Drag coefficients of plesiosaurs, ichthyosaurs and modern cetaceans
At equal Reynolds numbers (same body length and same flow velocity), the total drag coefficients of plesiosaurs (Cd) are higher than the estimated values for ichthyosaurs and modern cetaceans (Fig. 1a). The limbless bodies, however, display similar Cd in all three groups and are even lower-than-average in the long-necked plesiosaurs, indicating that the limbs are responsible for the observed high Cd. The limbs of plesiosaurs contribute to more than 20% of their total drag coefficient: up to 32.2% in the basal Meyerasaurus and averaging 25% in derived plesiosaurs, with no major differences between plesiosaur morphotypes. In parvipelvian ichthyosaurs the contribution of the limbs to Cd is 11.2–15.6%, compared to 8.7–14.3% in modern cetaceans. Some of the living taxa we include provide a functional reference for this analysis. Our computed drag coefficient for the bottlenose dolphin model (Cd = 0.00413 at Re = 107) for example, is consistent with the estimates from a gliding living dolphin33 (Cd = 0.0034 at Re = 9.1 × 106) and other static CFD simulations34 (Cd = 0.00413 at Re = 107). It is worth noting that these values are, as expected, lower than estimates obtained from kinematic models, as motion is not accounted for35. In a former study, drag coefficients for a plesiosaur (Cryptoclidus), two ichthyosaurs and various cetaceans were obtained from rigid models in water tanks36. However, the pressure drag component (Cp) was likely overestimated due to the proximity of the models to the air–water interface, and thus are not directly comparable to ours.
In all models across the various clades, velocity plots display a stagnation point at the anterior tip of the model, a thin velocity gradient along the body corresponding to the boundary layer, an area of higher velocity around the greatest diameter and a low velocity wake behind the body, characteristic features of a fully developed external flow (Fig. 1b, Supplementary Fig. 1). The acceleration of flow results in areas of low pressure (Supplementary Fig. 2), while high pressure areas are observed where stagnation occurs. Our CFD methodology has been previously validated against experimental data from slender torpedo-like shapes26 and has been shown to provide a reliable distribution of internal drag components29 essential when dealing with streamlined bodies35. In all our simulations, the proportion of frictional and pressure drag was consistent with the expected values for slender geometries31: most of the drag originated from skin friction with a minor pressure drag component (Supplementary Fig. 2). The relatively larger limbs of plesiosaurs (Supplementary Table 1) produce a small increase in skin friction (Supplementary Fig. 2a), but a large increase in the pressure drag coefficient (Supplementary Fig. 2b), indicating that the latter largely explains differences in total drag coefficient between the groups. These effects might be explained by the low local Reynolds number of the flippers (resulting from a small chord length) producing high local Cd relative to the rest of the body31, alongside interference drag (i.e. drag caused by the interaction of flow fields where limbs and body meet), which might be higher for larger flippers.
Effect of body shape and body size on drag-related costs of steady swimming
When comparing morphologies at the same volume (proxy for body mass) and the same velocity, to focus on the effect of shape alone, derived plesiosaurs produce on average 30% more drag than parvipelvian ichthyosaurs and modern cetaceans (Fig. 2a, Supplementary Table 3; two-sample t-tests p < 0.001). Drag-per-unit-volume represents the contribution of drag to the cost of transport (COTdrag), with COT being the mass-normalised effort required for sustained forward swimming37. As for the drag coefficient, these differences are observed only when the full morphology is considered and not in the limbless models, indicating that the differences are caused by the relatively larger limb sizes in plesiosaurs. The model with the lowest absolute value of COTdrag is Tursiops, against which all other taxa were normalised. The highest COTdrag was estimated for the basal plesiosaur Meyerasaurus, which generates about 69% more drag than a bottlenose dolphin of the same mass. Among derived plesiosaurs, drag values increase from 29.2% in Thalassomedon to 42.6% in Dolichorhynchops relative to an equal-mass Tursiops, and no substantial differences are observed between the short-necked and long-necked morphotypes. The estimates of COTdrag in parvipelvian ichthyosaurs are about 4% to 15% higher than for the Tursiops model, very close to our estimates for the modern cetaceans Orcinus and Megaptera, which have relatively large fins. Overall differences between parvipelvian ichthyosaurs and cetaceans are non-significant (two-sample t-test p = 0.63; Supplementary Table 3).
Our CFD-based analysis thus shows that the overall morphology of plesiosaurs produced higher drag than parvipelvian ichthyosaurs and modern cetaceans, meaning that all other things being equal, an ichthyosaur should endure longer swims at a given speed or cruise at a faster velocity than a plesiosaur of the same mass. It is, however, uncertain to what extent all other things were equal. Propulsive efficiency estimates from living caudal oscillators such as cetaceans are generally higher than those of underwater fliers such as penguins, turtles and sea lions (for data of efficiency in extant animals and their sources see Fish, 200638). However, plesiosaurs were quadrupedal swimmers, with no functional reference among living tetrapods, and recent work suggested that their propulsive efficiency was enhanced by fine-tuning of the fore and hind flippers14. Further, plesiosaurs have more surface area dedicated to producing thrust. Thus, whether a more efficient propulsion compensated the extra drag of the flippers in plesiosaurs is not yet known.
Our results show there is no correlation between the COTdrag and the body fineness ratio (FR), regardless of whether the limbs are included or not (Fig. 2c, e). These observations, although opposite to what is generally assumed for aquatic animals7, are consistent with previous analyses in ichthyosaurs29. The widely extended concept of a FR range for minimum drag comes from the study of aerodynamically engineered forms, and only applies to certain shapes when all other geometric parameters are kept constant26,27, but cannot be extended to all complex streamlined forms. Instead, COTdrag displays a strong positive correlation with the ratio of surface area to volume only in simulations with full morphology (Pearson’s product-moment correlation, r2 = 0.89, p = 4.11 × 10−9; Fig. 2f). This is consistent with the expected behaviour of flow over streamlined forms for which drag is mainly frictional31. The large hydrofoil-shaped limbs in plesiosaurs, necessary for their lift-based quadrupedal appendicular swimming13,14,39, contribute to a large fraction of the surface area without adding much volume (Supplementary Table 1). In contrast, parvipelvian ichthyosaurs and modern cetaceans, both caudal oscillators, have lower proportions of body surface dedicated to the limbs (Supplementary Table 1), which add very little drag relative to a limbless body.
When body size is incorporated into the analysis (i.e. assessing the combined effect of shape and size by simulating the flow of life-size models for a constant velocity of 1 ms−1), the group differences detected in the volume-scaled simulations disappear (Fig. 2b, Supplementary Table 3; all two-sample t-tests p > 0.05). In these conditions, the drag-related costs of steady swimming of plesiosaurs fall within the range observed in both modern cetaceans and ichthyosaurs. Normalised against a 2.85 m-long Tursiops, the COTdrag for derived plesiosaurs ranges from 0.42, estimated for the large elasmosaur Thalassomedon, to 1.41 in the medium-sized Dolichorhynchops. In the parvipelvians, COTdrag spans from 0.33 estimated for the large Temnodontosaurus, to 1.76 in a 2.5 m-long Stenopterygius. Cetaceans show a smaller lower limit, because they include the largest animal in our sample, a 16 m-long humpback whale, with a COTdrag of 0.13 compared to Tursiops. The estimated cetacean upper COTdrag limit is 1.54 for a 1.9 m Tursiops. On the other hand, comparisons of the total drag power (Pdrag, i.e., the non-mass normalised version of COTdrag) for the same speed of 1 ms−1 (Fig. 3), show a different trend. Pdrag is highest for Megaptera, higher than in any fossil taxa included in this study, and is lowest in Tursiops. Thalassomedon is comparable both in total drag power and COTdrag to the killer whale. Similarly, the thalassophonean pliosaurid Liopleurodon matches the elasmosaurian Hydrotherosaurus in having a similarly low mass-normalised COTdrag but requiring about 4× more total drag power than Tursiops. Smaller forms like the polycotylid Dolichorhynchops and the thunnosaurian Ophthalmosaurus resemble the extant bottlenose dolphin in having a relatively high COTdrag and low total power.
Thus, in contrast to the volume-normalised simulations, differences between animals at their life-size scale are mainly influenced by size. For example, medium-sized plesiosaurs and ichthyosaurs, such as Dolichorhynchops and Ophthalmosaurus, have values of COTdrag close to that of a dolphin, while large plesiosaurs like Thalassomedon are more like the parvipelvian ichthyosaur Temnodontosaurus and a modern Orcinus. It is worth noting that the inflow velocity of 1 ms−1, is a reference velocity used for comparative purposes, and is not equivalent to the optimal cruising speed (i.e. speed at which COT is minimum16). This parameter is known to vary little in nature, with most vertebrates displaying values of preferred speed between 1–2 ms−1 regardless of body size40,41,42, which means it is reasonable to assume all tested taxa, regardless of their size, were able to swim at this velocity. Using a different reference velocity (2 ms−1) has no effect on the relative values of drag per unit of volume and the mass-normalised drag power (Supplementary Fig. 3; Supplementary Data). A reduction of mass-normalised drag-related costs of cruising as body size increases is selectively advantageous, as energy savings can be used to extend foraging and mating range, increase swimming speed and fuel other activities42,43.
Our analysis shows that for highly aquatic tetrapods, size dominates over shape in affecting the drag-related costs of steady locomotion. This is because COTdrag (i.e., the balance of drag to volume) is highly sensitive to surface/volume proportion (Fig. 2f), and so is much influenced by isometry in streamlined animals.
Interplay between neck anatomy and body size in plesiosaur drag
Simulations at constant Reynolds number (i.e., comparing models at same total length and same flow velocity), show that necks up to 5× the length of the trunk do not increase substantially the total drag coefficient. Longer neck ratios up to 7× were found to impact the drag coefficient by as little as 3% (Fig. 4a). We estimated a 4–10% increase in skin friction drag coefficient for neck ratios of 3–7×, but also a comparable reduction in pressure drag resulting in almost no change in the total drag coefficient. A previous CFD-based study also found no differences in drag coefficient between plesiosaur models with variable neck proportions20, but further comparison is not possible because of great differences in the order of magnitude of Cd, the use of a different scaling reference area and the lack of information on skin and pressure drag20. Here, we have shown that long necks produce only a small increase in skin friction, although not as great as previously speculated25,30, and this is nullified by reduced pressure drag.
Next, we explored the impact of neck proportions on drag-related costs of swimming in simulations where the size factor is removed. We found that if trunk dimensions are kept constant while the neck is enlarged, the drag per unit of trunk volume does not change appreciably for neck ratios up to 2×. However, longer neck proportions did impact resistive forces. This was moderate for a 3× ratio, with 12% more drag per unit of trunk volume, but became more substantial for longer necks, with 22%, 35% and 59% excess drag for necks of 4×, 5× and 7× respectively (Fig. 4b). This means that elasmosaurine elasmosaurs, with necks commonly 3–4× the length of the trunk23 might have experienced higher drag than other plesiosaurs of similar trunk dimensions.
To test if the ‘long neck effect’ remains when body size is accounted for, we compared the relative amount of drag-per-unit-trunk-volume (D/Vtr) in a wide sample of plesiosaurs (Fig. 4c) at life-size scale for a constant velocity of 1 ms−1, including three species with neck ratios above 2×: Styxosaurus (2.76×), Hydrotherosaurus (3.18×) and Albertonectes (3.72×), the last being the elasmosaur with the longest reported neck44. Our results show great variability in D/Vtr. Small-bodied plesiosaurs such as Plesiosaurus, Meyerasaurus and Dolichorhynchops generated up to six times more D/Vtr than the largest plesiosaurs, Kronosaurus and Aristonectes (Fig. 4d, e). Comparisons per group show that both basal plesiosaurs and derived polycotylids, the groups with the smallest specimens, produced generally higher D/Vtr. Moreover, we did not find substantial differences between elasmosaurs and thalassophonean pliosauroids (Fig. 4e, Supplementary Table 4; all two-sample t-tests p > 0.05). Both groups had similarly low ranges of D/Vtr regardless of neck length, lower on average than in polycotylids. These results stand even if we exclude Aristonectes, which belongs to the aristonectines, an elasmosaur subfamily with reduced neck length23,45. Further comparisons by morphotype show no significant differences between short-necked pliosauromorphs (here arbitrarily including plesiosaurs with neck ratios below 2×) and long-necked plesiosauromorphs (Supplementary Table 4, all two-sample t-tests p > 0.05). The highest values of D/Vtr occur in animals with trunk lengths of 100 cm or less, followed by a steep decrease between 100–150 cm and a steadier decrease in longer trunks. This indicates a strong negative correlation between trunk dimensions and D/Vtr (Pearson’s product-moment correlation between the log-transformed variables, adjusted r2 = −0.92, p = 2.28 × 10−7). The curve that best describes this relationship is the power equation, D/Vtr = 69.76 × Trunk length−0.944 (Fig. 4e), an almost inversely proportional relationship, consistent with the streamlined nature of these animals for which skin friction drag is dominant.
Polycotylids and thalassophonean pliosaurs, both derived pliosauromorph plesiosaurs9,21, share the same general body proportions9,21,46, but the latter had larger bodies and therefore needed less power in relation to their muscles to move at the same speed. Elasmosaurs on the other hand, despite their disparate morphologies, were no different from thalassophonean pliosaurs in their drag-related costs of forward swimming (Fig. 4c–e) and therefore they were likely to have been equally efficient cruisers.
Earlier research suggested that, even if long necks did not add extra drag during forward swimming, speed in elasmosaurs would have been limited to avoid added drag when their necks bent20. However, when the neck is bent in living forms, the course of swimming changes, as does the flow direction, but the body remains streamlined in the direction of incoming flow. For example, sea lions perform non-powered turns initiated by the head in which the body glides smoothly in a curved position, limiting deceleration47. Further biomechanical research is needed to understand the role of plesiosaur necks in manoeuvrability and other aspects of swimming performance, as well as how these were influenced by shape and flexibility. The well-established idea that long-necked plesiosaurs were sluggish, slow swimmers7,30 is thus not supported here, not because long necks did not increase drag20, but because body size overrode this drag excess.
Long necks evolved in large-bodied plesiosaurs: implications for drag
We analysed trends of body size and neck proportion in a wider sample of sauropterygians, including plesiosaurian and non-plesiosaurian Triassic sauropterygians. Long necks (neck ratio > 3×) occur in taxa with trunk lengths > 150 cm, whereas most sauropterygians had neck ratios of ≤ 2× (Fig. 5a). The great plasticity of body proportions of sauropterygians before and after their transition to a pelagic lifestyle after the Triassic has been well documented21,23,46, but this is the first time that neck and body size have been explored in the context of swimming performance for such a wide sample. We show that overall, sauropterygians and particularly plesiosaurs, mainly explored neck morphologies with little or no effect on drag costs and did not enter morphospaces that were suboptimal for aquatic locomotion (i.e., corresponding to small trunks with long necks; Fig. 5a). In fact, ancestral state reconstruction for trunk length shows that the ancestor of elasmosaurs was likely around 180 cm long and had a relatively short neck with a ratio smaller than 2× (Fig. 5b, c). This indicates that large trunks preceded neck elongation in elasmosaurs and suggests that extreme proportions might have been favoured by a release of hydrodynamic constraints.
We next explored evolutionary rates of relative neck length and trunk length in sauropterygians. The pattern of trunk length evolution is consistent with a heterogeneous rates model, not a homogeneous Brownian motion model (log Bayes Factor48 (BF) > 5 in 100% of the sampled trees and > 10 in 92.5%, Supplementary Table 5). Analysis of non-transformed trunk data shows that through the evolution of Sauropterygia, there was a general increase in trunk length with some higher rates, in Triassic nothosauroids, Jurassic rhomaleosaurids and Cretaceous aristonectine elasmosaurs (Fig. 5d; Supplementary Fig. 4a). Additionally, analysis of the log10-transformed trunk data highlights variation in the small-to-medium size ranges and reveals high rates in Triassic eosauropterygians (Supplementary Figs. 5 and 6). The largest trunks evolved independently in two groups, thalassophonean pliosaurids and elasmosaurid plesiosauroids, with no evidence of high rates in the former. In the plesiosauroids, rates are not particularly high in the basal branches, but they are very high in derived aristonectines, and rates for the whole clade were significantly higher than the background rate in 40% of randomisation tests (Supplementary Fig. 7 and Table 6). A progressive increase in body mass over evolutionary time has been described for various clades of aquatic mammals49 and seems to be a common hallmark of the aquatic adaptation to marine pelagic lifestyles in secondarily aquatic tetrapods44. Whether body size reaches a plateau as is the case in cetaceans49 and what constraints influence the evolutionary patterns of size in plesiosaurs remains unexplored. Against this general trend, some derived plesiosaurs, such as polycotylids, saw a reduction in body size, which might have been related to pressures on niche selection, such as adaptation to specific prey, the need for higher manoeuvrability or other ecological factors. As shown earlier, small sizes require lower amounts of total power for a given speed, and therefore would be favoured if for example food resources were limited. This suggests that, in spite of the energy advantages of large size in terms of reduced mass-specific drag29 and metabolic rates49,50, which make it a common adaptation to the pelagic mode of life, other constraints limiting very large sizes were also at work50,51.
A heterogeneous evolutionary rates model for neck proportion is also strongly supported (log BF > 5 in 100% of the sampled trees and > 10 in 45%, Supplementary Table 5). Fast rates are consistently seen at the base of Pistosauroidea (including some Triassic forms and plesiosaurs) and, interestingly, also within elasmosaurs (Fig. 5e; Supplementary Fig. 4b). The neck proportions of elasmosaurs were found to evolve at a faster pace than the background rate in 90% of analyses (randomisation test p-value < 0.001 in 80% and < 0.01 in 10% of the sampled trees; Supplementary Fig. 7 and Table 6). Very fast rates in elasmosaurs are concentrated in the most derived branches (i.e., Euelasmosauridia from the late Upper Cretaceous52) and represent both rapid neck elongation in elasmosaurines and rapid neck shortening in weddellonectians (i.e., aristonectines and closely related taxa52). Additionally, various other independent instances of relative shortening of the neck occurred during the evolution of Sauropterygia, most notably in placodonts, pliosaurs and polycotylids, but these are not associated with high rates.
Our findings contrast with a previous study23 which did not identify any significant evolutionary rate shifts in the neck ratio across Sauropterygia. Here we use a larger number of taxa and a different model fitting approach, which might account for these discrepancies. The association between very long necks and large trunks, along with our flow simulations results and the evidence of high rates in the elongation of necks in elasmosaurines (Fig. 5e), suggests that neck elongation was facilitated by large body sizes. The question remains why neck ratios did not evolve longer than 4×. According to our data, hydrodynamic constraints might have operated against the selection of such long necks. However, it is possible that the primary function for which they were selected, which is still debated30,53, did not require necks with those characteristics. Neck anatomy is likely to be the result of a compromise between different functions/constraints, one of them being hydrodynamic, as shown by the results presented herein.