Nest depth may not compensate for sex ratio skews caused by climate change in turtles
J. M. Refsnider,
Department of Ecology, Evolution and Organismal Biology, Iowa State University, Ames, IA, USA
Jeanine M. Refsnider, Department of Environmental Science, Policy, and Management, University of California, Berkeley, 130 Mulford Hall, Berkeley, CA 94720-3114, USA. Tel: +1 510 643 7430; Fax: +1 510 643 5438
Maternal ability to match nest characteristics with environmental conditions can influence offspring survival and quality, and may provide a mechanism by which animals can keep pace with climate change. In species with temperature-dependent sex determination that construct subterranean nests, the depth of the nest may affect incubation temperatures, and thus offspring sex ratio. Maternal adjustment of nest depth may be a mechanism by which climate change-induced sex ratio skews could be prevented in globally imperiled taxa such as turtles. We experimentally manipulated nest depth within a biologically relevant range in nests of the model turtle species Chrysemys picta. We then quantified the effects of nest depth on incubation regime, offspring sex ratio and offspring performance. We found no effect of nest depth on six parameters of incubation regime, nor on resultant offspring survival, size or sex ratio. However, deeper nests produced hatchlings that weighed less, and were faster at righting themselves and swimming, than hatchlings from shallower nests. We suggest that cues used by females in adjusting nest depth are unreliable as predictors of future incubation conditions, and the adjustment in nest depth required to affect sex ratio in this species may be too great to keep pace with climate change. Therefore, maternal adjustment of nest depth seems unlikely to compensate for climate change-induced sex ratio skews in small-bodied, freshwater turtles.
Human activities are contributing significantly to global climate change, one result of which is a predicted increase in global temperatures of 1.1–6.4°C by 2100 (Solomon et al., 2007). A temperature increase of this magnitude is likely to have dramatic effects on species and ecosystems, but many of these outcomes are difficult to predict because they involve indirect effects of environmental changes on a wide variety of taxa and occur via complex pathways (Barnosky et al., 2012). Some thermally sensitive traits that are directly impacted by climate can have demographic consequences for populations. One example is temperature-dependent sex determination (TSD), in which offspring sex is irreversibly determined by the temperature experienced by developing embryos (Bull, 1980). TSD occurs in many reptile groups as well as some fishes and invertebrates (Bergerard, 1972; Valenzuela & Lance, 2004). In reptiles with TSD, the temperature range within which the complement of offspring sex within a clutch shifts from all of one sex to all of the other sex is generally narrow, and is often less than 1°C (Ewert, Jackson & Nelson, 1994). Consequently, population sex ratios are extraordinarily sensitive to temperature changes because a small shift in environmental temperature could dramatically alter offspring sex ratio (e.g. Schwanz et al., 2010). Human-induced climate change could severely impact reptiles with TSD by resulting in populations comprised of predominantly one sex (Janzen, 1994a; Mitchell et al., 2008), and therefore recognizing the potential effects of climate change on species with TSD is important if conservation efforts are to be effective.
Maternal nest-site choice and nest construction influence offspring survival and quality, as well as sex in species with TSD (reviewed in Refsnider & Janzen, 2010). For example, in birds, the amount of vegetative cover around a nest can reduce its visibility to predators (e.g. Martin & Roper, 1988; Stokes & Boersma, 1998), and the direction of a nest's opening may protect offspring from thermal stress (e.g. Walsberg & King, 1978; Hartman & Oring, 2003); in reptiles, the temperature within a nest can affect offspring performance (e.g. Miller, Packard & Packard, 1987; Van Damme et al., 1992; Shine et al., 1997). Maternal ability to match nest location and construction to environmental conditions is an important determinant of a female's reproductive success, and may be particularly critical in allowing females to shift nest characteristics to match changing environmental conditions and thereby continue to successfully reproduce. Matching nest characteristics with prevailing environmental conditions could occur via two, non-mutually exclusive mechanisms: microevolutionary responses of populations over several generations, or plastic responses among or even within individuals (Bulmer & Bull, 1982). The importance of both mechanisms in population responses to inter-annual climatic variation has been demonstrated in species with TSD (Morjan, 2003a; McGaugh et al., 2010; Refsnider & Janzen, 2012), but the relative importance of each mechanism is dependent upon the component of nest-site choice being examined. In reptiles with TSD, components of nest-site choice that could theoretically change in response to environmental conditions include date of nesting, shade cover over the nest site, nest microhabitat such as soil moisture and depth of the nest cavity (e.g. Morjan, 2003a; Doody et al., 2006a; Schwanz & Janzen, 2008). For example, in painted turtles, shade cover over the nest predicts sex ratio (Janzen, 1994b), and behavioral plasticity in maternal selection of shade cover over the nest appears to allow female turtles to match nest microhabitat to prevailing environmental conditions (Refsnider & Janzen, 2012). However, heritability in choice of shade cover appears insufficient to compensate for inter-annual climatic variations (McGaugh et al., 2010). Similarly, although nesting date is a plastic response based on the preceding winter's climate, it is not repeatable within individuals and therefore is unlikely to respond to selection (Schwanz & Janzen, 2008). Therefore, in both the shade cover and phenological components of nest-site choice in painted turtles, behavioral plasticity was a more important mechanism than microevolutionary changes in population responses to inter-annual climate variation.
Nest depth is another component of nest-site choice that could vary in response to prevailing environmental conditions, and thereby match incubation regime with climate. Modeling studies suggest that nest depth affects both incubation temperature and the magnitude of temperature variation (Georges, Limpus & Stoutjesdijk, 1994), and empirical data show that incubation temperature differs with nest depth in turtles (Roosenburg, 1996), lizards (Shine & Harlow, 1996) and crocodilians (Leslie & Spotila, 2001). Indeed, altering nest depth as global temperatures increase may be critical for some species; for example, tuatara (the sole representative of an ancient reptilian order) are predicted to produce all male offspring at current nest depths under maximum climate-warming scenarios (Mitchell et al., 2008). The mechanism (i.e. microevolutionary response or behavioral plasticity) behind such adjustments in nest depth have not been studied. For example, nests constructed at depths that were previously representative of the population average could, under severe climate-warming scenarios, experience lethally high temperatures, which would select for deeper, cooler nests and result in a microevolutionary change in mean population nest depth. Alternatively, if nest depth is behaviorally plastic, individual females might base nesting behavior on ambient temperatures during the period from spring emergence and onset of nesting, and construct deeper nests in warm years and more superficial nests in cool years.
Adjusting nest depth may not be a component of nest-site choice that will compensate for climate change in all cases, and therefore, it is important to understand the circumstances under which nest depth adjustment may or may not be possible. For example, in some reptiles, nest depth is not related to nest temperature (eastern fence lizard; Warner & Andrews, 2002), does not contribute to variation in sex ratio (freshwater turtles; Vogt & Bull, 1982) or cannot be altered because of low availability of nest sites with suitable microhabitats, such as appropriate soil type (tuatara; Mitchell et al., 2008). In other reptiles, despite climatic differences among geographically widespread populations, nest depths do not differ among populations (Australian water dragon; Doody et al., 2006a) or, when they do differ, the differences may be a function of female body size (painted turtle; Morjan, 2003b) rather than adaptations to local climate. Finally, even when nest depth changes in response to warming temperatures, such change may be insufficient to compensate for the magnitude of climatic warming (three-lined skink; Telemeco, Elphick & Shine, 2009). The preceding studies demonstrate that testing whether adjustment of nest depth could compensate for climate change-induced sex ratio skews is an important part of understanding how reptiles, which tend to be long lived, may respond to rapid environmental change. One additional note is that incubation conditions within a nest site affect numerous offspring traits in species with TSD other than sex, including size (Brown & Shine, 2004), growth rate (Brooks et al., 1991), metabolism (Van Damme et al., 1992), speed (Miller, 1993) and predator avoidance behavior (Burger, 1989). Because shifts in maternal nesting behavior in response to climate change may have simultaneous and perhaps unexpected effects on offspring size and performance (Refsnider, 2013), it is important to incorporate measures of offspring quality as well as sex ratio in studies on species that are particularly temperature sensitive.
We conducted a manipulative experiment to test the hypothesis that incubation temperature in naturally constructed turtle nests decreases with depth, which would indicate that adjustment of nest depth might compensate for inter-annual variation in climate. Our study is a first step toward evaluating the capacity of nest depth adjustment to compensate for inter-annual climatic variation. Although this study does not directly address the mechanism (i.e. microevolutionary response or behavioral plasticity) behind potential changes in nest depth in response to inter-annual climatic variation, we discuss the support our results provide for potential underlying mechanisms. Instead, we were interested in quantifying the effects of nest depth adjustment on offspring sex and performance, regardless of how such adjustment might occur, to determine whether this component of nest-site choice has any potential to compensate for inter-annual climatic variation, and by extension, for climate change. Because our study design required the euthanasia of a portion of hatchlings produced at the study site, we used a common species as a model to avoid negatively impacting any populations of rarer species. Nevertheless, our results should be of use to managers involved in conservation efforts for similar, imperiled turtle species. This study had two primary objectives. First, we used a long-term dataset to determine whether mean nest depth in a wild turtle population was correlated with annual climate. Second, we conducted a manipulative experiment to determine the effect of nest depth on thermal characteristics of the incubation environment within the nest cavity (hereafter incubation regime), offspring sex ratio and offspring performance.
We conducted this experiment using the western painted turtle, Chrysemys picta bellii, a small-bodied freshwater turtle in the family Emydidae. The painted turtle is widely dispersed across the USA and southern Canada, and the western subspecies occurs primarily west of the Mississippi River. Painted turtles inhabit a variety of aquatic habitats including rivers, lakes and ponds. In early summer, females emerge from wetlands to nest in open areas such as beaches and lawns. Incubation lasts ∼55–85 days depending on temperature (Ernst, 1971; F. Janzen, unpublished data). After hatching, neonates remain in the nest cavity through their first winter and emerge the spring following nest construction, at which time they travel terrestrially until reaching a wetland habitat (e.g. Paukstis, Shuman & Janzen, 1989).
Materials and methods
Long-term climate and nest depth trends
Our study site was a nesting beach on the northeastern side of a 1.5-ha island in the Mississippi River in Carroll County, Illinois, USA, at the Thomson Causeway Recreation Area. Data collection for the long-term study of nesting ecology is described in Schwanz et al. (2010). Briefly, we patrolled a nesting area in the South Potter's Marsh Campground hourly between 06:00 and 21:00 h from mid-May through early July. We observed nesting turtles from a distance until they completed the nesting process, at which time we briefly captured females for individual identification before releasing them. As in Schwanz et al. (2010), we excavated nests within 24 hours of construction to assess clutch size, egg mass and nest depth; we then re-covered nests with soil and left them to incubate in situ.
Nest depths were measured, using a straight-edge ruler, as the vertical distance from the soil surface to the base of the nesting cavity. We recorded depths of 2371 nests from 2000 to 2010, and we used these data to determine the mean population nest depth in each year. We focused on the grand mean of population nest depth from 2004 to 2009 (n = 1126) to determine our experimental treatments (see below) because turtles nesting in 2004–2010 are likely from the same age cohort. We used mean air temperature in May, acquired from the National Climate Data Center (http://www.ncdc.noaa.gov) for nearby Clinton, IA, as an indicator of climatic conditions in each year from 2000 to 2010.
Nest depth manipulation experiment
We conducted a nest depth manipulation experiment on a subset of painted turtle nests constructed in the North Potter's Marsh Campground in June 2010. Nests were located and processed as described above. However, in this experiment, we randomly assigned 44 nests to one of three nest depth treatments: shallow (n = 14), mean (n = 15) or deep (n = 15). The mean treatment was 8.7 cm, which was the grand mean nest depth for 1126 nests constructed during 2004–2009. The shallow and deep treatments were 6.7 and 10.7 cm, respectively, are equivalent to two standard deviations from the grand mean, and were selected to represent biologically relevant values that, while relatively extreme compared to most nests, were not outside the range of physically possible nest depths in this population. To achieve the assigned nest depth treatment, we either added soil to the bottom of the nest cavity (to decrease nest depth) or excavated additional soil (to increase nest depth) before replacing the eggs. In cases where the entire clutch would no longer fit in a nest because of artificial decreasing of the nest depth, we increased the size of the egg chamber horizontally (rather than vertically) until all eggs would fit. In addition to manipulating nest depths, we inserted a temperature logger (iButton, Embedded Data Systems, Lawrenceburg, KY, USA) among the eggs in the center of each nest. Loggers recorded nest temperatures hourly throughout incubation. Also, because canopy cover over nests affects incubation temperature (Morjan & Janzen, 2003), we took a hemispherical photograph over each nest and quantified canopy cover using Gap Light Analysis software (Frazer, Canham & Lertzman, 1999 as in Doody et al., 2006b). Finally, to minimize nest loss because of predation (which can range up to 95% of nests; Schwanz et al., 2010), we covered all nests with a 10-cm2 piece of wire mesh staked at each corner.
We retrieved temperature loggers and all surviving hatchlings in September 2010 (after hatching but before nest emergence). For each nest, we considered the incubation period to start on day 0 (the day of oviposition) and continue through day 70. For reptiles with TSD, the thermosensitive period is generally the middle third of embryonic development (Wibbels, Bull & Crews, 1994). As we did not directly observe when hatching occurred in any nest, we considered days 16–45 to encompass the thermosensitive period during which sex differentiation occurs. We then calculated six parameters related to incubation conditions for each nest (hereafter incubation regime): minimum and maximum incubation temperatures (the lowest and highest temperature recorded during the 70-day incubation period), mean temperature throughout the incubation period (days 0–70) and the thermosensitive period (days 16–45) and the mean daily range (for each 24-hour period, highest recorded temperature – lowest recorded temperature) for the incubation period and the thermosensitive period.
Offspring quality and sex ratio
After retrieving hatchlings, we calculated the survival rate of each nest as the number of live hatchlings retrieved, divided by the known clutch size. We carefully cleaned and dried hatchlings, and recorded their plastrons with a color scanner to facilitate individual identification. We then weighed and measured (straight carapace length) all hatchlings and housed clutchmates together in plastic deli cups containing moist soil over the winter in an incubator (Revco, Thermo Scientific, Asheville, NC, USA) at 4°C, conditions comparable to overwintering temperatures observed in wild nests from the Illinois population (Weisrock & Janzen, 1999). Starting in mid-March 2010, we gradually increased incubator temperature to 19°C over a 2-week period and kept hatchlings at 19°C thereafter.
We conducted three performance tests on each hatchling: (1) righting time (the time it took for a hatchling placed on its back to right itself); (2) sprinting time (the time it took a hatchling to walk 0.5 m); and (3) swimming time (the time it took for a hatchling to swim 1.0 m). The performance tests used here simulated a hatchling's journey from the nest site to wetland habitat, and then to suitable habitat within the wetland, following nest emergence in the spring. Therefore, we tested each hatchling in the order of righting time, immediately followed by sprinting time, and swimming time immediately following sprinting time. All hatchlings underwent two trials of three performance tests each, with the two trials separated by c. 2 weeks. In all three performance tests, we timed a hatchling's latency to begin moving (latency), the total time taken to complete the test from initial placement of the hatchling until the test was completed (total) and the time during which the hatchling was actively moving during completion of the test (active, or total – latency). We recorded all times to the nearest second using a digital stopwatch, and censored all tests at 180 s. During the performance tests, researchers were blind to the depth treatment of the nest from which hatchlings were produced.
Upon completion of the performance trials, we euthanized a subset of the hatchlings by a pericardial overdose of 0.5 mL of 1:1 sodium pentobarbital : water. To avoid negatively impacting the population, we euthanized and sexed up to six hatchlings per nest. Because most (66%) nests at our study site are unisexual (Janzen, 1994b), the sex ratio of a nest can be estimated by determining sex in a portion of the hatchlings from that nest (Schwanz et al., 2010), which reduces the number of individuals that must be euthanized. We assigned sex based on macroscopic examination of the gonads (Schwarzkopf & Brooks, 1985), classifying individuals lacking oviducts and possessing short gonads as males and those with complete oviducts and long gonads as females. After sexing, we preserved all specimens in 70% ethanol; the remaining hatchlings that were not sexed were released at the collection site in May 2011.
We conducted all statistical analyses in SAS 9.2 (SAS Institute, Cary, NC, USA). We first analyzed relationships between nest depth, female body size, year and annual climate using general linear regression. We then compared each of the six incubation regime parameters among the three nest depth treatments using one-way analysis of covariance (ANCOVA) in the General Linear Model (GLM) Procedure with canopy cover as a covariate. We tested for differences among nest depth treatments in mean hatchling mass and carapace length using one-way ANCOVA in the MIXED procedure with canopy cover and mean initial egg mass as covariates, and we tested for differences among treatments in per cent hatching success using one-way ANCOVA in the GLM Procedure with canopy cover as a covariate. Differences in nest sex ratios among nest depth treatments were compared using a chi-square goodness-of-fit test (Wilson & Hardy, 2002). Because shade cover predicts sex ratio in natural nests at the study site (Janzen, 1994b), we also assessed whether shade cover continued to predict sex ratio of experimentally manipulated nests. This was done using logistic regression in the GENMOD procedure, with depth treatment and shade cover as independent predictors. Finally, we analyzed differences among treatments in median hatchling performance (i.e. latency, active and total median times for righting, sprinting and swimming) using Kruskal–Wallis tests in the NPAR1WAY procedure, and corrected the P-values for multiple tests (m = 9) using the Bonferroni adjustment.
We analyzed the total depth of 2371 unmanipulated painted turtle nests from 2000 to 2010 (Fig. 1). Mean May temperature did not significantly increase over time between 2000 and 2010 at the study site (P = 0.96; R2 = 0.0003; Fig. 2a). Mean annual nest depth ranged from 8.48 to 9.11 cm, and there was a slight but statistically significant positive correlation between nest depth and mean May air temperature (F1,2369 = 19.58, R2 = 0.008, P < 0.0001; Fig. 2b). Mean annual female body size decreased over time (F1,9 = 7.73, R2 = 0.46, P = 0.02; Fig. 2c), but was not correlated with mean annual nest depth (P = 0.88).
In 2010, we manipulated depths in 44 nests. Flooding of the Mississippi River in August caused the complete loss of 23 nests; therefore, we have incubation regime data for 8 shallow, 9 mean and 4 deep nests. The hatchlings in one additional nest were crushed by construction machinery shortly before we retrieved hatchlings from the nests, but because we were able to determine the sex of the crushed hatchlings, they were included in the analysis of sex ratio differences among treatments. We retrieved 185 live hatchlings from the 20 surviving nests, and 165 of these hatchlings survived the overwintering period to be included in the performance tests.
None of the six parameters of incubation regime differed among nest depth treatments (all P-values > 0.10; Table 1). Shade cover did not differ over nests assigned to the three depth treatments (F2,17 = 0.20, P = 0.82), but it was a marginally significant predictor of nest sex ratio (χ2 = 3.3; 1 d.f.; P = 0.07; Table 2). Nest depth treatment did not influence per cent hatching success (F3,15 = 1.47; P = 0.26), mean hatchling carapace length (F2,135 = 0.90; P = 0.41) or offspring sex ratio (χ2 = 6.3; 2 d.f.; P = 0.10; Table 2). In contrast, nest depth was negatively correlated with hatchling mass (F4,148 = 17.15; r = 0.56; P = 0.04). Hatchlings from the three nest depth treatments did not differ in any measure of median sprinting time (all P-values > 0.05). However, hatchlings from deeper nests had shorter total righting times (χ2 = 13.0; 2 d.f.; P = 0.01), active righting times (χ2 = 16.6; 2 d.f.; P = 0.002), latency to swim times (χ2 = 20.8; 2 d.f.; P = 0.001) and total swimming times (χ2 = 13.6; 2 d.f.; P = 0.01; Fig. 3) than hatchlings from shallower nests.
Table 1. Mean values for six parameters of incubation regime in manipulated painted turtle (Chrysemys picta) nests at Thomson Causeway Recreation Area, Carroll County, Illinois in 2010
Shallow (6.7 cm)
Mean (8.7 cm)
Deep (10.7 cm)
(n = 8 nests)
(n = 9 nests)
(n = 4 nests)
Values shown are means (°C) ± one standard deviation (number of hatchlings included in analysis). None of these parameters differed among nest depth treatments. All flooded nests are excluded.
Minimum incubation temp
15.6 ± 1.3
17.1 ± 1.6
17.1 ± 1.4
Maximum incubation temp
38.2 ± 4.6
37.8 ± 3.4
36.8 ± 4.5
26.1 ± 1.7
25.9 ± 1.0
25.7 ± 1.3
Daily range of temp, incubation period
8.0 ± 0.7
7.4 ± 1.1
7.0 ± 1.0
Daily range of temp, thermosensitive period
8.7 ± 0.9
7.6 ± 1.5
7.5 ± 1.2
Table 2. Mean hatchling survival, hatchling mass, hatchling carapace length (CL), sex ratio and shade cover in manipulated painted turtle (Chrysemys picta) nests at Thomson Causeway Recreation Area, Carroll County, Illinois in 2010
Shallow (6.7 cm)
Mean (8.7 cm)
Deep (10.7 cm)
(n = 8 nests)
(n = 9 nests)
(n = 4 nests)
Values shown are means ± one standard deviation (number of hatchlings included in analysis); boldface indicates significant differences. All flooded nests are excluded.
% hatchling survival
76 ± 22 (75)
79 ± 15 (67)
92 ± 6 (43)
Hatchling mass (g)
4.3 ± 0.6 (75)
4.1 ± 0.6 (67)
4.0 ± 0.6 (43)
Hatchling CL (mm)
25.7 ± 1.5 (75)
25.2 ± 1.5 (67)
25.4 ± 1.4 (43)
22 ± 25 (76)
8 ± 10 (67)
30 ± 21 (53)
% shade cover over nest
48.2 ± 6.6
48.0 ± 12.4
50.2 ± 10.2
The strong influence of nest-site characteristics on offspring survival, quality and phenotype illustrates the importance of maternal ability to match nest location and characteristics with environmental characteristics. Matching nest-site characteristics to prevailing environmental conditions may be either behaviorally plastic or an evolved response, both of which are mechanisms by which organisms could track climate change and thereby mitigate some negative impacts, such as skews in sex ratios in species with TSD. One nest-site characteristic that is potentially adjustable to match environmental conditions is nest depth, whereby females may simply construct deeper nests in warmer years. To determine the effect of nest depth on incubation regime, offspring sex ratio and offspring performance, we experimentally manipulated depths of naturally constructed nests within a biologically relevant range in a model turtle species with TSD.
May air temperatures at our study site displayed no consistent trend over the course of the 11-year study, yet nests were deeper in years with warmer May temperatures. Similarly, C. picta nests in climatically warmer New Mexico are slightly deeper than C. picta nests at our cooler Illinois site (Morjan, 2003b). While population differences in nest depths might be an evolved response to climatic differences, the fact that mean nest depths in Illinois tracked air temperatures during the nesting season suggests behavioral plasticity in females' ability to adjust nest depth based on prevailing environmental conditions. Similarly, the study population is known to adjust timing of nesting in response to conditions during the previous winter (Schwanz & Janzen, 2008). Importantly, however, cues associated with past or current conditions might not be predictive of future incubation conditions. In our study, despite the fact that nest depth tracked temperatures during the May nesting season, May temperatures were not correlated with temperatures in July (P = 0.58), which is the approximate thermosensitive period at the study site (Janzen, 1994a). Therefore, air temperature at the time of nesting might be an unreliable indicator of the incubation regime a nest site will experience during the thermosensitive period, unless climatic conditions follow a predictable, linear trend and do not simply vary unpredictably among years. If nest depth is adjusted based on current conditions that are not necessarily predictive of future incubation conditions, then nest depth adjustment, even if it is behaviorally plastic, might not be a component of nest-site choice that could reliably compensate for climate change. Female turtles might instead use a more reliable cue to predict incubation conditions during the thermosensitive period, such as shade cover (Janzen, 1994b). Our study provides additional support for the importance of shade cover as a predictor of future incubation conditions in that, even after our manipulation of nest depth, shade cover was a stronger predictor of offspring sex ratio than nest depth.
Within the biologically realistic range of nest depths tested here, incubation regime did not differ among nest depth treatments. It is not surprising, then, that we observed no significant difference in the thermally sensitive trait of sex ratio among our nest depth treatments. There were also no differences in hatchling survival or carapace length among depth treatments, although hatchling mass declined as nest depth increased. The lack of difference in sex ratio among treatments suggests that, while nest depths may be adjusted slightly under thermally divergent climatic conditions, the adjustment is of insufficient magnitude to affect offspring sex ratio in this small-bodied species. Therefore, for nest depth to compensate for potential sex ratio skews produced by climate change in this species, females at the Illinois site would have to construct nests that are 2 cm deeper than the current population mean, which translates into depths greater than two standard deviations from the current population mean. Because maximum nest depth in turtles is constrained by female limb length (Tiwari & Bjorndal, 2000; Refsnider, 2012), selection for increased nest depth would likely necessitate a concomitant increase in female body size if females are to construct considerably deeper nests. Strong direct selection on adult survival can cause rapid shifts in reptile body size (e.g. Wolak et al., 2010), but indirect selection for increased female body size to construct deeper nests is likely much weaker; therefore, compensation for rapid climate change through selection for deeper nests is likely to be evolutionarily constrained by relatively weak selection for increased female size (Refsnider, 2012). Female body size does not appear to be a strong driver of nest depth in the study population for two reasons. First, female turtles are currently constructing nests farther from their maximum physical capacity than populations at the edge of the species' geographic range (Refsnider, 2012). Second, changes in mean nest depth over time were not correlated with mean body size of reproductive females. However, populations at both the northern and southern edges of the species' range are currently constructing nests near their maximum physical capacity (Refsnider, 2012), suggesting that an adjustment in nest depth to compensate for climate change would require an increase in female body size in at least some populations.
The large shift in nest depth that would be required to affect incubation regimes sufficiently to alter sex ratios in this species seems unlikely to occur in response to climatic warming for the reasons discussed above. Similarly, in a reptile where nest depth was not constrained by female size and females did adjust nest depth to match climatic conditions, the adjustment was insufficient to prevent sex ratio skews (Telemeco et al., 2009). Instead, other components of nest-site choice might be more likely to compensate for climate change, either through microevolutionary change or phenotypic plasticity. For example, nesting phenology is behaviorally plastic in response to short-term climatic fluctuations; however, the heritability of nesting date is too low to allow this trait to respond to selection in painted turtles (Schwanz & Janzen, 2008). In another emydid turtle, nesting dates have become substantially earlier over the last decade, but the result has been a sex ratio skewed toward males because the earlier onset of nesting has allowed production of an additional clutch annually, and the final clutch produced in a year now experiences cool, male-producing temperatures during the thermosensitive period (Tucker et al., 2008). Based on these studies of phenology, nesting date is not thought to be a strong compensatory mechanism for climate change in the species studied (Schwanz & Janzen, 2008; Mitchell & Janzen, 2010). Shade cover over the nest is another component of nest-site choice, and is known to affect sex ratio (Janzen, 1994b) and to shift in response to novel climatic conditions (Refsnider & Janzen, 2012) in the study species. Shade cover might be the component of nest-site choice most likely to compensate for climate change in freshwater turtles, and behavioral plasticity is at least one mechanism known to underlie choice of shade cover in painted turtles.
We found that deeper nests produced hatchlings that weighed less and were faster at both righting themselves and swimming than shallower nests. Although nest depth treatments did not differ statistically from each other in incubation regime, deep nests tended to be less variable than shallower nests (Fig. 4), a difference which may have been statistically significant if more nests in the deep treatment had remained after the flood event. As discussed above, flooding resulted in mortality of all but four nests in the deep treatment. The four remaining deep nests were similar in location, date of construction and shade cover, so the differences observed in offspring size and performance are not likely because of differences in nest microhabitat. However, maternal identity was unknown for these nests, so maternal or paternal effects on offspring performance cannot be excluded as an explanation for differences in offspring performance. In contrast, a common-garden experiment (i.e. individuals from different populations housed under common environmental conditions) on the same species from five populations across a geographic range found that faster hatchlings were produced from nests that were more variable in daily temperature (Refsnider & Janzen, 2012). Other studies have also found differing effects of fluctuating temperatures on both sex ratio (e.g. Georges et al., 1994; Neuwald & Valenzuela, 2011) and performance (e.g. Andrews, Mathies & Warner, 2000; Les, Paitz & Bowden, 2007; Du & Feng, 2008) of neonatal reptiles, even within the same population. For example, hatchling smooth softshell turtles (Apalone mutica) from an Iowa population were reported to swim faster as thermal variability during embryogenesis increased (Ashmore & Janzen, 2003), whereas a subsequent study of the same population found a less clear effect of such thermal variability on post-hatching swimming ability (Mullins & Janzen, 2006). Such diverse outcomes call into question the utility of generalizing about the effects of thermal variation experienced during embryonic development on the post-hatching performance of reptilian offspring. Results from our study suggest that the impacts of climatic warming on performance of hatching turtles are difficult to predict and likely will differ with latitude.
Although the depth of turtle nests varies geographically and tracks May temperature at our study site, results from our study suggest that adjustment of nest depth is not likely to compensate for climate change in painted turtles. First, the amount by which nest depth would have to increase to affect sex ratio in the study population is 2 cm greater than, or two standard deviations from, the current population mean. Selection for a shift of this magnitude is likely relatively weak, and in some populations is probably biologically unfeasible without a simultaneous, substantial increase in female body size. Second, the proximate cues used by females to adjust nest depth may not reliably predict future incubation conditions. Finally, shade cover over the nest was a stronger predictor of sex ratio than nest depth. Therefore, shifts in components of nest-site choice other than nest depth, such as selection of shade cover over nest sites, may be more likely to allow this species to match incubation conditions to a changing climate. Importantly, variation among reptiles with TSD in whether nest depth adjustment affects sex ratio or responds to climatic variation (e.g. Vogt & Bull, 1982; Warner & Andrews, 2002; Morjan, 2003b; Doody et al., 2006a; Mitchell et al., 2008; Telemeco et al., 2009) indicates that adjustment of nest depth does not have similar effects across taxa, and it should not be assumed that nest depth adjustment is a compensatory mechanism for climate change without evaluating its potential in particular species. In species where nest depth adjustment in response to inter-annual climatic variation has been observed, future research should focus on determining the relative importance of potential mechanisms underlying nest depth adjustment.
We thank T. Mitchell, H. Streby, J. Strickland and D. Warner for help with data collection; A. Bronikowski, P. Dixon, C. Kelly, E. Takle, H. Streby and two anonymous reviewers for helpful comments on the paper; and the US Army Corps of Engineers for access to the study site. This research was conducted in accordance with Institutional Animal Care and Use Committee protocol # 12-03-5570-J (Iowa State University). This study was funded by Sigma Xi Grants-in-aid-of-Research (to J. M. R.) and the National Science Foundation (DEB-064932 to F. J. J.).