Effect of water depth, velocity and temperature on the surfacing frequency of the bimodally respiring turtle Elseya albagula


*Correspondence author. E-mail: c.franklin@uq.edu.au


  • 1We apply a cost-benefit model to investigate whether bimodal turtles change their diving behaviour in response to changing energetic costs of aerial vs. aquatic respiration. This question is significant both in the context of the evolution of specialized respiratory structures in bimodal turtles, and the ecological role of respiratory partitioning.
  • 2Elseya albagula is a bimodally respiring turtle that can extract oxygen from water via specialized cloacal bursae and can vary its reliance on aquatic respiration. We examined the effect of water depth, velocity and temperature on the surfacing frequency of E. albagula through analysis of submergence durations (dive length). We applied a model of resource gain maximization to predict that as the cost of surfacing increases there will be longer dives made possible by the increased reliance on aquatic respiration. Diving behaviour of juvenile turtles was recorded in a large observation tank at varying water depths (50, 100 and 150 cm) and temperatures (20, 25 and 30 °C). Diving behaviour of adult turtles was also monitored in a flume at water velocities of 5, 15 and 30 cm s−1.
  • 3Dive length significantly increased with water depth, with dives at 150 cm twofold longer than dives at 50 cm. Dive length was also significantly influenced by temperature, with shorter dives at 30 °C than at 20 °C. In contrast, water velocity had no effect on dive length or on the proportion of a trial adult E. albagula spent using a velocity refuge.
  • 4Juvenile E. albagula possess the flexibility to adjust dive length in response to changes in water depth and temperature in a manner we would expect if energy use were being optimized. We advance the proposition that the driving force for the evolution of specialized aquatic gas exchange structures in E. albagula is the reduced cost of transport to and from the surface.


A key prediction of optimality theory is that animals should obtain resources in an ‘optimal’ manner with respect to net energy acquired (Andersson 1978; Kramer 1983). That is, as the cost of obtaining a resource increases, the use of that resource should decrease. Kramer (1983) applied this principle to the acquisition of oxygen by aquatic, bimodally respiring animals and went on to generate predictions about the relative use of aerial vs. aquatic oxygen by animals capable of obtaining oxygen from both air and water. Kramer (1983) postulated that bimodal respirers exhibit ‘optimal diving behaviour’ reflective of maximum respiratory efficiency, by utilizing whichever form of oxygen is cheaper to obtain under a given set of environmental conditions.

Kramer (1983) identified costs for diving animals that need to periodically surface for air. Although air is a superior respiratory medium to water, containing more oxygen and having a lower ventilation cost (Boutilier 1990), travelling to and remaining at the surface requires time (Bevan & Kramer 1987), energy (Feder & Moran 1985), and can increase the risk of predation (Kramer, Manley & Bourgeois 1983). According to Kramer's (1988) ‘theory of optimal breathing’, the use of aerial oxygen should decrease as the cost associated with surfacing increases. Similarly, when surfacing is not energetically costly, aerial respiration should be the primary mode of oxygen acquisition given the high ventilation cost of aquatic respiration (Kramer 1983, 1988; Boutilier 1990).

An investigation into the effect of water depth on the respiratory behaviour of Trionyx sinensis (Hua & Wang 1993) provided the first indication that freshwater turtles respond to an increase in the energetic cost of aerial respiration by adjusting respiratory partitioning. The soft shell T. sinensis reduced surfacing frequency and increased the frequency of pharyngeal movements (aquatic ventilation) when water depth, and therefore travel time to the surface, was increased. Another largely overlooked potential cost of surfacing is the energetic cost of ascending and maintaining a position in fast-flowing water (Fausch 1983; Facey & Grossman 1990). In contrast to variables such as oxygen and water temperature which have immediate physiological impacts on respiration of many turtle species (Gatten 1980; Ultsch & Jackson 1982; Herbert & Jackson 1985; Stone, Dobie & Henry 1992; Priest & Franklin 2002; Gordos, Franklin & Limpus 2003), flowing water exerts a physical or mechanical force against an animal trying to surface (Facey & Grossman 1990). High water velocity has also been found to induce micro-habitat selection in stream dwelling fish, whereby fish take refuge in low-velocity micro-habitats within the water stream to minimize energy expenditure and avoid displacement downstream (Vehanen et al. 2000). The extent to which water velocity influences surfacing patterns and micro-habitat selection in freshwater turtles is poorly understood (Reese & Welsh 1998; Tucker et al. 2001). One recent study (Gordos, Franklin & Limpus 2004) showed that water velocity affects surfacing frequency and micro-habitat selection of the bimodal respirer Rheodytes leukops under laboratory conditions.

Water temperature is well established as a factor influencing the diving behaviour of freshwater turtles under laboratory conditions, through its impact on both oxygen metabolism and oxygen availability (Gatten 1980; Ultsch & Jackson 1982; Herbert & Jackson 1985; Prassack, Bagatto & Henry 2001; Priest & Franklin 2002). The oxygen capacitance of water decreases as water temperature rises (Dejours 1994) intensifying the difficulty of meeting oxygen demands from water. This effect, when combined with raised metabolic demand for oxygen (Schmidt-Nielsen 1997), should lead to an aquatic ectotherm to preferentially utilize the higher concentration of oxygen in air as opposed to in water.

Elseya albagula (Thomson, Georges & Limpus 2006) is a freshwater chelonian whose distribution is limited to the Burnett, Mary and Fitzroy Rivers in Queensland, Australia. In the Burnett River, this species inhabits calm, well-oxygenated pools and slow-flowing reaches of the river either upstream or downstream of riffle zones (areas of fast-flowing water), in close association with undercut banks and submerged debris (Limpus, Limpus & Hamann 2002; Gordos et al. 2007), and at depths of up to 8 m (Limpus et al. 2002). The cloacal bursae of E. albagula are lined with highly vascularized, multi-branched papillae which are ventilated with water by muscular pumping (King & Heatwole 1994; Limpus et al. 2002). This morphology resembles that of R. leukops, a species thought to demonstrate the greatest aquatic respiratory ability among aquatic chelonians (Legler & Georges 1993). Elseya albagula is believed to have a medium to high reliance on aquatic respiration, with aerobic dives of greater than 3 h recorded in the field and aquatic oxygen consumption percent in the laboratory ranging from 20–40% (adults) up to 100% (hatchlings) (Mathie & Franklin 2006; Gordos et al. 2007).

The aim of this study was to investigate the effect of water depth, temperature and velocity on the diving behaviour of E. albagula. Applying the principles of Kramer's optimal breathing theory to a bimodal respirer, we predicted that as the cost of surfacing increased there would be an increase in submergence time (dive length), corresponding to reduced surfacing frequency. We also predicted that dive length would decrease as temperature increased; that is, we would observe more frequent trips to the surface at higher temperatures. For a bimodal respirer such as E. albagula variations in submergence time would be made possible by altered respiratory partitioning to favour aquatic respiration. Finally, we examined the use of an artificial velocity refuge by E. albagula to provide insight into micro-habitat selection in the wild. We discuss possible ecological driving forces for the evolution of aquatic respiration in E. albagula and the conservation implications of our findings.

Materials and methods

animal capture and husbandry

Seven juvenile E. albagula of mean body mass 133·3 ± 19·9 g were collected from the Burnett and Mary River systems in Queensland. Eight adult male E. albagula of mean body mass 2240 ± 115 g were captured with traps from the Walla weir on the Burnett River (24°98′98″S, 152°16′50″E; 74·5 AMTD). All turtles were housed at The University of Queensland in large holding tanks and experiments were conducted under ethical clearance (Approval No. ZOO/ENT/387/03/2004). Tanks were supplied with filtered, de-chlorinated, aerated water. Adult turtles were fed twice a week with a combination of meat, fish, fresh fruit and vegetables. Juveniles were fed twice a week with commercial turtle pellets. Turtles were allowed to acclimate for 1 month before experimentation began. Experiments were conducted at The University of Queensland's animal holding facility.

depth and temperature experiments

Trials investigating the effects of water depth and temperature on the diving behaviour of juvenile E. albagula were carried out within a 2·0 m3 dive tank in which water temperature could be manipulated using a spa heater, and depth by emptying or filling the tank with de-chlorinated water. The front of the tank was fitted with Perspex windows so that turtle behaviour could be videotaped. Water PO2 remained at near saturation throughout experimentation, to match the conditions in locations inhabited by this species in the wild. Turtles were provided with structures to hide beneath.

The effect of water depth was measured at three depths (50, 100 and 150 cm) at each of three temperatures (20, 25 and 30 °C), resulting in a total of nine treatments. Temperatures were selected based on those that can be experienced on a seasonal basis by turtles in the wild. Juvenile turtles were placed randomly in groups of either three or four in the dive tank 24 h prior to the start of each experiment, to allow the animals to become accustomed to the conditions of the dive tank (12L/12D light cycle). Turtles were then videotaped using a time lapse VCR for 24 h during which there was a 24 h light cycle. Turtles were not disrupted during recording.

Only resting dives were analysed in this study. A resting dive was classified as any dive where the turtle spent 1 min or greater sitting still on the bottom of the dive tank. Dives were measured from the time the turtle's nares first submerged until the nares re-surfaced. For each turtle and each experiment, mean dive length, maximum dive length and the percentage of time spent in resting dives were calculated. All data was log transformed, with the exception of the percentage of a trial spent in resting dives, which were arc-sine and square-root transformed prior to analysis to ensure normality of data. Relationships between dive length, water depth and temperature were investigated using two-way repeated measure anova (P < 0·05), with Holm-Sidak post hoc tests, using SigmaStat v. 5·0. All means are presented with their SEs unless otherwise stated.

velocity experiments

An oval-shaped, fibre-glass aquatic flume (4·0 × 1·6 × 0·65 m) was used for the velocity experiments. Two Minn Kota electric outboard motors created an adjustable, unidirectional velocity around the flume. A glass observation chamber (1·5 × 0·60 × 0·65 m) on one side of the flume allowed viewing of the turtles. Laminar velocity was created by walls of horizontal PVC pipes (diameter 5 cm) at the ends of the observation chamber. A Perspex insert with raised ridges was laid on the observation chamber floor to allow turtles to walk across the flume. The observation chamber was split into a high velocity and low-velocity region by a 10 cm high, clear plastic deflector ramp, laid across the width of the flume (Fig. 1). Water depth was held constant at 0·5 m. Water temperature was held constant at 25 °C. PO2 remained at near saturation throughout experimentation. Fluorescent lights set on a 12L/12D cycle were positioned above the flume.

Figure 1.

Flow profiles for treatment velocities. (a) 5 cm s−1 (b) 15 cm s−1 (c) 30 cm s−1. Values shown are in cm s−1.

Three experimental water velocities of 5, 15 and 30 cm s−1 were chosen based on observed current strengths from typical riverine pool and riffle zones in the Burnett River (E. Storey, pers. obs.) and in other rivers where species of the Elseya complex are found (Legler & Cann 1980). Within each of the overall velocity settings, measurements of velocity at multiple depths across the length of the observation chamber were taken with a flow-meter (Flow Probe Model FP101, Enviroequip, Sydney, Australia) to generate a flow profile for each experimental velocity setting (Fig. 1). Flow profiles showed that at each of the three main velocity settings, velocity was lower downstream of and under the water deflector, than upstream of it. The deflector thus provided a velocity refuge for turtles.

The order in which turtles underwent trials, as well as the velocity used for a trial, were randomized. Adult turtles were placed one at a time into the flume at 1700 on the day before the experiment began. At 0600 the next day, lights were switched on. At 0700 the experimental velocity was set. Early observations showed that turtles required 2 h to settle after the experimental velocity was changed. After 2 h turtles largely became restful and either sat still on the flume floor or walked slowly around the observation chamber. In nearly all trials, this resting behaviour was interspersed with brief periods of activity which consisted of swimming back and forth across the chamber. As turtles of all groups displayed the same low level of activity across all treatment speeds, activity level was not considered to be a covariate for analysis purposes. At 0900 video recording began. A closed circuit video camera (AVC, model 301A) positioned in front of the observation chamber, and time lapse VCR (National, model AG6010) recorded turtle behaviour over 8 h.

Turtles were allowed to rest for at least one week between trials. Video-tapes were analysed using a frame advance VCR (Sony, model SLV815) for the frequency and duration of surfacing events, resting dive length and percent of time spent outside of the velocity refuge. Dives were measured from the time a turtle's nares first submerged until the nares re-surfaced. Similarly, a surfacing event was defined as the time from when a turtle's nares breached the surface, until they were submerged again. A turtle was considered to be ‘not using the velocity refuge’ when sitting upstream of the water deflector or on top of it, and to be ‘using the velocity refuge’ when downstream of or under it.

Individual values for the proportion of a trial spent not using the velocity refuge were arc-sine and square-root transformed prior to analysis to ensure normality of data. These values, and individual values for median dive length, were then analysed using a one-way repeated measures anova (SigmaStat v. 5·0). In the case of a significant finding, a Holm-Sidak all pairwise multiple comparison post hoc test was used to identify which treatments were different from one another. All results are presented as mean ± SE with the exception of average dive length for the velocity experiments, where the median value was used as a more robust representative of average dive length.


depth and temperature experiments

Temperature had a significant effect on mean (F(2,6) = 46·944, P < 0·001; Fig. 2a) and maximum (F(2,6) = 224·553, P < 0·001; Fig. 2b) dive length of E. albagula. When averaged across all depths, mean dive length decreased from 39·4 ± 9·8 to 12·7 ± 2·4 min as temperature increased from 20 to 30 °C. Pairwise comparison showed all temperature treatments were significantly different from one another for averaged maximum dive length, with values of 153·8 ± 25·4, 63·9 ± 9·5 and 29·6 ± 4·9 min recorded at 20, 25 and 30 °C respectively (in all cases differences were significant after Holm-Sidak correction). The percentage of a trial spent diving was also significantly influenced by temperature (F(2,6) = 29·543, P ≤ 0·001), with the percentage decreasing from 82·4 ± 4·3% at 20 °C to 52·5 ± 8·3% at 30 °C (P < 0·001).

Figure 2.

Effect of water depth and temperature on diving performance. (a) Mean dive length, (b) Averaged maximum dive length. Significant differences between temperature treatments (after allowing for the effects of depth) are indicated by different letters in upper case. Within each temperature treatment, significant differences between depth treatments are indicated by different letters in lower case. Error bars represent one SE.

Water depth had a significant effect on mean dive length (F(2,6) = 16·751, P < 0·001; Fig. 2a). When averaged across the three temperature treatments, mean dive length approximately doubled from 17·6 ± 5·6 to 33·3 ± 6·5 min as depth increased from 50 to 150 cm (P < 0·001). Water depth had no significant effect on maximum dive length (F(2,6) = 2·447, P = 0·128; Fig. 2b). Depth did not have a significant effect on the percentage of a trial spent diving (F(2,7) = 1·968, P = 0·182) and there was no change in percentage diving as depth increased from 50 to 150 cm (63·3 ± 6·8% to 71·4 ± 5·0%, respectively; P < 0·148).

There was no significant interaction between depth and temperature for mean (F(8,6) = 0·625, P = 0·649) or maximum (F(8,6) = 1·786, P = 0·165) dive length, that is, the effect of different temperatures on dive length did not differ with the depth of those dives.

velocity experiments

There was no effect of water velocity on median dive length for adult E. albagula (df = 2, F(2,7) = 0·888 P = 0·915; Fig. 3a). When averaged across the three velocity treatments, median dive length was 21·3 ± 3·8 min. There was no effect of water velocity on maximum dive length (df = 2, F(2,7) = 0·745, P = 0·487; Fig. 3b). Water velocity did not affect the distribution of dives of different lengths, for example, the proportion of total dives that were short (0–10 min) did not differ significantly between velocity treatments (P > 0·05 in all analyses). At all velocities, the majority of dives were less than 20 min long. The majority of surfacing events undertaken by adult E. albagula were less than 60 s long. The median surface duration of adult E. albagula was not affected by water velocity (df = 2, F(2,7) = 0·282, P = 0·758).

Figure 3.

Effect of water velocity on: (a) median dive length, (b) Mean maximum dive length, (c) the average proportion of the trial during which E. albagula did not use the velocity refuge. Error bars represent one SE.

Water velocity had no effect on the proportion of the trial spent not using (i.e. outside of) the velocity refuge (df = 2, F(2,7) = 0·824, P = 0·453; Fig. 3c). When averaged across all velocity treatments, turtles spent approximately half (56·6 ± 16·9%) a trial not using the refuge.


If bimodal turtles obtain oxygen in a way that optimizes their energy use, we should observe a reduced use of aerial oxygen as the cost associated with surfacing increases. Similarly, when surfacing is not energetically costly, aerial respiration should be the primary mode of oxygen acquisition given the high ventilation cost of aquatic respiration (Kramer 1983, 1988; Boutilier 1990; Dejours 1994; Schmidt-Nielsen 1997). In this study, the responses of juvenile turtles to increases in water depth were those we would expect if energy use were being optimized. Average dive length of juvenile E. albagula approximately doubled as depth increased from 50 to 150 cm; that is, surfacing frequency was halved, irrespective of temperature. In contrast, diving behaviour was unaffected by increasing water velocity. Adult E. albagula exhibited similar median dive length (and thus surfacing frequency) across all water velocities, with the majority of dives less than 20 min long. From this we can conclude that water flowing at a velocity of up to 30 cm s−1 does not cue a behavioural response in this species, at least under laboratory conditions.

water depth

Surfacing for air can impose a variety of costs (Kramer & McClure 1980; Kramer et al. 1983; Feder & Moran 1985; Bevan & Kramer 1987; Hua & Wang 1993), one being the energetic cost of travel, the magnitude of which is linked to water depth. This study advances the finding that even at shallow depths in the laboratory, juveniles of a bimodal turtle species are capable of reducing surfacing frequency as depth increases. This is what we would expect if respiratory partitioning were being altered in favour of aquatic respiration and if energy use were being optimized. Despite the lower ventilation cost of air-breathing, for juvenile E. albagula in our study the increase in temporal and energetic costs of travel to the surface as depth increased may have outweighed the ventilation cost of aquatic respiration. Hua & Wang (1993) recorded the same response to increasing depth for the North American soft-shell, T. sinensis, a species capable of substantial aquatic respiration via the bucco-pharynx (Wang, Sun & Sheng 1989). Surfacing frequency of T. sinensis was reduced from 10·2 to 1·5 breaths h−1 when depth was increased from 15 to 45 cm. In contrast, the study by Gordos et al. (2004) on adult R. leukops employed an experimental design similar to the present study (experimental depths of 50, 100 and 150 cm), but found no effect of increasing depth on surfacing frequency. The authors attributed this to the fact that overall travel costs of aerial respiration under the experimental conditions remained low regardless of water depth, due to very infrequent surfacing by the highly aquatic R. leukops.

Body size was also probably influential in producing the marked response to depth changes observed for small (hatchling and juvenile) E. albagula. For a given distance travelled, small animals have a higher cost of transport than large animals (Azuma 1992). This factor would further lower the threshold at which aquatic respiration becomes the more efficient mode for juvenile E. albagula. The results of a study by Mathie & Franklin (2006) further support the capacity for and importance of respiratory partitioning in juvenile E. albagula. Mathie and Franklin showed that in aquatic normoxia, juvenile E. albagula had a higher percent aquatic oxygen consumption than adults as a result of a larger mass-specific cloacal bursae surface area. This allowed juvenile turtles to increase dive duration above that based on aerial respiration alone, to durations equivalent to those of adult turtles.

water velocity

If an animal is acquiring oxygen in an energy-efficient manner, water velocity should theoretically have an impact on respiratory partitioning if it makes surfacing more energetically expensive. In fast-flowing water, an animal would need to expend energy to maintain its position whilst surfacing and risk being swept away in the current, potentially losing its preferred position and sustaining physical injury. In the present study, surfacing frequency of adult E. albagula did not change as water velocity increased. Given that in the wild E. albagula inhabit depths of up to 8 m (Limpus et al. 2002; E. Storey, pers. obs.), it is possible that negligible travel costs in the shallow (50 cm) flume facilitated aerial respiration to such an extent that any cost applied by water velocity was outweighed. However, Gordos et al. (2004) recorded a 20-fold decrease in surfacing frequency for R. leukops also in 50 cm of water, when water velocity was increased from 5 to 30 cm s−1. This suggests that R. leukops, unlike E. albagula, has developed a behavioural response to flowing water which provides an energetic or other benefit. Therefore additional ecological factors (discussed later) may explain the lack of a response to velocity in E. albagula.

The extent to which water velocity influences habitat selection by freshwater turtles is little known (but see Reese & Welsh 1998; Donner-Wright et al. 1999; Tucker et al. 2001). The model explored here of micro-habitat selection based on energetic costs of position-holding was originally applied in studies of stream-dwelling fish. Some species of fish have been shown to avoid micro-habitats with high water velocities because of the metabolic cost required to hold position (Facey & Grossman 1990). Brown trout (Salmo trutta) sought velocity shelters after being exposed to high water velocity in an artificial flume (Vehanen et al. 2000). Rosenfeld & Boss (2001) showed that energetics could account for avoidance of riffles by cutthroat trout (Oncorhynchus clarkii). The observed ‘habitat selection’ behaviour of E. albagula in our flume study does not fit this model, at least under these experimental conditions. It is possible that E. albagula has not adapted an organized response to flowing water because its preferred habitat is not characterized by riffles. In the wild E. albagula is found in calm pools and slow-flowing areas, and often wedges itself beneath rocks and logs and frequents in undercut banks (Limpus et al. 2002; E. Storey, pers. obs.). Conversely, R. leukops has a preferred habitat of shallow, fast-flowing riffle zones (Legler & Cann 1980; Tucker et al. 2001) and showed a significant preference for a velocity refuge under laboratory conditions (Gordos et al. 2004).

water temperature

Water temperature is well established as an environmental variable that affects diving behaviour in freshwater turtles both in the laboratory (Gatten 1980; Ultsch & Jackson 1982; Herbert & Jackson 1985; Prassack et al. 2001; Priest & Franklin 2002; Gordos & Franklin 2003) and field (Gordos et al. 2003). Turtles are ectothermic, thus an increase in water temperature raises metabolic rate and oxygen demand (Schmidt-Nielsen 1997). This occurs concurrently with reduced oxygen availability in the water (Dejours 1994) which might otherwise be used for aquatic respiration by bimodal species. In the present study we consider it likely that a combination of these two factors triggered the observed response of significantly shorter average and maximum dives (more frequent surfacing) as temperature increased from 20 to 30 °C. The 10 °C increase saw a fivefold reduction in maximum dive length. These findings are consistent with those of Prassack et al. (2001), who observed that Apalone ferox, while capable of substantial aquatic respiration, had significantly longer dive times and higher oxygen uptake from the water at 15 °C than at 25 °C.

ecological and evolutionary considerations

A key advantage of Kramer's (1983) cost-benefit model of breathing is that it recognizes the importance of both habitat selection and species traits in influencing the relative costs of respiratory modes for a bimodal respirer. Gordos et al. (2004) postulated that the ability to respire aquatically allows R. leukops to exploit the high levels of dissolved oxygen available in riffle zones and permits energy conservation via reduced surfacing frequency. As previously explained, current opinion would not place E. albagula in the category of riffle specialist; rather it is believed that this species prefers well-oxygenated reaches of calm water near riffle zones. In addition, E. albagula and R. leukops have differing buoyancies, matched, as it were, to the conditions of their respective habitats. Gordos et al. (2004) and Priest & Franklin (2002) reported that R. leukops was negatively buoyant and had difficulty ascending to and remaining at the water's surface during laboratory experiments at depths between 50 and 150 cm. In the present study, E. albagula was neutrally buoyant and navigated horizontally and vertically with ease. Bimodal respirers will obtain more oxygen from air when neutrally buoyant as opposed to negatively buoyant (Kramer 1983). This may also explain the observed differences in the surfacing behaviour of the two species during identical velocity trials. The habitat selected by E. albagula coupled with species-specific traits helps to explain the lack of a definitive behavioural response to fast-flowing water for this species in the current study.

Juvenile E. albagula possess the flexibility to increase dive length, that is, reduce surfacing frequency, in response to changes in temperature and depth. The observed behaviour of reduced surfacing frequency as cost of travel to the surface increased is consistent with that predicted by a theory of optimal breathing. The presence of highly specialized aquatic gas exchange organs in E. albagula raises a question of selective pressure for the evolution of such structures. Given the presence of highly developed aquatic gas exchange organs which themselves incur metabolic and energetic costs (i.e. in development, blood supply and ventilation; Kramer 1983), a corresponding benefit outweighing these costs must exist. We could postulate from our results that the selective pressure is the ability for juvenile turtles to conserve energy through reduced surfacing. Natural selection should favour the most efficient mode of oxygen acquisition, leaving more energy available for growth and feeding (Kramer 1983). Even at shallow experimental depths of 50–150 cm, juveniles undertook longer, and what appeared to be aerobic dives (as indicated by short surface durations) as depth increased, effectively reducing the number of trips to the surface. This is what we would expect if an animal were altering its respiratory partitioning in favour of aquatic respiration. Considering that the natural riverine habitat of E. albagula is characterized by depths of up to 8 m (Limpus et al. 2002; E. Storey, pers. obs.), the energetic costs of aerial respiration particularly for juveniles may have been significant enough to drive evolution of a supplementary, alternative mechanism for oxygen acquisition in this species. Aquatic respiration may confer a critical advantage on juveniles, which have a high rate of mortality in the wild. Further study to produce quantitative measurements of respiratory partitioning in response to depth changes is needed to explore this hypothesis.

conservation implications

An important outcome of this study has been to advance the proposition that travel costs associated with aerial respiration have behavioural and perhaps even evolutionary consequences for E. albagula. Turtle populations in the Burnett River are likely to experience impacts of water resource development, namely dams and weirs. The impacts of these structures on freshwater turtle populations is still poorly understood (Vandewalle & Christiansen 1996; Gordos et al. 2007). However, water PO2, temperature, depth and velocity are all impacted by hydrological alteration of rivers (Ward & Stanford 1989). Laboratory and limited field studies (Gordos et al. 2003, 2007) have shown that these variables affect aquatic respiration and surfacing patterns of turtles, to a greater or lesser extent. Construction of dams and weirs converts natural riverine habitat of riffles and pools into deep, lentic reservoirs with associated declines in water quality and oxygen saturation (Ward & Stanford 1989; Dodd 1990). The advantage of aquatic respiration is negated in deep, oxygen-poor water. Furthermore, excessive depth is likely to present a problem for juvenile turtles if, as our results suggest, they are sensitive to energetic travel costs. Increased difficulty in meeting oxygen demands, combined with the host of other ecologically-damaging effects of river regulation (Dodd 1990; Vandewalle & Christiansen 1996), could have severe repercussions for fitness of juvenile turtles, and on E. albagula population viability as a whole.


The authors would like to thank the Queensland Parks and Wildlife Service for their assistance with turtle collection and husbandry advice. Turtle collection and experimentation was approved by the Queensland Parks and Wildlife Service (SPP-WISP01477903) and supported by the University of Queensland Ethics Committee (AEC-ZOO/ENT/595/04/URG). This research was funded by a University of Queensland Research Grant to CEF. We declare that the experiments comply with the current laws of Australia where the experiments were carried out. Authors thank Dr Simon Blomberg for statistical advice.