Oecologia

, Volume 171, Issue 1, pp 83–91

Detrimental influence on performance of high temperature incubation in a tropical reptile: is cooler better in the tropics?

Authors

    • School of Marine and Tropical BiologyJames Cook University
  • Simon Blomberg
    • School of Biological SciencesThe University of Queensland
  • Lin Schwarzkopf
    • School of Marine and Tropical BiologyJames Cook University
Physiological ecology - Original research

DOI: 10.1007/s00442-012-2409-6

Cite this article as:
Bell, K., Blomberg, S. & Schwarzkopf, L. Oecologia (2013) 171: 83. doi:10.1007/s00442-012-2409-6
  • 410 Views

Abstract

Global temperatures have risen over the last century, and are forecast to continue rising. Ectotherms may be particularly sensitive to changes in thermal regimes, and tropical ectotherms are more likely than temperate species to be influenced by changes in environmental temperature, because they may have evolved narrow thermal tolerances. Keelback snakes (Tropidonophis mairii) are tropical, oviparous reptiles. To quantify the effects of temperature on the morphology and physiology of hatchling keelbacks, clutches laid by wild-caught females were split and incubated at three temperatures, reflecting the average minimum, overall average and average maximum temperatures recorded at our study site. Upon hatching, the performance of neonates was examined at all three incubation temperatures in a randomized order over consecutive days. Hatchlings from the ‘hot’ treatment had slower burst swim speeds and swam fewer laps than hatchlings from the cooler incubation temperatures in all three test temperatures, indicating a low thermal optimum for incubation of this tropical species. There were no significant interactions between test temperature and incubation temperature across performance variables, suggesting phenotypic differences caused by incubation temperature did not acclimate this species to post-hatching conditions. Thus, keelback embryos appear evolutionarily adapted to development at cooler temperatures (relative to what is available in their habitat). The considerable reduction in hatchling viability and performance associated with a 3.5 °C increase in incubation temperature, suggests climate change may have significant population-level effects on this species. However, the offspring of three mothers exposed to the hottest incubation temperature were apparently resilient to high temperature, suggesting that this species may respond to selection imposed by thermal regime.

Keywords

Climate changeDevelopmental acclimationEctothermPerformancePhenotype

Introduction

Global temperatures have increased markedly since the eighteenth century, and are predicted to rise at an increasing rate over the next century (IPCC 2007). Temperature has profound effects on the biological processes of all animals (Birchard 2004). Ectotherms lack effective physiological mechanisms to control body temperature, and are therefore particularly sensitive to environmental conditions (Bogert 1949). Because they have been exposed to smaller variations in temperature over evolutionary time, tropical ectotherms are likely to be more strongly influenced by changes in environmental temperature than temperate species (Huey et al. 2009; Sunday et al. 2010).

Embryos require appropriate levels of temperature, moisture and gas exchange for successful development (Shine and Thompson 2006). However, early stage embryos lack the physiological and behavioural responses available to post-hatching animals, suggesting that ectothermic organisms may be most vulnerable to adverse thermal conditions at early stages of development (Birchard 2004; Shine and Thompson 2006). Thus, a warming climate has potentially significant implications for embryogenesis in ectotherms.

Many oviparous ectotherms lay their eggs underground (Booth 2006), presumably to provide protection from predation, and to avoid thermal and hydric extremes. The thermal buffering properties of soil attenuate heat flux, typically causing a complete absence of diel fluctuations in temperature from depths of between 30 and 50 cm (Ackerman and Lott 2004). Despite little variation in temperatures at even shallow soil depths, it is often implicitly assumed that there are a wide range of thermal regimes available, from which females can select suitable nest sites (see Wilson 1998; Kolbe and Janzen 2002; Brown and Shine 2004). Testing of this assumption is required if we are to understand the potential susceptibility of oviparous ectotherms to climate change.

The effects of climate change on ectotherms could potentially be mitigated by acclimation. Several studies have found that hatchlings prefer, or respond positively to, ambient temperatures corresponding to temperatures experienced during development (Blouin-Demers et al. 2000; Bronikowski 2000; Geister and Fischer 2007), suggesting that acclimation may pre-adapt individuals to post-hatching conditions (the Beneficial Acclimation Hypothesis; Leroi et al. 1994). The majority of existing studies that examine the plasticity of ectotherms to incubation regime test offspring at only one temperature (see review by Deeming 2004). Such an approach prevents the accurate assessment of differential temperature-performance relationships (temperature reaction norms), because it excludes the effects of phenotypic acclimation (Seebacher 2005; Deere and Chown 2006).

Plasticity in the ecological and life history traits of snakes make them good ‘model’ organisms in which to examine available nest temperatures, offspring responses to incubation temperature, and offspring acclimation to incubation temperature (Shine and Bonnet 2000). However, many past studies do not base incubation temperature treatments on ‘natural’ temperatures (Burger et al. 1987; Gutzke and Packard 1987; Lin et al. 2005; Ji et al. 2007; Du and Ji 2008). To determine the availability of suitable nest sites to oviparous reptiles in a wetland ecosystem in north-eastern Australia, we measured the magnitude of thermal heterogeneity at different soil depths, by measuring soil temperature profiles. We then used these temperature profiles, together with current projections for climate over the next century, as a basis to create appropriate incubation treatments. We examined the morphological and phenotypic responses of hatchling keelback snakes (Tropidonophis mairii) to incubation temperatures. Evidence of acclimation responses were examined by incubating eggs from each clutch at three temperatures, then testing the performance of all hatchlings at the three test temperatures.

Materials and methods

Study species

The freshwater keelback snake (Tropidonophis mairii, Gray 1841) is a small, natricine colubrid that occurs in wetland and woodland habitats throughout tropical Australia (Cogger 1986) (Fig. 1). They are oviparous, and natural nests occur in soil cracks and burrows at depths of between 10 and 20 cm (Webb et al. 2001; Brown and Shine 2004).
https://static-content.springer.com/image/art%3A10.1007%2Fs00442-012-2409-6/MediaObjects/442_2012_2409_Fig1_HTML.jpg
Fig. 1

The keelback snake (Tropidonophis mairii)

Data collection

Soil temperature

To determine the range of temperatures available to keelbacks in nest sites, 24 iButton© temperature data loggers were deployed on 12th February 2010 at the Townsville Town Common Conservation Park, northern Queensland (19°12′28.69″S, 146°44′25.17″E). Locations for data loggers were based on randomly generated vectors (using S-PLUS to generate values between −15 and 15, representing distances in metres for north/south and east/west bearings) from catch locations of gravid female keelbacks. At each location, we deployed iButtons© at depths that reflected the range in which natural nests have been observed (10–20 cm), as well as the temperature taken just under the soil’s surface (2 cm depth) for comparison. The iButtons© recorded temperatures hourly for a minimum of 2 weeks. The loggers were relocated on 3 March and 17 March 2010, thus providing temperature profiles for 24 locations adjacent to capture points, at a time of year when eggs were likely to be in the soil. The proximity of vegetation, water and shade were recorded for each location.

Incubation

Thirty-six gravid female keelbacks were captured during nightly surveys between 14 January and 25 February 2010 in the Townsville Town Common Conservation Park. The adults were maintained in ventilated 37 × 30 × 20 cm (L × W × H) plastic bins, provided with water ad libitum, a 650-ml hide containing a substrate of vermiculite and water (1:1 by mass), and a ‘heat mat’ for thermoregulation, and allowed to lay eggs. Containers were checked twice daily for eggs. Immediately following discovery, eggs were weighed and measured (maximum length and width in mm using callipers) before being placed in separate 300-ml plastic tubs containing 100 ml of vermiculite substrate moistened with 10 ml of water, and covered in plastic cling film to prevent excessive water loss. Eggs were then randomly assigned, in a split-clutch design, to one of three temperature treatments (24.9 °C ± 0.7, 26.6 °C ± 0.9, and 30.1 °C ± 0.5) until they hatched. Temperatures within each treatment fluctuated by the same magnitude as those recorded at the study site. The coolest temperature treatment (hereafter referred to as ‘cold’) reflected the average minimum daily temperature, at a 20-cm depth, of the coldest iButton location (24.8 °C ± 0.8). The intermediate temperature incubation treatment (‘mid’) reflected the average daily temperature, at a depth of 20 cm, across all iButton locations (26.2 °C ± 1.0). The ‘hot’ incubation treatment reflected the average maximum daily temperature, at 20-cm depth, of the warmest iButton location (29.7 °C ± 0.7). The hot treatment also reflected the predicted average temperature at a depth of 20 cm (or current ‘mid’ treatment) by 2090–2099, given current best estimates of climate change (IPCC 2007). Humidity in all three rooms was held constant at approximately 80 %. Eggs were checked daily and were re-arranged within each room every 2 days to avoid position effects. Following oviposition, females were released at their point of capture.

Upon hatching, neonates were weighed, measured [snout-vent length (SVL), tail length, head width and length] to the nearest millimetre using callipers, sexed by eversion of hemipenes and checked for the presence of ventral scale asymmetries, which can indicate developmental instability (Shine et al. 2005; Lowenborg et al. 2010). Once measured, neonates were placed in ventilated 650-ml containers, provided with water ad libitum and randomly assigned to one of the three temperature treatments. Following an acclimation period of 24 h, hatchlings were tested (see below) before being transferred to the next (randomly assigned) treatment room and allowed to acclimate for a further 24 h, so that each snake was tested over 3 consecutive days in each temperature treatment. Once testing was complete, the neonates were held in conditions similar to those described for gravid females above, until a first slough had been completed, and then they were released at the point of capture of the mother.

The effects of incubation treatment on hatchlings were assessed over a range of performance measures. Locomotor ability was assessed using video-recorded swim trials. Keelbacks live near water and often swim to escape predation and to forage for food (Webb et al. 2001). Snakes were placed in a circular swim chamber (outer diameter, 120 cm; inner diameter, 80 cm) filled with water to a depth of 4 cm. The water used in swim trials was left to stand in each treatment room for a minimum of 24 h prior to testing, and water temperatures were measured immediately prior to swim trials. The hatchlings were encouraged to swim by gently tapping their tails with a test tube brush. Each snake was allowed to swim for 5 min, and the number of times they were tapped, the number of laps completed and the time of the first escape attempt (whereby hatchlings attempted to crawl up the sides of the chamber) were recorded. The fastest time taken by each hatchling to cover a 1/8th section (47 cm) of the circuit was measured by examination of the videos, and used as a measure of burst speed.

Data analysis

Incubation

There were significant correlations between SVL, tail length, and head length/width of hatchlings, so a principal components analysis (PCA) was conducted on the covariance matrix to reduce the dimensionality of morphological traits. PC1 and PC2 were used as dependent variables in unbalanced, mixed-effects ANOVAs, with clutch-of-origin as a random, block effect and incubation treatment as a fixed effect. Body mass scaled differently from length and width measurements, and was therefore not included in the PCA. Variables that violated the assumptions of ANOVA were log-transformed prior to analysis.

We used linear mixed-effects models (LMM) and generalised linear mixed-effects models (GLMM) to analyse snake performance. Clutch of origin and snake ID (nested within clutch) were included as random factors. Incubation treatment and test temperature treatment (and their interaction) were included as fixed effects, and PC1 was included as a covariate. For the number of times tapped, a Poisson GLMM with a square-root link function was used. LMMs were used for burst speed and number of laps. Plots of the (weighted) residuals versus the fitted values were examined to check for heteroskedasticity, and normal quantile–quantile plots were examined to check for departures from normality of the residuals. Analyses were performed in R (R Development Core Team 2012), using the lme4 package (Bates et al. 2011). p values for the fixed effects in LMMs were calculated using Markov Chain Monte Carlo (MCMC) methods, using the languageR package for R (Baayen 2011). Significance of the fixed effects in GLMM was tested using likelihood ratio Chi-square tests. Post hoc comparisons of fixed effects were conducted using the package multcomp for R (Hothorn et al. 2008). Significance of random effects were assessed using likelihood ratio Chi-square tests.

Results

Soil temperature

The average temperature recorded across all 24 sites and depths between February and May 2010 was 26.2 °C (±2.3 SE). There was a significant difference in temperature (t test, t(12989) = 14.36, p < 0.001) and variance (F test, F(10318, 10788) = 7.29, p < 0.001) among sites exposed to the sun and those in the shade. Shaded sites (26.1 °C ± 1.1) were cooler and exhibited less variation in temperature than sites exposed to the sun (26.5 °C ± 3.1). Temperature and variation differed significantly among 2, 10 and 20 cm depths (means: Kruskall–Wallis, \( \chi_{(2)}^{2} = 158.7 \), p < 0.001; Variances: Levene’s Homogeneity of Variance test(2,35351) = 2,112.3, p < 0.001). Temperatures 2 cm underground (26.5 °C ± 3.5) were warmer and varied more than at 10 cm (25.9 °C ± 1.3) and 20 cm (26.2 °C ± 1.0) (Fig. 2).
https://static-content.springer.com/image/art%3A10.1007%2Fs00442-012-2409-6/MediaObjects/442_2012_2409_Fig2_HTML.gif
Fig. 2

Average daily soil temperatures at the study site at different soil depths

Incubation

Egg weight, width and length were not significantly different among incubation treatments [egg weight, mean = 3.0 g (cold, mid, hot), F(2,390) = 0.02, p = 0.98; egg width, mean = 25.7 mm (cold), 25.8 mm (mid), 26.1 mm (hot), F(2,390) = 0.37, p = 0.70; egg length, mean = 14.0 mm (cold), 14.1 mm (mid), 14.0 mm (hot), F(2,390) = 0.71, p = 0.49], so the random allocation of eggs among treatments was effective. Across all treatments, the number of female hatchlings (103) did not differ significantly from males (112) (Binomial test, n = 215, p = 0.59). Also, there were no significant differences in sex ratios among incubation treatments (Pearson, \( \chi_{(2)}^{2} = 2.80 \), p = 0.25). After hatching, hot-incubated snakes took significantly longer (11.4 days ± 0.9) to slough than snakes from mid (7.2 days ± 0.3) or cold (7.5 days ± 0.3) incubation treatments (Kruskall–Wallis, \( \chi_{(2)}^{2} = 6.65 \), p = 0.036; Table 1).
Table 1

Effects of incubation treatment on hatching success, morphology, behaviour and performance in hatchling keelback snakes (Tropidonophis mairii)

Variable

Incubation temperature, mean (SD)

Cold (n = 106)

Mid (n = 94)

Hot (n = 27)

Sex ratio (M:F)

44:57

46:44

9:15

Time to first shed (days)

7.5 (0.3)

7.2 (0.3)

11.4 (0.9)*

Hatchability

 Incubation time (days)

60.7 (2.2)*

51.0 (2.3)*

41.4 (1.4)*

 Hatching success (%)

79.2

72.9

21.4*

 Residual egg weight (g)

0.69 (0.23)

0.61 (0.28)

0.75 (0.39)

Morphology

 Mass (g)

2.7 (0.4)

2.7 (0.4)

2.5 (0.3)*

 SVL (cm)

15.6 (1.0)

15.3 (0.8)

14.7 (0.9)*

 Tail length (cm)

4.4 (0.5)

4.4 (0.5)

4.0 (0.4)*

 Head width (mm)

4.6 (0.1)*

4.6 (0.2)

4.5 (0.2)

 Head length (mm)

9.7 (0.3)

9.6 (0.3)

9.4 (0.4)

 Ventral scale asymmetries (%)

6.9

6.4

85.7*

 Spinal deformities (%)

3.0

3.2

71.4*

Performance

 Burst speed (cm/s)

22.0 (7.1)

19.9 (6.4)

8.0 (6.1)*

 Number laps

5.4 (0.1)

5.45 (0.1)

2.3 (0.3)*

 Number taps

12.0 (0.6)

13.7 (0.6)

6.9 (1.2)*

* Statistically different subgroup

Hatching success

Incubation time was inversely related to incubation temperature, with neonates from the hottest incubation temperature hatching earliest (41.4 ± 1.4 days), followed by mid (51.0 ± 2.3 days) then cold (60.7 ± 2.2 days) (ANOVA, F(2,189) = 3,253.6 p < 0.001). Hatching success decreased significantly with increasing incubation temperature (\( \chi_{(2)}^{2} = 107.8 \), p < 0.001), with a hatching success rate of 79.2 and 72.9 % in cold and mid incubation temperatures respectively, dropping to a success rate of 21.4 % in the hottest temperature treatment. The probability of a neonate hatching successfully from the hottest incubation temperature was not distributed randomly among mothers, with significant clustering in hatching success amongst clutches (three mothers had an average hatching success of 85 %, compared to a hatching success of 14 % on average for all other mothers, Poisson, \( \chi_{(4)}^{2} = 22.5 \), p < 0.001). Clutches in the hottest incubation temperature were more likely to have many successful hatchings or none at all, and significantly less likely to produce intermediate numbers of viable offspring than would be expected if hatching success was random (Fig. 3).
https://static-content.springer.com/image/art%3A10.1007%2Fs00442-012-2409-6/MediaObjects/442_2012_2409_Fig3_HTML.gif
Fig. 3

Frequencies of observed and expected hatchling numbers from clutches incubated in the warmest temperature treatment. Note that many more clutches had no hatchlings than expected in the hot treatment

Morphology

The number and proportion of ventral scale asymmetries differed significantly among incubation temperatures (\( \chi_{(2)}^{2} = 110.6 \), p < 0.001), with the hottest treatment showing the highest prevalence (85.7 %, 24/28), followed by cold (6.9 %, 7/101), then mid (6.4 %, 6/94). Similarly, the occurrence of other developmental abnormalities (typically post-cloacal kinking of the spine) was significantly different among incubation treatments (\( \chi_{(2)}^{2} = 111.1 \), p < 0.001). 71.4 % (20/28) of hot incubated snakes displayed some visible form of developmental abnormality, compared to just 3.2 % (3/94) and 3.0 % (3/101) of mid- and cold incubated snakes respectively.

PC1 (56.2 %) and PC2 (19.6 %) explained 75.8 % of the total variance among morphological variables (Fig. 4). The morphological variables with the highest loadings on PC1 were tail length (0.81) and SVL (0.79). Head width (0.79) had a considerably larger loading on PC2 than all other variables, thus PC1 and PC2 appear to reflect body ‘size’ and body ‘shape’ respectively. Body size (PC1) differed significantly among incubation treatments, after correcting for maternal effects (ANOVA, F(2,189) = 34.9, p < 0.001) (Fig. 3). Hatchlings from the hot incubation treatment were significantly smaller than both cold and mid incubation treatments, though cold and mid incubation temperatures did not differ from one another. Body shape (PC2), particularly head width, was also significantly different among the three treatments (ANOVA, F(2,189) = 10.9, p < 0.001). Cold and hot incubated hatchlings had a slimmer build than hatchlings from the mid incubation temperature.
https://static-content.springer.com/image/art%3A10.1007%2Fs00442-012-2409-6/MediaObjects/442_2012_2409_Fig4_HTML.gif
Fig. 4

Principal components analysis plotting body ‘size’ (PC1) and body ‘shape’ (PC2) in hatchling keelback snakes from hot (circles), mid (squares) and cold (crosses) incubation treatments. Dark bold markers indicate centroids ± standard errors

Hatchlings from the hot incubation treatment (2.46 g ± 0.30) weighed significantly less than hatchlings from the mid (2.68 g ± 0.44) and cold treatments (2.75 g ± 0.44) (ANOVA, F(2,188) = 9.0, p < 0.001).

Performance

Burst speed was significantly influenced by both incubation temperature (LMM, F(2,x) = 82.75, MCMC p < 0.0001 and the temperature at which trials were conducted (LMM, F(2,x) = 13.01, MCMC p = 0.0004) (Fig. 5a). The interaction between incubation temperature and trial temperature was not significant (LMM, F(4,x) = 0.58, MCMC p = 0.7742), after correcting for body size, maternal effects and ID effects.
https://static-content.springer.com/image/art%3A10.1007%2Fs00442-012-2409-6/MediaObjects/442_2012_2409_Fig5_HTML.gif
Fig. 5

Incubation treatment and test temperature effects on the performance of hatchling keelback snakes. a, b Measures of locomotor speed, c willingness to swim (note: more taps indicate snakes were less willing). The x axis represents the temperature at which snakes were tested. The y axis represents the means of the dependent variables. Hatchlings incubated at the cold temperature treatment are indicated by the dotted black line, mid temperature by the grey line and hot by the solid black line. Vertical bars standard errors

Body size was significantly positively related to burst speed (slope = 1.756, t(x) = 2.852, MCMC p = 0.0011). The proportion of variance associated with maternal effects was only 7 %, and this was not significantly different from zero (LMM, LR \( \chi_{(1)}^{2} = 2.6915 \), p = 0.1009). The proportion of variance explained by individual snake ID was higher and significant at 28 % (LMM, LR \( \chi_{(1)}^{2} = 15.676 \), p < 0.0001). After averaging over body size and the incubation × test interaction, the hot incubation treatment had significantly lower burst speeds than either the cold treatment (Tukey HSD test, z = 6.469, p < 0.0001) or the mid treatment (Tukey HSD, z = 5.187, p < 0.0001). The Cold treatment was not significantly different from the mid treatment (Tukey test, z = −1.234, p = 0.433).

An additional measure of swim speed, the number of laps completed in 5 min, also showed significant differences with incubation temperature (F(2,x) = 113.94, MCMC p < 0.0001) and performance test temperature (F(2,x) = 8.211, MCMC p < 0.0001), with no significant interaction between the two factors (F(4,x) = 1.952, MCMC p = 0.16159). See Fig. 5b. Again, there was a significant effect of body size, with larger individuals swimming further (slope = 0.3180, tx = 3.543, MCMC p < 0.0001). Maternal and individual components of variance were 7.6 % (LLMM, LR \( \chi_{(1)}^{2} = 8.524 \), p = 0.0035), and 16.6 % (LLMM, LR \( \chi_{(1)}^{2} = 17.983 \), p < 0.0001), respectively. Post hoc comparisons of means (averaged over the fixed factor interaction and the body size covariate) revealed that the hot incubated animals swam fewer laps than either the cold (Tukey HSD, z = −11.941, p < 0.0001) or mid temperature (Tukey HSD, z = 12.301, p < 0.0001) incubated animals, but that the cold and mid temperature incubation treatments were not significantly different (Tukey HSD, z = 0.452, p = 0.89). Post hoc analysis of the performance temperature treatments showed that there was a significant difference between hot and cold running temperatures (Tukey HSD, z = 2.919, p = 0.0098), but not between hot and mid (Tukey HSD, z = −1.967, p = 0.121) or between cold and mid (Tukey HSD, z = 0.951, p = 0.608), indicating that the effect of performance temperature on the number of laps swum in 5 min is less than the effect of incubation temperature on this response variable.

Willingness to swim, as measured by the number of taps required during testing (Brown and Shine 2002), showed a significant interaction between incubation temperature and performance temperature (LR \( \chi_{(4)}^{2} = 11.586 \), p = 0.0207; see Fig. 5c). Post hoc tests for the incubation temperature effect (averaged over the interaction term and body size covariate) revealed a significant difference between both hot and cold incubation treatments (Tukey HSD, z = −4.017 p < 0.001) and hot and mid incubation treatments (Tukey HSD, z = −5.120, p < 0.001) but not mid and cold incubation treatments (Tukey HSD, z = 1.757, p = 0.177). Thus, hatchlings from the hot incubation treatment were more willing to swim than hatchlings from both mid and cold incubation treatments. An equivalent analysis of the performance treatments revealed no significant difference between the mid and cold treatments (Tukey HSD, z = 0.415, p = 0.91) but differences between the hot and cold treatments (Tukey HSD, z = 6.614, p < 0.0001) and between hot and mid treatments (Tukey HSD, z = 6.198, p < 0.0001). Hence, the hot performance temperature required less taps than either the mid or cold performance temperature treatments.

Discussion

Clearly, keelback snakes show high levels of phenotypic plasticity in response to thermal regime; incubation temperature significantly affected 12 of the 15 traits examined. The temperature at which hatchlings were incubated had a stronger influence on performance than did test temperature across all performance measures, including locomotor ability and willingness to swim. Performance was “best”, i.e., offspring had fewer abnormalities, were larger, and could swim faster and for longer, if they had experienced cold and mid temperatures as embryos, suggesting that development is evolutionarily adapted to occur at these temperatures. Apparent adaptation to cooler temperatures is consistent with a surprisingly low optimal incubation temperature in this tropical reptile (Brown and Shine 2006). The only characteristic that was not “best”, or of apparently higher fitness, in offspring incubated in the two cooler treatments was willingness to swim. Smaller, slower offspring from warmer temperatures swam more readily when coaxed compared to offspring from other treatments. This may be due to covariation among physiological traits, such that high speed is a trade-off against higher endurance (or willingness to swim) (Vanhooydonck et al. 2001). However, many studies dispute the existence of significant trade-offs between speed and endurance (Huey et al. 1990; Secor et al. 1992, 1995; Brodie and Garland 1993; Goodman et al. 2007). Alternatively, perhaps increased willingness to swim is a behavioural adaptation that mitigates smaller size and slower swimming speeds. For all other characters, both behavioural and morphological, cooler temperatures appeared to produce ‘fitter’ offspring.

In our study, individuals exposed to cooler incubation temperatures (similar to cooler ranges of temperatures available in the wild) showed higher performance in all incubation temperatures. The lack of interaction between incubation temperatures and test temperatures indicates that developmental acclimation did not pre-adapt individuals to post-hatching environments. Cooler incubation temperatures are often better in terms of hatchling fitness in ectotherms (Partridge et al. 1995; Atkinson and Sibly 1997; Shine 1999; Stillwell and Fox 2005; Brown and Shine 2006). However, our study does not really support the ‘cooler is better’ hypothesis of thermal acclimation, because we tested only the range of temperatures actually available to female snakes in the wild. Thus, we did not demonstrate that colder temperatures in general produce fitter hatchlings, rather we showed that, within the range of temperatures available to females, cooler temperatures were better for successful embryonic development. These results are, therefore, also consistent with the local adaptation hypothesis of thermal acclimation (Caley and Schwarzkopf 2004), because females apparently select cool nest sites, at least in other regions of Australia (Brown and Shine 2006).

Preference for relatively cool nest sites may increase the sensitivity of T. mairii to global increases in environmental temperatures. The 3.4 °C difference in average soil temperatures between the hottest and coldest sites (at a depth of 20 cm) suggests that the availability of suitable nest sites may be limited, even in relatively complex environments. Global average surface temperatures are forecast to rise by up to 4 °C by 2100 (IPCC 2007). Should surface temperatures increase as predicted, then not one of the 24 sites at which soil temperature profiles were recorded would provide optimal conditions for development of T. mairii. However, despite the apparently small variation in soil temperature, gravid keelbacks may still be able to locate cool nest sites, and therefore they may be able to compensate behaviourally for the influences of higher environmental temperatures. Reptiles may adjust behaviour to select thermally appropriate sites that enhance their fitness, and expand or alter their geographic range (Huey et al. 2003). For example, grass snakes (Natrix natrix) have extended their range in Europe into high latitudes by taking behavioural advantage of heat sources such as manure piles (Lowenburg 2010). Alternatively, compensation may occur through temporal shifts in the breeding season, such as laying eggs earlier in the spring. However, the importance of moisture to the viability of eggs, and the correlation of moisture levels with season, may restrict the extent to which keelbacks could alter the timing of breeding (Brown and Shine 2004). Given the detrimental impacts on hatchling viability of relatively minor differences in incubation temperature, further investigation into the presence of cooler microclimates may be critical in predicting how this species responds to climate change.

Another way that snake populations may respond to global temperature increases is evolutionarily (Bradshaw and Holzapfel 2006; Kearney et al. 2009). Although all the clutches were highly sensitive to incubation temperature, inconsistent hatching success among dams in the hot incubation treatment suggests there may be genetic differences in the thermal tolerance of individuals within populations. The presence of apparently ‘heat tolerant’ dams suggests this population may be resilient to some of the impacts of climate change. Identifying the genetic basis for these differences may provide insight into the extent of resilience in other populations of keelbacks, and potentially, other species of snake.

While we have described the likely short-term consequences of increased incubation temperature, little is known about how specific phenotypic changes lead to differences in post-hatching survival and, ultimately, fitness. Survival can be negatively, positively or uncorrelated to incubation temperatures within turtles, lizards and crocodiles (Deeming 2004). Limited evidence from mark–recapture studies on keelbacks suggests that large sizes at hatching enhance survival (Brown and Shine 2004). Despite a lack of information on long-term compensation effects, and the relationship between specific traits and survivorship, a relatively small increase in incubation temperature can considerably reduce viability and performance of hatchling keelbacks. Hot-incubated hatchlings had a lower hatching success rate, were smaller and had more developmental abnormalities than hatchlings from the mid and cold incubation treatments. Climate change is predicted to cause local reptile extinctions of 39 % worldwide by 2080 (Sinervo et al. 2010). Further studies examining the links between phenotype and fitness, the availability of suitable nest sites and the mechanisms behind the high variability in phenotypic response to incubation temperature, are critical to understanding how oviparous reptiles will respond to a changing climate.

Acknowledgments

We thank Teresea Lambrick, Gus McNab, Ross Gottlieb, Joel Voiselle, John Llewelyn and Matthew Crowther for their assistance in the field and laboratory. The work was carried out under a permit from the Environmental Protection Agency and Queensland Parks and Wildlife Service (WITK06346309), and James Cook University Animal Ethics Committee (A1469). This research was supported financially by a grant from the National Climate Change Adaptation Research Facility.

Copyright information

© Springer-Verlag 2012