therefore appears not to have been co-ossified with the synsacrum; in Prophaethon even the cristae iliacae dorsales of the alae praeacetabulares ilii are tightly co-ossified with the crista spinosa synsacri, whereas the cristae iliacae dorsales are widely separated in extant Phaethontiformes (Mayr 2015).Judging from the narrowness of the synsacrum and the width of the right os coxae, the pelvis of Clymenoptilon was mediolaterally compressed as in Prophaethon, whereas the pelvis of extant Phaethontiformes is very wide.The ala praeacetabularis ilii is dorsoventrally deeper than in extant Phaethontidae and proportionally shorter than in Prophaethon (even though the rostral-most portion of the ala appears to be broken in the C. novaezealandicum specimen); its ventral margin is markedly concave.The foramen ilioischiadicum is large, as it is in Prophaethon.Further as in Prophaethon, there is a small tuberculum praeacetabulare.Unlike in extant Phaethontidae but also as in Prophaethon, a foramen obturatum is absent.

Phylogeny
In addition to an overall resemblance to other tropicbirds in the morphology of the skull, the wing bones, and the coracoid, Clymenoptilon novaezealandicum is assigned to the Phaethontiformes on the basis of several derived features, including (1) a sulcus along the lateral surface of the mandible, (2) the fossa temporalis forming a very deep notch that is bounded by a long, caudolaterally directed processus postorbitalis and a laterally projected, dorsoventrally deep and ridge-like processus zygomaticus, (3) a synsacrum exhibiting less than 14 co-ossified vertebrae, (4) a coracoid with a large and mediolaterally wide extremitas omalis, and (5) a proximodistally long processus extensorius of the carpometacarpus that has a broadly rounded tip.The long nostrils are a plesiomorphic feature shared with Prophaethon, with which Clymenoptilon agrees in the very narrow pelvis (see below).As in other Phaethontiformes, the humerus has a long crista deltopectoralis.
The holotype of C. novaezealandicum is distinctly larger than the putative phaethontiform from the Waipara Greensand reported by Mayr & Scofield (2016), with the proximal width of the humerus, from the tuberculum dorsale to the crista bicipitalis, being 24.4 versus 17.0 mm and the width of the humerus shaft being 8.9 versus 6.0 mm.In size, the unnamed smaller specimen (CM 2010.108.4)corresponds to Lithoptila abdounensis and Zhylgaia aestiflua from the late Paleocene of Morocco and Kazakhstan, respectively.The original description of CM 2010.108.4 was based on a partial proximal humerus and the proximal end of a carpometacarpus.After the study was published, more fragments of the specimen came to light, which were kept separate from the material described by Mayr & Scofield (2016).This additional material includes the cranial extremity of the scapula and the proximal-most portion of the humerus with the caput humeri (Fig. 5K, L).Apart from its smaller size, CM 2010.108.4 differs from C. novaezealandicum in the more pointed processus extensorius of the carpometacarpus, so that at least two differently sized phaethontiform species occurred in the mid-Paleocene of the Waipara Greensand.
The affinities of the Moroccan taxon Phaethusavis cannot be convincingly established without the discovery of further material, but we note that the holotype of P. pelagicus differs from the humerus of other phaethontiforms in the more proximally located apex of the crista deltopectoralis.The fragmentary representation of Zhylgaia, of which only partial humeri and a tarsometatarsus are known, limits meaningful comparisons with Clymenoptilon, and in the following we focus on the interrelationships of the three well-represented taxa Clymenoptilon, Lithoptila and Prophaethon (Bourdon et al. 2008a hinted at the possibility that Lithoptila may be a junior synonym of Zhylgaia, but the tarsometatarsus of "Tshulia littoralis", which was identified as that of Zhylgaia aestiflua by Mayr & Scofield 2016, differs from the tarsometatarsus of Lithoptila in being more elongated and in the shape of the distal end; compare Mayr & Scofield 2016: fig. 2F, G with Bourdon et al. 2008a: fig.2M, O).
The very narrow pelvis distinguishes Clymenoptilon and Prophaethon from extant Phaethontiformes, in which the pelvis is much wider (Fig. 4J; the pelvis of Lithoptila is unknown).A narrow pelvis is absent in the Eurypygiformes and may be an apomorphy of a clade including Clymenoptilon and Prophaethon.On the other hand, early diverging members of the Aequornithes also have a narrow pelvis, and the pelvis of Prophaethon closely resembles that of the procellariiform Diomedeidae (albatrosses; Mayr 2015).
In its relative length, the short pelvis of Clymenoptilon corresponds to that of extant Phaethontidae, whereas the pelvis of Prophaethon is proportionally much longer than that of Clymenoptilon.This indicates different locomotory characteristics of both taxa, and because a long and narrow pelvis is typically found in birds that use their hind limbs for aquatic propulsion, such as the Gaviiformes, Podicipediformes and Diomedeidae, Prophaethon may have been more aquatic than Clymenoptilon, foraging at or below the sea surface.However, a well-founded functional interpretation of the pelvis morphology of the new taxon is impeded by the absence of hind limb bones.
Pelvis morphology may support a sister-group relationship between Clymenoptilon and Prophaethon.On the other hand, Prophaethon and crown group Phaethontiformes exhibit a distinct, synovial nasofrontal hinge, which is caudally bordered by a dorsally bulging portion of the frontal bone.This derived morphology (by outgroup comparison with the Eurypygiformes and most other neornithine birds) is absent in Clymenoptilon and Lithoptila, which suggests that Prophaethon and Phaethon form a clade to the exclusion of the former two taxa.In this case, the narrow pelvis of Clymenoptilon and Prophaethon is likely to be plesiomorphic for the Phaethontiformes.
In the known skeletal elements, C. novaezealandicum shows an overall resemblance to Lithoptila abdounensis, but Clymenoptilon differs from both Lithoptila and Prophaethon in that the foramen magnum and the condylus occipitalis are almost of the same width.In Lithoptila, Prophaethon, Phaethon and most other extant birds, the condylus occipitalis is much smaller than the foramen magnum, but an equally small foramen magnum and large condylus occipitalis to those of Clymenoptilon occur in the Pelagornithidae (Mayr 2008, Bourdon et al. 2010).Here it is considered likely that the small foramen magnum and large condylus occipitalis are plesiomorphic traits that support a position of Clymenoptilon outside a clade formed by Lithoptila, Prophaethon and Phaethon.
Clymenoptilon also has a proportionally shorter ulna than extant Phaethontiformes, which indicates that the fossil taxon was less aerial.No complete ulna of Lithoptila is known, but based on several partial specimens Bourdon et al. (2008a, table 1) considered the bone to have been significantly longer than the humerus.If these length estimates are correct, the short ulna may be another plesiomorphic feature distinguishing Clymenoptilon from Lithoptila and Phaethon (the ulna of Prophaethon has not been described).

Biogeography
After the pelagornithid Protodontopteryx ruthae described by Mayr et al. (2021), Clymenoptilon novaezealandicum is the second formally named seabird species from the Waipara Greensand to show affinities to taxa from the early Paleogene of the Northern Hemisphere.The new find confirms the coexistence of the Phaethontiformes and Pelagornithidae in the Southern Hemisphere, with both taxa having already been known to occur together at various Northern Hemispheric sites (Mayr 2022, p. 236).By contrast, and except for the Sphenisciformes (penguins), all seabird taxa that today predominate in marine environments of the Southern Hemisphere-that is, species of the Suliformes (frigatebirds, gannets, cormorants and allies), Charadriiformes (gulls, skuas and allies), and Procellariiformes (albatrosses, petrels, and allies)-have not been reported from the Waipara Greensand so far.
The pelagornithid Protodontopteryx ruthae exhibits a more plesiomorphic morphology than its early Paleogene Northern Hemispheric relatives, and the Pelagornithidae may have originated in the Southern Hemisphere (Mayr et al. 2021).If C. novaezealandicum can be conclusively shown to be the sister taxon of a clade including Lithoptila, Prophaethon and crown group Phaethontiformes, it also hints at a Southern Hemispheric centre of origin of phaethontiforms.
Paleocene Phaethontiformes had widely disparate geographic occurrences in New Zealand (Clymenoptilon), Northern Africa (Lithoptila), Europe (Prophaethon), western Asia (Zhylgaia), and North America (undetermined taxon).These birds therefore seem to have rapidly dispersed across the globe, and potential geographic barriers, such as less productive marine zones around the equator, apparently did not constitute dispersal barriers.The wide distribution of the Phaethontiformes and Pelagornithidae in the early Paleogene stands in sharp contrast to the geographic restriction of coeval diving seabirds, with taxa such as the Vegaviidae and Sphenisciformes having been, and still being in the case of the Sphenisciformes, restricted to the Southern Hemisphere.Apparently, different factors limited the dispersal of aquatic and pelagic seabird taxa in the early Paleogene, but it has yet to be determined why, for example, penguins were not able to disperse into the Northern Hemisphere.
on the true right bank of the Waipara River, Canterbury, New Zealand.Located in situ at Site S2 (Fig. 1), about 40 m away from the east end of this site; New Zealand Map Grid Coordinates: 43 03 0 27ʺ S, 172 35 0 47ʺ E (New Zealand Fossil Record Number M34/f1077 [skull] and M34/f1078 [postcranial remains]).Collected in December 2020 (skull) and February 2021 (postcranial remains) by LL; late early Paleocene (late lower Teurian) to early late Paleocene (middle upper Teurian; the range is based on the dinoflagellate cyst assemblage, with the presence of Apteodinium 'trifolliculum' (sensu Browne

Figure 1 .
Figure 1.Locality map and stratigraphic chart of the Waipara Greensand in the Canterbury region, New Zealand.The three main sites where Paleocene bird fossils (mainly stem group Sphenisciformes) were found are denoted as S1, S2 and S3, following Mannering & Hiller (2008).

Figure 4 .
Figure 4. Pelvis of Clymenoptilon novaezealandicum, gen.et sp.nov.from the Waipara Greensand in New Zealand (holotype, UC 22048) in comparison to that of Prophaethon shrubsolei from the early Eocene London Clay of the Isle of Sheppey, Essex, UK (holotype; NHMUK a 683) and the extant Phaethon lepturus (SMF 9969).A, B, Right os coxae of C. novaezealandicum in lateral view; in B, the surrounding matrix was digitally brightened, the dotted line indicates the reconstructed caudal margin of the foramen ilioschiadicum.C, D, Pelvis of C, P. shrubsolei and D, P. lepturus in lateral view.E-G, Right os coxae of C. novaezealandicum in dorsal view; in F, the surrounding matrix was digitally brightened.H, Synsacrum of C. novaezealandicum in ventrolateral view.I, J, Pelvis of I, P. shrubsolei and J, P. lepturus in dorsal view; the dotted line in J indicates the outline of the left ilium; in I, the surrounding matrix was digitally brightened.Abbreviations: api, ala praeacetabularis ilii; fac, foramen acetabuli; fem, proximal portion of left femur; fii, foramen ilioischiadicum; pub, pubis.The scale bars equal 10 mm.