Introduction

Organisms that inhabit seasonal environments require particular traits and strategies to deal with unfavorable environmental conditions (Plummer et al., 2003; Roff, 2002; Seidel, 1978; Stearns, 1992; Withers & Cooper, 2010). Diverse physiological and behavioral adaptations have been described to deal with seasonality (Barros et al., 2020; Volaire et al., 2023). Withers and Cooper (2010) suggested that it is possible to distinguish two basic types of strategies: cryptobiosis (ametabolism) and dormancy (hypometabolism). The first involves a severe and measurable metabolic depression, in which the vital rates of organisms are reduced to a minimum. This includes anhydrobiosis (lack of loss of water from the body), cryobiosis (partial freezing of the body), osmobiosis (osmotic depression), and anoxybiosis (lack of oxygen). The second includes dormancy or hypometabolism, which involves a moderate reduction in metabolic rates (can be minimal or sometimes undetectable), as well as associated behavioral traits. There are special types of dormancies such as those expressed in early development stages: diapause and quiescence (associated with low or high temperatures) and dormancies for adult stages of individuals (torpor): hibernation (promote by low temperatures) and aestivation (in high temperatures).

Aestivation is considered a strategy to overcome the lack of water during the dry and/or hot season of the year (Withers & Cooper, 2010). The most obvious effect of the dry season is the reduction of available water and reduced relative humidity, or the complete absence of the aquatic habitat during some part of the year. For some aestivating species, the life cycle involves spending some period of the year in the water and other period in terrestrial environments (Hudson & Franklin, 2002; Li et al., 1996). For example, aquatic and ectothermic vertebrates in seasonal environments, such as lungfish, amphibians, crocodilians, and turtles, often face long periods without water (or with low atmospheric humidity) (Secor & Lignot, 2010). There is also some evidence of turtles (Kinosternon sonoriense) using ephemeral aquatic habitats or remaining in the muddy remaining ponds during dry season (Ligon & Sotne, 2003), also other mud turtles (K. creaseri) endure using reduced pools during dry season (Macip-Ríos et al., 2018).

Among the adaptations to water loss in ectotherms are the ability to conserve water in the body, the construction of burrows, tissues, and fluids to facilitate aestivation (e.g., hibernaculum, mucus or excretions in lungfish and amphibians), as well as other behavioral and physiological traits (e.g., migration, activity and/or metabolism reduction) (Glass et al., 2009; Plummer et al., 2003; Steen et al., 2007; Withers & Cooper, 2010). Morphological traits to deal with desiccation and water loss have been examined in aestivating ectotherms such as snails (Emberton, 1994; Giokas et al., 2005; Perrott et al., 2007), generating even some discrepancies about the functionality of morphological features such as the operculum, the size and the shape of the shell to avoid dehydration (see Arad, 1993; Emberton, 1994). In general perspective, morphological traits as opposed to physiological and behavioral adaptations for aestivation have received less research attention (Murphy et al., 2016; Secor & Lignot, 2010).

In turtles, at least two primary behavioral responses to the absence of environmental water have been reported; some individuals of the population migrate (Gibbons et al., 1983), while others aestivate (Roe et al., 2008). Few things have been mentioned about the morphology associated with aestivation (Wygoda & Chmura, 1990). It is possible that some morphological characters are associated with aestivation in turtles; for example, a shell that closes completely can reduce evapotranspiration (Berry, 1997; Dodd, 2001; Murphy et al., 2016). Similarly, reductions in skin surface area can retard evaporative water loss (Stone & Iverson, 1999). However, these are little-explored aspects of turtle morphology and function.

Aestivation has been reported for several turtle lineages, including both pleurodires (side-necks) and cryptodires (hidden-necks) (Peterson & Stone, 2000; Roe et al., 2008; Withers & Cooper, 2010), but the origin and evolution of this trait in turtles remains unknown. Although aestivation has been described in some species (Ligon & Peterson, 2002; Peterson & Stone, 2000; Withers & Cooper, 2010), it is still an understudied phenomenon from the evolutionary and comparative points of view (Peterson & Stone, 2000). However, basic information for many turtle species does exist on whether or not they aestivate, as well as their average and maximum aestivation times (Abe, 1995; Ernst & Barbour, 1992). Here, we compiled a global database on aestivation in turtles, conduct a comparative examination of macroevolutionary trends in turtle aestivation, and tested the association of shell morphology with aestivation times. We hypothesized (1) that the origin of aestivation in turtles has an early ancestral root and has evolved several times in the ancestors of current species and (2) morphological variation in shell traits of extant turtles could explain differences in the aestivation duration.

Materials and Methods

Data Collection

We compiled a comparative dataset including aestivation status (aestivate/do not aestivate), shell morphometric data, and aestivation time from 225 freshwater turtle and tortoise species. Aestivation data (presence/absence and times) were gathered through information available in the scientific literature (Supplementary data, Table S1). Morphometric data were collected from the carapace and plastron of 1143 museum or living specimens (when available) (see Table S2; main institutions and source of specimens).

To compile our dataset on aestivation presence/absence in chelonians, we considered aestivation as present in a species if it was reported in at least one population or subspecies in the literature. Tentatively, we recorded aestivation as absent when published studies or personal communications with qualified researchers supported that hypothesis. Maximum and mean aestivation time (henceforth called MAEST and MEST) were compiled, and a phylogenetic ANOVA was implemented under the function phylANOVA with the package phytools (Revell, 2012) in R (R Core Team, 2021) to search for differences in aestivation times among turtle families.

Many species (n = 27) assessed for the status of aestivation (see Table S1) were not present in the phylogenetic time tree used to assemble the final dataset for the main evolutionary analysis (Thomson et al., 2021; see below). Therefore, these species were not included in the phylogenetic comparative analyses, instead were included to illustrate the summary information on the global status of aestivation in chelonians.

Shell Morphological Traits

The shell morphological characteristics that we measured included the following: minimum straight-line carapace length (CL; mid-line of cervical scute to cleavage between supracaudal scutes), straight-line plastron length (PL; midline length), straight plastron width (PW; length of the seam that connects abdominal and pectoral plastral scutes), straight plastral lobe width (LOBE; the length of the seam that connects femoral and abdominal scutes), carapace width (CW; width at the junction of the fifth and sixth marginal scutes), shell height (SH; the maximum vertical height from plastron to carapace), and bridge length (Bridge; the length from axillary to inguinal notch). All measurements were made to the nearest 0.1 mm using dial calipers.

In addition, we calculated the S and F indices, two metrics related to the shape of the turtle’s shell (Domokos & Várkonyi, 2008) and a third metric, the surface area-to-volume ratio of the carapace (SA/V), known to be associated with rates of heat exchange with the environment (Boyer, 1965). We calculated these metrics using some of the shell measurements with the following formulae: S = (SH × CW)/CL2)1/3 and F = (CL + CW)/(2 × SH) (Domokos & Várkonyi, 2008). The SA/V ratio was calculated by approximating the carapace as half of an ellipsoid. The surface area (SA) and volume (V) of the turtle’s carapace was calculated using the following formulae: SA = 4π[(ApBp + ApCp + BpCp)/3]1/p/2 and V = 2/3 (ABC); where A was half of CL, B was half of CW, C was half of SH, and p was a standardized, fixed value of 1.6075 (Berlant & Stayton, 2017). In general, high scores of S correspond with a more spherical shell whereas a high value of F corresponds with a flatter shell (Ana et al., 2015). A higher SA/V ratio corresponds to a higher relative surface area of the carapace (Boyer, 1965). We compared average values by species for all indices.

Evolutionary Analyses

We reconstructed ancestral states for aestivation status using the phylogeny of Thomson et al. (2021), which is based on a multilocus DNA sequence data for 279 extant turtle species and calibrated with 22 fossil species. We inferred ancestral states for presence or absence of aestivation using the corHMM function in the corHMM v2.5 R package (Beaulieu et al., 2017, 2021). This function calculates the maximum likelihood estimates of transition rates between states and then uses these values for determining state probabilities for internal nodes of the tree and can also incorporate “hidden” rate changes (e.g., different velocities of evolutionary rates among clades (Boyko & Beaulieu, 2021).

We constructed five models by modifying the transition (Q) matrix. The first considered all transition rates as equal (ER model). The second allowed for transition rates between states to differ (ARD model). These two models assume that the process generating the different states at the tips and ancestral nodes is homogenous across all branches of a phylogenetic tree, which may be a major simplification of biological reality (Ibáñez et al., 2021). The generalized hidden Markov model (Beaulieu & Donoghue, 2013; Boyko & Beaulieu, 2021) implemented in corHMM v2.5 relaxes this assumption by allowing more than one process to affect trait evolution across a phylogeny. This is achieved by constructing > 1 rate categories (i.e., transition matrices) and allowing them to vary across the tree by parameterizing the transitions among rate categories (Ibáñez et al., 2021). We constructed three different models, each with two rate matrices (R1 and R2, rate.cat = 2). We specified a model with two equal rate (ER/ER model) matrices, another with two all rates different (ARD/ARD model) matrices, and finally a mixed ER/ARD model. In all cases, rate category transitions (R1 → R2 and R2 → R1) were allowed to differ. Akaike’s information criterion corrected for small sample size (Burnham & Anderson, 2004) was calculated from the log-likelihoods to compare the fit of the models. Each model was run 100 times (nstarts = 100). All analyses mentioned were performed using R v4.0.5 (R Core Team, 2021).

Correlating Aestivation Time and Shell Morphology

Given that turtle species in our study have a shared evolutionary history, we performed phylogenetic comparative analyses. Aestivation times (MAEST and MEST) for each species (Supplementary data, Table S1) were square root-transformed to fulfill normality assumptions and standardize values. We obtained mean values of ten shell morphological variables (CL, PL, PW, Lobe, CW, SH, Bridge, S index, F index, SA/V) for each species with aestivation present (Supplementary data, Table S2). We log10-transformed these data and then eliminated body size effect on shell variables by calculating phylogenetic residuals using the ‘phylo.resid’, a function in the ‘phytools’ package (Revell, 2012). To explore whether the aestivation times (dependent variables) can be explained by shell morphological variables (predictive variables); we performed phylogenetic generalized least squares (PGLS) model analyses employing the ‘pgls’ function in ‘caper’ (Orme et al., 2018) and ‘ape’ (Paradis & Schliep, 2019) R packages. We constructed saturated and simplified models that included all sampled turtle species, and the two models by representative families in our dataset. We based our model selection (including evolutionary models) on two criteria: first, Akaike Information Criterion corrected for small sample size (AICC), which had the lowest value for AICC and the highest Akaike weight (AICw; Burnham & Anderson, 2004); second, we considered the delta AICC values (ΔAICC) criterion, where better models are indicated by ΔAICC values less than 2 (Symonds & Moussalli, 2011). See Supplementary data S4 for details.

Results

Aestivation Distribution

Our dataset indicates that just under half (44.4%) of 225 species of turtles aestivate. (Fig. 1, Supplementary Table 1). Within pleurodire turtles, the aestivation was present in all three families: Chelidae (43.8% of species), Podocnemididae (25% of species) and Pelomedusidae (88.9% of species), while in cryptodiran turtles, the families with evidence of aestivation were the Trionychidae (16.7% of species), Kinosternidae (76.6% of species), Emydidae (34.1% of species), Testudinidae (56.4% of species) and Geoemydidae (29.8 of species) (Fig. 1). No evidence (but there is anecdotal data that suggest some type of dormancy; George, 2010; Howey & Dinkelacker, 2007) of aestivation is reported for 4 of 5 extant chelydrid species (Fig. 1).

Fig. 1
figure 1

Summary of the number of turtle species with presence or absence of aestivation by family (based on the report of 225 species, Supplementary Table S1)

Full size image

Considering the average comparisons in MAEST and MEST among turtle families with more than two species in our dataset (i.e., excluding the Podocnemididae and Trionychidae), we found no significant differences associated with aestivation times among turtle families (MAEST and MEST: phylANOVA, F = 1.61, P = 0.99; for both). However, the longest aestivation times (both MAEST and MEST) were observed for Chelidae, Pelomedusidae, Geoemydidae, and Kinosternidae (Fig. 2), and the shortest aestivation times were observed for emydids and testudinids (Fig. 2). Aestivation time was also a variable trait across species, with very long times in some species such as Pelomedusa subrufa and Kinosternon integrum, with 304 and 279 days of MAEST respectively, and very short times (e.g., 61 maximum days) in the species Mauremys leprosa, Gopherus berlandieri, and Terrapene carolina (Supplementary Table 1). MAEST also varied among species of the same genus, such as Kinosternon sonoriense (61 days) vs. K. alamosae and K. oaxacae with around 273 days each (Supplementary Table 1).

Fig. 2
figure 2

Aestivation times in days (mean ± SE) by families of extant turtle species (based on the 100 species exhibiting aestivation, Supplementary Table1). MAEST maximum aestivation time, MEST mean aestivation time

Full size image

Inference of Ancestral States

Our binary state reconstruction using the Thomson et al. (2021) phylogeny revealed that aestivation does not have a single ancestral origin, being an independent trait that arose separately in the two major turtle clades (i.e., Cryptodiran and Pleurodiran) (Fig. 3). The best theoretical model with the fewest independent gains and losses was the ARD model (all rates different) (Table 1).

Fig. 3
figure 3

Maximum likelihood inference of ancestral states in aestivation status in extant chelonians based on the molecular phylogeny of Thomson et al. (2021), allowing for different transition rates between states (ARD model). Pie diagrams at each node represent likelihood estimates for ancestral states. Black and white circles on the tips show aestivation status for extant species

Full size image
Table 1 The fit of alternative models of aestivation evolution in all sampled turtles, based on the phylogenetic hypotheses of Thomson et al. (2021)
Full size table

Notably, two of the following best-fitting models (model with two equal rates: ER/ER; mixed model: ER/ARD) incorporated hidden Markov processes with a matrix category transition at the two major turtle clades, consistent with the ARD model. In keeping with the inferred ancestral reconstruction of aestivation, two models (ARD and ER/ARD) showed that the loss rate of aestivation was much higher than the rate of aestivation maintenance (Supplementary data, Table S3). In contrast, the ER/ER model suggested a balanced rate of loss and maintenance of aestivation but was unsupported (ΔAICc = 0.25). Out of these, the ARD model was the least parameterized.

Under the ARD model, the history of aestivation in pleurodire turtles was particularly variable. Independent losses of this trait occurred multiple times, with at least 14 important losses in chelids occurring in most ancestral lineages of Phrynops, Emydura and Elseya; and six generalized and recurrent losses in the ancestors of most extant podocnemidids. The aestivation gains in pleurodires occurred mostly in ancestral clades of extant pelomedusids (seven gains) and in some few lineages of Chelodina and Mesoclemmys (Chelidae). The evolution of aestivation in cryptodires was also highly variable with contrasting patterns of loss and maintenance in the ancestral testudinid clades (at least three times more gains than losses 21:7), and prominent losses occurring in most ancestral clades of trionychids (14 losses), emydids (at least 30 losses), and geoemydids (at least 35 losses). However, the maintenance or gains of aestivation in cryptodiran turtles occurred mainly in ancestral clades of extant semiaquatic kinosternids (19 gains), Terrapene for emydids, and the terrestrial lineages of Testudo, Kinixys, Psammobates, and Pyxis for testudinids.

Association of Aestivation Time with Shell Morphology

A total of 188 PGLS models were examined to disentangle whether morphology is a good predictor of aestivation time in turtles (Supplementary data, Table S4). Based on the selection criteria used (weight and Akaike´s delta) we selected 25 models (Table 2). We did not find significant models that explain aestivation time in all turtle lineages (Table 2); although, two models “MEST ~ F index” and “MAEST ~ F index” showed an association with marginal values close to a 95% confidence interval (S = 6.32, P = 0.07; and S = 8.89, P = 0.07, respectively; Table 2). Additionally, we found in all the analyses a similar statistical response of the two dependent variables (MEST and MAEST; Table 2). Therefore, the consecutive analyses by family were only based on the dependent variable MAEST. Ten models were significant for the MAEST in Chelidae (Table 2). To avoid redundancy, we describe only the two best models here (details in Table 2). First, the best model was “MAEST ~ SH + LOBE + PL” (full model R2 = 0.888; P = 0.0003; Fig. 6A) showing MAEST has a positive and significant association with the shell height (S = 21.430; P = 0.001; Table 2; Fig. 4A) and plastral lobe width (S = 20.250; P = 0.003; Table 2; Fig. 4B), and a negative and significant association with plastron length (S =  − 47.231; P = 0.001; Table 2; Fig. 5C). The second-best model was “MAEST ~ SA/V” (R2 = 0.504; P = 0.008; Table 2) showing MAEST has a negative and significant association with surface area to volume ratio of the carapace (S = -− 30.423; P = 0.008; Table 2; Fig. 5B). Finally, although not among the top best models, we consider it important to describe the model “MAEST ~ F index”, which shows a third, negative and significant association of MAEST with an indicator of the flat shape of the shell (S = -− 23.480; P = 0.018; Table 2; Fig. 5A).

Table 2 Best-fit models for the effects of morphological variables on estivation time in turtles, from pgls analyzes, based on two criteria of selection: Akaike weight criterion and Akaike’s delta criterion
Full size table
Fig. 4
figure 4

Positive associations between maximum aestivation time (MAEST) and shell morphological traits in chelonian families. A, B Chelidae, C Pelomedusidae, D, E Kinosternidae and F Geoemydidae. SH shell height, LOBE straight plastral lobe width, PW straight plastron width, BRIDGE bridge length, PL straight plastron length

Full size image
Fig. 5
figure 5

Negative associations between maximum aestivation time (MAEST) and shell morphological traits in chelonian families. A–C Chelidae, D Pelomedusidae and E, F Testudinidae. F index flat shell scores, SA/V surface area-to-volume ratio of the carapace, PL straight plastron length, BRIDGE bridge length, CW carapace width, PW straight plastron width

Full size image

Models examining MAEST in the Emydidae and Pelomedusidae were not significant (Table 2); although, in the Pelomedusidae two models (“MAEST ~ PW + PL”; Fig. 6B and “MAEST ~ BRIDGE + CL”; Table 2) MAEST showed positive and significant associations with straight plastron width (Fig. 4C) and straight carapace length and two negative and significant associations with bridge length (Fig. 5D) and straight plastron length (all, P ≤ 0.03; Table 2).

Fig. 6
figure 6

Positive associations between maximum aestivation time (MAEST) and the conjunction of two or more shell morphological traits in chelonian families. A Chelidae, B Pelomedusidae, C Kinosternidae, D Testudinidae, and E Geoemydidae

Full size image

The best fit model for the Kinosternidae was “MAEST ~ BRIDGE + CW + PL” (R2 = 0.387; P = 0.038; Fig. 6C). Suggesting that MAEST had positive and significant associations with bridge length, carapace width, and plastron length (all, P ≤ 0.02; Table 2, Fig. 5D, E). For the Testudinidae the best model was “MAEST ~ SH + PW + LOBE + CW” (R2 = 0.547; P = 0.015; Fig. 6D). Where MAEST had negative and significant associations with shell height (S = − 30.733; P = 0.008; Table 2) and straight plastron width (S = − 111.09; P = 0.002; Table 2; Fig. 5F), and carapace width (S = − 130.54; P = 0.002; Table 2; Fig. 5E). Finally, the best fit model for the Geoemydidae was “MAEST ~ LOBE + PL + CL” (R2 = 0.464; P = 0.046; Fig. 6E). This suggests that MAEST had positive and significant associations with straight plastral lobe width (S = 72.270; P = 0.046; Table 2) and straight plastron length (S = 107.465; P = 0.036; Table 2; Fig. 4F).

Discussion

Evolutionary History of Aestivation in Turtles

Aestivation has been an adaptive dormancy among numerous species of invertebrates and vertebrates (Navas & Carvalho, 2010). For example, the pulmonate snail Oxystyla pulchella and the Australian water-holding frog Cyclorana platycephala can survive long dormant periods 23 years, Baker, 1934; 5 years, Van Beurden, 1980, respectively. In turtles, the longest recorded aestivation time is for the Yellow Mud Turtle Kinosternon flavescens with two years of dormancy (Rose, 1980). Although the duration of aestivation is highly variable, the understanding of aestivation in ectothermic vertebrates is still a little explored field, and the evolutionary relationships between aestivating and non-aestivating species has been slightly established in few studies (Criscione & Köhler, 2016). In turtles, the relationships between phylogeny and behavioral traits have been little studied, compared to evolutionary studies of morphological or anatomical traits (e.g., Ascarrunz & Sánchez-Villagra, 2022; Cordero & Vlachos, 2021; Ibáñez et al., 2021). Our compilation of the occurrence of aestivation combined with a robust phylogeny provided the basis for a better understanding of the evolutionary history of a behavioral trait involved in eco-physiological adaptations, a subject little explored in turtles. The evolutionary reconstruction indicated that aestivation is a derived trait in the two major turtle clades and occurring in many independent evolutions. Aestivation seems to have been secondarily lost multiple times within the pleurodires in two of three families and within the cryptodirians in four of six families. The loss of aestivation is most evident for clades that are largely or completely aquatic (e.g., Chelydridae, Trionychidae, Emydidae, Chelidae), compared to clades that evolved toward terrestriality, such as in Testudinidae, or diversified in ephemeral aquatic habitats (e.g., Kinosternidae, Pelomedusidae).

Our results on the evolution of aestivation suggested a synapomorphic state shared by all extant turtle taxa, although with various lineages showing different trajectories (e.g., compare the podocnemidids vs. pelomedusids, or trionychids vs. kinosternids, respectively; Figs. 1, 3). The most recent paleontological hypothesis suggests that the stem ancestor of all turtles, the early proto turtle Eunotosaurus africanus (260 Ma), was a terrestrial species that dug burrows (Lyson & Bever, 2020; Lyson et al., 2016), and whose original osteological modification during the Permian Period was principally an adaptation for fossorial life, and the current protective function of the turtle shells might be an exaptation (Lyson & Bever, 2020; Lyson et al., 2016). This suggests that morphological and behavioral evolution were essential processes in the burrowing ancestor of all turtles in order to tolerate stressful conditions in terrestrial environments. Hence, at least some ability to tolerate hot and dry conditions by retreating into a burrow would be adaptive. Thus, these first proto-turtles had at least behavioral adaptations to avoid desiccation or extreme temperatures. Most turtles can tolerate a short period of lethargy if the environment is harsh, but at some point that lethargy would be supplemented by other adaptations (e.g., aestivation) that allowed the turtles to survive longer periods of environmental stress. Here, our hypothesis is that the first evolutionary event of aestivation probably arose in fully shelled turtles (i.e., pleurodires and cryptodires) during the Jurassic Period as our reconstruction suggests.

Fossil records of chelonian burrows in the Cretaceous period could have been used for aestivation in the extinct species Bauruemys elegans (Silva et al., 2022). However, it is impossible to determine a specific function accurately, since these turtle burrows could also be used for other functions such as thermoregulation, reproduction, breeding or nesting. However, the fossil species B. elegans is an extinct genus of podocnemids in which our ancestral reconstruction seems to indicate one of the largest aestivation losses for pleurodires.

The evolution of aestivation seems to be particularly variable in the extant Testudininoidea group (i.e., Emydidae, Testudinidae, and Geoemydidae), with cases of aestivation loss and retention, even within the same family (e.g., Emydidae). Despite deep divergences among branches of the turtle tree within families, the behavioral component to aestivate was retained and persists in descendants of ancient lineages that have the ability to aestivate. Regarding the turtle descendant species in which aestivation is apparently absent, the inferences indicate that this aestivation capability is not entirely erased in the evolution of turtle clades.

Shell Morphology and the Aestivation Time

Our results suggest that shell morphology is correlated with the length of aestivation time across different families of chelonians. This result seems to be consistent with other studies, since the shape, size, and other morphological dimensions of the turtle shell are important factors related to ecological and physiological performance (Butterfield et al., 2021; Murphy et al., 2016). However, it is necessary to consider this result with caution, since aestivation time in animals depends not only on the contribution of morphological traits, but also on the effect generated by several factors such as the quality of the aestivation shelters, intrinsic adaptations of each species (i.e., physiological and behavioral traits, see below) and environmental conditions.

Desiccation is considered one of the primary sources of mortality for aestivating, migrating or dispersing freshwater turtles (Finkler, 2001). Desiccation risk depends on an individual’s rate of evaporative water loss (EWL) and critical thermal tolerance, two important physiological traits of water balance and thermoregulation in ectothermic animals (Chessman, 1984; Costanzo et al., 2001; Le Galliard et al., 2021). Differences in EWL are associated with behavioral restrictions such as dispersion distance to new water bodies or terrestrial aestivation sites, time spent aestivating, and ultimately the individual’s risk of mortality from prolonged drought periods (Finkler, 2001; Seidel & Reynolds, 1980). In turtles, two morphological features that may determine inter- and intraspecific differences in EWL rates and drought sensitivity are body size and shell morphology (Murphy et al., 2016). Small turtles (e.g., juveniles or smaller species) have a higher surface area to volume ratio and therefore would be expected to lose water at a faster rate than would larger turtles (Murphy et al., 2016). Likewise, shell morphology, including the relative amount of exposed skin, also likely influences EWL rates (Costanzo et al., 2001; Stone & Iverson, 1999). Species with reduced plastrons seem to have higher EWL rates than species with more extensive shells and species capable of fully enclosing themselves within their shell (e.g., box turtles and mud turtles) should have very low EWL rates (Murphy et al., 2016; Stone & Iverson, 1999).

Morphological characteristics of the shell that influence EWL could also influence aestivation time. Rather than a single morphological characteristic exhibiting a correlation between aestivation time and shell traits across turtles, our results suggest that the effects of shell morphology on aestivation times appear to be more family-specific. Relationships between morphological indices (of shell shape) and aestivation time were only significant within chelids. Chelids with flatter shells (> F index) or with a higher SA/V tend to aestivate for less time. We believed that the shell shape of chelids probably facilitates water loss and increases rapid and progressive warming, reducing the duration of prolonged aestivation. However, more research will be needed to prove or reject our hypothesis.

Aestivation in Chelonians and Future Research

Before 2000 there was a major controversy over the role of the physiological capacities of turtles during aestivation. Authors such as Gregory (1982), Grigg et al. (1986), and Ultsch (1989) suggested that physiological abilities have little or no relevance for aestivation, emphasizing mainly behavioral definitions. Subsequent studies have shown that aestivation in turtles can be a natural or lab-induced process with associated physiological and behavioral characteristics (Peterson & Stone, 2000). Therefore, Seidel’s (1978) original definition of aestivation was: “a behavioral strategy accompanied by physiological adjustments” seems more accurate. In chelonians, aestivation is probably a facultative trait with remarkable plasticity in physiological and behavioral responses to drought (Henen et al., 1998; Kennett & Christian, 1994; Peterson & Stone, 2000), which may be different between species, populations or even between individuals (Morales-Verdeja & Vogt, 1997; Peterson & Stone, 2000).

This variation is particularly evident in the kinosternids (a lineage well known for aestivating) between aestivating and non-aestivating experimental samples (Ligon & Peterson, 2002; Seidel, 1978; Wygoda, 1979). In species such as Kinosternon hirtipes, aestivation can have serious physiological and behavioral limits (Seidel & Reynolds, 1980), and it is considered a limited or non-aestivating species (Ligon & Peterson, 2002). In contrast Kinosternon flavescens is a species strongly adapted for aestivation (Ligon & Peterson, 2002; Peterson & stone, 2000). This implies that not all turtles have the same aestivation capabilities or efficiencies. Thus, there are some species (probably the case of K. hirtipes) that use migration to more stable water bodies as a primary response to drought (Gibbons et al., 1983). Decisions to aestivate or emigrate may depend on body condition or energy stores (Derickson, 1976; Roe et al., 2008). Alternately, Peterson and Stone (2000) suggested that drought-migrant and drought-aestivating species are probably influenced by different genetic factors.

Among the aestivation responses in turtles are: (1) inactivity and/or sheltering during drought (Ligon & Peterson, 2002; McKnight & Ligon, 2020); (2) tolerance of long-term dehydration and concomitant increases in body fluid concentrations (anhomeostasis; Peterson, 1996; Peterson & Stone, 2000); (3) reduced rates of evaporative water loss (Chessman, 1984; Seidel & Reynolds, 1980); (4) reduced metabolic rate (Kennett & Christian, 1994; Seidel, 1978); and (5) storage of excretory wastes to minimize loss of water (Ligon & Peterson, 2002; Peterson & Stone, 2000). Further exploration of the evolution of physio-behavioral traits of aestivation in turtles will allow us to better understand this adaptive strategy.

A large proportion of the planet’s biodiversity and habitats is projected to be susceptible to the threat of climate change in the future (Ledig et al., 2010; Malcolm et al., 2006). It is possible that species of turtles with reported aestivation strategies are already dealing with the impacts of climate change in some areas. Model species such as Kinosternon sonoriense and K. flavescens, faced with the desertification of southwestern North America over the last 12,000 years (Fredrickson et al., 1998) may be under particularly strong selection for aestivation ability (Ligon & Peterson, 2002; Peterson & Stone, 2000). The shift toward an increasingly arid climate undoubtedly will have a profound impact on aquatic biodiversity (Ligon & Peterson, 2002). Given those possible changes, some questions arise: If the thermal and hydric quality of the habitat is affected by global warming, for turtles with little or no possibility of migrating, would aestivation be selected for in those environments? Will the limits of aestivation ability be reached in the near future? In cases of aestivating species dealing with prolonged periods of drought, would increased body size be an adaptive response? These and other questions can only be answered by conducting population studies that address the multiple ecological and physio-behavioral relationships of the aestivation process through time. Finally, based on experimental and field data, we need to evaluate the effects of hypothesized climate change on present and future projections on the potential distributions of each species of turtle, including its optimal habitat for aestivation (e.g., Butler et al., 2016). Undoubtedly current and future turtle conservation strategies would benefit from the future studies suggested here.