Spinosaurus is not an aquatic dinosaur

A predominantly fish-eating diet was envisioned for the sail-backed theropod dinosaur Spinosaurus aegyptiacus when its elongate jaws with subconical teeth were unearthed a century ago in Egypt. Recent discovery of the high-spined tail of that skeleton, however, led to a bolder conjecture that S. aegyptiacus was the first fully aquatic dinosaur. The ‘aquatic hypothesis’ posits that S. aegyptiacus was a slow quadruped on land but a capable pursuit predator in coastal waters, powered by an expanded tail. We test these functional claims with skeletal and flesh models of S. aegyptiacus. We assembled a CT-based skeletal reconstruction based on the fossils, to which we added internal air and muscle to create a posable flesh model. That model shows that on land S. aegyptiacus was bipedal and in deep water was an unstable, slow-surface swimmer (<1 m/s) too buoyant to dive. Living reptiles with similar spine-supported sails over trunk and tail are used for display rather than aquatic propulsion, and nearly all extant secondary swimmers have reduced limbs and fleshy tail flukes. New fossils also show that Spinosaurus ranged far inland. Two stages are clarified in the evolution of Spinosaurus, which is best understood as a semiaquatic bipedal ambush piscivore that frequented the margins of coastal and inland waterways.


Introduction
In 1915, Ernst von Stromer announced the discovery in Egypt's Western Desert of the elongate jaws and partial skeleton of a large sail-backed predator Spinosaurus aegyptiacus (Stromer, 1915). Other bones found nearby (Stromer, 1934) contributed to his initial reconstruction of S. aegyptiacus as a sail-backed, piscivorous biped (Stromer, 1936), shortly before all of these bones were destroyed in World War II (Nothdurft et al., 2002;Smith et al., 2006). Over the last 30 years, additional skull and postcranial bones came to light in western Morocco in beds of similar age to those in Egypt (Russell, 1996;Dal Sasso et al., 2005;Smyth et al., 2020;Ibrahim et al., 2020a). Central among these 1. reverted to a quadrupedal stance on land, as shown by a trunk-positioned center of mass (Ibrahim et al., 2014;Ibrahim et al., 2020b), ostensibly knuckle-walking with long-fingered, long-clawed forelimbs; 2. functioned in water as a capable, diving pursuit predator using an expanded tail as a 'novel propulsor organ' (Ibrahim et al., 2020b) or as a 'subaqueous forager' (Fabbri et al., 2022); and 3. fossils would be found exclusively in coastal or deep-water marine habitats, like all large-bodied secondarily aquatic vertebrates, and would not be expected to be found in freshwater inland environments.
We test these three central propositions. Critique of the aquatic hypothesis thus far has focused on an alternative functional explanation for the high-spined tail as a display structure and largely qualitative functional interpretations of its skeletal anatomy (Hone and Holtz, 2021). Biomechanical evaluation of the aquatic functionality of S. aegyptiacus remains rudimentary. The propulsive capacity of the tail in water was judged to be better than terrestrial counterparts by oscillating miniature plastic tail cutouts in water (Ibrahim et al., 2020b), a limited approximation of the biomechanical properties of an anguilliform tail (Lighthill, 1969;van Rees et al., 2013;Gutarra and Rahman, 2022) that failed to take account of the bizarre anterior half of the animal. The center of body mass, a critical functional parameter, has been estimated for S. aegyptiacus three times, each estimate pointing to a different location ranging from the middle of the trunk (Ibrahim et al., 2014;Ibrahim et al., 2020b) to a position over the hind limbs (Henderson, 2018). Quantitative comparisons have not been made regarding the size or surface area of the limbs, hind feet, and tail of S. aegyptiacus to counterparts in extant primary or secondary swimmers.
Thus, adequate evaluation of the aquatic hypothesis requires more realistic biomechanical tests, quantitative body, axial and limb comparisons between S. aegyptiacus and extant primary and secondary swimmers, and a survey of bone structure beyond the femur and shaft of a dorsal rib. Such tests and comparisons require an accurate 3D digital flesh model of S. aegyptiacus, which, in turn, requires an accurate skeletal model. Hence, we began this study by assembling a complete set of CT scans of the fossil bones for S. aegyptiacus and its African forerunner, Suchomimus tenerensis .

Aquaphilic terminology
Aquatic status is central to the 'aquatic hypothesis.' The hypothesis holds that S. aegyptiacus is the first non-avian dinosaur bearing skeletal adaptations devoted to lifestyle and locomotion in water, some of which inhibited terrestrial function. The contention is that S. aegyptiacus was not only a diving pursuit predator in the open-water column, but also a quadruped on land with long-clawed forelimbs poorly adapted for weight support. A later publication seemed to downgrade that central claim by suggesting that any vertebrate with 'aquatic habits,' such as wading, submergence, or diving, had an 'aquatic lifestyle' (Fabbri et al., 2022). That broadened usage of 'aquatic lifestyle,' however, blurs the long-standing use of aquatic as applied to lifestyle (Pacini and Harper, 2008). We outline below the traditional usage of aquaphilic terms, which we follow.
We enveloped the skeletal model in flesh informed by CT scans revealing the muscle volume and air spaces in extant reptilian and avian analogs. To create a 3D flesh model for S. aegyptiacus (Figure 2A and B), internal air spaces (trachea, lungs, air sacs) were shaped and positioned as in extant analogs. We created three options for internal air volume based on extant squamate, crocodilian, and avian conditionsand assigned densities to body partitions based on local tissue types and air space. We calculated the surface area and volume of the flesh model as well as its component body parts.
We posed this integrated flesh model in bipedal, hybrid-and axial-powered poses, the latter two based on the swimming postures of extant semiaquatic reptiles (Grigg and Kirshner, 2015; Figure  2B). We calculated center of mass (CM) and center of buoyancy (CB) to evaluate the habitual two-or four-legged stance of S. aegyptiacus on land ( Figure 1A), the depth of water at the point of flotation ( Figure 2D), and the neutral position of the flesh model in deeper water (Figure 2A and B). Using biomechanical formulae (Lighthill, 1969) and data from extant alligators (Fish, 1984), we estimated the maximum force output of its tail, which was used to calculate maximum swimming velocity at the surface and underwater( Figure 3A). We also evaluated its stability, maneuverability, and diving potential in water ( Figure 3B), with all of these functional capacities compared to extant large-bodied aquatic vertebrates.
We turned to extant analogs to consider the structure and function of similar spine-supported sails over the trunk and tail in lizards and the form of tail vertebrae in tail-powered secondary swimmers ( Figure 4). We also considered the relative size (surface area) of appendages in a range of secondary swimmers (Figure 5), and how the surface area of foot paddles and tail scale in crocodylians ( Figure 6).
Lastly, we turned to the spinosaurid fossil record to look at the habitats where spinosaurid fossils have been found. We reviewed their distributionto determine whether spinosaurids, and S. aegyptiacus in particular, were restricted to coastal, marine habitats like all large secondarily aquatic vertebrates. We updated spinosaurid phylogeny in order to discern major stages in the evolution of spinosaurid piscivorous adaptations and sail structures (Figure 8)

Spinosaurid skeletal models
Our skeletal reconstruction of an adult S. aegyptiacus is just under 14 m long ( Figure 1A), which is more than 1 m shorter than previously reported (Ibrahim et al., 2014). Major differences are apparent when compared to the 2D graphical reconstruction of the aquatic hypothesis (Ibrahim et al., 2020b). The length of the presacral column, depth of the ribcage, and length of the forelimb in that reconstruction were overestimated by ~10, 25, and 30%, respectively, over dimensions based on CT-scanned fossils. When translated to a flesh model, all of these proportional overestimates (heavier neck, trunk, forelimb) shift the center of mass anteriorly (see 'Materials and methods').
The hind limb long bones (femur, tibia, fibula, metatarsals) in S. aegyptiacus lack the medullary cavity common to most dinosaurs and theropods in particular. When first discovered, the infilled hind limb bones in S. aegyptiacus were interpreted as ballast for swimming (Ibrahim et al., 2014). However, the infilled condition is variable as shown by the narrow medullary cavity in a femur of another individual slightly larger than the neotype (Russell, 1996;NMC 41869). Furthermore, the bone infilling is fibrolamellar and cancellous, similar to the infilled medullary cavities of other large-bodied terrestrial dinosaurs (Vanderven et al., 2014) and mammals . In contrast, dense pachystotic bone composes the solid and sometimes swollen bones of many secondarily aquatic vertebrates that use increased skeletal density as ballast (Houssaye, 2009).
( Figure 1C and J), and large air-filled pneumatic spaces are present in the centra and neural arches of cervical vertebrae (Evers et al., 2015;Figure 1B). Collectively, these less dense, internal marrow-and air-filled spaces in S. aegyptiacus more than offset the added density of infilled medullary space in the relatively reduced hind limb long bones ( Figure 1A). Hind limb bone infilling is better explained as compensation for the reduced size of the hind limb long bones that must support a body mass at the upper end of the range for theropods. Bending strength increases by as much as 35% when the medullary cavity is infilled (see Appendix 1).

S. aegyptiacus flesh model form and function
We added flesh to the adult skeletal model and divided the flesh model into body partitions adjusted for density. Muscle volume was guided by CT cross-sections from extant lizards, crocodylians, and birds ( Figure 2B), and internal air space (pharynx-trachea, lungs, paraxial air sacs) was modeled on lizard, crocodilian, and avian conditions ( Figure 2C; see 'Materials and methods,' Appendix 2). Wholebody and body part surface area and volume were calculated, and body partitions were assigned density comparable to that in extant analogs (see 'Materials and methods'). For biomechanical analysis, we positioned the integrated flesh model in bipedal stance ( Figure 1A) as well as hybrid-and axial-powered swimming poses (Grigg and Kirshner, 2015;Figure 2A and B).
The CM and CB of the flesh model were determined to evaluate habitual stance on land and in shallow water ( Figure 1A), the water depth at the point of flotation ( Figure 2D), and its swimming velocity, stability, maneuverability, and diving potential in deeper water ( Figure 3). No matter the included volume of internal air space, CM is positioned over the ground contact of symmetrically positioned hind feet ( Figure 1A, red cross). Thus, S. aegyptiacus had a bipedal stance on land as previously suggested (Henderson, 2018), contrary to trunk-centered CM of the aquatic hypothesis (Ibrahim et al., 2020b). Consistent with a bipedal stance, the manus is adapted for prey capture and manipulation (elongate hollow phalanges, scythe-shaped unguals) rather than weight support ( Figure 1A and D).
Adult S. aegyptiacus can feed while standing in water with flotation occurring in water deeper than ~2.6 m ( Figure 2D). In hybrid or axial swimming poses, trunk air space tilts the anterior end of the model upward (Figure 2A and B). With density-adjusted body partitions and avian-like internal air space, the flesh model of S. aegyptiacus has a body mass of 7390 kg and an average density of 833 kg/m 3 (see 'Materials and methods'), which is considerably less than the density of freshwater (1000 kg/m 3 ) and saltwater (1026 kg/m 3 ) or the average density of living crocodylians (1080 kg/m 3 ; Grigg and Kirshner, 2015).
Swimming velocity at the surface and underwater in extant lizards and crocodylians is powered by foot paddling and axial undulation (hybrid swimming; Frey and Salisbury, 2001) and at moderate to maximum (critical) speeds by axial undulation alone (axial swimming) (Fish, 1984;Grigg and Kirshner, 2015). We used Lighthill's bulk momentum formula to estimate maximum surface and underwater swimming velocity for the flesh model of S. aegyptiacus (Lighthill, 1969). Assuming a fully compliant Alligator-like tail (tail amplitude 0.24/body length, tail wavelength 0.57/body length, and tailbeat frequency 0. 25 Hz;Fish, 1984;Sato et al., 2007), tail thrust (P t ) and maximum velocity (U) can be determined (P t = -164.93 + 1899.1U -896.35U 2 ). Assuming turbulent conditions, a body drag coefficient of 0.0035 was estimated for a Reynolds number of 752,400 at a swimming speed of 1.0 m/s. The total power from estimates of drag increased threeto fivefold to account for undulation of the tail, near-surface wave formation, and increased sail drag when underwater ( Figure 3A). The addition of the sail increases the drag on the body of S. aegyptiacus by 33.4%. The intersection of the thrust power curve and drag power curves, where the animal would be swimming at a constant velocity, indicates slow maximum velocity at the surface (~0.8 m/s) and only slightly greater when submerged (~1.4 m/s) ( Figure 3A). Maximum tail thrust in S. aegyptiacus is 820 Watts (683 N or 154 lbs), a relatively low value for the considerable caudal muscle mass in this large theropod (Snively and Russell, 2007;Mallison et al., 2015). Only a minor amount of caudal muscle power, however, is imparted to the water as thrust during undulation. As a result, maximum velocity is only 1.2 m/s, an order of magnitude less than extant large-bodied (>1 m) pursuit predators. These species (mackerel sharks, billfish, dolphins, and killer whales) are capable of maximum velocities of 10-33 m/s (Tinsley, 1984;Fish, 1998;Fish and Rohr, 1999;Iosilevskii and Weihs, 2008).  Stability and the capacity to right are important in water. When positioned upright in water, the trunk sail of S. aegyptiacus is emergent ( Figure 3B, position 1). The flesh model, however, is particularly susceptible to long-axis rotation given the proximity of CM and CB, with stable equilibrium attained when floating on its side ( Figure 3B, position 3). Righting requires substantial torque (~5000 Nm) that is impossible to generate with vertical limbs and a tail with far less maximum force output (~700 N). This stability predicament remains even with the smallest internal air space. The absence of vertical stability and righting potential in water stands in stark contrast to the condition in extant crocodylians and marine mammals (Fish, 1998;Grigg and Kirshner, 2015).
Maneuverability in water (acceleration, turning radius, and speed) wanes as body length increases (Domenici, 2001;Parson et al., 2011;Domenici et al., 2014;Hirt et al., 2017;Gutarra and Rahman, 2022), which is further compromised in S. aegyptiacus by its rigid trunk (see below) and expansive, unretractable sail. In contrast, large-bodied secondary swimmers capable of pursuit predation in open water have fusiform body forms with a narrow caudal peduncle for efficient tail propulsion (ichthyosaurs, cetaceans; Motani, 2009), control surfaces for reorientation, and narrow extensions (bills) to enhance velocity in close encounters with smaller more maneuverable prey (Maresh et al., 2004;Domenici et al., 2014). Besides some waterbirds, semiaquatic pursuit predators are rare and include only the small-bodied (<2 m), exceptionally maneuverable otters that employ undulatory swimming (Fish, 1994).
Diving with an incompressible trunk requires a propulsive force (F g ) greater than buoyancy. For S. aegyptiacus, in addition, a depth of ~10 m is needed to avoid wave drag ( Figure 3A, bottom). The propulsive force required to dive is ~17,000 N:  Figure 2D), significantly reducing tail thrust. The now common depictions of S. aegyptiacus as a diving underwater pursuit predator contradict a range of physical parameters and calculations, which collectively characterize this dinosaur as a slow, unstable, and awkward surface swimmer incapable of submergence.

Axial comparisons to aquatic vertebrates and sail-backed reptiles
Axial flexibility is requisite for axial-propulsion in primary or secondary swimmers. However, in S. aegyptiacus, trunk and sacral vertebrae are immobilized by interlocking articulations (hyposphenehypantrum), an expansive rigid dorsal sail composed of closely spaced neural spines, and fused sacral centra ( Figure 1A). The caudal neural spines in S. aegyptiacus stiffen a bone-supported tail sail by an echelon of neural spines that cross several vertebral segments, which effectively resist bending at vertebral joints ( Figure 4A). The caudal centra in S. aegyptiacus have nearly uniform subquadrate proportions along the majority of the tail in lateral view, rather than narrowing, spool-shaped centra in crocodylians and other secondarily aquatic squamates ( Figure 4D), which increases distal flexibility during tail undulation. These salient structural features of the tail suggest that it functioned more as a pliant billboard than flexible fluke.
No primary or secondary vertebrate swimmer has a comparable drag-magnifying, rigid dorsal sail including sailfish, the dorsal fin of which is fully retractable and composed of pliable spines in membrane (Domenici et al., 2014). In contrast, the distal tail of secondary swimmers such as crocodylians (Grigg and Kirshner, 2015), mosasaurs (Lindgren et al., 2013) and cetaceans (Fish, 1998) is expanded with pliable soft tissues free of bone to form a flexible caudal paddle or fluke ( Figure 4C).
Spine-supported, torso-to-caudal sails aligned with median cranial crests, in contrast, have evolved multiple times for intraspecific display rather than aquatic propulsion among extant lizard (agamids, iguanians, chameleons). Semiaquatic sailfin and basilisk lizards ( Figure 4B), for example, do not use their sails while swimming, spend very little time submerged, and are not aquatic pursuit predators (Hone and Holtz, 2021).
Caudal centra proportions in most secondary swimmers, as mentioned above, grade from subquadrate to spool-shaped in the distal half of the tail to increase flexibility and undulatory amplitude ( Figure 4D), whereas those in S. aegyptiacus maintain relatively uniform proportions along the tail. This uniformity of subquadrate proportions in S. aegyptiacus should not be confused with a more derived piscine pattern of uniform, short, disc-shaped centra that has evolved in parallel in mosasaurs (Lindgren et al., 2013; Figure 4D, Appendix 2).

Appendicular comparisons to vertebrate secondary swimmers
Appendage (fore and hind limb) surface area in secondary swimmers is minimized to reduce drag because terrestrial limbs are inefficient aquatic propulsors. Appendage surface area in S. aegyptiacus, in contrast, is substantially greater than in reptilian and mammalian secondary swimmers and even exceeds that of the terrestrial predators Allosaurus and Tyrannosaurus ( Figure 5).
Interdigital webbing is used by some secondary swimmers to increase the area of the foot paddle (Fish, 2004). Extant crocodylians use their limbs in paddling only at launch and slow speed before tucking them against the body (Grigg and Kirshner, 2015). Crocodylian interdigital webbing, which is better developed and always present in the hind foot ( Figure 6C and D), only modestly increases surface area (<20%). Across a range of body size, we show that crocodylian paddle area scales isometrically ( Figure 6F; see Appendix 3). The crocodylian foot paddle, thus, becomes even less effective as a propulsor with increasing body size. A crocodylian of spinosaurid size, nonetheless, would have a foot paddle area an order of magnitude greater than is possible in S. aegyptiacus ( Figure 6E). Even a fully webbed hind foot in S. aegyptiacus ( Figure 6A), for which there is no hard evidence to establish as likely, is far too small to have functioned either for significant aquatic propulsion or for stabilizing control.

Paleohabitats and evolution
Most Spinosaurus fossils come from marginal basins along northern Africa in deltaic sediment laid down during an early Late Cretaceous transgression ( Figure 7A, sites 1, 2). These deposits, however, also include the majority of non-spinosaurid dinosaur remains, all of which may have been transported to some degree from inland habitats to coastal delta deposits. Because fossil transport is one way   (downstream), documenting the inland fossil record is key to understanding true habitat range. We recently discovered fossils pertaining to Spinosaurus in two inland basins in Niger far from a marine coastline ( Figure 7A, site 3). They were buried in fluvial overbank deposits alongside terrestrial herbivores (rebbachisaurid and titanosaurian sauropods) (see Appendix 4).
Large-bodied semiaquatic reptiles, in contrast, frequent coastal and inland locales today and in the past. Sarcosuchus imperator, among the largest of semiaquatic reptiles (~12 m length; Sereno et al., 2001), lived in the same inland basin as S. tenerensis. The fossil record supports our interpretation of Spinosaurus as a semiaquatic bipedal ambush predator that frequented the margins of both coastal and inland waterways.  The large body size of S. aegyptiacus and older related species such as S. tenerensis also mitigates against an aquatic interpretation for the former as it would constitute the only instance among vertebrates where the evolution of a secondarily aquatic species occurred at body size greater than 2-3 m. The profound changes involved in fully reentering the aquatic realm from a land-based lifestyle seem more likely to occur at relatively small body size. All other large-bodied secondarily aquatic vertebrates (e.g., ichthyosaurs, plesiosaurs, metriorhynchoid crocodylomorphs, protostegid turtles, mosasaurs, sirenians, whales) evolved the adaptations requisite for an aquatic lifestyle at small body size, increasing in body size once fully established within the marine realm (Domning, 2000;Thewissen et al., 2009;Polcyn et al., 2014;Moon and Stubbs, 2020;Motani and Vermeij, 2021).
Phylogenetic analysis of an enlarged dataset for spinosaurids clarifies piscivorous adaptations in the earliest spinosaurids (stage 1, ~130 Ma) that enhance prey capture in shallow water and heighten visual display (Figure 8; Appendix 5). In the skull, these include an elongate snout tipped with a dental rosette for snaring fish, retracted external nares to inhibit water intake, and a prominent nasal crest (Charig and Milner, 1997;. The ornamental crest over the snout is accompanied by the evolution of a postcranial sail of varying height supported by neural spines of the posterior dorsal, sacral and caudal vertebrae (Stromer, 1915;Allain et al., 2012;. The earliest spinosaurids, in addition, have 'cervicalized' anterior trunk vertebrae to enhance ventroflexion and the effective length of the neck, presumably as an adaptation to feeding in water (Hone and Holtz, 2021). Using the second dorsal vertebrae of the terrestrial predator Allosaurus for comparison, the homologous vertebra in spinosaurids shows marked modification (anterior face is convex, prominent ventral keel for muscular attachment, neural spine is reduced, zygapophyses large Figure 8. Calibrated phylogeny of spinosaurids (Barremian to Cenomanian, ~35 My). Updated phylogenetic analysis of spinosaurids resolves two stages in the evolution of piscivory and display. We show key cranial adaptations in the skull and highlight changes at the anterior end of the trunk to enhance neck ventroflexion (second dorsal vertebra in lateral and anterior views). Bottom, the fully terrestrial theropod Allosaurus fragilis (Madsen, 1976); middle, the baryonychine spinosaurid Suchomimus tenerensis (MNBH GAD70); top, the spinosaurine Spinosaurus aegyptiacus (BSPG 1912 VIII 19). con, condyle; dr, dental rosette; en, external naris; hy, hypopophysis; k, keel; ncr, nasal crest; ns, neural spine; pa, parapophysis; tp, transverse process. and planar). Giraffids, for a similar purpose, have 'cervicalized' the first thoracic vertebra to facilitate dorsiflexion and effective neck length (Lankester, 1908;Danowitz et al., 2015;Müller et al., 2021). Neural spines over the trunk and tail are heightened to varying degrees in all spinosaurids including baryonychines ( Figure 1F and I). The features cited above to enhance piscivory and display appear to be shared by all currently known spinosaurids and comprise the adaptations we identify as Stage 1 (Figure 8).
Baryonychines (e.g., Suchomimus, Baryonyx, Ceratosuchops) have a low, cruciate nasal crest and swollen brow ridges over the orbits for display or agonistic purposes. These may comprise features unique to this subclade of spinosaurids. Spinosaurines, on the other hand, exhibit further specializations for piscivory and display ( Figures 1A and 8, stage 2). Piscivorous adaptations include spaced teeth with smooth carinae for puncturing efficiency, smaller, more retracted external nares to inhibit water intake, more prominent muscle attachments on the ventral aspect of cervicodorsal vertebrae for ventral lunging, and scythe-shaped manual unguals for slicing (Sues et al., 2002;Dal Sasso et al., 2005;Ibrahim et al., 2014). Adaptations for enhanced display include a heightened cranial crest, low cervical sail, and a hypertrophied torso-to-caudal sail.

Discussion
In 1915 Ernst Stromer highlighted the remarkable adaptations in the jaws and neural spines of S. aegyptiacus for piscivory and ostentatious display, respectively, citing modern analogs for both (Stromer, 1915). Nothing close to this morphology had ever been described among nonavian dinosaurs at that time. More recently, a sequence of investigators have gone further, attempting to fathom the manner in which the lifestyle of this large predatory dinosaur engaged coastal waters. All have been hamstrung by the scarcity and fragmentary nature of the specimens, as all of Stromer's Egyptian fossils were destroyed in World War II. Indeed, the unveiling of a new partial skeleton from Morocco (Ibrahim et al., 2014) and its tail 6 years later (Ibrahim et al., 2020b) generated hypotheses for semiaquatic and aquatic interpretations, respectively.
The superficially eel-like morphology of the tail, viewed as a 'novel propulsor organ,' provided the inspiration for the 'aquatic hypothesis,' which envisioned S. aegyptiacus as a tail-propelled, diving predator 'that pursued and caught its prey in the water column' (Ibrahim et al., 2020b). Conversely, as would be requisite for status as a secondarily aquatic reptile, its terrestrial capabilities were regarded as seriously diminished by a trunk-positioned center of body mass (Ibrahim et al., 2014;Ibrahim et al., 2020b) that would require a quadrupedal stance on land and the use of long-clawed forelimbs not at all designed for weight support. Presented as support for the aquatic hypothesis, Fabbri et al used bone compactness to assert that S. aegyptiacus was a 'subaqueous forager' with diving bona fides (Fabbri et al., 2022).
The aquatic hypothesis, nonetheless, requires far more than proving its tail was a high-powered source of propulsion or its bones a bit more compact. In order to conclude that S. aegyptiacus was an aquatic diver and pursuit predator, one also must understand its buoyancy, stability, velocity, maneuverability, and diving performance in water. Those calculations require an accurate flesh rendering, which in turn is built over an accurate skeletal model. Therefore, we began with CT scans of the fossils to piece together an accurate skeletal model, discovering major discrepancies with the original 3D skeletal model (Ibrahim et al., 2014) and the 2D skeletal silhouette used by the aquatic hypothesis (Fabbri et al., 2022). Comparisons to the 2D model with the more accurate tail show that skeletal regions anterior to the hips are enlarged in length and depth beyond the dimensions of our CT-based reconstruction, shifting the CM in the resulting flesh model forward from the hips to the trunk. Trunk length was increased in both previous models of S. aegyptiacus due to unnatural ventroflexion of the dorsal column that also spread further the neural spines of the sail (Figure 9). When neotype (CT-scanned) or rebuilt holotype dorsal vertebrae of S. aegyptiacus are rearticulatd in an osteological neutral pose, the shorter torso has a straighter column with less spread neural spines. The ribcage, in addition, is not as deep, based on the preserved rib pieces of the holotype and neotype and the nearly complete ribcage known for S. tenerensis ( Figures 1F and 9). These proportions effectively reduce the volume of the trunk in our flesh model ( Figure 2). The flesh model used by the aquatic hypothesis, likewise, underestimated the muscle mass at the base of the tail, judging from our study of CT scans of crocodylians and a range of other reptiles (Díez Díaz et al., 2020). These differences are far from trivial when considering centers of mass and buoyancy in S. aegyptiacus. The dinosaur, in fact, stood back up on its hind legs like all other theropods.
With a more accurate flesh model in hand, we embarked on a range of biomechanical tests of its performance in water, determining that it fell short in all critical measures by huge margins. S. aegyptiacus failed spectacularly by factors from four-to tenfold for maximum swimming speed on the surface or underwater ( Figure 3A), for the capacity to right and remain stable or maneuver underwater ( Figure 3B), and for generating the force needed to overcome buoyancy and fully submerge. S. aegyptiacus was an unstable, slow swimmer without the capacity to submerge. These are stiff biomechanical hurdles for the aquatic hypothesis to overcome.
We thought of other comparative means to test the aquatic hypothesis, plotting S. aegyptiacus against various extant and extinct secondarily aquatic amniotes to consider appendage area  Ibrahim et al., 2020b). On one side of each length comparison, one or two blue lines are shown that register the alternative reconstructions. The opposing end of each length comparison either has a single blue line (when comparisons match, both 100%) or a red line as well for the shorter one (<100%): A blue line on the right or top sides of each comparison is used for registration. The opposing side has a blue line if reconstructions agree on length (100%), or a blue line for the length estimate in this study and a red line for that of the aquatic hypothesis. In all disparate comparisons, the reconstruction in this study is shorter (percentage given). Skeletal reconstructions (A, G) are aligned by the anterior and posterior margins of the ilium and measured to the cervicodorsal junction (C10-D1); the pelvic girdle (B) is aligned along the ventral edge of the sacral centra and base of the neural spines and measured to the distal ends of the pubis and ischium; the cervical column (C) is aligned at the cervicodorsal junction (C10-D1) and measured to the anterior end of the axis (C2); the scapula and components of the forelimb (humerus, ulna, manual digit II, manual phalanx II-1) (D) are aligned at the distal end of the blade and their proximal ends, respectively, and measured to the opposing end of the bone; the components of the hind limb (femur, tibia, pedal digits I, III) (E) are aligned at their proximal ends and measured to the opposing end of the bone; and anterior trunk depth (F) is aligned along the ventral edge of the centrum and neck of the spine of D6 and measured to the ventral edge of the coracoid. II-1, manual phalanx II-1; Ili, ilium; F, femur; H, humerus; Ish, ischium; l, left; Pe, pes; Pu, pubis; Ma, manus; r, right; RaU, radius-ulna; Sc, scapula; TF, tibia-fibula.
( Figure 5), the size of foot and tail paddles in crocodylians (Figure 6), tail structure ( Figure 4C and D), and the habitats occupied by large-bodied secondarily aquatic vertebrates ( Figure 7A). S. aegyptiacus fails all of these comparative tests as well because it resembles other theropod dinosaurs in limb size, other reptiles that use midline sails for display, and semiaquatic reptiles in the diversity of coastal and inland habitats occupied.
Although additional fossils of S. aegyptiacus and other spinosaurids will surely come to light, the overall skeletal proportions and form of S. aegyptiacus are largely known. Although many fine points on the structure and function of this interesting clade of predators will surely continue to engender controversy. The aquatic hypothesis, for that reason, is unlikely to survive as a plausible lifestyle interpretation. What then is our lifestyle interpretation for S. aegyptiacus? Our study and that of Hone and Holtz, 2021 envision S. aegyptiacus as a bipedal, semiaquatic dinosaur using ambush predation of large fish while wading into shallow coastal and riverine waters.
Thirteen principal conclusions can be drawn from this study, all of which may be tested: 1. Adult S. aegyptiacus had a body length of under 14 m with the axial column in neutral pose. 2. The reduced hind limb long bones in neotypic skeleton of S. aegyptiacus are infilled likely as an adaptation to weight support on land rather than functioning as ballast to increase density in water. 3. The segment-crossing caudal neural spines in S. aegyptiacus suggest that its tail functioned more as a pliant billboard than flexible fluke. 4. S. aegyptiacus, like S. tenerensis and other spinosaurids, was bipedal on land with its CM positioned over its hind feet. The long-clawed forelimbs of S. aegyptiacus were not used in weight support on land. 5. S. aegyptiacus could wade into shallow water for feeding with flotation occurring at water depth greater than ~2.6 m. 6. An adult flesh model of S. aegyptiacus has a body mass of ~7400 kg and average density of ~830 kg/m 3 , which is considerably less than the density of saltwater (1026 kg/m 3 ). 7. S. aegyptiacus was incapable of diving, given its buoyancy and incompressible trunk. Full submergence would require 15-25 times the maximum force output of its tail, depending on estimated lung volume. 8. S. aegyptiacus was unstable in deeper water with little ability to right itself, swim, or maneuver underwater. Maximum power from its tail, assuming it could undulate as in Alligator, is less than 700 N, which would generate a top speed of ~1 m/s, an order of magnitude slower than extant large-bodied pursuit predators. 9. All extant and extinct large-bodied (>2 m long) secondarily aquatic vertebrates are strictly marine, whereas fossils pertaining to Spinosaurus have been found in inland basins distant from a marine coast. 10. Transition to a semiaquatic lifestyle, as occurred in the evolution of spinosaurid theropods, can occur at any body size. Transition to an aquatic lifestyle among tetrapods, in contrast, has only occurred at relatively small body size (<3 m) with subsequent radiation once in the marine realm into larger body sizes. 11. S. aegyptiacus is interpreted as a semiaquatic shoreline ambush predator more closely tied to waterways than baryonychine spinosaurids. 12. Spinosaurids flourished over a relatively brief Cretaceous interval (~35 My) in circum-Tethyan habitats with minimal impact on aquatic habitats globally. 13. Two phases are apparent in evolution of aquatic adaptations among spinosaurids, the second distinguishing spinosaurines as the most semiaquatic of non-avian dinosaurs.

Skeletal reconstruction
The composite skeletal reconstruction of S. aegyptiacus is based principally on bones of holotypic and neotypic specimens supplemented by associated and isolated bones from Cenomanian-age formations in Egypt, Morocco, and Niger ( Figure 1A). The two most important specimens include the subadult partial skeleton composing the holotype (BSP 9012 VIII 19) from the Western Desert of Egypt (Stromer, 1915;Smith et al., 2006) and a subadult partial skeleton designated as the neotype from the Kem Kem Group in Morocco (FSAC-KK 11888;Ibrahim et al., 2014;Ibrahim et al., 2020b). A third referred specimen from Egypt was also considered (BSPG 1922 X45, 'Spinosaurus B';Stromer, 1934). These are the only associated specimens known for S. aegyptiacus on which to base the skeletal reconstruction, the relative size calculated from overlapping bones ( Table 1). Of these three specimens, only the bones of the neotype are preserved, all of which have been CT-scanned except for recently discovered bones of the tail (Ibrahim et al., 2020b). Noteworthy isolated specimens have been recovered from the Kem Kem Group in Morocco, including a large snout and manual phalanx used to gauge maximum adult body size. We incorporated all CT-scanned bones of the neotype and reconstructions (based on lithographic plates and photographs) of bones of the holotype and referred specimen from Egypt. For unknown bones without sequential adjacency as a guide, other spinosaurids were consulted for shape and proportion. All digital bones were articulated in osteologically neutral pose (Stevens and Parrish, 1999), a standardizing criterion for comparing vertebrate skeletons (Mallison, 2010;Vidal et al., 2020). In the case of overlapping bones, priority was given to the neotype (blue) followed by the holotype (red) and referred specimens (yellow). Bones without representation among specimens attributed to S. aegyptiacus are shaded gray ( Figure 1A).

Skeletal reconstructions compared
We compared our digital skeletal model of S. aegyptiacus to the recently published 2D silhouette skeletal reconstruction in the aquatic hypothesis (Ibrahim et al., 2020b), both of which are based primarily on holotypic and neotypic specimens ( Figure 9). We registered the reconstructions to each other by superimposing the four longest complete bones of the neotype (femur, tibia, ilium, ischium). Significant differences are apparent in several dimensions with major implications for the calculation of CM and CB.
When aligned at the hip, sacral and caudal columns have nearly identical length, but the presacral column is significantly longer (~10%) in the reconstruction of the aquatic hypothesis. The extra length of the presacral column is located in the neck between C2-10 and torso between D4-13. The trunk in our digital skeletal model is also not as deep as that in skeletal silhouette drawing, as can be seen by aligning the skeletons along the dorsal column ( Figure 9C). The contour of the belly marked by the gastral basket and the coracoids of the pectoral girdle extend farther ventrally (~25%) than the ends of the pubes, unlike our digital reconstruction or that of most other silhouette reconstructions for non-avian theropods. The length of the ribcage in our model is consistent with the only well-preserved spinosaurid ribcage known to date (S. tenerensis, MNBH GAD70). Finally, the forelimb in the skeletal silhouette drawing is ~30% longer than that in our digital reconstruction ( Figure 9A). The neotype is the only associated specimen of S. aegyptiacus preserving bones from the forelimb (partial manual digit II). The preserved manual phalanges are slender with deeply cleft distal condyles, which allows reference of additional phalanges of similar form from the Kem Kem Group ( Figure 1D). Our reconstruction of the manus is based on a recently described forelimb of the close relative Irritator (=Angaturama; Machado and Kellner, 2009;Aureliano et al., 2018). The proportions of more proximal forelimb segments and the pectoral girdle are based on the holotypic specimens of Baryonyx and Suchomimus. The forelimb in S. aegyptiacus is robust and long relative to other non-avian theropods, although considerably shorter than in some previous reconstructions (Ibrahim et al., 2014). Figure 10. CT scans inform cross-sectional muscle mass in Spinosaurus aegyptiacus. Muscle mass reconstructions of the axial column at five points (A-E) in S. aegyptiacus are compared to CT scan cross-sections of Struthio (Snively and Russell, 2007) and Alligator (Wedel, 2003;Mallison et al., 2015).
The longer presacral proportions, deeper torso, and longer forelimb of the skeletal reconstruction and flesh model used by the aquatic hypothesis cantilever show significant additional body mass anterior to the hip joint. That additional front-loading appears to be the main factor generating their mid trunk location for CM (Ibrahim et al., 2020b), which constitutes the basis for regarding S. aegyptiacus as a quadruped on land (Ibrahim et al., 2014).

Flesh reconstruction of axial musculature
To estimate the volume of axial musculature in S. aegyptiacus (Figure 10), we referenced CT-based studies on the ostrich (Struthio; Wedel, 2003;Snively and Russell, 2007;Persons et al., 2020) and alligator (Alligator; Cong et al., 1998;Mallison et al., 2015). To estimate caudal muscle mass, we used CT scans of various reptiles including the sail-backed basilisk lizard, Basiliscus plumifrons ( Figure 11, . For epaxial muscle mass in S. aegyptiacus, we estimated its vertical extent as twice centrum height, measuring upward from the base of the neural spine. The transverse width of epaxial musculature was estimated to be a little less than that of the hypaxial muscles, widest ventrally and tapering to the midline dorsally. For hypaxial muscle mass in S. aegyptiacus, we estimated its vertical depth at approximately twice chevron length in the anterior tail and 1.5 times chevron length in mid and posterior portions of the tail. We estimated the transverse width of hypaxial muscles as twice the length of the transverse processes. CT cross-sections of extant reptiles show that considerable muscle mass is present beyond the distal end of caudal transverse processes in anterior and middle portions of the tail ( Figure 10).
Several cross-sections from Basiliscus plumifrons (crested basilisk) provided valuable insights on the distribution of axial muscles in a lizard with a dorsal-to-caudal sail ( Figure 11, . Epaxial musculature in the trunk and tail comprises less than one-third of total axial muscle volume ( Table 2). Caudal neural spines project beyond the epaxial musculature to support the sail to a greater extent in mid and distal portions of the tail. At the base of the tail (CA4), approximately one-third (29.3%) of the neural spine projects dorsally supporting the sail. At mid tail (CA15), approximately three-quarters (76.8%) of the neural spine projects dorsally supporting the sail. The hypaxial musculature extends well below the distal end of the chevrons. At the base of the tail (CA4), the chevron lies internal to approximately one-half (52.6%) of hypaxial muscle depth, with the remainder (47.4%) distal to the end of the chevon. Farther along the tail (CA10-15), the chevrons are proportionately longer, supporting approximately two-thirds (67%) of hypaxial muscle depth with approximately one-third of hypaxial musculature beyond the distal end of the chevron. These cross-sections confirm the presence of considerable muscle mass ventral to the distal end of the chevrons in anterior through mid portions of the tail (Figures 10 and 11).
We divided the flesh model into six parts (axial body-presacral, axial-caudal, dorsal-to-caudal sail, forelimbs, hindlimbs, lungs) in order to assign appropriate densities ( Table 6). Densities were assigned to body parts based on their estimated composition, using values for tissues ranging from fat (900 g/l) to compact bone (2000 g/l). The average whole-body density for S. aegyptiacus, 833 g/l ( Table 6) compares favorably to whole-body density estimates for various non-avian dinosaurs (800-900 g/l). We have compiled various functional dimensions of the adult flesh model ( Table 7), and we divided Table 2. Axial muscle area in the crested basilisk. Area measurements of epaxial and hypaxial musculature along the axial column in the crested basilisk Basiliscus plumifrons (FMNH 112993 Table 3. Axial muscle and transverse process length in the tail of the crested basilisk. Transverse processes versus muscle width in the tail cross-sections in the crested basilisk Basiliscus plumifrons (FMNH 112993). Measurements are from the midline to the distal end of the transverse process (or centrum margin when there is no process) and to the lateral surface of the tail. CA, caudal. the flesh model into 10 body parts, for which we list volumes and external surface areas (excluding cut surfaces) ( Table 8).
We registered CM as the horizontal distance from the apex of the acetabulum (x-coordinate) and the vertical distance from the ground surface under the sole of the foot (y-coordinate) (Table 9,no. 4). This is the fourth estimation of CM for S. aegyptiacus, and we argue here the most accurate. We prefer a measure from the acetabulum rather than the distal tail tip, which as in S. aegyptiacus is often a matter of speculation given the rarity of completely preserved caudal columns (Hone, 2012). For the acetabulum, we recommend using its 'apex' rather than its 'cranial end' (Ibrahim et al., 2020b) for three reasons. First, the apex of the acetabulum is a more easily recognized landmark than the poorly defined anterior edge (or rim) of the acetabulum. Second, the apex rather than the 'cranial end' of the acetabulum is a more functionally intuitive point from which to measure CM, given its proximity to the rotational point for body mass centered over the hind limbs. And third, the dorsal (proximal) articular end of the femoral head is close to the apex of the acetabulum, and so the length of the femur and the distance that CM lies farther forward can be directly compared (CM located anteriorly beyond femoral length excludes stable bipedal posture with a relatively horizontal dorsosacral column).
With avian-like air space ('maximum'), CM is positioned only 15.3 cm anterior to the apex of the acetabulum and clearly over the pedal phalanges of the foot for a bipedal stance ( Table 10). The smallest air space option modeled on lizards ('minimum,' only 4% of body volume) generates the heaviest torso and displaces CM anteriorly 13.2 cm to a distance of 28.5 cm from the apex of the acetabulum ( Table 10). In this location, CM is still ~12 cm short of the midpoint along the length of the femur (~40 cm; femoral length is 81 cm in adult S. aegyptiacus). In this worst-case scenario regarding internal air volume, CM is still positioned over the pedal phalanges of the hind limb. Our flesh model does not support an obligatory quadrupedal pose on land for S. aegyptiacus.     Table 9. Center of mass (CM) calculations for S. aegyptiacus. There have been four estimates for the location of CM in flesh models of S. aegyptiacus using four different points of origin as a reference. Because three were based on an adult flesh model, we convert the one study based on a subadult (number 3) to reflect its position in an adult flesh model: "x" is the distance anterior tothe origin, and "y" is the height above the ground, both in cm.  • Source data 2. Data matrix as tnt file for TNT.

Data availability
All data generated or analysed during this study are included in the manuscript and appendices. Data has been deposited to MorphoSource and provided as Source datas 1 and 2.
The following previously published dataset was used: Author ( Appendix 1

Long bone infilling in S. aegyptiacus
In S. aegyptiacus, the medullary cavities in the long bones of the hind limb are reduced in diameter or infilled altogether by non-pachystotic bone. This condition is unusual among non-avian dinosaurs and was initially thought to be an adaptation for decreasing buoyancy (Ibrahim et al., 2014;Fabbri et al., 2022). Enhanced bone strength is an alternative explanation for solid hind limb long bones in S. aegyptiacus, a very large bipedal theropod with reduced hind limbs. Similar long bone infilling occurs in other large-bodied bipedal and quadrupedal dinosaurs and mammals (Vanderven et al., 2014;. To better understand this alternative explanation, we compare the increase in strength of a femur with a solid shaft compared to one of identical size with a hollow shaft using the section modulus (SM) as an indicator of the resistance to bending (Farlow et al., 1995). For a solid circular crosssection, the SM is given by π 4 R 3 , where R is the outer radius. When R has a value 1.0, SM is π 4 , or 0.785 m 3 . The SM for a cylinder with a hollow core is calculated by subtracting the SM of an inner cylinder comprising the hollowed space from that of the outer cylinder.
An inner radius of 0.5 generates a hollow cylinder that decreases the solid cylinder volume with 25% and generates an SM of π . This is a reduction of ~13% of SM. However, many theropods have hollow cores within long bones that comprise 50% or more of the volume of the long bone. For a hollow core equal to 50% of outer cylinder volume, the inner radius is 0.  Fig. 9). Photographs and line drawings of MNBH IGU11 from the Echkar Formation (Cenomanian) of Niger in ventral (A) dorsal (B), right lateral (C), and posterior (D) views. Cross-hatching indicates broken bone. ana, articular surface for the neural arch; hy, hypapophysis; k, ventral keel; nc, neural canal; pa, parapophysis; pc, pleurocoel. Scale bar, 5 cm.
In 2019, a fragmentary snout preserving most of the right and left maxillae of Spinosaurus sp. was recovered from a second locality southeast of Agadez also in the Echkar Formation ( Figure 7B). The material also includes a partial right dentary (MNBH EGA1) and subconical crowns and long tapering roots (MNBH EGA2). The jaw bones, which closely resemble S. aegyptiacus although possibly a new species, are comparable in size to the subadult holotypic and neotypic skeletons, or about 75% of the size of the large snout recovered in Morocco (Dal Sasso et al., 2005). They were found in overbank deposits near the remains of rebbachisaurid and titanosaurian sauropods and evidence of a vertebrate fauna common to Cenomanian sites across northern Africa (including Carcharodontosaurus, lungfish tooth plates, sawfish rostral teeth, etc.).