High-latitude neonate and perinate ornithopods from the mid-Cretaceous of southeastern Australia

Dinosaurs were remarkably climate-tolerant, thriving from equatorial to polar latitudes. High-paleolatitude eggshells and hatchling material from the Northern Hemisphere confirms that hadrosaurid ornithopods reproduced in polar regions. Similar examples are lacking from Gondwanan landmasses. Here we describe two non-iguanodontian ornithopod femora from the Griman Creek Formation (Cenomanian) in New South Wales, Australia. These incomplete proximal femora represent the first perinatal ornithopods described from Australia, supplementing neonatal and slightly older ‘yearling’ specimens from the Aptian–Albian Eumeralla and Wonthaggi formations in Victoria. While pseudomorphic preservation obviates histological examination, anatomical and size comparisons with Victorian specimens, which underwent previous histological work, support perinatal interpretations for the Griman Creek Formation femora. Estimated femoral lengths (37 mm and 45 mm) and body masses (113–191 g and 140–236 g), together with the limited development of features in the smallest femur, suggest a possible embryonic state. Low body masses (<1 kg for ‘yearlings’ and ~20 kg at maturity) would have precluded small ornithopods from long-distance migration, even as adults, in the Griman Creek, Eumeralla, and Wonthaggi formations. Consequently, these specimens support high-latitudinal breeding in a non-iguanodontian ornithopod in eastern Gondwana during the early Late Cretaceous.


Localities, Geological Settings, and paleoenvironments
The two new femora (LRF 0759 and LRF 3375) were recovered from subterranean exposures of the Griman Creek Formation (Rolling Downs Group, Surat Basin) near the town of Lightning Ridge in central-northern New South Wales, Australia (Fig. 1). Both specimens derive from laterally extensive, but discontinuous, clay-rich horizons (informally, the 'Finch Clay' facies) within the Wallangulla Sandstone, collectively interpreted as a lowland fluvial system punctuated by large freshwater lakes 9 . Rivers likely drained north-northeast into the epicontinental Eromanga Sea. Intermittent connections between some of these lakes and the inland sea is evinced by the rare occurrence of marine vertebrates (e.g., sharks and plesiosaurs); their rarity, however, and the occurrence of exclusively freshwater invertebrate taxa (e.g., gastropods and unionid bivalves) indicate that such marine connections lay distal to the study area at Lightning Ridge 9 . Recent detrital zircon analyses constrained the minimum depositional age of the Griman Creek Formation at Lightning Ridge to the early to mid-Cenomanian (100.2-96.6 Ma) 9 . During the Cenomanian, the area was located at a paleolatitude of ~60°S 20 , approaching the paleoantarctic circle. The paleoclimate at these latitudes was one of high precipitation (humid-perihumid) 28 and mild temperatures, given the diversity of Testudines and Crocodylomorpha [29][30][31][32] . The area possibly had a mean annual average temperature (MAAT) of ~14 °C, based on the minimum thermal tolerance of modern crocodylians 33 .
The specimens from the Eumeralla Formation (Aptian-Albian, Otway Group; NMV P186004, NMV P198900) come from the Dinosaur Cove locality, Otway Basin, Victoria, Australia. Recent fission track-calibrated biostratigraphic work places this locality within the early Albian Crybelosporites striatus Zone 34 . Specimens from the slightly older Wonthaggi Formation (Barremian-Aptian, upper Strzelecki Group; NMV P198982, NMV P216768) derive from the Flat Rocks locality, Gippsland Basin, Victoria, Australia, which has been assigned to the late Aptian upper Cyclosporites hughesii Zone 35 . Both localities were deposited at a paleolatitude of ~70°S 20 within extensional terrains of the Australian-Antarctic rift valley. Paleoenvironmental reconstructions suggest floodplains, freshwater lakes, and braided streams dominated the valley lowlands 12,[36][37][38] . The paleoclimate in this region was wet (humid-perihumid), highly seasonal and potentially much cooler (MAAT = −6 to +10 °C) 36 than that of the Griman Creek Formation. A cold paleoclimate, including winter freezing, is supported by extremely low δ 18 O values of meteoric fluids, the presence of possible cryoturbation structures (earthy hummocks and involutions), and the leaf size and physiognomy of local flora 28,36,39-42 . institutional abbreviations. AM, Australian Museum, Sydney, New South Wales, Australia, LRF, Australian Opal Centre, Lightning Ridge, New South Wales, Australia, MV, Museums Victoria, Melbourne, Victoria, Australia (formerly, National Museum of Victoria (NMV)).

Methods
Previous histological work on a subset of nine ornithopod femora from the Eumeralla and Wonthaggi formations documented between 0 and 8 cyclical growth marks (CGMs), regarded as a record of annual growth 43,44 . In addition, the presence of an external fundamental system (EFS) in some of the largest specimens marks the cessation of appreciable growth and the onset of skeletal maturity 25 . The smallest of these femora (NMV P216768; length≈48 mm) lacks CGMs and displays an abrupt transition from fibro-lamellar to parallel-fibred tissues hypothesized to be a 'hatching line' 25 , similar to that seen in a neonate sauropod from Madagascar 45 . A 'hatching line' is a distinct transition of bone texture observed at the point of hatching in squamates, crocodilians, and some birds, and the point of birth in some mammals. As a result, the size of NMV P216768 was chosen as arbiter for femora of animals zero to one year old and a threshold of < 60 mm was proposed for femoral length 25 . Body mass estimation. Body mass was estimated using the R package MASSTIMATE v. 1.4 49,50 , which now incorporates the developmental mass extrapolation (DME) approach based on Erickson and Tumanova 51 (see Supplementary Information for R script). DME is the most appropriate mass estimation approach for juveniles as it cannot be assumed that intraspecific growth-related patterns will follow adult-based interspecific circumference-mass relationships such as those originally proposed by Anderson et al. 52 . To conduct the DME, first the mass of an adult representative (BM adult ) was calculated using its minimum circumference and the bipedal corrected equation from Campione et al. 50 . This value was then scaled based on the proportion between the cube of the juvenile femoral lengths (FL juvenile  NMV 177935 was chosen as the adult proxy for the DME (BM adult ) as it is the largest femur (208 mm) found to display an external fundamental system 25 .
Age estimation. The ontogenetic stage or age of fossil vertebrates is usually investigated histologically 53 .
Periodic variation in bone textures observed in histological analyses can indicate seasonal growth and cyclical growth marks (CGMs), which can act as a proxy for the individual's age. For this approach to be viable, the microtexture of the target bone must be well preserved. In instances of pseudomorphic preservation, such as the opalized Griman Creek Formation specimens in this study, bone microtexture is lost in preservation and, therefore, alternatives to histology must be employed.
In order to estimate the ages of the Griman Creek Formation ornithopod femora, we first explored the relationships between age and circular growth mark (CGM) circumference in Australian small ornithopod femora published in the supplementary information of Woodward et al. 25 . Three specimens with multiple CGMs (NMV P221151, NMV P186326, NMV P177935) and two smaller specimens (NMV P216768, NMV P208495) were included. Both NMV P221151 and NMV P186326 followed an evident linear pattern and were thus expressed as an OLS model, whereas NMV P177935 follows a non-linear power function and was modeled via a nonlinear least squares (NLS) model.
To maximize the sample used for age estimation, an overall OLS linear regression was generated between age and CGM circumferences in all available specimens ( Table 1). The validity of this overall estimation model was evaluated by calculating its 95% prediction errors, which were compared to the OLS and NLS models derived from single specimens. Ages for the focal specimens (LRF 0759, LRF 3375) were predicted based on their surface circumferences, working on the assumption that CGMs represent the surface circumferential dimensions of the bone at the point in time when they were created 54,55 . While the CGM circumferences of specimens from www.nature.com/scientificreports www.nature.com/scientificreports/ Woodward et al. 25 were measured digitally from thin section images, the dimensions of our focal specimens were measured by hand; circumferences of these femora were calculated as ellipses using mediolateral and anteroposterior diameters as the elliptical axes. Diameters were taken with digital calipers at the narrowest available section of the femur. Ellipse circumference was approximated using one of the Ramanujan formulations based on h, where a is half of the anteroposterior diameter and b is half of the mediolateral diameter. The circumference (C) can then be approximated through, Due to the preservation of LRF 3375, only the anteroposterior diameter could be taken and thus a circular circumference was calculated.

Results
Description of material from the Griman creek formation, new South Wales. LRF 0759 (maximum length = 16 mm) and LRF 3375 (maximum length = 16 mm) are partial proximal right femora. LRF 0759 is broken just below the fourth trochanter ( Fig. 2), whereas LRF 3375 is broken obliquely just below the base of the lesser trochanter so that the medial portion of the femoral shaft is also missing (Fig. 2b). The two specimens differ noticeably in the form and orientation of the femoral head. In LRF 0759, the femoral head is rounded-triangular in anterior and proximal views and is angled, ~65°, to the diaphysis (Fig. 2), resulting in a ventral position for the femoral head relative to the greater trochanter. In LRF 3375, the femoral head is more 'tongue'-shaped in anterior view and forms an angle of ~110° to the diaphysis (Fig. 2). The femoral head in LRF 3375 is sheared in a vertical plane, but terminates dorsal to the greater trochanter. Other features in both specimens are similar. A distinct fossa trochanteris is absent. The lesser trochanter is significantly lower than the greater trochanter and is separated from it by a shallow cleft (Fig. 2). The lateral contour of the greater trochanter is gently convex in proximal view. The preserved portion of the femoral diaphysis is straight in lateral and medial views and the low, arcuate fourth trochanter (present only on LRF 0759) appears to be within the proximal half of the femur (Fig. 2). A broad depression on the medial surface of the femoral diaphysis, adjacent to the fourth trochanter, marks the insertion site for the M. caudofemoralis longus (Fig. 2).
Description of material from the eumeralla formation, Victoria. NMV P186004 (length = 47 mm) and NMV P198900 (length = 55 mm) are complete, well-preserved left femora (Table 2; Fig. 3). Both femora are similar in most respects, except for the shape of the proximal end. In NMV P186004, the femoral head is angled approximately 90° to the diaphysis and the femoral head appears to be at the same level as the greater trochanter, www.nature.com/scientificreports www.nature.com/scientificreports/ although damage to the proximal end makes this uncertain (Fig. 3). By contrast, the femoral head of NMV P198900 is angled at approximately 80° to the diaphysis, the relatively low angle accentuated by the rather sinuous profile of the femur in anterior view (Fig. 3). As a result, the greater trochanter is higher than the femoral head ( Fig. 3). In NMV P198900, the femoral head and greater trochanter are separated by a shallow, saddle-shaped fossa trochanteris (Fig. 3). The form of the fossa trochanteris in NMV P186004 is unclear due to damage to the proximal end of the femur. The lesser trochanter terminates below the level of the greater trochanter and the two are separated from each other by a short, deep cleft. The femoral diaphyses are bowed in lateral view, although NMV P198900 displays a sigmoidal (rather than straight) anterior profile owing to the medial tilt of the proximal end. The fourth trochanter is large and pendant (but missing the distal extremity) and within the proximal half of the femur. The M. caudofemoralis longus insertion scar forms a distinct depression on the medial surface of the femoral shaft, adjacent to the fourth trochanter. The distal end is mediolaterally expanded to form the medial and lateral condyles. In NMV P198900 the medial condyle is the larger of the two, whereas the opposite is true for NMV P186004. The lateral margin of the lateral condyle is convex in distal aspect (i.e. a medially inset lateral condylid is absent). The distal flexor fossa is broad and U-shaped in distal view (Fig. 3), whereas the extensor fossa on the anterior surface of the femur is weakly developed and not visible in distal aspect (Fig. 3).
Description of material from the Wonthaggi formation, Victoria. NMV P198982 (length = 61 mm) and NMV P216768 (length = 48 mm) are nearly complete right femora lacking their distal articular ends (Fig. 4). NMV P198982 is diagenetically compressed anteroposteriorly, particularly at the proximal end. As a result, the femoral head appears bladelike in proximal view and oriented at approximately 80° to the diaphysis (in anterior view). The greater trochanter is higher than the femoral head proximally, but distortion and damage on both features obscures the transition between them (Fig. 4). In NMV P216768, damage to the proximal end renders the form and orientation of the femoral head and the form of the fossa trochanteris unclear. The proximal end of the lesser trochanter is lower than the greater trochanter, with a shallow cleft between them. The lateral contour of the greater trochanter is flat in NMV P216768 (in proximal aspect) but cannot be accurately observed in NMV P198982 due to distortion.
In both specimens, the femoral diaphysis is straight in anterior view and gently bowed in lateral view. The fourth trochanter is proximally positioned but incomplete in both specimens. A depression for the M. caudofemoralis longus is found medially, immediately adjacent to the base of the fourth trochanter. In NMV P198982, where the distal end is partially preserved, the flexor fossa forms a broad, open 'U' in distal aspect. A distinct lateral condylid is present, and is medially inset from the lateral edge of the lateral condyle (Fig. 4). The extensor fossa is shallow and not visible in distal aspect. The corresponding distal portion is missing in NMV P216768. femur length, body mass, and age estimates. The reconstructed lengths of the two Griman Creek Formation specimens, LRF 0759 and LRF 3375, were estimated at 36 mm (upper maximum value of 59 mm) and 39 mm (upper maximum value of 62 mm) long, respectively (Table 3; Fig. 5a). Length estimates for two larger incomplete small-ornithopod femora from the Griman Creek Formation (AM F105673, AM F127930) were also included (Fig. 5a). These were 82-125 mm and 106-150 mm long, respectively: substantially larger than the focal specimens. Mass estimates, calculated via DME, indicate that LRF 0759 and LRF 3375 represent the smallest individuals thus far sampled. Based on the point femoral length estimate and the ~25% mean per cent prediction error of the adult proxy, the body mass of LRF 0759 was likely between 113 and 191 g and that of LRF 3375 was likely between 140 and 236 g. Mass estimates for the juveniles from the Eumeralla and Wonthaggi formations ranged from 251-424 g for the smallest (NMV P186004) to 788-1332 g for the largest (NMV P179561) ( Table 3).
Exploration of growth-related changes between CGM circumferences and ages reveal substantial differential scaling among the three NMV specimens (Fig. 5b). Nevertheless, at the size of the focal Griman Creek Formation specimens, the three models occur within the 95% prediction intervals of the overall OLS model supporting its use as a general age-estimation model. Given the overall OLS model, the ages of LRF 0759 and LRF 3375 are estimated at -0.68 years (-2.62 to 1.27 based on the 95% prediction error) and -0.50 years (-2.42 to 1.43), respectively (Fig. 5b). The complete list of length, mass, and age estimates is presented in Table 3.   www.nature.com/scientificreports www.nature.com/scientificreports/ Discussion taxonomic considerations. Early work on the ornithopod femora from the Eumeralla and Wonthaggi formations indicated as many as four morphotypes 23 , although this was later revised to two 24 . Nevertheless, four taxa have been named from these deposits: Leaellynasaura amicagraphica, Qantassaurus intrepidus, Atlascopcosaurus loadsi, and Diluvicursor pickeringi 23,24,37 . Molnar and Galton 56 identified two femoral morphotypes from the Griman Creek Formation and craniodental remains suggest that up to three small-bodied non-iguanodontian ornithopods were present 22 . Larger-bodied iguanodontians were also present in the Griman Creek Formation but are absent from the Eumeralla and Wonthaggi formations 9,22,57 . Our results support the occurrence of at least two morphotypes in the Griman Creek Formation based on differences in the inclination of the femoral head (Fig. 2). Two morphotypes are also collectively recognized from the Wonthaggi and Eumeralla formations, based on the presence/absence of a medially inset lateral condylid (Table 2). However, these morphological distinctions are used merely for convenience, as femoral head inclination and the presence/absence of a lateral condylid are variably represented within the sample making the taxonomic definition of these morphotypes highly ambiguous.  www.nature.com/scientificreports www.nature.com/scientificreports/ The pervasiveness of diagenetic alteration among the specimens from the Wonthaggi and Eumeralla formations similarly renders any taxonomic conclusions, based on these divisions, dubious 24 . Nevertheless, such morphological variations imply that several taxa are likely represented by the remains.
Characters of femora described in this study differ from those of sauropods and theropods. The well-defined femoral head and lesser trochanter are unlike those of many sauropods and thyreophorans, including known juvenile specimens, where both features are relatively ill-defined (e.g., Burns et al. 58 ). The tongue-shaped lesser trochanter, which is separated by a narrow cleft from the greater trochanter, is unlike that in theropods (e.g., Griffin 59 ; Hutchinson 60 ). Other theropod (and saurischian) synapomorphies, such as a distinct fovea capitis, a deep sulcus for the ischiofemoral ligament 61 , a dorsolateral trochanter, and a trochanteric shelf, are also absent, although the absence of the latter two may be ontogenetically variable among certain theropods 59 . Although finer taxonomic assessments are impossible, all femora described in this study can be constrained to Ornithopoda based on: a bowed femoral shaft, a proximally positioned and pendant-shaped fourth trochanter, a shallow (or absent) distal extensor fossa, and an 'open' distal flexor fossa (e.g., Norman et al. 62 ; Rich and Rich 23 ; Rich and Vickers-Rich 24 ), although not all of these features are preserved in each specimen due to the variable degrees of completeness. The two smallest specimens (LRF 0759, LRF 3375), which do not preserve most of these features because of their incompleteness and/or inferred juvenile state, possess additional characters widespread among non-iguanodontian ornithopods, including: a proximally positioned fourth trochanter and a finger-like (not blade-like) lesser trochanter that is closely appressed (rather than widely separated) to the greater trochanter. In these specimens, the straight proximal margin between the caput and trochanteric ends of the proximal femur (i.e. absence of a fossa trochanteris, the presence of which is a typical cerapodan feature 63 ) is interpreted to be the result of the early developmental stage represented by these individuals.   www.nature.com/scientificreports www.nature.com/scientificreports/ ontogenetic interpretations. Due to the absence of bone microtexture within our focal Griman Creek Formation specimens, age estimations are based on comparisons with other Australian small ornithopod femora that were histological investigated by Woodward et al. 25 . Woodward and colleagues found that the growth sequences of Eumeralla-and Wonthaggi-formation small ornithopod femora displayed weakly asymptotic trends, with the annual increase in CGM circumference appearing to plateau as maturity was reached. Bone texture also reflected this growth pattern, with a shift from predominantly fibro-lamellar and poorly organized parallel fibers to a primarily parallel-fibred texture after the third CGM. Overall, there was little variation in growth rate between specimens, with the exception of NMV P180892. This much larger specimen was included to test the hypothesis that the smaller examples were in fact juveniles of a larger form. However, the presence of small femora from apparently mature individuals (exhibiting an EFS; maximum length 208 mm) within the sample, and the fact that this larger specimen was yet to reach maturity in spite of its large size (lacking an EFS at an estimated length of 315 mm), indicates that this taxon could be distinguished from the rest of the sample for its differential growth regime. With such small specimens exhibiting an EFS, the Australian small ornithopods studied in Woodward et al. 25 appear to have been diminutive, even as adults.
The smallest femora within our sample are inferred to be from either neonate (NMV P216768, NMV P186004) or perinate (LRF 3375, LRF 0759) individuals based on their overall size and the presence of an apparent 'hatching line' in NMV P216768 25 . Age estimates from the regression based on CGM circumferences ( Fig. 5b; Table 3) further support this conclusion. The smallest known Victorian specimen (MNV P208159) reported by Rich and Vickers-Rich 24 presumably also represents a perinate. As the two Griman Creek Formation specimens (LRF 3375, LRF 0759) were incomplete, the diameter measurements are likely to exceed the true narrowest section of the femur. Therefore, point length, mass, and age estimates may well represent overestimates. In the absence of direct histological evidence, femora larger than NMV P216768, but less than 60 mm in length, were considered to be between neonate and one year old, informally referred to here as a 'yearling' . Mass estimates for presumed perinates are around half that of the neonate NMV P216768, which further support this identification (Table 3). Although slightly larger than these specimens, the body mass of NMV P198900 overlaps with the range of other neonates (NMV P216768, NMV P186004) and may represent a third neonatal specimen ( Table 3). The largest of the juvenile femora, NMV P198982, may also represent an individual less than one year old, here referred to as a 'yearling' .
Although missing its distal half, the perinate femur LRF 0759 has a low, crescentic fourth trochanter and is unlike the typical pendant-shaped trochanter seen in more mature non-iguanodontian ornithopod individuals 62,64 including the neonate NMV P216768. Notably, this feature is similar to the low, triangular fourth trochanter typically seen in Hadrosauroidea and some non-hadrosauroid iguanodontians [64][65][66][67] . A low, triangular fourth trochanter is also present in perinatal Hypacrosaurus stebingeri 68 but was apparently absent in perinatal Saurolophus angustirostris 69 . The similarity in fourth trochanter morphology between the Griman Creek Formation specimens and most iguanodontians hints at a possible heterochronic shift in the evolutionary history of ornithopods, in which iguanodontians paedomorphically (specifically via neoteny, reduction in developmental rate) retain juvenile characteristics into adulthood. Independent of the functional nature of the pendant fourth trochanter in adult non-iguanodontian ornithopods 64 , a hypothesized association between the loss of this feature and quadrupedality in ornithischians (sensu Maidment and Barrett 70 ) presents the intriguing possibility that the evolution of quadrupedality results from such heterochronic transitions, at least in ornithopods. Similar shifts, associated with limb proportions, have been proposed for the evolution of quadrupedality in sauropods 71,72 . This hypothesis suggests that juvenile ornithopods are, if anything, less bipedal than their adult counterparts (but see Dilkes 73 ) or, at least, limited in their degree of function at these early stages of development. Limited functionality is supported by the absence of a distinct fossa trochanteris in LRF 0759, LRF 3375, and NMV P198982 and the absence of insertion scars for the M. iliotrochantericus on the posterior surface of the femoral head (except for NMV P198900 and NMV P186004) and the M. caudofemoralis brevis on the fourth trochanter. Their absence or limited development in the present material implies the gradual acquisition of these traits through repetitive muscle action as the individual grew 74 . ornithopod nesting environments at high palaeolatitudes. The presence of perinate and neonate ornithopods from high-palaeolatitudes in south-eastern Australia invites palaeoenvironmental comparisons with similar nesting sites from the Northern Hemisphere. To date, evidence of ornithopod breeding at high-palaeolatitudes in the Northern Nemisphere is limited to hadrosaurid remains discovered in U.S.A., Canada, and Russia. The late Maastrichtian Kakanaut Formation (70-75°N palaeolatitude) 75 in north-eastern Russia produces hadrosaurid and other non-avian dinosaur eggshell fragments 8 . A low-energy lowland depositional environment was inferred from fine-grained sediments and the presence of fish remains within the fossiliferous lens. Ectothermic tetrapods, such as crocodilians and turtles, appear to have been absent. In the Campanian units of the Wapiti Formation (western Canada; ~65°N palaeolatitude), hadrosaurid nesting is evinced by hatchling-sized skeletal elements 19 . The palaeoenvironmental setting was interpreted as a low-energy fluvial deposit, punctuated by floods and coal-forming wetlands 19,76 . Squamates and turtles were present, perhaps reflecting a milder climate as resulting from the proximity to the Western Interior Seaway. The most northerly evidence of dinosaur breeding comes from the Maastrichtian Prince Creek Formation in northern Alaska, U.S.A. (~85°N palaeolatitude) 7 . Hatchling-sized hadrosaurids, along with small-bodied non-iguanodontian ornithopods and dromaeosaurid teeth, were recovered from channel-lag, overbank, and pond deposits formed in lowland riverine, floodplain, and deltaic environments [77][78][79] . Amphibians and reptiles (excepting dinosaurs) were absent, although a partial turtle carapace was recovered from earlier, Cenomanian deposits on the Alaskan North Slope 7,79,80 .
High-latitude ornithopod breeding sites in the Northern Hemisphere were located in lowland settings whereas, at lower palaeolatitudes, nesting sites occupied both dry upland and wet lowlands regions 19 www.nature.com/scientificreports www.nature.com/scientificreports/ nesting strategies among hadrosaurid ornithopods. Whether such latitude-dependent strategies occur in non-iguanodontian ornithopods, however, remains unknown. Neonate tooth crowns attributed to the elasmarian Talenkauen santacrucensis 21 constitute the sole record of a 'nestling' non-iguanodontian ornithopod outside of Australia. These teeth derive from the Campanian-Maastrichtian Cerro Fortaleza Formation in southern Argentina, which was variably deposited in fluvial, fluvial-palustrine, and coastal floodplain environments [82][83][84] . Palaeogeographic reconstructions place southern Argentina at ~50°S 20 , considerably more equatorial than the Australian localities studied herein.
In general, the nesting sites of high-palaeolatitude ornithopods, including hadrosaurids from the Northern Hemisphere and non-iguanodontian ornithopods from the Southern Hemisphere, were restricted to moist, lowland settings. This limited record is undoubtedly biased, not least by the Australian fossil record, where appropriate strata-both in terms of age and depositional setting-are constrained to a handful of localities. Nevertheless, if the apparent association between ornithopod nesting sites and lowland environments truly reflects ornithopod nesting choices at high latitudes, it may be explained by temperature constraints on egg incubation. In oviraptorosaurs, estimates suggest bird-like incubation temperatures (~35-40 °C) 88 . Incubation temperatures have not been estimated for ornithopods, however, the porosity of their eggshells suggests the use of covered nests buried in organic matter, where bacterial respiration can provide the main source of heat 89 . The necessity for a stable nest microclimate for incubation purposes may have limited high-latitude ornithopods to breeding in lowland environments, which would have generally been warmer and less prone to extreme temperature fluctuations than upland settings 90 . implications for overwintering strategies. Migration was proposed as an overwintering strategy for high-palaeolatitude dinosaurs, although current evidence contends that most were year-round residents 1,2,8,14 . High-latitude zones receive above-average sunlight and can be highly productive during the summer, making these regions attractive breeding destinations for migratory animals and perennial residents. Migrators that breed at high latitudes are fast and efficient travelling animals, but migratory strategies are contingent on precocial young capable of joining their parents on the journey to lower latitudes prior to the winter season. Such migratory strategies are particularly prevalent among birds, which are able to travel rapidly and efficiently via flight 91 . In contrast, migration is rare among terrestrial vertebrates and long distance seasonal high-latitude migration is only known to occur in the caribou (Rangifer tarandus) 92 . In addition to developmental constraints, migration is limited by body size and its positive allometry with locomotory speed and energetic efficiency 1,7,93 . The smallest extant terrestrial vertebrate migrators are small antelopes with an adult body mass of 20 kg 93 . Osteohistological evidence suggests that Australian high-latitude ornithopods shared similar growth dynamics with their more temperate counterparts, such as Orodromeus 94 , growing rapidly in their formative years, an adaptation presumed to have improved their chances against predation and environmental pressures 26 . However, at body masses <1 kg, 'yearling' ornithopods fall well below this threshold (Tables 2 and 3). Small-bodied Australian ornithopods appear to have grown at a moderate rate and attained skeletal maturity between five to seven years and maximum body masses of approximately 20 kg (Table 3) 25 . This small body size and relatively low growth rate make seasonal migration an unlikely overwintering strategy for these small dinosaurs, thus it is most likely that they were obligate high-latitude residents.