Green and loggerhead marine turtles (Chelonia mydas and Caretta caretta, respectively) are the 2 marine turtles nesting commonly along Mediterranean coasts (Casale and Margaritoulis 2010). Whereas green turtle nesting sites are restricted to the east of the Oriental Basin of the Mediterranean Sea, the distribution of nesting sites of loggerheads encompasses most of the Oriental Basin. The main aggregations occur along the Cypriot, Turkish, and Greek coasts at the east and in Libya at the west of the Oriental Basin. Although the Cypriot, Turkish, and Greek populations have been well studied (Casale and Margaritoulis 2010), the Libyan population has not, and it may be one of the most important populations in the Mediterranean region (Saied et al. 2012).

Loggerhead nesting populations in the Mediterranean are genetically separated into 3 groups (Garofalo et al. 2009): one at the west of the Oriental Basin nesting in Calabria (Italy), one at the east of Turkey, and a central one nesting in Greece, western Turkey, Lebanon, Cyprus, and Libya (Fig. 1). The population nesting in Libya emerged as the oldest population in the Mediterranean, dating from the Pleistocene ∼ 65,000 yrs ago (20,000–200,000) (Clusa et al. 2013). This suggests that the Libyan population might have settled in the Mediterranean Basin before the end of the last glacial period. The remaining populations from the central part (Greece, western Turkey, Lebanon, and Cyprus) originated as a result of a more recent Holocenic expansion after the last glacial era some 12,000 yrs ago (Clusa et al. 2013). The Libyan population belongs to the central genetic group but nests in a thermal environment that is very different from the other populations (Figs. 2 and 3) because it is the only major nesting along the southern part of the Mediterranean. The differences between Libyan and Turkish populations give the opportunity to test whether this species has been able to adapt rapidly (< 12,000 yrs) in different thermal environments (Figs. 2 and 3).

View Figure

• Download as Powerpoint
• Download Figure

View Figure

• Download as Powerpoint
• Download Figure

View Figure

• Download as Powerpoint
• Download Figure

Virtually all biological rates of ectothermic animals are affected by temperature, including development times (Gillooly et al. 2002) and growth rates (Gillooly et al. 2001). Egg incubation temperatures affect the duration of embryogenesis (Miller 1985), the probability of embryo survival (Van Damme et al. 1992), the sex for species with temperature-dependent sex determination such as C. caretta (Pieau 1996), and the morphology and body size at hatching (Du and Ji 2003). In addition, long-term effects of incubation temperature on the physiology and behavior of hatchlings have been observed (Burger 1989; Sibly and Atkinson 1994). Thus, the fitness of developing embryos and hatchlings is strongly dependent on the temperature within the nest during incubation.

Recently, a general way to model sigmoidal embryo growth with variable incubation temperature obtained from field data was proposed and tested with data from C. caretta from Dalyan beach, Turkey (Girondot and Kaska 2014a). We apply this model to new data of nest temperatures gathered in 2009 in Libya to test if embryos of both populations react differently to incubation temperature. Finally, we discuss how the different temperature-dependent life history traits could interact with each other to mitigate or enhance the effect of climate change.

METHODS

Field Data.

In Turkey, temperatures of 21 loggerhead turtle nests were examined via electronic temperature recorders during the 2010 nesting season on Dalyan Beach. Tiny talk temperature recorders (resolution ± 0.37°C; Orion Components Ltd) were used to measure temperature. They were placed into the middle of the nests (∼ 45-cm depth) during oviposition or on the same night or following morning in the case of relocated nests. For each of the 21 monitored nests of C. caretta, temperatures were recorded every 90 min from the time of egg deposition to the emergence of hatchlings.

In Libya, temperatures were recorded with 32 electronic temperature recorders (LogTag HAXO-8 Humidity and Temperature Data Logger; LSTechnology) during the 2009 nesting season (July–September) in 14 loggerhead nests on several beaches (Al Ghbeba, Althalateen, Shash, Al-Arbaeen, and west of Althalateen) located to the east of Sirte city, Libya, in the Gulf of Sirte (Fig. 1). One to 4 data loggers were placed in nests found in the morning during beach monitoring. All loggers were programmed to record temperature every 15 min. The dates of emergences of hatchlings were also recorded. Some nests were equipped several days after nesting occurred for logistical reasons, and missing temperatures were reconstructed as described in the next section. Data logger measures have been compared with temperatures recorded by mercury thermometers, and no systematic deviation was observed.

Filling the Temperature Gaps.

Based on the observation that nest temperature is correlated with air temperature (AT) and with the sea surface temperature (SST) in front of the beach (Girondot and Kaska 2014b), we collected such data for the years 1979–2014 from the European Centre for Medium-Range Weather Forecasts (ECMWF) data sets, which provide temperature every 6 hrs (UTC) for several decades at 0.1° resolution (Dee et al. 2011). SST and AT at a height of 2 m and the closest available to the nesting sites were extracted for the entire nesting period (Fig. 3).

We modeled the average daily temperature using a generalized linear mixed model with Gaussian distribution and identity link function:

where NT_j,i is the daily average temperature of the nest j on day i (I = 0 for nest deposition), μ is a constant term, and α and β are the coefficients associated with the linear relationship between NT_x and SST and AT lagged by dSST and dAT days, respectively. The analysis was performed for dSST and dAT from 0 to 10 d, and the best combination was discriminated using the Akaike information criterion (AIC), which is a measure of the quality of fit, which penalizes the quality of the fit for too many parameters in the model (Burnham and Anderson 2002). When several combinations had similar AIC support (ΔAIC < 2), the solution with the lowest dSST + dAT was retained. Nest temperature is also affected by metabolic heating, which is the increase of temperature within the nest compared with adjacent sand temperature resulting from metabolic reactions within the nest. The parameter γ is the maximum difference of temperature between sand and nest at the end of incubation, and d_i is the proportion of incubation time for the nest (0 is for day 0, and 1 is for the end of incubation). Random effect was the nest identity, which includes its position on the beach, and a temporal autocorrelation structure of order 1 was used (AR1 model). Average daily amplitudes of temperatures (AmpT) have been calculated for all days and all nests using daily maximum [from noon to noon] − daily minimum [from 12 to 12] (Eccel 2010a).

Using the fitted μ, α, β, and γ values and the daily amplitude of temperatures, it was possible to reconstruct the missing temperatures following a similar method used by Eccel (2010b) available in the R packages HelpersMG (Girondot 2017). In short, time series of temperatures are modeled using series of sinus functions in which a maximum at day i is observed for (x = T_max, y = NT_j,i + AmpT/2) and the minimum for (x = T_min, y = NT_j,i − AmpT/2). Only time sequences with less than 5 d missing were retained to limit the influence of the reconstructed time series of temperatures on the final results. As a consequence, data from 3 nests from Libya were discarded.

Growth Rate Thermal Reaction Norm.

The model of embryo growth integrates in a single framework both the growth rate dependency on temperature and the embryo growth based on growth rate (Girondot and Kaska 2014a). The growth rate dependency on temperature and the embryo growth model were fitted using maximum likelihood to best describe the observed hatchling size according to time series of temperature within nests (Girondot and Kaska 2014a). This method is summarized briefly, and some changes to the original method are emphasized.

Biological temperature-dependent rate models based on Arrhenius's and Eyring's equations were formulated by Sharpe and DeMichele (1977). The original formulation of Sharpe and DeMichele was modified by Schoolfield et al. (1981) to remove the very high correlations of parameter estimators:

where r(T) is the mean development rate at temperature T(time⁻¹), T is temperature in K (298 K = 24.85°C), R is the universal gas constant (J K⁻¹ mol⁻¹) (the original model defined R in cal K⁻¹ mol⁻¹, but this has been converted to SI units), ρ_{(298 K)} is the development rate at 24.85°C assuming no enzyme inactivation (time⁻¹), is the enthalpy of activation of the reaction that is catalyzed by the enzymes (J mol⁻¹), is the temperature in K at which the enzymes are ½ active and ½ low-temperature inactive, ΔH_L is the change in enthalpy associated with low-temperature inactivation of the enzymes (J mol⁻¹), is the temperature in K at which the enzymes are ½ active and ½ high-temperature inactive, and ΔH_H is the change in enthalpy associated with high-temperature inactivation of the enzymes (J mol⁻¹). To ensure that , a new variable ΔT has been set up with . Thus, the fitted variables were , ΔT, ΔH_H, ΔH_L, , and ρ_{(298 K)}.

This model can be simplified taking into account only 4 parameters (Schoolfield et al. 1981):

For equation 2, the fitted variables were , ΔH_H, , and ρ_{(298 K)}.

Change of Embryo Size with Time.

The growth of embryos was modeled using a modification of the Gompertz model proposed by Laird (1964):

where X(0) is the size or mass at nesting time (time = 0), r(T) is the growth rate at the beginning of the curve, and K is the carrying capacity with X(t)  = K. Note that hatching occurs generally before the time when the embryo reaches a size, or mass K. The K parameter can be viewed here simply as a way to slow down the growth at the end of incubation (Girondot and Kaska 2014a).

The dynamic of X(t) is governed by the Gompertz differential equation:

At oviposition, the gastrula is approximately a disk 1.7 mm in diameter, and this size was used as X(0) (Kaska and Downie 1999).

Parameter Fitting.

r(T) can be calculated with the 4 or 6 parameters of the Schoolfield et al. (1981) model and an incubation temperature T. Knowing X(0), K, and a time series of r(T), the pattern of change of size for this nest was evaluated using the Runge-Kutta method of order 4 for the approximation of solutions of ordinary differential equations. The mean straight carapace length (SCL) at hatching for the turtles in Mediterranean loggerhead nests is 39.33 mm (SD = 1.92 mm) (Girondot and Kaska 2014b).

Estimation of parameters was performed using maximum likelihood with a Gaussian distribution of SCL and an identity link. A matrix of variances–covariances for fitted parameters was estimated using the inverse of the Hessian matrix (matrix of second-order derivatives for all couples of parameters), and a confidence interval of r(T) was estimated using the delta method (Bates and Watts 1988).

Comparison of Libyan and Turkish Populations.

First, r(T) was fitted separately to data from nests from the Libyan and Turkish populations, and second, r(T) was fitted to the data grouped in a single data set. We used the AIC and the Akaike weight to test whether a single model for both populations was sufficient to describe the data. The Akaike weight gives the relative statistical support of several models tested on the same data set (Burnham and Anderson 2002).

RESULTS

Nest Temperature Reconstruction. — To reconstruct missing parts of nest temperature from Libya, daily average temperatures were modeled using a generalized linear mixed model (Girondot and Kaska 2014b). The best AIC was for dSST = 3 d and dAT = 0 d (AIC = −634.93). SST, AT, and metabolic heating were all significant factors (α = 0.19, p = 2.26 × 10⁻⁵; β = −0.024, p = 2.53 × 10⁻⁵; and γ = 2.05, p = 5.31 × 10⁻¹⁶, respectively) and were used to predict daily average temperature during the missing incubation period.

A sinusoidal function was applied to generate periodic hourly-based temperature time series. The amplitude of daily thermal fluctuations was low with little variability (mean, 0.34°C ± 0.17°C SD), the lowest and highest temperatures being observed at T_min = 11 hrs 28 min UTC (SD = 2 hrs 7 min) and T_max = 23 hrs 52 min UTC (SD = 1 hr 54 min), respectively (local time is UTC − 5 hrs). Time series with missing temperatures were completed by the modeled ones.

Growth Rate Dependent on Temperature.

Temperatures recorded within nests were much lower in Turkey (mean, 29.18 ± 2.18 SD; range, 22.7–34.1) than in Libya (mean, 29.98 ± 0.93 SD; range, 27.7–32.6). The range of temperatures for each location is defined as the informative temperatures (Girondot and Kaska 2014a); it means that only the portion of r(T) within this range should be discussed. The temperatures recorded in Sirte in 2009 were not exceptionally high, and the ones recorded in Dalyan in 2010 were not exceptionally low based on the air data temperatures and sea surface data temperatures recorded for the past 35 yrs (Dee et al. 2011) (Fig. 3).

Two equations relating temperature and embryo growth rate (equations 1 and 2) were compared using the AIC for the Libyan data. The simplest form (equation 2) of the Schoolfield et al. (1981) model was strongly supported (AIC = 67.48; Akaike weight = 0.81) compared with the 6-parameter equation (equation 1; AIC = 70.49; Akaike weight = 0.18). The same result was obtained for the Turkish population (Girondot and Kaska 2014a). Only the 4-parameter equation (equation 2) will be used further in this work.

To test whether a difference between the responses of embryos from Libya and Turkey to temperature (Fig. 4) could be detected, all the data were grouped and fitted with a common 4-parameter model (Fig. 5). Then this model with a single set of parameters for both locations (AIC = 151.3) was compared with the model with 1 set of parameters for each location (AIC = 157.8). The probability that a single set of parameters is sufficient to explain the combined data was highly supported (Akaike weight = 0.96) (Fig. 6). The growth rate reached its maximum at 31.5°C, and the growth rate at 20°C was only one-fifth of the one observed at 31.5°C.

View Figure

• Download as Powerpoint
• Download Figure

View Figure

• Download as Powerpoint
• Download Figure

View Figure

• Download as Powerpoint
• Download Figure

DISCUSSION

Nest temperatures for nests incubated in Libya were modeled to fill missing temperatures at the beginning of the incubation period. As previously demonstrated for nests incubated in Turkey (Girondot and Kaska 2014b), the relative contribution of AT and SST to explain nest temperature differs both quantitatively and temporally. Whereas the effect of SST lagged by 3 d, the effect of AT was not delayed in Libya. These results differ from those obtained in Turkey, where the effect of SST lagged by 1 d and the effect of AT lagged by 2 d (Girondot and Kaska 2014b). However, because solar irradiation is moderated by reflectance, albedo, thermal capacitance, and thermal conductivity, this difference could be the result of beach-specific physical characteristics. In addition, we found that the relative contribution of SST in front of the beach compared with AT was higher in explaining nest temperature. This finding is logical because water is a better heat conductor than air (Girondot and Kaska 2014b). Also, metabolic heating within the nest was significant: nests were 2.05°C warmer than the surrounding sand at the end of incubation. This result is consistent with previous studies on metabolic heating for sea turtles that range from 0.07°C to 5°C (Booth and Astill 2001; Broderick et al. 2001; Godley et al. 2001; Zbinden et al. 2006). Daily patterns of thermal fluctuations were reconstructed to represent as best as possible the time series of nest temperatures. Because nests of loggerhead turtles are incubated at a 45-cm depth (depth of the center of the nest), it was not surprising that the amplitude of thermal fluctuation was low. Thus, we found that extreme daily temperatures are observed in the late morning and near midnight (for minimum and maximum, respectively). The times when extreme daily temperatures occur during a day vary depending of latitude and season according to earth orbital parameters (Eccel 2010a, 2010b). However, setting extreme temperatures at the same moment during a day is not a strong hypothesis for nests incubated at low latitudes and when analysis is performed for a few months (the incubation period of monitored nests ranged from the beginning of July to the end of September).

Temperature during incubation of ectothermic animals can have profound consequences on the fitness of individuals, and selection acts to adapt the response of embryos to temperature (Angilletta et al. 2003). Here we study 2 genetically close populations that are separated by at most only 12,000 yrs but that experience very different thermal nesting habitats. We show that we cannot detect difference in the response of growth of embryos to temperature for both these populations, whereas temperatures within nests were different. In both populations, the growth rate reached its maximum at 32°C (Figs. 4 and 5) or 31.5°C (Fig. 6) and decreased below and above, but this effect is masked by a large confidence interval. Particularly, we show that nests incubated at very high temperatures have longer incubation duration than those incubated at 31.5°–32°C (the optimum temperature). Our results raise several questions about the adaptability of this population to change in temperature in the context of global change. We did not detect a difference between the 2 Mediterranean Sea populations separated for less than 12,000 yrs but nesting in very different thermal habitats. These populations have not had enough time to adapt to the different thermal contexts. It should be noted also that the high longevity of marine turtles hampers rapid adaptation (Tucek et al. 2014).

Negative effects of climate change could be mitigated by shift of nesting phenology to fit when a suitable thermal environment for the development of embryos is available. Such a phenotypic plasticity in the timing of nesting has been reported to be another temperature-dependent life history trait for sea turtles: several studies have shown that warmer sea surface temperatures at the foraging and breeding sites result in earlier nesting events for the loggerhead sea turtle (Weishampel et al. 2004, 2010; Mazaris et al. 2009; Ben Hassine et al. 2011; Lamont and Fujisaki 2014) and the green sea turtle (Weishampel et al. 2010). On the other hand, a delay in the first nesting events related to warmer sea surface temperatures at the foraging and breeding sites has been reported for the leatherback sea turtle (Neeman et al. 2015) and the green sea turtle (Dalleau et al. 2012).The effect of temperature on the temporal parameters of nesting phenology could be indirectly mediated by the availability of resources at the foraging sites. If warmer conditions are associated with fewer resources being available, females will spend more time accumulating reserves before starting migration to breeding and nesting sites. Therefore, first nesting events will be delayed. Temperature also affects the rate of biological processes such as oogenesis, vitellogenesis, and metabolism. Consequently, higher temperatures will promote a departure of females from the foraging sites earlier in the season and then an earlier first nesting event. In such a situation, there are no rules, and each population will react in different ways, depending on the thermal conditions at the foraging, breeding, and nesting sites and the distance between them.

Marine turtles exhibit all temperature-dependent sex determination with eggs incubated at lower temperatures, producing more males, and eggs incubated at higher temperatures, producing more females (Hulin et al. 2009). It has been proposed that the larger the range of temperatures producing both sexes (TRT), the more the population can adapt sex ratio to a climate change (Hulin et al. 2009). The TRT has been studied in 8 populations of C. caretta, and the population of Greece exhibits the second-lowest TRT after the South African population (Hulin et al. 2009), perhaps as a result of the founder effect when this population colonized the Mediterranean Sea (Clusa et al. 2013). The nearby population nesting in the Kuriat Islands (Tunisia) has low protein diversity (Chaieb et al. 2010). It should be noted that individuals from Atlantic origin still enter the Mediterranean Sea, but they seem not to participate to reproduction (Carreras et al. 2011). Thus, Mediterranean populations of C. caretta may have difficulties adapting rapidly to climate change and its effect on the sex ratio. It would be very interesting to predict the primary sex ratio for both the Turkish and the Libyan populations studied here, but the sex ratio thermal reaction norm for Libyan population is unknown. Thus, we are forced to hypothesize that Turkish and Libyan populations have the same sex ratio thermal reaction norm, and the test will not be conclusive because of the circularity in hypothesis testing.

This temperature-dependent temporal adjustment of nesting leads to 3 scenarios: embryos will be exposed to 1) warmer conditions, 2) colder conditions, or 3) the same thermal environment during incubation. In the first situation, negative effects of climate change on the development of embryos are exacerbated, leading to a low survival and a female-biased hatchling sex ratio. In the second situation, negative effects of climate change are mitigated because embryos are not exposed to lethal temperatures and a more balanced sex ratio is probably produced. In the last scenario, negative effects of climate change are neutralized. Consequently, this system, with 4 temperature-dependent life history traits (embryonic survival, sex ratio, incubation duration, and nesting phenology), should be correctly parametrized to quantify the effect of climate change on sea turtle populations, but further work is needed to obtain reliable parameters.