Surface and Subsurface Swimming of the Sea Snake Pelamis Platurus

This paper reports several aspects of the swimming movements of the yellow-bellied sea snake Pelamis platurus Linnaeus. P. platurus is a totally aquatic pelagic species that commonly occurs in oceanic drift line communities. This snake is known to dive as deep as 50 m and spends about 87 % of its time submerged (Rubinoff, Graham & Motta, 1986a,b). Adaptations for locomotion include a paddle-shaped tail and a ventral body keel that extends from just behind the jaw to the vent (Fig. 1; Pickwell, 1972). The long lung of this snake reduces swimming costs by providing positive buoyancy at the surface. Recent work has demonstrated that P. platurus regulates lung volume prior to diving (Graham, Gee & Robison,1975), dives with a specific lung volume that enables it to achieve near-neutral buoyancy at depth (Graham, Gee, Motta & Rubinoff, 1986), and is able to prolong its dives by using its skin for aquatic respiration (Graham, 1974). Most observations of movements by this species have been made on individuals at the water surface (Pickwell, 1972; Kropach, 1973, 1975). There have been no kinematic analyses or extended observations of subsurface swimming movements.

The first objective of this paper is to describe and compare the surface and subsurface swimming movements of P. platurus, one of few anguilliform swimmers that routinely swims both at the surface and at depth. Surface swimming imposes greater energetic costs to an organism because of the drag associated with surface wave formation (Hertel, 1966). Very little is known about how an organism’s style of anguilliform swimming may change at depth. We have determined this for P. platurus by contrasting the slow, horizontal swimming movements typical of snakes at or just below the surface with those of snakes swimming at 7–10 m depth in the Hydraulics Laboratory tank at Scripps Institution of Oceanography (SIO). Seymour, Spragg & Hartman (1981) showed that inclination of P. platurus resulted in the displacement of lung gas towards the highest point in the body. Although hydrostatic pressure would compress lung volume at depth, we tested the hypothesis that the lung volume of subsurface-swimming snakes remained sufficient, if displaced to either end of the body, to bring about changes in the body posture and profile and in regional body lift. Another objective is to compare the near-surface, steady anguilliform swimming of P. platurus with that of other aquatic snakes (Natrix,Nerodia) and the eel (Anguilla). Comparisons of this nature may reveal how the morphological specializations for aquatic life noted for sea snakes have influenced the locomotory mechanisms of this family relative to other phylogenetically distant anguilliform swimmers.

Specimens of Pelamis platurus (45–70 cm total length, L) were captured by dip net in the Gulf of Panama, transported by air to the Physiological Research Laboratory at SIO and maintained in seawater aquaria (25–27°C) for up to 3 months on a diet of frozen and live fish.

Measurements of the body shape and dimensions of P. platurus were made to evaluate specializations for anguilliform locomotion and to aid in hydrodynamic analyses. The following morphological measurements were made using both frozen and fresh snakes: total length (L), wet body mass, surface areas of the head, ventral keel (Fig. 1), paddle tail and of the total body, and vertical body thicknesses at five locations; the head, the 25, 50 and 75 % points of the body length (B25, B50, B75) and the mid-point of the tail (T50). Dunson & Robinson (1976) previously estimated the skin surface area of P. platurus. In the present study, surface areas were estimated gravimetrically by tightly applying aluminium foil, with a known area: weight ratio, to body surfaces. The foil was smoothed, fitted to the body and then trimmed to obtain a fully-contoured fit. The ventral keel originates behind the jaw and extends to the vent. Its base is formed by ventral fusion of the ribs, and its boundary with the body is usually demarcated by the transition from a convex to concave vertical profile (Fig. 1). Prior to wrapping the body and keel areas, the lung of each snake was filled to the known mean volume for surface-floating snakes (i.e. 8·8% of body weight, Graham et al. 1975) and then sealed by ligation. It was easier to fit the foil if the body was first suspended vertically and frozen. Foil from different sections of the roll was weighed to verify a similar area:weight ratio. As a control for the foil method, eplicate surface area measurements on the same snake were performed and these revealed an estimation precision of within 10%. The total body surface area estimates obtained for these snakes were found to agree with area data presented by Dunson & Robinson (1976, fig. 3), who skinned and traced their study specimens.

To examine the effect of submergence on lung volume and gas displacement, a 60 cm snake was anaesthetized with Halothane and its lung was inflated with 20 ml of air via the trachea, which was then ligated. Aluminium foil bands were then placed around its body at several positions to measure circumference at surface pressure. The snake was then loosely suspended and tethered inside an open-ended, 75 cm long × 8 cm diameter, clear Lucite tube which was then submerged to a position in front of the deep tank observation window (8·5 m). By tilting the tube from the horizontal towards the vertical, the snake could be orientated ‘head up’ or ‘tail up’ so that changes in the relative swelling and flotation of the caudal and throat regions could be noted.

A Locam high-speed camera equipped with a 120 mm Schneider lens was used to make 16 mm overhead films of snakes swimming at velocities of 15 and 32 cm s⁻¹ and at 1–5 cm depths in a 75 cm long × 29 cm wide working section of a controlled-speed water tunnel. All filming was carried out at water temperatures from 23 to 25°C. The camera was positioned l·5m above the working section of the tunnel. Two 650-W tungsten floodlights were used and camera shutter speed was 1/300 s with an f setting of 2·8. The water tunnel design is similar to that described by Prange (1976). Its total volume is approximately 4001 and a variable-speed, 12-V trolling motor was used to circulate water. Water velocities were measured with a General Oceanics (Model 2035) flow meter. Because the working section surface was open to the air, small standing surface waves were created by the flow. These were smaller at 15 than at 32 cm s⁻¹. Ektachrome video news film was used and all films were shot at either 50, 100 or 150 frames s⁻¹. Calibration of camera film speed by time dots revealed a framing rate error of less than 1 %.

Kinematic data for P. platurus swimming at 15 and 32cms⁻¹ were used to estimate total power generated, kinetic energy lost to the wake and thrust power, according to Lighthill’s (1969, 1970) bulk momentum model. The long, slender body of P. platurus is laterally compressed, lacks substantial posterior tapering and terminates in a broad, flat tail and is thus highly amenable to this analysis. Power estimates from the Lighthill model were compared to theoretical drag values estimated from hydrodynamic equations.

Video tapes of the subsurface swimming movements of P. platurus were made through the underwater window (0·51×0·46m) located at a depth of 8·5m in the 10m SIO tank. The tank has a diameter of 3·1 m. Snakes released into it could be easily observed from the window, and previous studies showed that after 24 h most snakes behaved normally and had dive durations similar to those of individuals tracked at sea (Rubinoff et al. 1986a). A 2×2m plastic frame containing a rope grid with approximately 10 × 10 cm squares was submerged in the tank and placed opposite the window to allow estimation of subsurface swimming speeds and ascent or descent angles and rates. During video sequences individual snakes were followed for as long as possible. Tank water temperature ranged from 26°C at the surface to 21 °C at 10 m. Water was filtered and 10 1000-W metal halide lights set 3·5 m over the tank simulated mid-day light levels over the entire depth range. Subsurface video recordings were analysed with a stop action recorder and colour monitor.

Body measurements were made on seven P. platurus (Table 1) ranging in total length from 49·6 to 70-0 cm (weight range 36·4·106⁻5 g). Tail length of these snakes ranged from 10 to 12 % (mean ± s.D. = 11 -2 ± 0·7 %) of total length, which is similar to values reported by Kropach (1973). Mean (±S.D.) percentages for the separate nody areas of the seven snakes are: head 4·6 ±0·5%, body 73·2 ±2·2%, keel 15·6 ± 3·0 % and tail 5·8 ±0·9%. The total body surface area estimates for these snakes generally increase with body size, ranging from 206 to 348 cm², and are similar to values given by Dunson & Robinson (1976, fig. 3). The thickness data show that the body of P. platurus does not taper posteriorly.

Observations and video recordings were made of snakes swimming at the still surface of the SIO deep tank, where some snakes floated motionless while others routinely swam as fast as 20 cm s⁻¹. Snakes swimming on the surface held their heads at or just below the surface and, owing to the long and buoyant lung, their bodies were slightly out of the water while the tail was just below the surface. Even though the body emerged slightly, surface waves were not formed and the body was stable. There were no obvious differences in the anguilliform swimming motions of snakes on the surface of the deep tank and those swimming just under the surface in the water tunnel.

Snakes forced to swim in the water tunnel always swam within 5 cm of the surface with their bodies horizontal and their heads and tails completely submerged (Fig. 2). Whenever a snake raised its head to breathe, it was swept back in the flow, but once its head was resubmerged it would recover its position. Films showed that the snake’s ventral keel shifts laterally with each undulation and tends to flare outwards (giving the snake a concave vertical surface on the outer edge, Fig. 1) at maximum Ag. The flat paddle tail occasionally rolled slightly to the side during a full sweep; however, the stabilizing effect of the ventral keel prevented snakes from rolling or twisting.

The pattern of four propulsive half waves along the body (Fig. 2) was observed at both 15 and 32cms⁻¹. Inspection of kinematic data for each segment of the 51 cm snake (Table 2) reveals that all variables were increased in posterior segments. While λg, As and 6 remained nearly the same at both velocities, values of Ws and cs were nearly doubled and tail amplitude declined at 32 cms⁻¹. Estimated from Lighthill’s model, the total power generated by a 51cm P. platurus is 3·641×10⁻⁴ J s⁻¹ at 15 cms⁻¹ and 29·877×10⁻⁴J s⁻¹ at 32cm s⁻¹ (Table 3). After correction for kinetic energy losses at each speed, the Froude efficiencies are 79 % at 15 cms⁻¹ and 81 % at 32 cms⁻¹. As expected, theoretical drag estimates for this snake at the two speeds and under both laminar and turbulent surface boundary conditions (Table 4) are less than the thrust power estimates made from Lighthill’s bulk momentum model.

In contrast to the largely two-dimensional nature of body motions during surface swimming, snakes swimming below the surface typically, but not always, adopted a dorsal-ventral body curvature, usually with the anterior body level or slightly sagging and the tail somewhat elevated relative to the head (Fig. 3). Depending upon velocity, subsurface swimmers commonly had their tails up and the axis of their bodies tilted from 10 to 50° relative to their horizontal plane of progression. The average (±S.D.) velocity of snakes swimming level but in the ‘tail-up’ mode at depth was 2·0 ± 0·7 cms⁻¹ (range 1·0–3 0, N = 27). This is significantly slower (one-tailed i-test) than the velocity of snakes swimming level but without their tails elevated (4·3 ± 2·0cms⁻¹, range 1·0–8·0, N = 26).

Tail elevation by P. platurus during submerged swimming results from this snake’s vertebral flexibility and the ascent of gas into the elevated end of its long lung. Observations with the anaesthetized snake suspended at 8·5 m showed than when the tail was elevated the underlying lung expanded, and the caudal region became positively buoyant. Conversely, lowering the tail to a point below the head resulted in caudal flattening and sinking. Deep tank video recordings additionally showed that snakes swimming in the ‘tail-up’ mode, but in a more or less horizontal plane, had a large gas swelling just anterior to the vent where the tip of the lung is located (Fig. 3). Video recordings also showed that the swelling would disappear and reappear as snakes made the transition from ‘tail-up’ to ‘tail-down’ and then back to ‘tail-up’ swimming, which occurred routinely in the course of slight (1–3 m) depth changes as snakes swam around the tank.

The keel of a diving P. platurus becomes accentuated by the combined effects of tail elevation, the bending of the body and hydrostatic pressure (i.e. lung compression and concomitant skin stretching), resulting in a largely foil-shaped trunk. Depending upon the degree of tail-up swimming being used by a snake, a section of the ventral keel actually becomes vertical and positioned at the trailing edge of the body (Figs 3, 4). As a result of this, torsional and twisting body motions are incorporated into the basic undulatory movement pattern, and the vertical body and tail appear to act together as an extended caudal fin or wing (Fig. 4). These, along with the horizontal trunk segment, generate swimming thrust. Comparison of the frames in Fig. 3 shows that relative to the mid-body region, the anterior trunk and head both remain level and exhibit low yaw. Also, and unlike the pattern observed in horizontal swimming, the amplitudes of the tail and vertical body segment are low relative to the mid-body region.

How does the horizontal locomotion of P. platurus compare to that of other slendei anguilliform swimmers? Analyses of anguilliform swimming have been made for the eel, Anguilla, by Gray (1933) and for the water snakesNatrix (Taylor, 1952; Hertel, 1966) and Nerodia (Jayne, 1985). For Anguilla, which seldom leaves water, and P. platurus, which has a completely aquatic existence, natural selection has refined both the nature and degree of development of specializations for aquatic locomotion. P. platurus, for example, cannot propel itself on land; females are ovoviviparous and young are born at sea. By contrast, the occurrence in water of snakes such as Natrix and Nerodia is the result of selection for a behavioural plasticity that includes the capacity to be amphibious. These snakes remain specialized for terrestrial locomotion and, depending upon ecological circumstances, may or may not enter water. In water, Natrix frequents shallow areas where it both swims and crawls, but, unlike P. platurus it seldom enters deep, open-water bodies except to cross them, and it rarely floats motionless at the water surface. Natrix will, however, enter water and dive to escape predators (Schmidt & Inger, 1962).

Fig. 5 compares the body undulation patterns made by these anguilliform swimmers, and reference to Fig. 2 shows that P. platurus has more and smaller segment oscillations along its body. A similar pattern has been shown for another sea snake, Enhydrina, by Aleyev (1977, fig. 25). The relative caudal amplitudes (expressed as %L) estimated for animals in Fig. 5 are: Natrix (L = 24cm) 43 %L, (105 cm) 41 %L; Nerodia 41 %L; Anguilla 39 %L. These are all greater than that of the 51 cm P. platurus (26 and 19%L at 15 and 32cm s⁻¹, respectively). Smaller amplitude segment and tail oscillations reflect the greater posterior body and tail thicknesses of P. platurus which, together with a ventral keel, provide more stability during swimming. The caudal tapering of Natrix and Nerodia markedly reduces their posterior virtual mass, and this results in a high caudal amplitude and probably energy loss (Webb, 1978), as well as a twisting of the body (Jayne, 1985) to obtain thrust. Photographs and side-view diagrams of swimming Natrix and Nerodia show that they often swim at the water surface with their heads above water and their trunks and tails submerged. Common features of this swimming mode are the production of a large surface wake and, depending upon velocity, high amplitude body and caudal oscillations and the tendency for the body to roll (Hertel, 1966, figs 232, 246; Jayne, 1985, fig. 3). This contrasts with the flat, slightly buoyant swimming posture of P. platurus. The tendency for the trunk and tail of both Nerodia and Natrix to sink below the surface is due to their relatively short lungs (about 55 %L vs 90 %L in P. platurus, Jayne, 1985; Graham et al. 1975).

Using kinematic data we calculated (Table 5) the total power, kinetic energy loss, and net swimming thrust power of a 24 cm Nerodia swimming at 12·4, 24·6 and 35·9cms⁻¹ (Jayne, 1985) and a 7cm Anguilla swimming at 4cms⁻¹ (Gray, 1933; Webb, 1975). Data for Nerodia were obtained during steady swimming (Jayne, 1985), and the selected range of velocities permits relevant comparisons to data for P. platurus. (Snake morphological data needed for these calculations [e.g. values for k and d, see Table 3] were obtained from preserved specimens in the herpetological collection at San Diego State University.)

Power production by Anguilla and P. platurus are first compared with reference to the U²·⁸ power scaling ratio predicted from theory for turbulent flow (Table 4). The swimming eel has a total power of 0·055× 10⁻⁴Js⁻¹ at a velocity of 4cms⁻¹ (Table 5). That of P. platurus at 15cms⁻¹ is 3·641×10⁻⁴ J s⁻¹ (Table 3), and at 4cms⁻¹ it would be 0·09× 10⁻⁴Js⁻¹ (i.e. 3·641/[15²·⁸/4²·⁸]), which is 36% higher than the estimated power of Anguilla at this velocity. Even without correcting for the body size differences (P. platurus weighs about 45 g, Anguilla about 5 g) and Froude efficiencies, this calculation shows that the thrust powers of Anguilla and P. platurus are somewhat similar at 4cms⁻¹, and thus supports Seymour’s (1982) conclusion that the metabolic cost of swimming for a sea snake is similar to that of an eel. We have no means of knowing whether the magnitude of the difference between these two power estimates would be biologically significant.

Table 5 shows that the total swimming thrust power of the 24 cm Nerodia at 12·4, 24·6 and 35·9 cm s⁻¹ is much less than that of the 51 cm P. platurus swimming at 15 and 32cms⁻¹ (Table 3). (These comparisons also ignore size differences; a 24cm Nerodia weighs 5·1 g.) However, Nerodia has proportionately higher kinetic energy losses at all speeds and correspondingly lower Froude efficiencies. The reduced surface area of the posterior body sections of Nerodia relative to P. platurus must account for its greater caudal amplitude and its higher propagated wave velocity and both of these factors lower Froude efficiency (Webb, 1978). In view of the larger amplitude motions of the body and tail of Nerodia, it is not surprising that its estimated wake energy loss is proportionately greater than that of P. platurus. However, these estimated losses may still be low because, when swimming at the water surface, it would encounter a three- to five-fold increase in drag due to surface wave formation (Hertel, 1966), which would further affect its thrust requirement and locomotor efficiency. Finally, the body of this snake is nearly round and, as described above, its tail tapers steadily; virtual mass at the trailing edge of its body becomes quite small compared to that of P. platurus. Webb (1978) pointed out that estimates of both total swimming power (T) and kinetic energy lost to wake displacement (Tk) for tapered-body forms are usually conservative.

Additional insight into the relative efficiencies of the swimming patterns and body designs of Nerodia and P. platurus can be gained through a comparison of swimming thrust values that have been in some way normalized for the length and mass differences between these two snakes. Because of their size difference the Reynolds numbers (Re) of these snakes hardly overlap and meaningful comparison of size-corrected thrust coefficients or other kinematic properties as a function of Re, as was done by Yates (1983), is not possible. Rather we compare Nerodia and Pelamis by dividing swimming thrust power (Tp) values for each snake at each speed by the product of segment virtual mass (nip) and body length (L) (data provided in Tables 3 and 5). The resulting size-adjusted swimming thrust indices range from 42 to 1602, and when these are compared relative to swimming speed (Fig. 6) it is apparent that, relative to P. platurus, Nerodia is a less efficient swimmer that requires both a greater swimming thrust and a larger increase in thrust with velocity. It is recognized that Lighthill’s bulk momentum model does not adequately describe all active hydromechanical elements of anguilliform swimming (Lighthill, 1983). Future development of a resistive-reactive theory for anguilliform swimming will increase our abilities to compare the hydromechanical forces in these phylogenetically distant yet convergent swimmers.

Our observations of the subsurface swimming motions of P. platurus indicate that proximity to the surface severely limits the scope of swimming motions available to this species. When away from the confining surface, this snake can, by shifting its body attitude, displace the bulk of its pulmonary gas volume from one end of the lung to the other. Assuming that this action is not solely an artefact of confinement in the SIO tank (i.e. diver verification of this swimming mode for snakes in their natural environment is needed), P. platurus appears to use this mechanism to achieve nearly effortless and rapid changes in regional static lift, and thus exercise subtle control of the direction of its swimming thrust.

The ability to shift lung gas has significance for the energetics of subsurface swimming, including angles and rates of depth change, as well as for buoyancy control and pulmonary oxygen storage at depth. Deep tank video recordings and direct observation revealed that the switch from ‘head-up’ to ‘tail-up’ swimming always coincided with slight depth changes as snakes circuited the tank. It is important to emphasize that even if a snake swimming below the surface was neutrally buoyant, it could, by inclining its body, concentrate a sufficient amount of gas in its posterior (saccular) or anterior (tracheal) lung area to establish a positive static lift at one end of its body. Thus when a snake elevates its head and swims upwards, lung gas shifts anteriorly, the saccular lung deflates and the tail sinks. Similarly, when the snake levels off at its new depth and then slightly dips its head, the saccular lung will again fill and the tail will rise. Logically, a swimming snake’s head should always initiate an upward or downward course change, with the body following. However, the shifts in regional lift accompanying this action in subsurface swimming P. platurus can be attributed to specialized structural features of its lung, including its length, the compliance of the posterior sections and a capacious terminal segment.

Subsurface swimming with a vertical tail and posterior body section and the accentuation of the ventral keel by body curvature appear to impart vertical and horizontal stability to the snake in a manner analogous to the median and paired fins of a fish. Depending upon the buoyancy state of a snake, its tail may tend to raise it towards the surface, and this is countered by both forward swimming thrust and the flat surface of the head which acts as a stabilizing plane. Although in the opposite direction, this is analogous to the manner in which the combined functions of swimming and pectoral lift counter the tendency of negatively buoyant sharks and scombrids to sink.

This work was supported by grants from the Tupper Foundation and the Smithsonian Institution Scholarly Studies Program. We acknowledge the Ministerio de Desarrollo Agropecuario of the Republic of Panama for permission to export sea snakes to the United States for scientific study. We acknowledge the following individuals who assisted in various aspects of this work: J. Bryant, L. Chin, L. Cruz, R. Etheridge, S. Feldkamp, D. Gibson, G. Kooyman, K. Mandernack, O. Vallarino and A. Velarde. Studies at the SIO deep tank were done with the assistance and cooperation of C. S. Coughran and J. D. Powell. We thank Drs Clifford Hui, George Pickwell, George Yates and Paul Webb and two reviewers for valuable comments on earlier drafts of this manuscript.