Heat storage in Asian elephants during submaximal exercise: behavioral regulation of thermoregulatory constraints on activity in endothermic gigantotherms

Scaling relationships suggest that gigantic animals have difficulty dissipating metabolic heat. Resting metabolic heat production, M, is proportional to body mass, m, as M∝m_b^0.75 in mammals and M∝m_b^0.83 in reptiles, while the total skin surface area, A, is proportional to m_b^0.67 (Benedict, 1932; Benedict, 1936; Calder, 1984; Schmidt-Nielsen, 1984). Therefore, as m_b increases, M increases at a faster rate than A (M/A∝m_b^0.08–0.16). Further, the distance, x, that heat must be transferred from metabolically active organs in the body core to the skin surface increases with body size (x∝m_b^0.33). Gigantic animals are thus predisposed to storage of metabolic heat in core tissues.

Gigantic size presents both opportunities and challenges in thermoregulation. On the one hand, large body size permits the maintenance of fairly constant core body temperatures in animals with low reptile-like metabolic rates by means of gigantothermy (Paladino et al., 1990). Gigantothermy is a combination of ectothermic homeothermy (Spotila et al., 1973), the maintenance of a relatively high body temperature by vascular adjustments in tissue insulation (Fig. 1), with inertial homeothermy resulting from large body mass (McNab and Auffenberg, 1976). Gigantothermy has been demonstrated in leatherback turtles (Dermochelys coriacea) with adult body mass up to 900 kg and hypothesized for an ~3700 kg migratory dinosaur, Edmontosaurus, assumed to have a low reptile-like metabolism (Paladino et al., 1990; Spotila et al., 1991). On the other hand, gigantothermy combined with a high mammal-like metabolic rate and activity likely results in heat production rates exceeding heat loss rates (Spotila et al., 1991; O'Connor and Dodson, 1999). In tropical environments, it has been suggested in both elephants and endothermic dinosaurs that substantial amounts of heat storage in tissues might result in a potentially lethal rise in core body temperature (Spotila et al., 1991; O'Connor and Dodson, 1999). Indeed, studies of both wild and domesticated elephants have found that activity can result in a rapid increase in core temperature, ranging from 0.5 to 6.0°C (Baldwin, 1974; Toscano et al., 2001). However, by concentrating activity at night, when radiant environmental heat is at a minimum (Elder and Rodgers, 1975; Guy, 1976; Douglas-Hamilton et al., 2005; Kinahan et al., 2007; Graham et al., 2009; Joshi, 2009), elephants may behaviorally reduce the amount of metabolic and environmental heat stored in tissues and facilitate activity without risk of a lethal increase in core body temperature.

Mathematical heat transfer models have provided valuable estimates of thermoregulatory constraints on gigantotherms with low reptile- or high mammal-like metabolism (Spotila et al., 1991; O'Connor and Dodson, 1999). Few living animal models are available for validating computer models of gigantothermy. However, well-trained captive elephants are large, tractable and utilize vascular adjustments to regulate heat transfer from the body core to the skin surface (Phillips and Heath, 1995; Weissenböck et al., 2010). Therefore, captive elephants are an excellent model for thermoregulatory constraints on activity in endothermic gigantotherms. In the present study, we use active Asian elephants (Elephas maximus) as a model animal for endothermic gigantothermy, much as leatherback turtles have been used as a model for ectothermic gigantothermy (Paladino et al., 1990).

Specifically, we test the hypothesis that there is a functionally significant relationship between heat storage and locomotion in elephants, and that this relationship is sensitive to the level of radiant environmental heat. Several studies have reported on heat storage and energy budgets of resting elephants in both outdoor (Hiley, 1975; Rowe, 1999; Weissenböck et al., 2012) and indoor conditions (Williams, 1990). However, no published studies have examined thermoregulation of active elephants in quasi-natural context where they are subjected to a wide range of environmental heat loads. Specifically, we expect to shed light on the thermoregulatory constraints on activity in endothermic gigantotherms, and use these results to model and discuss behavioral regulation of heat storage resulting from prolonged activity in elephants and a similarly sized dinosaur, Edmontosaurus.

Two adult female Asian elephants Elephas maximus Linnaeus 1758 (Panya and Jean; Table 1), at the Audubon Zoo in New Orleans, LA, USA, were used for all of the measurements. Both elephants were managed in free contact with keepers, and were very tractable and well trained. Their feeding schedule was unaltered and water was available ad libitum except during the exercise trials. All methods were approved by the Indiana State University and the Audubon Zoo Institutional Animal Care and Use Committees. As is commonly the case when working with large, rare animals, sample size was limited by availability. Thus, our results are strictly valid only for these two elephants, but we have no reason to believe they are not typical.

All of the exercise trials during the June 2009 study period were started before 10:00 h (15:00 UTC) or after 17:20 h CDT (22:20 UTC). During the February and November 2009 study periods, all trials were started before 10:00 h CST (16:00 UTC). Whenever possible both elephants were exercised simultaneously. At the start of each exercise trial, the elephants were led by their keepers from the exhibit area into the barn or to a shaded area outside of the elephant exhibit, where thermograms and rectal temperatures were recorded. Thermograms of the elephants and the track were recorded using a FLIR ThermaCAM PM575 (FLIR, Portland, OR, USA) radiometric thermal imaging camera (sensitivity of ca. ±0.1°C) fitted with a 45 deg lens. Images were taken from a distance of ~5 m from the elephant. The mean (±s.d.) temperature (°C) of the elephants' skin, T_r (Fig. 2), and the track surface temperatures, T_g (°C), were measured using ThermaCAM Researcher Professional version 2.7 software (FLIR). The add polygon tool was used to outlining the elephants' body and track surfaces and determine the mean (±s.d.) T_r and T_g in each thermogram. Thermal radiation was converted to surface temperatures (°C) using the settings tools and concurrently recorded ambient air temperature and relative humidity and assuming radiating surface emissivities were 0.96 for elephant skin and 0.93 for asphalt (Gates, 1980).

As a measure of core body temperature, rectal temperature, T_b (°C; Fig. 3), was measured for 5 to 7 min prior to the start of each exercise trial using a 35 cm thermocouple probe attached to a COMARK N9002 (COMARK, Hitchin, UK) thermocouple thermometer (accuracy ±0.1°C). We attached an accelerometer (SENSR, GP1 programmable accelerometer, Elkader, IA, USA) on the right rear leg of the elephants to provide a time stamp for determining walking speed. The elephants walked either one lap (945 m) or two laps (1614 m) around a closed circuit at speeds ranging from 0.56 to 1.25 m s⁻¹ (Table 2). The length of the track was measured using a Bushnell (Yardage Pro Compact 800, Bushnell, Overland Park, KS, USA) range finder, and average speed was computed by dividing track length by time needed to complete the circuit. At the end of each trial thermograms and rectal temperatures were again recorded.

Environmental conditions needed for heat transfer calculations (Table 3) were recorded during exercise trials. Parameters were measured every 5 min, and averaged and recorded every 30 min using a HOBO® Micro-Station (Onset Computer Corporation, Bourne, MA, USA). The station was located in an exposed position on the periphery of the exercise track. A shielded thermistor measured ambient air temperature (T_a; ±0.2°C), a capacitive humidity sensor measured relative humidity (RH; ±2.5%; used to correct thermograms) and a cup anemometer measured sustained wind speed (u; ±1.1 m s⁻¹). A silicon pyranometer measured (±10 W m⁻²) global solar radiation, and a second silicon pyranometer shielded from direct sunlight by a shade ring measured diffuse solar radiation. Direct solar radiation was determined by subtracting diffuse from global radiation and dividing by the cosine of the zenith angle of the sun. Zenith angle at the time of each exercise trial was determined using the online solar position calculator available from the US Department of Commerce (www.esrl.noaa.gov/gmd/grad/solcalc). The sensors were set at elephant head height (i.e. 2 to 2.75 m above the ground).

The Porter and Gates (Porter and Gates, 1969) model assumed an isothermal core body. Because of their gigantic size and exposure to a wide variety of environmental conditions, it is likely that variations in effective tissue insulation, I (Fig. 1), influence the size (i.e. mass) of the isothermal core tissues in elephants. Therefore, following the preliminary analyses of the thermal energy budget, we estimate the influence of a variable I on the rate of heat storage by varying core mass from 50 to 100% of total body mass. Similarly, following preliminary analyses of the thermal energy budget, we estimate the influence of behaviorally regulating heat transfer and storage by switching from diurnal to nocturnal activity.

Seasonal mean (±s.d.) Q_n (Table 4) values were then estimated from means of R_s and Q_a from individual elephants by pooling estimates for each trial and dividing by the number of exercise trials conducted during a season.

Conductive heat loss, K (W) (Gates, 1980), refers to heat exchanged through the area of contact between the plantar surface of the elephant's feet and the supporting ground surface. During locomotion, the bottoms of the elephants' feet are only in contact with the ground a small portion (i.e. the stance phase) of a stride. No skin temperatures of the bottom of the elephant's feet were recorded. Published estimates indicate that K through the plantar skin surface accounts for ≤2% of the total heat loss in a standing elephants (Hiley, 1975; Williams, 1990; Rowe, 1999). In addition, during the June trials, track temperature was hotter than core body temperature, which would cause a conductive heat gain. Therefore, we omitted conductive heat transfer K due to its minor contribution to overall thermal balance.

To have predictive precision, mathematical heat transfer models must include an accurate description of all relevant variables and be simple enough to be intellectually tractable, which is challenging given the complexity of biological systems (Bakken and Gates, 1975). We modeled a linear first approximation of the increase in core body temperature resulting from continuous locomotion (Fig. 6) and discuss the thermoregulatory constraints on activity proposed in elephants (Fig. 6A) and endothermic dinosaurs (Fig. 6B) (Spotila et al., 1991; O'Connor and Dodson, 1999).

We estimated the thermoregulatory constraints on continuous activity in endothermic gigantotherms by rearranging the heat storage equation (Eqn 10) to solve for the change in core body temperature, dT_b, where dt(X/3430)m_b=dT_b. In this model, seasonal rates of heat storage, X (Table 4), remained constant for the duration of locomotion; however, the duration of activity, dt, varied between 1 and 8 h. To include the influence of effective insulation (Fig. 4), the fraction of core body mass storing heat, m_b, was varied from 50% during nocturnal activity (Fig. 3B) to 75% in full sun during February and November (Fig. 3A) and 100% in full sun during June (Fig. 3A). The elephant model makes four general assumptions. First, active elephants experienced the seasonal mean rates of heat transfers in the full sun and nocturnal environments (Fig. 3, Table 4). Second, a starting core body temperature of 35.3±0.44°C, the mean (±s.d.) pre-exercise T_b recorded in elephants during the June trials. Third, as core body temperature increased during activity, the Q₁₀ effect on metabolic heat production was matched by an approximately equivalent increase in heat transfer from the elephants to the environment. Fourth, 43°C was considered a potentially lethal core body temperature (Simon, 1999).

Similarly, we model and discuss the thermoregulatory constraints on activity proposed in endothermic and ectothermic Edmontosaurus. The Edmontosaurus model uses the same arrangement of Eqn 10 to determine dT_b during activity. Our Edmontosaurus model was based on endothermic and ectothermic Edmontosaurus with a body mass of 3655 kg (Spotila et al., 1991). Energetics models predicted that the 30% lower metabolic rate of ectothermic Edmontosaurus resulted in an energetic advantage over endothermic Edmontosaurus during migrations (Spotila et al., 1991). Estimates of active metabolic heat production in endothermic and ectothermic Edmontosaurus (Spotila et al., 1991) were ~4 and 30% less than values recorded in elephants, respectively (Langman et al., 1995; Langman et al., 2012). The Edmontosaurus model makes the same four general assumptions as the elephant model with two additions. First, we used the same spherical radiant heat transfer model (O'Connor and Spotila, 1992) used to describe radiant environmental heat (Q_a; Eqn 6) in elephants, which assumes that the rates of heat transfer and heat storage were not affected by differences in body shape between elephants and Edmontosaurus. Second, the solar absorptance of Edmontosaurus skin was similar to that of elephants (Table 1). However, the solar absorptance of reptile skin, ≈90%, is generally greater than that of mammals (Gates, 1980). Therefore, we discuss the influence of skin with a greater solar absorptance on heat storage during activity in ectothermic Edmontosaurus.

Statistical analyses were performed on seasonal means (±s.d.) of data pooled from both elephants (Fig. 2, Tables 2, 3). Animal thermoregulation studies are often carried out in an environmental room and locomotion studies are generally performed using a motorized treadmill. Both approaches provide the experimenter with control of experimental conditions. However, environmental rooms and treadmills suitable for elephants are rare. Therefore, seasonal means of data pooled from both elephants (Fig. 2, Tables 2, 3) were chosen as the level of statistical analysis. In addition, the small sample size of two elephants reduced the power of between-subjects statistical analyses, i.e. d.f.=1. Active metabolic heat production, M_ex, reported in elephants ranging in body mass from 1435 to 3545 kg were similar (Langman et al., 1995; Langman et al., 2012) and skin surface area to body mass ratio in the two elephants in the present study were similar (Table 1). Therefore, between-elephant variations in M_ex and rates of heat transfer were assumed to have a negligible effect on seasonal data, presented as means ± 1 s.d.

Seasonal mean M_ex (Table 2) was calculated from v_f (Eqn 2); therefore, ANOVA was used to determine whether statistically significant variations in v_f occurred between seasons. Similarly, ANOVA was used to determine whether statistically significant variations in pre-exercise T_b and T_r (Fig. 2), post-exercise increase in T_b and T_r (Fig. 2), pre- and post-exercise temperature gradient T_b–T_r (Fig. 2), and environmental variables, including solar radiation falling on a plane perpendicular to the sun's rays (S_n), atmospheric long-wave radiation (R_a), long-wave radiation from the ground surfaces (R_g) and wind speed (u; Table 3), occurred between seasons (Fig. 2). When statistically significant seasonal variations were found, a post hoc Tukey's honestly significant difference test was used to determine where statistically significant between-seasons variation occurred (Zar, 1999). The level of statistical significance was set at P≤0.05 for all analyses. Graphing and statistical analyses were performed using KaleidaGraph 4.03 (Synergy Software, Reading, PA, USA).

Over the course of the experimental trials, mean walking speeds recorded in the two Asian elephants were similar, 0.96±0.17 and 1.09±0.15 m s⁻¹ in Panya and Jean, respectively. Seasonal mean walking speeds, v_f (Table 2), were similar to the range of walking speeds (from ~1.0 to 1.3 m s⁻¹) that minimize the energetic and biomechanical costs of locomotion in elephants (Langman et al., 1995; Langman et al., 2012; Hutchinson et al., 2003; Hutchinson et al., 2006; Ren and Hutchinson, 2008; Genin et al., 2010). Seasonal variations in mean walking speeds (Table 2) were not statistically significant (ANOVA, d.f.=2, F=1.10, P=0.35). Therefore, seasonal estimates of active metabolic heat production, M_ex, based on walking speed (Eqn 2) were similar (Table 2).

Pre- and post-exercise core body temperature recorded in individual elephants increased with increasing ambient air temperature (Fig. 2). Over the range of air temperatures from 8 to 34.5°C, statistically significant (ANOVA, d.f.=2, F=7.1, P<0.003) seasonal variations in pre-exercise core body temperature were recorded. The mean pre-exercise core body temperature recorded during the June trials of 35.3±0.4°C was significantly greater than the pre-exercise core body temperature of 34.5±0.6°C recorded in the November trials (Tukey's all pairs comparison, P≤0.003).

Core body temperature in the two Asian elephants increased following all exercise trials (Fig. 2). Statistically significant (ANOVA, d.f.=2, F=6.06, P≤0.006) seasonal variations in mean post-exercise increases in core body temperature were recorded in elephants, with values of 0.38±0.19°C (mean trial time of 23.9±2.9 min), 0.48±0.21°C (mean trial time of 18.9±4.5 min) and 0.70±0.39°C (mean trial time of 21.7±5.5min) during February, November and June trials, respectively. The post-exercise increase in core body temperature recorded during the June trials was significantly greater than that during the February trials (Tukey's all pairs comparison, P≤0.005).

Pre- and post-exercise mean skin temperature in individual elephants increased with increasing ambient air temperature (Fig. 2). Over the range of air temperatures from 8 to 34.5°C, statistically significant seasonal variations in pre-exercise mean skin temperature were recorded (ANOVA, d.f.=2, F=85.1, P<0.0001). The pre-exercise mean skin temperature recorded in the June trials of 35.5±1.2°C was significantly greater than the pre-exercise mean skin temperatures of 26.0±2.9 and 24.9±2.3°C recorded in the February and November trials, respectively (Tukey's all pairs comparison, P≤0.001). Following exercise, mean skin temperature increased by 0.6±2.1, 1.4±1.7 and 1.7±1.3°C during the November, February and June trials, respectively. Seasonal mean post-exercise increases in mean skin temperature were not statistically significant (ANOVA, d.f.=2, F=1.32, P=0.28; Fig. 2).

Statistically significant seasonal variations in the temperature gradient between core body and skin (T_b–T_r; Fig. 2) occurred both pre- (ANOVA, d.f.=2, F=78.4, P≤0.001) and post-exercise (ANOVA, d.f.=2, F=73.3, P≤0.001). During the cool-weather November and February trials, T_b–T_r was >9°C. However, during the hot-weather June trials, heating of the elephants' skin surface by radiant environmental heat resulted in the reversal of the temperature gradient, i.e. conductance of heat from the skin surface to the core body tissues (Figs 1, 2). Indeed, during the June trials, T_b–T_r of −0.2±1.2 and −1.2±1.1°C were recorded pre- and post-exercise, respectively.

When active in full sun, exercising elephants experienced rates of total heat gain of ~9754, 5988 and 5577 W in the June, February and November trials, respectively (Fig. 3A). During slow walks, the two elephants in the present study had estimated 2- to 2.5-fold increases in the rate of metabolic heat production, M_ex (Eqn 2), above resting rates (Langman et al., 2012). Estimates of respiratory evaporative heat loss, E_b (Eqn 3), were 4.7, 12.4 and 12.8% of M_ex during June, February and November trials, respectively (Table 2). Our estimates of E_b were similar to previous estimates of 10 to 12% of resting metabolic heat production in elephants exposed to ambient air temperatures that ranged from ~12.5 to 25°C (Benedict, 1936; Williams, 1990). Our lowest estimate of E_b recorded during the June trials was the result of hotter ambient air temperatures that averaged 31.4±1.7°C (Table 3) and high absolute humidity.

Dry active metabolic heat production, M_ex–E_b, was similar between seasons, varying on average by less than ~448 W (Table 2, Fig. 3). However, over the range of air temperatures from 8 to 34.5°C, we recorded statistically significant 1.3- to 2.4-fold seasonal variations in the primary variables that influence radiant environmental heat, Q_a (Eqn 6), including: direct solar radiation, S_n (ANOVA, d.f.=2, F=24.7, P<0.0001); atmospheric long-wave radiation, R_a (ANOVA, d.f.=2, F=27.4, P<0.0001); and long-wave radiation from the ground surface, R_g (ANOVA, d.f.=2, F=23.5, P<0.0001; Table 3). Seasonal variations in radiation (Table 3) influenced the rate and direction of net radiant heat transfer, ±Q_n (Eqn 4, Table 4). Therefore, during exercise in full sun, M_ex–E_b (Table 2) remained relatively constant between seasons, but because of large seasonal variations in ±Q_n (Table 4), M_ex–E_b was equivalent to ~59, 89 and 100% of the rate of total heat gain during the June, February and November trials, respectively (Fig. 3A). Conversely, ±Q_n was equivalent to 41, 11 and 0% of the rate of total heat gain during the June, February and November trials, respectively.

When active in full sun, exercising elephants experienced rates of total heat loss of ~1091, 2479 and 2644 W during the June, November and February and trials, respectively (Fig. 3A). Because of large variations in ±Q_n, convection, C_ex (Eqn 9, Table 4), was often the only avenue of heat loss available in exercising elephants (Fig. 3A). During the June trials a statistically significant (ANOVA, d.f.=2, F=8.8, P<0.001) 3- to 4-fold decrease in wind speed, u (m s⁻¹; Table 3), in combination with a small, ~3.9±1.7°C temperature gradient between elephants' skin and ambient air temperature (T_r–T_a; Fig. 2), resulted in a 2.3-fold decrease in convective heat loss in exercising elephants (Table 4). During the June trials C_ex dissipated ~27% of −Q_n (Fig. 3A, Table 4). Therefore, in addition to 100% of M_ex–E_b (Table 3), the remaining 73% of Q_n (~2899 W; Table 4) was stored in peripheral body tissues. Conversely, during the November and February trials, when ±Q_n was reduced, the mean rates of total heat loss (+Q_n+C_ex; Fig. 3A, Table 4) dissipated by ~37 to 43% of M_ex–E_b (Table 3), respectively.

In all seasons, a fraction of M_ex–E_b (Table 2, Fig. 3) could not be transferred to the environment; as a result, core body temperature increased following all exercise trials (Fig. 1). During activity in full sun, the rate of heat storage, X (Eqn 10), in active elephants was sensitive to ±Q_n (Table 4). The walking speed (Table 2), and thus M_ex–E_b, in active elephants remained similar between seasons (Table 2, Fig. 3). However, ±Q_n in June in full sun was 36 to 48% greater than during the February and November trials, respectively (Table 4). Because M_ex–E_b remained relatively constant between seasons, the 37-fold increase in Q_n likely influenced the higher rate of heat storage recorded during the June trials (Fig. 3A, Table 4). Assuming that post-exercise increases in core body temperature represent uniform heating in 100% of core tissues, the rate of heat storage varied 2.2-fold between seasons (Table 4), and was equivalent to ~68 to >100% of M_ex–E_b (Fig. 3A).

The large size of the elephants' body core, and changes in effective tissue insulation (Figs 1, 4), likely results in regional heterothermy, which in turn influences the rate of heat storage (Fig. 3). Asian elephants adjusted effective tissue insulation in response to changing environmental conditions (Fig. 4). Over a range of ambient air temperatures, from 8 to 34.5°C, effective tissue insulation decreased ~8-fold, from 0.004 to −0.0005°C W⁻¹. At ambient air temperatures greater than ~30°C, skin temperatures were often hotter than core body temperature (Fig. 2), which indicates heating of the skin surface by direct solar radiation and a minimization of tissue insulation (Fig. 4).

Indeed, in hot conditions at the start of exercise trials, skin temperature was often equal to core body temperature (Fig. 2, Fig. 5A) due to vasodilatation and perfusion of peripheral tissues. Walking in full sun resulted in solar radiation heating the elephants' skin surface (Fig. 5B). During the June hot weather exercise trials, at an ambient air temperature of 31.4±1.7°C, the pre-exercise skin temperature of 35.5±1.2°C was approximately equal to core body temperature, 35.3±0.4°C, indicating vasodilatation and full perfusion of peripheral tissues (Figs 1, 2). Following exercise in full sun, core body and mean skin temperatures increased to 36.0±0.3 and 37.3±1.1°C, respectively (Fig. 3). The post-exercise increase core temperature of 0.7°C was the result of metabolic heating. However, the 1.8°C increase in skin temperature was more than metabolic heating could produce and was likely a consequence of the heating of the elephants' skin by solar radiation (Fig. 5B). Therefore, when walking in full sun during the hot weather June trials, in addition to the storage of 100% of M_ex–E_b, the rate of heat storage was influenced by solar heating in peripheral tissue layers. Conversely, in the cool weather November and February trials, at an ambient air temperature of 13.7±1.7 and 16.2±4.0°C, respectively, post-exercise increases in core temperature of 0.48±0.21 and 0.38±0.19°C in less than 100% of the core tissue mass stored ~56 to 63% of M_ex–E_b (Fig. 3A).

During the hot weather June trials, statistically significant (P<0.0001) increases radiant environmental heat (Table 3) resulted in Q_n accounting for ~41% of the total heat gain (Fig. 5A). In all seasons, nocturnal activity in the absence of solar radiation would facilitate net radiant heat loss in active elephants (Q_n; Fig. 3B, Table 4). During June, in the nocturnal environment with an air temperature of 31.4±1.7°C, in the absence of solar radiation, M_ex–E_b would be the only avenue of heat gain in active elephants (Fig. 3B). Therefore, the total amount of heat loss by net radiant and convective modes, +Q_n+C_ex, would dissipate a greater fraction of M_ex–E_b. During nocturnal activity, an estimated 36% of M_ex–E_b in hot weather June trials and >90% of M_ex–E_b in cooler weather February and November trials could be dissipated by +Q_n+C_ex. The greater dissipation of M_ex–E_b during nocturnal exercise would result in either a smaller portion of core tissue mass increasing in temperature (Fig. 1) or a smaller increase in core temperature (Fig. 2).

The primary thermoregulatory challenge for endothermic gigantotherms is the dissipation of active metabolic heat production in a hot environment (Spotila et al., 1973; Paladino et al., 1990; O'Connor and Dodson, 1999; Spotila et al., 1991). Indeed, as a result of allometric constraints on heat loss and seasonal variations radiant environmental heat (Tables 3, 4), during exercise trials that lasted 21.8±4.7 min, ~0, 38 and 44% of M_ex–E_b in Asian elephants (Table 2) was transferred to the environment during the June, February and November trials, respectively (Fig. 3A). Conversely, ~56, 62 and 100% of M_ex–E_b was stored in core tissues during the November, February and June trials, respectively (Fig. 3A). Over the range of ambient air temperatures, from 8 to 34.5°C, following all short duration exercise trials, increases in core body temperature from between 0.2 and 1.4°C were recorded in elephants (Fig. 2, Table 2). Similarly, at an ambient air temperature of ~26°C, arterial blood temperature, as an indicator of core body temperature, increased by ~5°C to a potentially lethal temperature in an elephant following 15 min of locomotion (Baldwin, 1974). Our results support the hypothesis that there is a functionally significant relationship between heat storage and locomotion in elephants, and that this relationship is sensitive to the level of radiant environmental heat.

Asian elephants used vascular adjustments in peripheral blood flow (Phillips and Heath, 1995; Weissenböck et al., 2010) to regulate effective tissue insulation and the conductance of heat from the body core to the skin surface (Figs 1, 4). In hot conditions, at ambient air temperatures above ≈30°C, effective tissue insulation, I, was minimized, i.e. I≤0.0°C W⁻¹ (Fig. 4). At that point, the direction, i.e. heat gain or loss, is influenced by radiant environmental heat, particularly solar radiation (Fig. 3A,B, Tables 3, 4). In a hot environment in full sun, in addition to 100% of M_ex–E_b, solar radiation warmed the skin surface to a temperature greater than core body temperature (Figs 2, 5), which resulted in a fraction (~44%) of net radiant heat gain, −Q_n, contributing to heat storage in 100% of core body mass (Fig. 3A). Conversely, in a cool environment, during the February and November trials, at ambient air temperatures of 13.7±1.7 and 16.2±4.0°C, respectively, I increased ~7.7-fold (Fig. 4) and therefore the mass of core tissues storing heat (m_b; Eqn 10) during activity (Fig. 3A,B) was influenced by increased I (Fig. 4).

In hot environments, elephants likely use behavioral mechanisms such as nocturnal activity to regulate the rate of heat storage, X (Fig. 3B). It is not surprising that some elephants in tropical environments behaviorally select nocturnal activity (Elder and Rodgers, 1975; Guy, 1976; Douglas-Hamilton et al., 2005; Kinahan et al., 2007; Graham et al., 2009; Joshi, 2009). Although other factors, such as minimizing human contact, also favor nocturnal activity (Douglas-Hamilton et al., 2005; Graham et al., 2009), avoidance of high radiant heat loads (Kinahan et al., 2007) is likely a major consideration favoring nocturnal activity. Our results indicated that in full sun, at an ambient air temperature of 31.4±1.7°C, greater than 100% of M_ex–E_b was stored in core tissues as a result of solar heating of skin (Fig 2, Fig. 3A, Fig. 5). In a hot environment, avoiding solar radiation by nocturnal locomotor activity increased the total rate of heat loss, Q_n+C_ex, by ~92%, and reduced the rate of heat storage in active elephants by ~57% (Fig. 3B). Nocturnal activity reduced the fraction of M_ex–E_b stored in tissues from 100 to ~64% (Fig. 3B).

When active in full sun, even during the cool weather November and February trials with mean ambient air temperatures of 13.7±3.4 and 16.2±4.0°C, respectively, Asian elephants had rates of active metabolic heat production, M_ex–E_b (Table 2), that were 2.2- to 2.7-fold higher than total heat loss, ±Q_n+C_ex (Table 4). During activity in full sun in the hot weather June trials at mean ambient air temperatures of 31.4±1.4°C, M_ex–E_b (Table 2) was ~5.3-fold higher than convective heat loss, C_ex, which was the only functional heat loss mechanism available to active elephants (Table 4).

In tropical environments, heat storage in active elephants and endothermic dinosaurs might result in a potentially lethal rise in core body temperature (Spotila et al., 1991; O'Connor and Dodson, 1999). Elephants can experience ambient air temperatures that range from 0°C to above 40°C (Sukumar, 1989; Kinahan et al., 2007), and some elephants travel long distances. Home range size for elephants varies from ~50 km² to more than 2500 km², and some populations of elephants engage in seasonal migrations of up to 140 km (Guy, 1976; Lethold, 1977; Sukumar, 1989; Lindeque and Lindeque, 1991; Tchamba, 1993; Thouless, 1995; Joshi, 2009). In the present study, a maximum post-exercise increase in core body temperature of 1.4°C and a maximum core body temperature of 36.3°C were simultaneously recorded in an individual elephant following a 1.6 km walk at an average speed of 1.0 m s⁻¹ at an ambient air temperature of 32.7°C. A continuous diurnal walk of 16 km, which lasted ~3.5 h, was recorded in wild African elephants walking at a speed of 1.3 m s⁻¹ (Guy, 1976). At an ambient air temperature of 31.4±1.7°C, we estimate that a potentially lethal increase core body temperature of ~8.0°C, i.e. to a core body temperature of 43°C (Simon, 1999), could occur in elephants after ~4 h of continuous locomotion in full sun (Fig. 6A). We estimate that choosing nocturnal activity, in the absence of solar radiation, would relax the thermoregulatory constraints on activity in endothermic gigantotherms and allow elephants approximately 7 to 8 h of continuous locomotion before the onset of lethal core body temperature (Fig. 6A).

In addition, because bathing and wallowing are the only means available to increase evaporative heat loss (Lillywhite and Stein, 1987), migratory routes in elephants are in often in close proximity to water (Lindeque and Lindeque, 1991; Tchamba, 1993; Joshi, 2009). Immediately following exercise events in hot conditions, one of the elephants, Panya, often chose to enter the pool and remain partially submerged for several hours (Fig. 7). Despite a relatively warm water temperature of approximately 30°C, the pool provided a behavioral option that also increased the rate of convective cooling ~8-fold. Further studies in elephants that regularly walk longer distances than we present here (Table 2) are necessary to confirm or deny the validity of our heat storage model in active elephants (Fig. 6A).

Edmontosaurus is the archetype of presumably migratory dinosaurs (Bell and Snively, 2011). During the Late Cretaceous, Edmontosaurus was widely distributed as far north as paleoarctic Alaska, above 70°N latitude, and seasonal migrations of 2000 to 3000 km were likely necessary to avoid exposure to prolonged (>3 months) darkness and reduced availability of food (Clemens and Nelms, 1993; Bell and Snively, 2011). The climate of the likely southern destinations for Edmontosaurus, located between 51 and 41°N latitude, was similar to present-day New Orleans, LA, USA, characterized by hot, humid summers and mild winters (Dodson, 1971). Therefore, it is highly likely that active Edmontosaurus experienced environmental conditions similar to those we recorded during the June trials (Table 3).

As was the case with elephants in a hot environment, our model indicates that after ~3.5 h of locomotion in full sun, heat storage might result in a potentially lethal rise in core body temperature in endothermic Edmontosaurus (Fig. 6B). An ectothermic metabolism would have reduced the rate of heat storage in Edmontosaurus by approximately 18%. However, the thermoregulatory advantage of reduced heat production in active ectothermic gigantotherms is questionable. We estimate that ectothermic Edmontosaurus would experience a dangerously high increase in core body temperature after ~4 h of diurnal locomotion (Fig. 6B). In addition, our model assumes that the skin of Edmontosaurus had a solar absorptance similar to that of elephants, ≈79% (Table 1). Although we have no way of confirming the absorptance of dinosaur skin, the skin of reptiles generally has a higher absorptance for solar radiation, ≈90% (Gates, 1980). In ectothermic Edmontosaurus, a higher rate of solar absorptance would increase the rate of heat storage and accelerate the onset of a lethal core temperature by ~0.5 to 1 h. Therefore, in a hot environment, the duration of activity prior to the onset of a lethal core temperature would be approximately equal in an endothermic Edmontosaurus and an ectothermic Edmontosaurus with reptile-like solar absorptance.

There is no evidence for or against nocturnal activity in Edmontosaurus. Gigantothermy would have permitted nocturnal activity even if they were ectothermic. As for elephants, our model indicates that endothermic Edmontosaurus could have reduced a potentially lethal rate of heat storage by behaviorally selecting nocturnal activity. We estimate that nocturnal activity would reduce the rate of increase in core body temperature in endothermic Edmontosaurus by approximately 56% (Fig. 6B). By choosing nocturnal activity in a hot climate, Edmontosaurus could likely postpone the onset of lethal core body temperature by ~4 to 5 h (Fig. 6B). At that point, the challenge for Edmontosaurus would be dissipation of stored heat either by wallowing or finding a shade microclimate. The models we present do not prove or dispute endothermic or ectothermic metabolism in Edmontosaurus. However, like behavioral choices made by elephants, the models we present do support the hypothesis that nocturnal activity and avoiding solar heat loads might have reduced potentially lethal increases in core body temperature in endothermic dinosaurs.

Although tail anatomy indicates that Edmontosaurus was not morphologically adapted for swimming, its preferred habitat was in close proximity to water (Bakker, 1986). Thus, like elephants, Edmontonsaurus might have used periodic bathing and wallowing as a behavioral thermoregulatory option to increase heat loss during activity in a hot environment. In addition, fossil evidence of gigantic feathered dinosaurs of northeastern China has recently been reported (Xu et al., 2012). Like polar Edmontosaurus, these feathered dinosaurs inhabited cold environments for at least a portion of the year. While the distribution of feathers on the body of dinosaurs is not clear from the fossil record, the potential use of feathers as insulation from the cold or as solar shields in the heat (Bartholomew, 1966; Bakken, 1981) remains a possibility.

The authors would like to thank Mr Joe Forys, the Curator of the Asian Domain, and the elephant staff at the Audubon Nature Institute in New Orleans, LA, USA, including: Alison Dietz, Gwynyn Miller, Patricia Parker and Jamie Orth. The authors would also like to thank Michael Angilletta, Diana Hews, Steve Lima, Michael O'Connor and an anonymous reviewer for their comments and criticisms on earlier versions of this manuscript.

 
• A

  total skin surface area (m²) calculated from the relationship A=0.1m_b^0.67

 
• A₁

  cross-sectional surface area of a spherical elephant (m²)

 
• A₂

  50% of the total skin surface area (m²)

 
• C

  convective heat loss (W)

 
• C_ex

  convective heat loss during exercise (W)

 
• c_p

  heat capacity of air (1003.5 J °C⁻¹ kg⁻¹)

 
• dt

  change in time (s)

 
• dT_b

  change in core body temperature (°C)

 
• E

  evaporative heat loss (W)

 
• E_b

  respiratory evaporative heat loss (W)

 
• E_r

  evaporative heat loss from skin (W)

 
• h_c

  convection coefficient (W m⁻² °C⁻¹)

 
• I

  whole-animal effective tissue insulation (°C W⁻¹)

 
• K

  conductive heat loss (W)

 
• k

  thermal conductivity of air (2.47 to 2.65×10⁻² W m⁻¹ °C⁻¹)

 
• M

  resting metabolic heat production (W)

 
• m_b

  body mass (kg)

 
• M_ex

  wet exercise metabolic heat production (W)

 
• M_ex–E_b

  dry exercise (active) metabolic heat production (W)

 
• Q_a

  environmental radiation absorbed by a spherical elephant (W)

 
• Q_n

  net radiant heat transfer (W)

 
• Q_n,night

  net radiant heat transfer at night (W)

 
• Q_n,sun

  net radiant heat transfer in full sun (W)

 
• r

  reflectance of the asphalt track surface (15%)

 
• R_a

  long-wave thermal radiation from the atmosphere (W m⁻²)

 
• R_g

  long-wave thermal radiation from the track surface (W m⁻²)

 
• RH

  relative humidity (%)

 
• R_s

  radiant heat loss from a spherical elephant (W)

 
• s

  diffuse short-wave solar radiation scattered in the atmosphere

 
• S_h

  direct short-wave solar radiation falling on a horizontal plane (W m⁻²)

 
• S_n

  direct short-wave solar radiation perpendicular to the body (W m⁻²)

 
• T_a

  ambient air temperature (°C)

 
• T_b

  core body (rectal) temperature (°C)

 
• T_E

  temperature of exhaled air (°C)

 
• T_g

  radiant ground temperature (°C)

 
• T_I

  temperature of inspired air (°C)

 
• T_r

  mean radiant skin temperature (°C)

 
• u

  sustained environmental wind speed (m s⁻¹)

 
• V

  respiratory minute volume (l s⁻¹)

 
• v_f

  walking speed (m s⁻¹)

 
• W_E

  water content of exhaled air (kg H₂O l⁻¹ air)

 
• W_I

  water content of inhaled air (kg H₂O l⁻¹ air)

 
• x

  distance from body core to skin surface (m)

 
• X

  heat storage in core tissues (W)

 
• α₁

  percent absorptance of elephant skin for short-wave solar radiation

 
• α₂

  percent absorptance of elephant skin for long-wave thermal radiation

 
• ε

  emissivity of elephant skin (0.96) and asphalt (0.93)

 
• λ

  latent heat of water vaporization (≈2.43×10⁶ J kg⁻¹ H₂O)

 
• ρ_E

  density of air (0.0011–0.0012 kg l⁻¹).

 
• σ

  Stefan–Boltzmann constant (5.67×10⁻⁸ W m⁻² K⁻⁴).