Global Change Biology (2015) 21, 2980–2988, doi: 10.1111/gcb.12918

Climate change overruns resilience conferred by temperature-dependent sex determination in sea turtles and threatens their survival  TOMILLO1,2, MERITXELL GENOVART1, FRANK V. PALADINO2,3, P I L A R S A N T I D R I AN J A M E S R . S P O T I L A 2 , 4 and D A N I E L O R O 1 1 Population Ecology Group, Institut Mediterrani d’ Estudis Avancßats, IMEDEA (CSIC-UIB), Miquel Marques, 21, 07190 Esporles, Mallorca, Spain, 2The Leatherback Trust, Goldring-Gund Marine Biology Station, Playa Grande, Costa Rica, 3 Department of Biology, Indiana-Purdue University, Fort Wayne, 46805 IN, USA, 4Department of Biodiversity, Earth and Environmental Science, Drexel University, Philadelphia, 19104 PA, USA

Abstract Temperature-dependent sex determination (TSD) is the predominant form of environmental sex determination (ESD) in reptiles, but the adaptive significance of TSD in this group remains unclear. Additionally, the viability of species with TSD may be compromised as climate gets warmer. We simulated population responses in a turtle with TSD to increasing nest temperatures and compared the results to those of a virtual population with genotypic sex determination (GSD) and fixed sex ratios. Then, we assessed the effectiveness of TSD as a mechanism to maintain populations under climate change scenarios. TSD populations were more resilient to increased nest temperatures and mitigated the negative effects of high temperatures by increasing production of female offspring and therefore, future fecundity. That buffered the negative effect of temperature on the population growth. TSD provides an evolutionary advantage to sea turtles. However, this mechanism was only effective over a range of temperatures and will become inefficient as temperatures rise to levels projected by current climate change models. Projected global warming threatens survival of sea turtles, and the IPCC high gas concentration scenario may result in extirpation of the studied population in 50 years. Keywords: adaptive significance, Dermochelys coriacea, global warming, genotypic sex determination, leatherback turtles, sex ratios, temperature-dependent sex determination Received 9 December 2014; revised version received 6 February 2015 and accepted 18 February 2015

Introduction Prevailing evolutionary theory developed by Charnov & Bull (1977) suggests that environmental sex determination (ESD) is beneficial over genotypic sex determination (GSD) when environments are patchy, and there is differential fitness between the sexes. The Charnov– Bull model has explained the adaptive significance of ESD in numerous invertebrate taxa (Blackmore & Charnov, 1989; McCabe & Dun, 1997) but seldom in vertebrates (Conover, 1984; Warner & Shine, 2008). In reptile species, temperature-dependent sex determination (TSD) is the prevalent form of ESD and predominates over GSD (Janzen & Krenz, 2004). However, the adaptive significance of TSD in most reptile groups remains unclear (Shine, 1999; Janzen & Phillips, 2006). There are different patterns by which temperature determines sex, and variability is also found in the Correspondence: Pilar Santidri an Tomillo, tel. +34 971 611756, fax +34 971 611761, e-mail: [email protected]

2980

transitional range of temperatures that produces both sexes (TR) and the pivotal or threshold temperature (PT) that results in 1 : 1 sex ratio (Lang & Andrews, 1994; Pieau, 1996). Increasing temperatures over the population-specific PT increases the production of female offspring (type Ia) in sea turtles (Standora & Spotila, 1985) and male offspring (type Ib) in tuataras (Mitchell et al., 2010) and some crocodilians (Deeming & Ferguson, 1988). In some species of lizards, turtles and crocodiles, high and low temperatures produce females, whereas male offspring are produced at intermediate temperatures (type II) (Valenzuela, 2004). Temperature-dependent sex determination and GSD have evolved several times in vertebrate species (Janzen & Krenz, 2004), have been gained and lost in particular groups (Valenzuela, 2004) and can co-occur in the same species (Pen et al., 2010). As TSD was first identified in a reptile in the 1960s (Charnier, 1966), numerous studies have proposed possible adaptive significance in this group, such as phylogenetic inertia, group adaptation, inbreeding avoidance and/or differential fitness © 2015 John Wiley & Sons Ltd

C L I M A T E C H A N G E O V E R R U N S T E M P E R A T U R E R E S I L I E N C E 2981 (Shine, 1999; Janzen & Phillips, 2006). However, only one study has demonstrated conclusively an adaptive significance of TSD in a short-lived lizard based on the differential fitness in lifetime reproductive success between the sexes (Warner & Shine, 2008). Thus, nearly 50 years after temperature was revealed as a sex-determining mechanism in reptiles, its adaptive significance in the vast majority of groups remains an evolutionary enigma. Increased incubation temperatures associated with the projected impacts of climate change will likely exacerbate biased sex ratios or produce single-sex offspring  in species with TSD (Ospina-Alvarez & Piferrer, 2008; Mitchell et al., 2010). In sea turtles, primary sex ratios are often female-biased (Standora & Spotila, 1985; Godfrey et al., 1996; Broderick et al., 2001) and sex ratios over 90% female have been reported at some locations (Godfrey et al., 1999; Godley et al., 2001; Sieg et al., 2011). Although there are differences among species and populations, the width of the TRT in sea turtles is only about 1 °C, and in most populations, the PT ranges between 29 and 30 °C (Chevalier et al., 1999; Wibbels, 2003). Climate change projections show that air temperatures by the end of the 21st century will likely be 1–4 °C warmer than today. Thus, hatchling production at many nesting sites could become 100% female by the end of the century. Rising temperatures can also reduce hatchling production in sea turtles due to the effect of high temperatures on egg and hatchling survival (Fuentes et al., 2010; Santidri an Tomillo et al., 2012). Recently, Lalo€e et al. (2014) assessed the viability of a sea turtle population under climate change scenarios and predicted an increase in the number of nesting females. However, their model only allowed for changes in sex ratios due to temperature and did not consider the associate effect on egg and hatchling survival. We have developed an alternative theoretical framework to explain the adaptive significance of TSD in a reptile species with temperature-dependent survival of early stages, the leatherback turtle. In the eastern Pacific leatherback turtles that nest in Costa Rica, incubation temperatures >30 °C increase mortality of eggs and hatchlings (Santidri an Tomillo et al., 2014) and result in 100% female ratio (Binckley et al., 1998). We hypothesized that an increasing production of female hatchlings with temperature would be evolutionarily advantageous to sea turtle populations because it helps to compensate for the increased mortality of eggs and hatchlings due to rising temperatures. Our main aim was to demonstrate that TSD is an effective mechanism to counteract the increased mortality of early life stages under warming conditions and to test the effectiveness of this mechanism under climate change scenarios. © 2015 John Wiley & Sons Ltd, Global Change Biology, 21, 2980–2988

Materials and methods We reconstructed nest temperatures between 1920 and 2012 based on air temperatures to assess fluctuations in nest temperatures over time. Then, we built a stage-based stochastic population model and ran simulations to assess population responses in mean population growth rate (ks) and number of nesting females to increasing nest temperatures and compared the results to those of a virtual population with fixed sex ratios as if it had GSD. We used number of nesting females for visual representation of population trend because information on other life stages is scarce, and population trends in sea turtle studies are often based on adult females identified on the nesting beaches. Based on reconstructed temperatures, we considered two scenarios that kept populations stable with initial mean nest temperatures as current means (30.4 °C) or historical means (29.9 °C). Then, we used the stage-based stochastic population model to assess the impact of anthropogenic climate change on the TSD and GSD populations and assessed the effect of increasing temperatures to levels projected under the low, medium and high gas concentration scenarios.

Reconstruction of nest temperatures We reconstructed nest temperatures for leatherback turtle clutches between 1920 and 2012 using the regression equation that best explained the relationship between nest and air temperatures that we identified at Playa Grande (10°20 N, 85°51 W), Costa Rica over nine nesting seasons (2004–2005 to 2012–2013, n = 926 clutches). We compared the mean temperature in the nest (Tnest) during the incubation period (ca. 60 days) to the mean air temperature (Tair) of the month the clutch was laid and the following month based on the average incubation period. We used the equation: Tnest = 2.119 + 1.209Tair, (R2 = 0.64, P < 0.001). For the reconstruction of past nest temperatures, we used monthly air temperatures since 1976 from the Daniel Oduber International airport at Liberia (~50 km from the site). For years between 1920 and 1976, we obtained monthly air temperatures for the area between 9–11°N and 85–87°W from the Comprehensive Ocean Atmosphere Data Sets (COADS) (Woodruff et al., 2011).

Sex ratios and emergence success We estimated monthly emergence success and sex ratios based on the mean monthly temperature randomly selected by the model. These temperatures correspond to mean temperatures during the middle trimester of incubation, which is the thermosensitive period that determines sex in sea turtles (Binckley et al., 1998). We used emergence success (percentage of eggs in a clutch that results in hatchlings that emerge from the nest) as a measure of early mortality because it combines hatching success (percentage of eggs that hatch) and emergence rate (percentage of hatchlings that emerge) and provides a more significant ecological measure of hatchling production (Wallace et al., 2007). We estimated emergence success (E) for each month (i) based on mean incubation temperature (Tnest) using the regression equation obtained from

 T O M I L L O et al. 2982 P . S A N T I D R I AN the empirical analyses on leatherback clutches over nine nesting seasons at Playa Grande (R2 = 0.752, P < 0.001); Ei ¼ 4:838 þ 0:449Tnesti  0:009ðTnesti Þ2 . To estimate sex ratios, we used the TSD curve previously described for this population (Binckley et al., 1998) in which, temperatures lower than 29.0 °C produce 100% male offspring and >30 °C, 100% female offspring. Within the transitional range of temperatures (TR: 29–30 °C), sex ratios were calculated as percentage female by 0.1 °C increments (Binckley et al., 1998).

Population model We built a stage-based stochastic population model, similar to models previously used in sea turtles (Crowder et al., 1994) with projection time interval of 1 year. We used the model for our subject population with TSD and a virtual GSD population with fixed sex ratios at 1 : 1. A postbreeding life cycle of the leatherback turtle was structured by defining four stages: stage one from the time hatchlings reach the water after leaving the nest to age 1 year; stage two for juveniles between ages one and 3 years; stage three for subadults between age 3 years and maturity, and stage four for adults that are sexually mature. We considered annual survival rates as s1 = 0.0625, s2 = 0.435 and 0.579 for TSD and GSD populations, respectively, s3 = 0.893 and s4 = 0.893; s1 was obtained from Spotila et al. (1996), s2 was calculated as the value needed to maintain the population stable, which was slightly different for the TSD and GSD simulated populations, we considered stage three (s3) to have the same survival rates as stage four (s4) because of the large body size attained in this stage (Jones et al., 2011). Survival rates of adults (s4) were based on survival rates estimated for leatherback turtles nesting at St. Croix, US Virgin Islands (Dutton et al., 2005). For each stage in the population matrix, we calculated the probability of surviving and remaining in the same stage (P) and the probability of surviving and passing to the next stage (G) (Crowder et al., 1994). A mean population growth rate (ks) was estimated when running the population matrix models. Stage four was the only reproductive stage, and fecundity (F) was set at an average value of 409 eggs in a season (62 eggs per clutch, 6.6 clutches). The probability of a female reproducing was 0.27, because leatherback turtles in this population nest every 3.7 years (Reina et al., 2002). Emergence success was a function of nest temperature in all simulations, but sex ratios were dependent on temperature only in the simulations for the TSD population. We fixed sex ratios at 1 : 1 in the virtual GSD population. Mean (SD) nest temperatures based on field data for October, November, December and January were 29.5  0.9, 30.2  0.9, 30.7  1.0 and 31.3  0.8 °C, respectively. We simulated environmental stochasticity by randomly selecting one within 1000 random values of temperature within the SD limits for each month. As there was a positive correlation in nest temperature between consecutive months, we applied a correlation coefficient identified from the empirical data (r = 0.514) to the temperature selected for October to calculate mean temperatures for the following months. We estimated mean emergence success and sex ratios per month based on

those temperatures. The monthly distribution of nests in the season was set at average values of 13%, 27%, 35% and 25% for clutches laid in October, November, December and January, respectively, based on field data (Santidri an Tomillo et al., 2012). The probability of hatchlings surviving on the walk from the nest to the water was 0.87 (Santidri an Tomillo et al., 2010). We simulated environmental stochasticity in some parameters that depended on temperature but did not include stochasticity in other vital rates that were not directly affected by temperature. This allowed us to focus on the effect of temperature on the population to determine the effectiveness of TSD as a mechanism to maintain sea turtle populations. This is a conservative approach as including stochasticity in other vital rates would increase extinction risk.

Simulations We used R version 3.0.1 (R Core Team, 2013) and the popbio library (Stubben & Milligan, 2007) to run all simulations.

Scenario 1. We considered two initial populations (TSD and GSD) stable under current conditions of nest temperatures (30.4 °C) with 500 nesting females per year, corresponding to a mid-size nesting population. We ran 11 simulations for each population experiencing, respectively, mean temperatures as: current means, +0.1, +0.2, +0.3, +0.4, +0.5, +0.6, +0.7, +0.8, +0.9 and +1.0 °C. We ran 1500 trajectories for each simulation over 200 years to test the population responses in mean population growth rate (ks) and number of nesting females. Each simulation corresponded to an increase in mean temperature of 0.1 °C with recreated stochasticity within the SD limits. We considered the same SD in all cases. Scenario 2. Because current conditions may not keep populations stable even in the absence of other anthropogenic pressures, we considered a second scenario under which initial conditions in nest temperatures were 0.5 °C cooler than current values (mean = 29.9 °C), based on our reconstructions of nest temperatures, as well as differences in mean temperature reconstructed for other sea turtle populations over 100 years (Hays et al., 2003). We repeated the aforementioned simulations considering that the initial conditions that kept the population stable were 0.5 °C cooler than current conditions. Survival of juveniles to keep the population stable under these conditions were s2 = 0.436 and 0.551 for TSD and GSD, respectively. Climate change scenarios. We used the same population model to run simulations under future scenarios of climate change. We used the four representative concentration pathways (RCPs) corresponding to low (RCP2.6), medium (RCP4.5 and RCP6.0) and high (RCP8.5) gas concentration scenarios from the Fifth Assessment Report of the Intergovernmental Panel on Climate Change (IPCC, 2013) that corresponded to rises in temperatures over current means as +1, +2 and +3.7 °C, respectively, by year 2100. We also ran 1500 trajectories for each simulation over 200 years and tested the population responses in mean population growth rate (ks) and © 2015 John Wiley & Sons Ltd, Global Change Biology, 21, 2980–2988

C L I M A T E C H A N G E O V E R R U N S T E M P E R A T U R E R E S I L I E N C E 2983 number of nesting females. We tested the effects of increased temperatures on the TSD and GSD populations that were kept stable under the scenarios of stability (scenarios 1 and 2) previously mentioned.

Results Maximum and minimum mean nest temperatures based on 10-year moving averages ranged by ~1 °C (Fig. 1). Minimum mean temperatures were reconstructed for years before the 1930s and were ~0.5 °C cooler than current conditions. There were two main warming events in the 1930s–1940s and the 1970s– 1980s, coinciding with observed trends in global air temperatures (Fig. 1). Temperature-dependent sex determination populations were more resilient to increased nest temperature than populations with fixed sex ratios (Figs 2 and 3). TSD populations mitigated the negative effect of high temperatures by increasing production of female offspring and therefore, future fecundity. Under the scenario of initial mean temperatures at 30.4 °C, ks declined in both TSD and GSD populations as the mean temperature increased, but ks was always lowest in the GSD population (Fig. 2a). However, ks was still lower than 1 under most temperatures. Under the scenario of initial mean temperatures at 29.9 °C in the GSD population ks gradually declined as mean temperature increased. However, ks in the TSD population first increased, remaining close to 1 and only declined at

temperatures above 0.7 °C warmer than the initial conditions (Fig. 2b). Likewise, in the TSD population with initial temperatures at 29.9 °C, the number of nesting females first increased with increasing mean temperatures, while numbers declined in the other three populations (Fig. 3). Number of nesting females after 200 years was always lowest in the GSD population that was kept stable under current temperatures, followed by the GSD population that was stable under historical means, the TSD population that was stable under current values and finally, the TSD population that was stable under historical means (Fig. 3). In the latter case, the number of females grew after 200 years for simulations up to +0.4 °C, remained stable for simulations +0.5–+0.7 °C and started declining at +0.8 °C (Fig. 3). Number of turtles declined in all other populations and was always lower in the GSD populations than in the TSD ones. Primary sex ratios became more female-biased and emergence success (percentage of eggs in a clutch that result in hatchlings that emerge from the nest) was reduced as temperature increased to +1.0 °C from the scenarios of stability under current and lower mean temperatures (30.4 and 29.9 °C, respectively) (Fig. 4). Sex ratios in the TSD populations increased from ~0.86 to ~0.96 (for mean temperatures of 30.4 and 31.4 °C, respectively) and from ~0.76 to ~0.92 (for mean temperatures of 29.9 and 30.9 °C, respectively). Mean emergence success decreased from 0.46 to 0.35 (mean temperatures of 30.4 and 31.4 °C, respectively) and

(a)

(b)

(c)

(d)

Fig. 1 Reconstructed nest temperatures for leatherback turtle clutches at Playa Grande, Costa Rica. Thick black line represents 10-year moving averages over reconstructed nest temperatures between 1920 and 2012. Nest temperatures correspond to clutches that were laid in (a) October, (b) November, (c) December and (d) January. © 2015 John Wiley & Sons Ltd, Global Change Biology, 21, 2980–2988

 T O M I L L O et al. 2984 P . S A N T I D R I AN (a)

(b)

from 0.51 to 0.41 (mean temperatures of 29.9 and 30.9 °C, respectively). Temperature-dependent sex determination only conferred resilience within a range of temperatures and became ineffective under all climate change scenarios resulting in population declines (Figs 2c and 5). Under the low gas concentration scenario, there was a greater negative effect on the GSD populations than in the TSD ones (Fig. 5). However, the differences disappeared as temperatures increased under the medium and high gas concentration scenarios. There was a similar effect of temperatures on the TSD and GSD populations under the high gas concentration scenario and all populations were extirpated in approximately 50 years (Fig. 5).

Discussion

Evolutionary significance of TSD in sea turtles

(c)

Fig. 2 Population response to increased mean nest temperature for two virtual populations of leatherback turtles, one with temperature-dependent sex determination and one with fixed sex ratios. Initial mean temperatures that kept population stable were (a) 30.4 °C corresponding to current mean values and (b) 29.9 °C corresponding to historical means, representing most likely conditions at the beginning of the 20th century. Each simulation had a mean temperature 0.1 °C higher than the previous one, starting from the conditions that kept the population stable. Panel (c) shows population response to climate change in the aforementioned populations with the conditions of stability of panels a and b, corresponding to low, medium and high gas concentration scenarios. Values in all panels corresponded to the mean population growth rate (ks) over 200 years. Stochastic simulations were done for 1500 population trajectories. Variances are not shown for facilitating clarity and interpretation.

Temperature-dependent sex determination confers resilience to high temperatures in sea turtles. When high temperatures increase mortality of early stages, sea turtles increase production of female offspring, increasing future fecundity and mitigating the negative effect of high temperatures on the population growth. We have showed the adaptive value of linking sex determination to temperature in a species with temperature-dependent survival of eggs and hatchlings. This constitutes an alternative but complementary theory to the Charnov–Bull model that contributes to unravel the evolutionary significance of TSD in vertebrates and why it persists in so many lineages today. Because TSD type II occurs in the three reptile groups that have TSD (turtles, lizards and crocodilians), Pieau (1996) suggested that TSD type II may be the ancestral form of TSD in reptiles, from which types Ia and Ib originated. As viable development can only occur over a range of temperatures (Birchard, 2004), TSD type II seems the most plausible mechanism in species with temperature-dependent survival in which females are always produced at high and low temperatures as egg mortality increases towards the extremes. The relationship between mortality and sex ratios may explain why females, and not males, are produced at the extreme temperatures in species with TSD type II. For instance, in Australian Agamidae (type II), embryo mortality increases above the high and under the low 100% female-producing temperatures (Harlow, 2004). In TSD types Ia and Ib, mortality and sex ratios may be linked at only one of the extremes of the viable range of temperatures, because the high or low temperatures that would increase mortality at the other end of the viable range are not experienced in the natural habitat. In the leatherback turtles of our subject population, low © 2015 John Wiley & Sons Ltd, Global Change Biology, 21, 2980–2988

C L I M A T E C H A N G E O V E R R U N S T E M P E R A T U R E R E S I L I E N C E 2985

Fig. 3 Stochastic projections of number of nesting females for TSD and GSD populations responding to increasing mean nest temperatures. Initial conditions kept population stable at (a) current mean temperatures (30.4 °C) and (b) temperatures 0.5 °C lower than current conditions (29.9 °C), representing most likely conditions at the beginning of the 20th century. Each simulation had a mean temperature 0.1 °C higher than the previous one, starting from the conditions that kept the population stable. Lines represent mean values for 1500 population trajectories over 200 years.

nest temperatures (

Climate change overruns resilience conferred by temperature-dependent sex determination in sea turtles and threatens their survival.

Temperature-dependent sex determination (TSD) is the predominant form of environmental sex determination (ESD) in reptiles, but the adaptive significa...
618KB Sizes 0 Downloads 6 Views