Tromelin Island: Influences on Chelonia mydas Incubation Temperature and Reproductive Traits in Light of Climate Change

Transcription

1 Mémoire de fin d études Pour l obtention du Diplôme d Agronomie Approfondie (DAA) Spécialisation Halieutique, Dominante Ressources et Ecosystèmes Aquatiques Tromelin Island: Influences on Chelonia mydas Incubation Temperature and Reproductive Traits in Light of Climate Change Présenté par : Théa JACOB Soutenu le : 11/09/2009

2 Mémoire de fin d études Pour l obtention du Diplôme d Agronomie Approfondie (DAA) Spécialisation Halieutique, Dominante Ressources et Ecosystèmes Aquatiques Tromelin Island: Influences on Chelonia mydas Incubation Temperature and Reproductive Traits in Light of Climate Change Présenté par : Théa JACOB Soutenu le : 11/09/2009 Devant le Jury M. BOURJEA Jérôme, Ifremer. M. CICCIONE Stéphane, Kélonia. M. LE GALL Jean-Yves. M. LE PAPE Olivier, Agrocampus Ouest. M. SABATIE Richard, Agrocampus Ouest.

3 Acknowledgements I really wish to thank Stephane Ciccione (Kelonia) for giving me the opportunity to undertake this research project and experience life on Tromelin Island, for his availability all along this internship and for his precious advice. Thanks also to Jerome Bourjea (Ifremer) for his great input and friendly support throughout the project. Without supervisors such as Stephane and Jerome, this project would never have come to be. Thank you also to my teachers, Olivier Le Pape and Richard Sabatié, for their availability throughout the entire process. Special thanks to Etienne Rivot, for his patience, support, and for sharing his love of statistics with so much faith. Thanks to Jean-Yves Le Gall, one of the first turtle investigators in Tromelin, for agreeing to be an assessor of my work. My thanks to Meteo-France, TAAF and the French Air Army for allowing me to work on Tromelin. Thanks again to Meteo France for the provision of environmental data, and to the Meteo France crew on Tromelin. Special thanks to some of the Tromelinese: Eugene and Jean-Michel, for their great cooking and cheerful company, and to Christophe and Francis for making life with nothing a memory to be kept fondly. Thanks to Emmanuel Cordier of Reunion Island University, for his time and for sharing his granulometry expertise with me. Thanks also to Genevieve Lebeau for providing laboratory know-how. Further thanks to Sylvaine Jegot from CIRAD, for her advice regarding the statistics in this report. A big thanks to all those at Kelonia, especially Murièle, Claire, Gerard, Emilie and Marie for all their help and providing such a good work environment. Special thanks to Alain for his friendship and for always having a smile on his face. A most gracious thank you to Eric Sida-Chetty, for his generosity and excellent nature. Also to the Sida family, especially Josiane, Jean-Luc and Sapote, as well as to Gerard and Gilles for making Reunion Island a home away from home. I of course also thanks my family and friends, for their love and the moments shared during all these student years.

4 CONTENTS 1 INTRODUCTION MATERIAL AND METHODS Study Site Field Protocol Biological Parameters Temperature Recording Climatic Factors Local Factors Analysis Statistics Sex Ratio Evaluation and Prediction RESULTS Biological Parameters and Population Trend Assessment Climatic and Local Factors Influences on Incubation Temperatures Climatic Factors Sand Temperature at Nest Depth Initial Incubation Stage Sex Determinate Stage Local Factors Mortality Sex Ratio Model Past and Future Predictions DISCUSSION Population Assessment and Evolution Incubation Temperatures: Climatic Influences and Internest Variability Mortality Predictions Climate Change Effects and Adaptation Possibility CONCLUSION AND FUTURE RESEARCH REFERENCES...31 APPENDICES...37

5 LIST OF FIGURES Figure 1 - Yearly number of tracks on Tromelin Island from 1987 to Figure 2 - Average monthly C. mydas tracks (± SD) and percentage activity in Tromelin ( )...7 Figure 3 a) Nests positions and b) daily average temperatures of control and C. mydas study nests on Tromelin, March-May Figure 4 - a) Control and air temperatures and b) control temperatures and rainfall from March to May 2009 in Tromelin...10 Figure 5 - a) Four day delayed and smoothed air temperature and control nest temperature, and b) linear trend lines of delayed and smoothed air temperature and control nest temperature in Tromelin, March-May Figure 6 - Control nest and air temperatures residuals from linear regressions (Tromelin, March-May 2009)...11 Figure 7 - Sand temperature at nest depth and C. mydas nest temperatures during first third of incubation in Tromelin, Figure 8 - Loess smoothing on a) Nest 1 temperature and b) control nest temperature during the first third of incubation of C. mydas nests in Tromelin with c) the resultant residuals...13 Figure 9 - C. mydas nest and control nest temperatures during the second third of incubation in Tromelin, Figure 10 - Loess smoothing on a) Nest 1 temperature and b) control nest temperature during second third of incubation of C. mydas nests in Tromelin with c) the resultant residuals...15 Figure 11 - a) Average first third of studied C. mydas nest temperatures plotted against longitudinal position to b) classify site position groups on Tromelin, March-May Figure 12 - Average first third temperatures of C. mydas nests on Tromelin plotted against a) site position and b) average grain size, March-May Figure 13 - Grain size and site position of C. mydas nests on Tromelin, March-May Figure 14 - Linear regression between sand temperature at nest depth and ambient temperature in Tromelin Island, March-May Figure 15 - Control nest recorded and predicted temperatures over the span of a) 12 months and b) 11 years compared to recorded data by Meteo-France at depths of 0.5 and 1m in Tromelin...19 Figure 16 - Average monthly and yearly temperatures on Tromelin Island ( )...20 Figure 17 - Predicted percentage of females produced by C. mydas clutches within each month over the last 4 decades ( ) in 5 year averages on Tromelin Island 21 Figure 18 - Average predicted a) C. mydas nest temperatures and b) percentages of produced C. mydas females and nesting activity for the years , and 2080 on Tromelin...22

6 LIST OF TABLES Table 1 - C. mydas biological parameters and juvenile production values for Tromelin Island...7 Table 2 A comparison of the observed and calculated average abundances of individual C. mydas during different periods on Tromelin...8 Table 3 - Pyper and Peterman results performed on residuals of control nest temperature and smoothed air temperature (delayed by 0 to 5 days) (Tromelin, March-May 2009)...12 Table 4 - Average monthly temperatures measured at nest depth in Tromelin, March-May 2009, compared to predicted temperatures at nest depth...19 Table 5 Average temperature during sex determination stage and its sex ratio outcome as recorded and predicted for C. mydas nests studied on Tromelin Island, March-May Table 6 - Comparison of average C. mydas sex ratios during Le Galls (1988) and the author's (March-May 2009) studies in Tromelin...21 Table 7 - Percentage of C. mydas females produced within clutches in Tromelin using the sex ratio prediction model and future climate scenarios...22 LIST OF EQUATIONS Equation 1- Calculation of ICC for juveniles (Bjorndal et al., 2000)...4 Equation 2- Sex ratio prediction model...5 Equation 3- Calculation of metabolic heating (Broderick et al., 2001)...5 Equation 4- Nesters abundance calculation...6 Equation 5 - Calculation of Produced Juveniles...8 Equation 6 - Prediction of temperature at nest depth in Tromelin using ambient Temperatures...19 Equation 7 - Prediction of nest temperatures during the second third of incubation...20

7 LIST OF APPENDICES APPENDIX A - Distribution and Life Cycle of C. mydas.. 39 APPENDIX B - Map and Aerial Picture of Tromelin..40 APPENDIX C - Vegetation Line on Tromelin s beach and Tournefortia Argentea.. 41 APPENDIX D - CCL Measurement and Tagging...42 APPENDIX E - Sex Ratio Prediction Model, adapted from Miller & Limpus, APPENDIX F - Results of Pyper and Peterman Tests on First Third Nest by Nest Residuals Vs. Corresponding Control Nest Residuals APPENDIX G - Results of Pyper and Peterman Tests on Second Third Nest by Nest Residuals Vs. Corresponding Control Nest Residuals.. 45 APPENDIX H - Nests Temperatures Summary...46 APPENDIX I Granulometry Results from Gradistat Software APPENDIX J - Results of Linear Regression between Grain Sizes ~ Site and Residuals Distribution for Model Validation...49 APPENDIX K - Results of Linear Regression from Average First Third Incubation Temperature ~ Site and Residuals Distribution for Model Validation APPENDIX L - Results of Linear Regression from Average Second Third Incubation Temperature ~ Site and Residuals Distribution for Model Validation APPENDIX M - Linear Regression Results from Metabolic Heating ~ Clutch Number and Eggs Weight and Residuals Distribution for Model Validation..52 APPENDIX N - GLM Binomial Results from Emerged/Not Emerged ~ Average Incubation Temperature APPENDIX O - Results of Linear Regression from ICC ~ Average Incubation Temperature and Residuals Distribution for Model Validation. 54 APPENDIX P - Grain Size Distribution Dynamic. 55

8 GLOSSARY CCL: Curved Carapace Length CITES: Convention on International Trade in Endangered Species of Wild Fauna and Flora CMS: Convention on Migratory Species CTE: Constant Temperature Equivalent FAO: Food and Agricultural Organisation GLM: Generalised Linear Model GPS: Global Positioning System ICC: Index Corporal Condition IPCC: Intergovernmental Panel on Climate Change IUCN: International Union for Conservation of Nature NRC: National Research Council SCL: Straight Carapace Length SD: Standard Deviation SWIO: South West Indian Ocean TAAF: Terres Australes et Antarctiques Françaises TRT: Transitional Range of Temperature TSD: Temperature-dependent Sex Determination

9 1 INTRODUCTION Chelonia mydas, the green turtle, is one of seven existing species of marine turtles. This species has been classified as globally endangered (IUCN, 2001) due to dwindling numbers initially caused by hunting, human disturbance and habitat destruction (Frazier, 1975; 2003; Jackson, 1997; Balazs & Chaloupka, 2004). The island of Tromelin is an important nesting site for green turtles in the South West Indian Ocean (SWIO) (Lauret-Stepler et al., 2007), with the beach providing one of only a few sites that remain almost completely undisturbed. As C. mydas populations are already under threat, impact of further pressures, such as climate change, needs to be investigated. Green turtles experience great differences between life stages in terms of habitat, diet, migration (Appendix A) and the role of temperature, with sexual maturity reached late in comparison to other marine turtles (Davenport, 1997). Mature C. mydas influence the structure and dynamics of their feeding habitats by continually cropping seagrasses to stimulate new leaf growth (Aragones, 2000). As little food is consumed during migration and breeding (Mortimer & Carr, 1987), new energy is input into the nesting beaches from feeding grounds, playing a vital role in maintaining nesting site ecosystems (Bouchard & Bjorndal, 2000). Juvenile turtles also provide energy to predatory creatures both on land and in the nearby reef ecosystems (Bouchard & Bjorndal, 2000). With the onset of changing global weather regimes, this important role of C. mydas within the trophic energy system comes under threat. Low worldwide populations (Jackson, 1997) of C. mydas emphasises the importance to understand the potential effects of climate change on their life cycle. Marine turtles exhibit the characteristic of Temperature-dependent Sex Determination (TSD) where sex of hatchlings is determined by temperature during the second third of incubation (Yntema & Mrosovsky, 1980; Mrosovsky & Pieau, 1991; Van de Merwe et al., 2005) rather than a genetic coding. A specific temperature during this third called pivotal temperature (Mrosovsky & Pieau, 1991), produces an even ratio of sexes (Godfrey et al., 1997) with higher or lower temperatures producing larger proportions of females and males respectively. Temperatures outside deviations from the pivotal temperature can produce a unisex clutch. This occurs outside the Transitional Range of Temperature (TRT), conservatively reported as 3 C and centred on the pivotal temperature (Mrosovsky, 1994). Exhibition of TSD denotes the importance of investigating influences on C. mydas incubation temperatures, as climatic variables may alter long-term evolved sex ratios. Characteristics of natural nesting beaches also need further attention to derive all potential influences on nest temperatures within C. mydas nests. Air temperature has been linked with sand temperatures at several nesting sites (Janzen, 1994; Hays et al., 2003; Hawkes et al., 2007), providing consensus on the gravity of climate change impacts for marine turtles. Due to a small TRT range for C. mydas (Mrosovsky, 1994), even slight changes in nest temperature can skew sex ratios to extremes. This has already been reported for the major nesting site of Ascension Island as a result of climate change, where nest temperatures have been reported as 0.49 C higher than in the mid 19th century (Hays et al., 2003). The conservation status of this species as well as their exhibition of TSD indicates that a greater knowledge of the potential impacts of climate change on C. mydas populations is required. 1

10 To investigate the impact of global warming on the nesting population of Tromelin, a three month research period (March, April, May) was undertaken on the island of Tromelin during Green turtle nest temperatures and sand temperatures at nest depth were recorded over this period. A capture recapture program was also performed and tracks on the beach were counted daily. Initially, the following report explores biological parameter findings for the Tromelin population to assess its health and trends. Climatic and nest specific factors are then investigated to uncover any influence of these on incubation temperatures of C. mydas clutches in Tromelin. These results are employed to derive prediction models of nest temperatures and sex ratios, to assess the influence of global warming on this population s reproductive traits. Thus, three main hypotheses are investigated within this report: 1. Climatic factors influence the temperature within nests during incubation, thus sex ratios of brooding clutches 2. Local factors can alter sex ratios and mortality within nests, allowing intrabeach variation of these two factors 3. Climate change will affect sex ratios of brooding clutches in Tromelin. 2

11 2 MATERIAL AND METHODS 2.1 Study Site Tromelin Island (Appendix B) is located in the South West Indian Ocean (SWIO) (15 33 S, E) and is one of five French Esparse Islands. Covering an area of 1.2km 2 with a maximum altitude of 7m, 1600m of its beach is conducive to turtle nesting, with the rest covered in boulders. The beach gives way to a line of vegetation consisting solely of Tournefortia argentea (Appendix C). The island is managed by Terres Australes et Antarctiques Françaises (TAAF) and, since 1954, has been used by Meteo-France as a weather observation point. Some manmade structures are present on the island consisting mainly of a Meteorological station. Since weather observations began, four Meteo-France employees have inhabited the island on a monthly rotational basis. Due to the low human disturbance on Tromelin, the nesting green turtle population has remained largely undisturbed by a human presence. Implemented by Ifremer, track numbers have been recorded daily since March 1986 by Meteo- France workers in cooperation with Ifremer and Kelonia. Using this data, peak nesting season was determined as between November and February, with approximately 7178 ±3053 tracks recorded on the beach annually over the last 19 years (Lauret-Stepler et al., 2007). 2.2 Field Protocol Biological parameters - Population Assessment Research was conducted on Tromelin Island over a three month period (March, April, May) in The entire laying beach was monitored from 12/03/09 to the 07/06/09 each night between 20h and 06h. Encountered turtles were tagged using metallic self-piercing Monel Tags (Style 56, National Band and Tag Co.) attached between the first and second scale on the anterior left flipper (Appendix D). Tagging was performed post-oviposition or when a turtle was returning to sea. The Curved Carapace Lengths (CCL) of these turtles were recorded as well as their activity (laid, not laid). Turtle CCL s were taken with a measuring tape (±0.5cm). CCL was measured between the anterior point at midline (nucal scute) and posterior notch between supracaudals at midline (Appendix D). Each morning the laying beach was surveyed and the activity (laid, not laid) of each track was recorded to ensure all mounts were taken into account. Data was pooled to assess the average interesting interval (the period between consecutive successful nesting attempts), average number of clutches per female ( = number of nests/individual females during study period) and the percentage laying success ( =Total No# of Nests/Total No# of Tracks). - Egg and Hatchling Parameters The number of eggs per clutch was recorded for study nests, and for 20 nests egg length and weight was gathered during oviposition. Length was taken as the distance between the two farthest points of an egg and was measured to the second decimal place using callipers. Weight was also measured to two figures of significance using a Professional Mini Pocket Scale, model EC-500 (±0.01g). Eggs were placed back into the egg chamber before oviposition had finished. 3

12 Once covering of the nest by the female was complete, a net with a diameter of 70cm was placed around its position and buried approximately 40cm deep into the sand. This was to reduce the potential destruction of clutches by predators and nesters, and to trap emerged hatchlings. Netting holes were 1cm 2 to reduce impairment of natural thermal, gas and moisture flow through the sand. Nests were checked hourly between sunset and sunrise after the 50 th day of incubation. Once emergence had occurred, 30 haphazardly chosen hatchlings were weighed, and their Straight Carapace Length (SCL) recorded using electronic scales and callipers (previously outlined). SCL was measured between the same points as CCL. Index Corporal Condition (ICC) was calculated by the following equation: Equation 1- Calculation of ICC for juveniles (Bjorndal et al., 2000). The fate of eggs and juveniles (unhatched; predated; emerged) were recorded during nest excavation, 72h after the last emergence. As emergence from the nest is dependent on group participation (Carr & Hirth, 1961; Balazs & Ross, 1974) juveniles found alive in the neck of the nests were considered as emerged but those alive in the egg chamber were considered as dead. Due to the difficulty of determining infertility in the field, emergence success was calculated as a percentage of clutch size (Total Emergence Success = (N emerged + alive in neck)/total clutch number) Temperature recording In order to monitor temperature within the egg chamber, VEMCO Minilog-T V3.09 thermometers were placed into 29 random turtle nests during oviposition. Thermometers recorded temperature hourly and were placed centrally in the nest after the 50 th egg has been droped. A thermometer was placed at 75cm deep on the beach to record sand temperature at nest depth (Hays et al., 1993) during the study period to act as a temperature control ( Control Nest ). Thermometers were removed 72 hours after the last hatchlings had emerged. The control nest was excavated after all emergences had occurred and temperature data was gathered using VEMCO Minilog-T software Climatic Factors Hourly values of air temperature, rainfall and sand temperature at both 50 and 100cm were supplied by the weather station on Tromelin, courtesy of Meteo-France Local Factors Sand was collected at the top of the egg chamber. Samples were rinsed and placed within an oven at 70 C for 48h to dry. Particle sieving was done using Retsch Siev Analysis AS 200 for 15 minutes. Grain size data was analysed using the Folk and Wald method with Gradistat software (Blott & Pye, 2001). 4

13 Nest shading was evaluated by recording the percentage of sun exposure within the constructed net (0.38m 2 ) hourly during daylight. A daily average was calculated and reduced to binominal data, where >50% was regarded as shaded and below this percentage as not shaded. The depth to the last egg excavated was recorded for all nests. The GPS point of each nest was taken. 2.3 Analysis Statistics Data was analysed using the statistical software R. Air temperature was smoothed using Tukey s running median smoothing method, to remove extreme fluctuations and limit diurnal ranges. When working on time series, Pyper and Peterman correlation test (on residuals) were performed. This method removes the incidence of autocorrelation among time series (Pyper & Peterman, 1998). Autocorrelation lag=5 was used for all tests, as it was the higher autocorrelation lag value for all time series. Residuals were taken from either linear regression or Loess smoothing, according to raw data. Loess smoothing coefficients were chosen for each data set (i.e. first third, second third of incubation) to gain the residuals from the most suited trend line. All linear models performed were validated post-analysis by looking at the normality of residual distributions. Full results and details from linear regressions as well as residual distributions can be seen in Appendices J,K,L,M & O. Local factors were tested for their influence on emergence success using a Generalised Linear Model (GLM) binomial model (Emerged/Not Emerged) Sex Ratio Evaluation and Prediction Based on a pivotal temperature of 28.8 C (Godley et al., 2002), and a TRT range of 3 C (Mrosovsky, 1994), sex ratio was predicted using the mean temperature during second thirds of incubation. Temperatures less than 27.3 C were assumed to produce 100% males and those above 30.3 C were assumed to produced 100% females with a linear relation existing between the two extremes (Miller and Limpus, 1981; Booth and Astill, 2001a) (Appendix E). Equation 2- Sex ratio prediction model; y = mean second third temperature, x = % females produced. Predictions of sex ratio employ the results from linear regression between air temperatures and sand temperatures at nest depth. The average metabolic heating was taken for all study nests and added to this linear model equation to account for temperature rises in the sex determinate stages of incubation. Metabolic heating was calculated by: Equation 3- Calculation of Metabolic Heating (Broderick et al., 2001a); T N = Nest Temperature, T C = Control nest Temperature and = minimum observed difference between the two. 5

14 3 RESULTS 3.1 Biological Parameters and Population Trend Assessment Very little research has been conducted on turtles in Tromelin, apart from daily track counts recorded by Meteo-France over the last 20 years (Figure 1). Number of Tracks (*1000) Year Figure 1 - Yearly number of C. mydas tracks on Tromelin Island from 1987 to Track counts in Tromelin have been reported to be decreasing since counting commenced (Figure 1), although not at a statistically significant rate (Lauret-Stepler et al., 2007). A recent review of this data using the more robust method of spline regression found that the amount of tracks has been significantly declining at a rate of 1.6% each year since 1987 (Bourjea et al., unpublished). Despite a conclusive decline in green turtle track abundance on the island, several factors must be determined before this can be attributed to a decrease in the C. mydas nesting population on Tromelin. The quantity of tracks is linked to nesting success and the number of clutches per female as per Equation 4. Equation 4 - Nesters abundance calculation; TNF=Total Nesting Females, NT=Number of Tracks, =average nesting success, =average number of clutches per female over the study period. The biological parameters included in the above equation have not been re-examined since the work on Tromelin by Le Gall (1988). Le Gall (1988) published results from a tagging and recapture program from November to February between 1981 and These findings included biological parameters contained in the above equation and summarised in the table below (Table 1). In order to assess changes in population trends and characteristics, results from Le Gall (1988) were compared to those obtained during research in 2009 (Table 2). 6

15 Table 1 - C. mydas biological parameters and juvenile production values for Tromelin Island. Average 1981/82 to 1983/84 Tromelin Island November to February (Le Gall, 1988) 2009 March to June ± Standard Deviation (SD) CCL NA 108 ±5 (N=121) Average No# Clutches per Female ±1.1 (N=74) Laying Success Rate 56% 49% (N=489) Inter-nesting Interval (days) ±3.46 (N=55) Average Clutch Size 135 ±2 119 ±18 (N=29) Total Emergence Success Rate 36% 60% 80% ±13.71 (N=29) As Le Gall s results were derived from the peak nesting season in Tromelin Island, some of the raw data is not comparable. Average clutch size and laying success rate can differ naturally between seasons (Mortimer & Carr, 1987; Mazaris et al., 2008), with the extent of these variations for Tromelin not known. The short duration of the 2009 study period would result in an underestimated number of clutches per female, developing an optimistic count of individual females. To compare individual females and juvenile production values between the author and Le Gall s study, daily track counts between 1987 and 2006 on Tromelin were used to gain mean monthly counts and a percentage of activity for each month (Figure 2) Average Track Count Jan (19.92) Feb (17.26) Mar (11.02) Apr (6.39) May (4.56) Jun (3.10) Jul (3.21) Aug (3.25) Sep (3.06) Oct (3.67) Month (Percentage of Activty) Nov (9.03) Dec (15.87) Figure 2 - Average monthly C. mydas tracks (± SD) and percentage activity in Tromelin ( ). Le Gall s period of study was over the peak season of nesting (Nov, Dec, Jan) which encompasses 45% of activity, whilst the study in 2009 (Mar, Apr, May) includes 22% of activity (Figure 2). The data was extrapolated to account for differences in activity between seasons, to permit comparisons of individual green turtle numbers and juvenile production of the two study 7

16 periods. Extrapolations were made using percentage of activity, assuming negligible differences in laying success over seasons and years. Therefore results may not be accurate but allow for a relative comparison. Juvenile production for the 2009 study was calculated according to the following equation (Le Gall et al., 1985): Equation 5 - Calculation of Produced Juveniles; JP=Juvenile Production, =Total Number of Tracks, =average nesting success, =average clutch number, =average emergence success. An average of the track counts from for the study period (March, April, May) was determined to calculate average number of individual females (Equation 4), assuming stable biological parameters over these years. This was to reduce the impact of natural variation between years, typical of green turtles (Bjorndal et al., 1999; Chaloupka, 2001). Table 2 - A comparison of the observed and calculated average abundances of individual C. mydas during different periods on Tromelin. (1981/ /84) ( ) Individual Females Off-Season (Mar, Apr, May) 461* 222 Peak Season (Nov, Dec, Jan) * Juvenile Production Off-Season (Mar, Apr, May) 147,121* 46,095 Peak Season (Nov, Dec, Jan) 300,000 93,993* * Extrapolated results The amount of individual turtles during the study period in 2009 was only 49% of what was expected according to extrapolated results from Le Gall (1988) (Table 2). This signifies a decrease in the population of nesters on Tromelin as suggested by Figure 1. The level of juveniles emerged in 2009 was a maximum of 31% of those in the same season during (Table 2) (using average clutch number + SD). These comparisons indicate the rise in emergence success has not compensated for the decreasing turtle population. The decrease of the green turtle track counts on Tromelin may therefore be indicative of a decrease in the population. A decline in the abundance of adult nesters could be impacting the ability to reproduce sufficient hatchlings to stabilise the population. Once again, these results should be taken cautiously as several assumptions of continuity and stability are made upon biological parameters. Despite these uncertainties, green turtle populations are classified as endangered worldwide (IUCN, 2001). C. mydas capacity to recover to natural levels may be further reduced if new pressures are introduced such as climate change. Reproduction efficiency, a determinate of population stability and health, is highly influenced by incubation temperature of developing juveniles (Yntema & Mrosovsky, 1980; Mazaris, 2008). The following sections will investigate the influence of climatic and local factors on incubation temperatures and what impact these may have on reproductive traits, such as sex-ratios and mortality, for the Tromelin C. mydas population. 8

17 3.2 Climatic and Local Factors Influences on Incubation Temperature The following section analyses fluctuations in sand and incubation temperatures of 29 monitored nests in Tromelin, from March to May 2009 (Figure 3a & b). Primarily, analysis of temperature fluctuations concentrates on sand temperatures at nest depth to investigate a potential correlation with climatic variables. Nest temperatures are then tested against control temperature to establish any relationship, indirectly analysing climatic variables influences on C. mydas nests. As temperature during the last stage of incubation does not influence sex ratio (Broderick et al., 2001a) and is impacted by both metabolic heating and hatchling piping, it was not included in any analysis. The effect of local nest parameters (depth, shading, etc.) are also analysed to discover nest-specific influences on incubation temperatures. a) Temperature (C ) b) Control Nest All other colours: Nests Temperatures Mar 14 Mar 24 Apr 03 Apr 13 Apr 23 May 03 May 13 Figure 3 - a) Nests positions and b) daily average temperatures of control and C. mydas study nests on Tromelin, March-May Date Nest temperature profiles (Figure 3b) during the second third of incubation show a rise in temperature attributed to metabolic heating (Broderick et al., 2001a). Hatching of juveniles causes the drop at the end of the incubation period. Despite a range in temperatures between nests, a pattern in nest variations can be seen, unrelated to the aforementioned biological influences. This indicates a communal outside influence on all nest temperatures. Relationships between temperature data analysed in the following section were explored using Pyper and Peterman correlation tests on residuals, as explained in section Statistics. 9

18 3.2.1 Climatic Factors Sand Temperature at Nest Depth This section assesses the effects of ambient temperature and rainfall on sand temperature at nest depth. Temperature ( C) a) Control Nest Temperature Air Temperature Temperature ( C) b) Control Nest Rainfall Rainfall (mm) Mar 14 Mar 24 Apr 03 Apr 13 Apr 23 May 03 May 13 May 23 Mar 14 Mar 24 Apr 03 Apr 13 Apr 23 May 03 May 13 May 23 Date Figure 4 - a) Control and air temperatures and b) control temperatures and rainfall from March to May 2009 in Tromelin. Date Air temperature, control nest temperature and rainfall during the study period in Tromelin can be seen in Figure 4. Graphic interpretation is difficult due to large fluctuations in air temperature; however, control nest temperature seems to follow variations and the decreasing tendency of air temperature. As temperature travels downward through interstitial spaces within the sand, control temperature variations show a delay in response to ambient fluctuations (Figure 4a). Further, drops in control nest temperature appear associated with incidences of rain (Figure 4b). Rainfall is naturally correlated to air temperature in tropical areas, with instances of rain associated to drops in ambient temperatures. This interaction incorporates rainfall into interactions between air and sand temperatures at nest depth. As moisture within the nest was not recorded, the independent impact of rain on nest temperatures from water percolation into nest chambers could not be statistically assessed in this study. To reduce the extremity of fluctuations, air temperature was smoothed using Tukey s running median smoothing method. This smoothed air temperature was given a delay effect to investigate and compensate for the thermal inertia gradient on Tromelin Island. 10

19 Temperature ( C) a) Control Nest Temperature Air Temperature Temperature ( C) b) Control Nest Temperature Air Temperature Mar 14 Mar 24 Apr 03 Apr 13 Apr 23 May 03 May 13 May 23 Mar 14 Mar 24 Apr 03 Apr 13 Apr 23 May 03 May 13 May 23 Date Figure 5 - a) Four day delayed and smoothed air temperature and control nest temperature, and b) linear trend lines of delayed and smoothed air temperature and control nest temperature in Tromelin, March-May Date A delay of four days clearly illustrates a relationship between ambient and sand temperatures at nest depth as shown in Figure 5a. To statistically assess this relationship between the two time series, a residual analysis must be performed. A Pyper and Peterman correlation test was chosen to account for autocorrelation within each variable. The residuals used for this test were derived from linear regressions performed on both control nest and smoothed air temperatures (Figure 5b). The decreasing linear trends were highly significant for the smoothed air temperature delayed by four days and control nest temperature (p< 2.2e-16, r 2 = 0.80; p< 2.2e-16, r 2 = 0.83, respectively). Figure 6 displays the residuals for the control nest temperature and smoothed air temperature with a four day delay. Control Nest Air Temperature Residuals Mar 14 Mar 24 Apr 03 Apr 13 Apr 23 May 03 May 13 May 23 Date Figure 6 - Control nest and air temperatures residuals from linear regressions (Tromelin, March- May 2009). 11

20 Synchronous fluctuations of the residuals for control nest temperature and smoothed air temperature with a four day delay are shown in Figure 6. To investigate the delay in control nest fluctuations and statistically determine the most appropriate lag time, a Pyper and Peterman test was repeated using residuals of the smoothed air temperature with delays between 0 and 5 days (Table 3). Table 3 - Pyper and Peterman results* performed on residuals of control nest temperature and smoothed air temperature (delayed by 0 to 5 days) (Tromelin, March-May 2009). LAG p1 r 2 p0 N Nstar LAG LAG e LAG e e LAG e e LAG e e LAG e e *p1= statistical significance accounting for autocorrelation; r 2 = correlation coefficient accounting for autocorrelation; p0= statistical significance discounting for autocorrelation; N= effective sample size discounting for autocorrelation; Nstar= effective sample size accounting for autocorrelation Analyses on residuals of the tested variables reveals highly statistically significant correlation between air and control nest temperatures (Lag 2-5), with a delay of four days being the most statistically accurate. This infers an impact of ambient temperature on sand temperature at nest depth when accounting for thermal inertia. According to these results, any further analysis involving air temperature will be performed using this smoothing and delay. Results of the LAG 4 residuals test returned an r 2 of 0.67, indicating other variables have an impact although these could not be analysed. Graphic interpretation of rain and control nest temperatures (Figure 4b) implies that rain is included in these other variables. Control nest temperature will therefore be taken for further analyses as a reflection of air temperature impact, and a hypothetical impact of other climatic variables (principally rainfall), on green turtle nests. Any relationship between sand temperature at nest depth and green turtle nests would reflect an influence of climatic variables on C. mydas nest temperatures Initial Incubation Stage Metabolic heating occurs at all stages of incubation as embryos synthesis tissues and develop (Carr & Hirth, 1961). This biological heat has been reported as negligible during first third with significant increases during the last two stages of incubation (Broderick et al., 2001a). It is therefore assumed that analyses on the first third of incubation will not introduce metabolic heat as a confounding factor. The following section investigates the relationship between sand temperatures at nest depth and temperatures within nests during the first third of incubation. 12

21 Nests First Thirds Control Nest Temperature ( C) Mar 14 Mar 19 Mar 24 Mar 29 Apr 03 Apr 08 Date Figure 7 - Sand temperature at nest depth and C. mydas nest temperatures during first third of incubation in Tromelin, To allow heat passed from the laying turtle to dissipate and thermometers to calibrate to surrounding sand temperature, the first day of temperature recording was not included in any analyses. Although nests and control nest differ in temperature ranges (Figure 7), the similar fluctuations confirm the confounding factor of metabolic heat during this stage of incubation is negligible. As climatic variables impact control nest temperature, a correlation between control nest and green turtle nests temperatures during initial incubation stages would represent an impact of these variables on C. mydas nests. Loess smoothing was performed on each nest and the corresponding control nest temperatures to gain residuals. First third temperatures displayed relatively long-term fluctuations (Figure 7), thus a high smoothing was required (smoothing coefficient alpha=1.5). Residuals from Loess smoothing were examined using a Pyper and Peterman correlation test. An example of Loess smoothing on both Nest 1 and control nest temperatures, with the resulting residuals, are presented below in Figure 8. Temperature ( C) a) Nest 1 First Third Loess smoothing (alpha=1.5) b) Control Nest Loess Smoothing (alpha=1.5) Residuals c) Residuals Nest 1 First Third Residuals Control Nest Incubation Days during First Third Incubation Days During First Third Incubation Days During First Third Figure 8 - Loess smoothing on a) Nest 1 temperature and b) control nest temperature during the first third of incubation of C. mydas nests in Tromelin with c) the resultant residuals. 13

22 Nest 1 (Figure 8a) and control nest temperatures (Figure 8b) display an obvious similarity in their residual fluctuations (Figure 8c). Each nests temperature residuals obtained from Loess smoothing and examined through Pyper and Peterman tests were all significantly correlated with control nest residuals (p<0.05). Correlation coefficients ranged from 0.71 to 0.98 (Appendix F). This represents a high impact of ambient temperature and other incorporated climatic variables on C. mydas nests during initial stages of incubation. To examine whether there is an effect of climatic factors (principally ambient temperature and rainfall) during the middle third of incubation, where metabolic heating occurs at a higher rate (Broderick et al., 2001a), further analyses were performed. This will depict if climate has the potential to impact sex ratios of C. mydas clutches in Tromelin Sex Determinate Stage A rise in nest temperatures due to metabolic heating (Broderick et al., 2001a) can be seen in Figure 9, where control nest temperature declines due to ambient temperature decrease (Figure 5a). Nests Second Thirds Control Nest Temperature ( C) Apr 01 Apr 06 Apr 11 Apr 16 Apr 21 Apr 26 Date Figure 9 - C. mydas nest and control nest temperatures during the second third of incubation in Tromelin, To reduce the effect of biological heating on nest temperatures, residuals of nest and control nest temperatures were derived using a Loess smoothing method (Figure 10a & b respectively, example on Nest 1). Temperature during the second third, raised by metabolic heating, required a lower smoothing than first third (smoothing coefficient alpha=0.75) to gain residuals only reflective of climatic impacts (Figure 10c, example on Nest 1). 14

23 Temperature ( C) a) Nest 1 Second Third Loess Smoothing (alpha=0.75) b) Control Nest Loess Smoothing Residuals c) Residuals Nest 1 Second Third Residuals Control Nest Incubation Days during Second Third Incubation Days during Second Third Incubation Days During Second Third Figure 10 - Loess smoothing on a) Nest 1 temperature and b) control nest temperature during second third of incubation of C. mydas nests in Tromelin with c) the resultant residuals. A Pyper and Peterman test was performed on resulting residuals to test for a correlation between sand temperature at nest depth and the 29 study nests during second thirds of incubation. All nests returned significant p values (p<0.05) with correlation coefficients ranging from 0.69 to 0.96 (Appendix G). This indicates an influence of air temperatures and hypothetically, other climatic variables, on nest temperatures during the second third of incubation. As this second third comprises the period of sex determination, air temperature trends and inclement weather can therefore influence the sex ratio of brooding clutches Local Factors The variation in temperatures between nests (first third average temperature range: C C, Appendix H), as seen previously in Figure 3, may be an indication of local nest parameter impacts. This range of temperature indicates internest variability that could also influence sex ratios. Local factors influences may therefore become increasingly important as ambient temperature rises due to climate change. These local nest parameters are explored in the following section to investigate the potential of intrabeach variability of nest temperatures. Analysis of local factor influences was primarily focused on the first third of incubation to reduce the impact of metabolic heating as a confounding factor. Emergences showed an obvious pattern to occur earlier at the most northerly nests and later as longitude increased. As incubation period is determined by incubation temperatures (Booth and Astill, 2001b), the decision to explore any correlation between nest temperatures and longitude was based on field observations. To graphically investigate any temperature differences between nests according to geographical positioning, first third mean temperatures of nests were plotted against longitude (Figure 11a) (reflection of nest disposition along the beach; Figure 11b). 15

24 Temperature ( C) a) 13 Site 1 b) Site 2 Site Longitude ( ') Figure 11 - a) Average first third of studied C. mydas nest temperatures plotted against longitudinal position to b) classify site position groups on Tromelin, March-May The local factors of grain size (classified from medium coarse sand to very coarse sand; Appendix I), bottom nest depth, hours of sun and site position were first graphically compared to average first third nest temperatures. Only site position and grain size seem to graphically correlate to first third average nest temperatures (Figure 12a & b respectively). Temperature ( C) a) b) Temperature ( C) Site Grain size (µm) Figure 12 - Average first third temperatures of C. mydas nests on Tromelin plotted against a) site position and b) average grain size, March-May To investigate any possible interaction between local factors, these variables were further graphically explored. Grain size appears to strongly graphically correlate with site position of study nests (Figure 13). 16

25 Grain size (µm) Site Figure 13 - Grain size and site position of C. mydas nests on Tromelin, March-May The variability of granulometry between sites was statistically tested using a linear regression and returned a highly significant result (p=9.06e-10, r 2 =0.80; Appendix J). Negative coefficients indicated that average grain size increased as nests were positioned more northerly (Appendix J). As site position was therefore assumed to incorporate the variable of grain size, only the former was used in further statistical analysis. A linear regression was performed on average nest temperatures within the first third of incubation against the local factors of site position, bottom nest depth and hours of sun. Of all tested factors, only site position returned a significant result (p=2.71e-06, r 2 =0.63; Appendix K). Site positions were negatively correlated to average first third temperatures (i.e. grain size was positively correlated to nest temperatures; Appendix K). Thus C. mydas nests located further north displayed higher temperatures during the first third of incubation than those positioned more south (Figure 11b). These factors were also tested against average temperature during the second third of incubation to examine the effect of these variables on sex differentiation. As metabolic heating occurs in the second stage of incubation, biological factors of total clutch number and egg weight were also included in analyses. Grain size was again excluded due to its incorporation into the variable of site position. As previous site grouping was based on the unconfounded temperature within the first third, these were retained for analysis on second third temperatures. Only site position returned a significant result (p=2.46e-06, r 2 =0.63; Appendix L) again, with negative coefficients. To explore the effect of clutch parameters on metabolic heating during second third (independent of other local variations), average metabolic heating (Equation 3) for each nest within the second third of incubation was tested against the biological factors of clutch number and egg weight. No significant results were returned (Appendix M). 17

26 3.3 Mortality Average incubation temperatures influence on emergence success of a clutch was tested using a GLM binomial model. No correlation was found (Appendix N). ICC was tested against average temperatures throughout incubation time and clutch number. No correlations were found (Appendix O). 3.4 Sex Ratio Model Natural sex ratios for green turtles on Tromelin Island may be skewed through the onset of global warming as climate can influence nest temperatures. A linear regression was used to establish an equation for the effect of ambient temperature on sand temperature at nest depth (Figure 14). Sand Temperature ( C) Air Temperature ( C) Figure 14 - Linear regression between sand temperature at nest depth and ambient temperature in Tromelin Island, March-May The linear regression returned a significant result, with a high r 2 value (p<2.2e-16, r 2 = 0.88), implying a strong relationship between ambient temperature and sand temperature at nest depth. Using this correlation, predictions will be made of trends in nest temperatures and sex ratios over both monthly and yearly intervals (i.e. long term predictions). To firstly investigate the validity of this relationship, the recorded control nest temperatures during the study period were predicted (Table 4) using the values reported by the linear regression model (Equation 6). 18

27 Equation 6 Prediction of temperature at nest depth in Tromelin using ambient temperatures; y = temperature at nest depth, x = air temperature. Table 4 - Average monthly temperatures measured at nest depth in Tromelin, March-May 2009, compared to predicted temperatures at nest depth. Monthly Average Temperatures Recorded at Nest Depth ( C) Monthly Average Temperatures Predicted at Nest Depth ( C) March April May Although the model does not calculate exact sand temperatures at nest depth, the margins of error are low (-0.25 to 0.06 C), allowing a relatively accurate prediction of sand temperature from air temperature. To further validate this model, Meteo-France in-ground temperatures at 50cm and 1m depths were plotted for July 08-June 2009 (Figure 15a) and over the years of (Figure 15b) against predicted and recorded (March, April, May) control nest temperatures. Temperature ( C) a) In-ground Temperature at 50cm In-ground Temperature at 1m Predicted Sand Temperature at 75cm Recorded Sand Temperature at 75cm Temperature ( C) b) In-ground Temperature at 50cm In-ground Temperature at 1m Predicted Sand Temperature at 75cm Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Months (July 08-June 09) Figure 15 - Control nest recorded and predicted temperatures over the span of a) 12 months and b) 10 years compared to recorded data by Meteo-France at depths of 0.5 and 1m in Tromelin. Years Similar fluctuations can be seen between recorded and predicted values over seasons (Figure 15a) with differences in temperature assumed as a consequence of thermometer placement on the island. Meteo-France thermometers are located inland under harder ground, whereas control nest temperatures are recorded on the nesting beach. Predicted temperatures were also seen to follow similar fluctuations to Meteo-France thermometers over an 11 year span (Figure 15b). To predict nest temperatures during the second third of incubation using Equation 6, the increase of temperature due to metabolic heating must be included. The average metabolic heating (Equation 3) for all nests during the sex determinate stage of incubation was 1.03 C (±0.23 C). This was added to the constant in Equation 6 to create a prediction model of nest temperatures during the sex determination stage of incubation (Equation 7). 19

28 Equation 7 - Prediction of nest temperatures during the second third of incubation, y = mean nest temperatures during second third of incubation, x = air temperature. To validate this model, a prediction of study nests average temperatures during the second third of incubation was calculated. This average second third predicted temperature (Equation 7) for study nests was then used to determine average sex ratio (Equation 2) of nests (Table 5). Table 5 Average temperature during sex determination stage and its sex ratio outcome as recorded and predicted for C. mydas nests studied on Tromelin Island, March-May Study Nests Average Study Nests Range Predicted Average Temperature 2 nd Third ( C) % Females Produced % Males Produced The predicted average sex ratio value is highly comparable to that derived from recorded information. Although the percentage of females is underestimated, the model is still applicable and can be used to predict monthly average sex ratios for the Tromelin nesting site Past and Future Predictions Inspection of ambient temperatures over the last four decades ( ) showed an increasing in average yearly temperature of approximately 0.75 C over 40 years (Figure 16). Temperature ( C) Average Yearly Air Temperature Linear Trend Years Figure 16 - Average yearly temperatures on Tromelin Island ( ) Employing the second third nest temperature prediction model (Equation 7), the average percentage of females produced monthly over five year spans (Equation 2) was calculated from air temperature data recorded between 1970 and 2009 (Figure 17). 20

29 % Females Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Years Figure 17 - Predicted percentage of females produced by C. mydas clutches within each month over the last 4 decades ( ) in 5 year averages on Tromelin Island (all negative values represent male unisex clutches). Rising ambient temperatures have lead to an increase in the predicted percentage of females produced by brooding clutches. Some months produce unisex male clutches (July to October) although predicted sex ratios during these months show an increasing trend. To further illustrate the effect of air temperature rises on C. mydas nests in Tromelin, predicted sex ratios from years studied by Le Gall (1988) and revisited in section 3.1 Biological Parameters, were evaluated and compared against recent years (Table 6). Table 6 - Comparison of average C. mydas sex ratios during Le Galls (1988) and the author's (March-May 2009) studies in Tromelin. % Females 1981/ / / /09 March, April, May November, December, January Hudson and Jones (2002) examined the effects of climate change using the Hadley Centre Regional Climate Model. Research employed the A2 emissions scenario where greenhouse gas emissions are based on a heterogeneous world with a continuously increasing global population and regionally oriented economic growth (Nakicenovic, 2000). The A2 scenario is more fragmented and slower in emission predictions than other global change scenarios designated by the IPCC in 2000 (Nakicenovic, 2000). The model reported an increase in air temperatures over the southern African region (0-45 S; 5-55 E) by 3.7 C during summer (December, January, February) and 4 C in winter (June, July, August) by The forecast ambient temperature rises from 2002 to 2080 (Hudson & Jones, 2002) were input into the prediction model using the average mean monthly temperatures from This average over five years was taken to reduce the impact of interannual temperature variability on 21

30 evaluated rises in air temperature. Table 7 displays predictions for average sex ratios for green turtle clutches in Tromelin during the early seventies, the turn of the century and the year The annual percentage of females produced in Tromelin over these years was also calculated. This was based on percentages of activity throughout the year (Figure 2). Table 7 - Percentage of C. mydas females produced within clutches in Tromelin using the sex ratio prediction model and future climate scenarios. Summer Winter Annually* Average %Female December January February June July August Prediction *Based on percentages of activity throughout year. Average nest temperatures during second third of incubation and percentage of produced female juveniles were calculated monthly for the above three time spans. The lower value of 3.7 C was applied to mean monthly temperatures not specified in the report by Hudson and Jones (2002) (Figure 18a & b). This allows graphical representation of sex ratio differences between months and decades. Temperature ( C) a) Mensual Mean Mensual Mean Prediction 2080 Presumed Pivotal Temp %Females Produced b) Mensual Mean %Females Mensual Mean %Females Prediction %Females 2080 %Frequentation of the Beach ( ) % Track Counts (Mean between ) Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Months Figure 18 Average predicted a) C. mydas nest temperatures and b) percentages of produced C. mydas females and nesting activity for the years , and 2080 on Tromelin. Months According to climate change predictions (Hudson & Jones, 2002) and nest temperatures during second third predicted in this report, average monthly nest temperatures will not reach the lethal level of 34 C (Bustard, 1970). Higher incubation temperatures in 2080 (Figure 18a) will result in no months producing on average unisex male clutches (Figure 18b). Moreover, periods of male skewed clutches are associated with lower percentages of nesting activity (Figure 18b). All clutches throughout the year will be highly female skewed in 2080 according to the sex ratio prediction model (Figure 18b). 22

31 4 DISCUSSION 4.1 Population Assessment and Evolution Track counting is the most easily available and repeatable method for discerning estimates of population health (Meylan, 1982). However, the changes in personal that perform the task on Tromelin coupled with a possible lack of training denote that track counts may not be accurate. An assumed consistency in precision does however allow track counts on Tromelin beach to be relied on to evaluate the trend of beach frequentation. Due to missing data during the last three years, only track counts between 1987 and 2006 were used to assess this trend. The significant decrease in track counts of -1.6% per annum (Bourjea, unpublished data) is a reflection of a change in population parameters. An evolution in biological parameters (i.e. greater reproduction efficiency) over the last three decades would result in a decrease in track counts, yet not be reflective of a decrease in nesting population size. The number of clutches per individual females was underestimated in 2009 due to the short duration of the research period, a 70% nests laid observation rate and a 9% rate of tag loss. This resulted in an optimised calculation of individual female nesters during the years of Even with an overestimated nesting population size, the amount of individual turtles calculated for this period is only 49% of those that mounted the beach in the early 1980 s (Table 2). Percentiles of activity were taken as representative of number of nests laid (i.e. stable laying success rate ) for extrapolation over different seasons (Table 2). As biological parameters can alter over years and between seasons (Mortimer & Carr, 1987; Mazaris, 2009), these results should be referred to as implicative of trend, not as an accurate value. Lower numbers of individual turtles, found with optimistic calculations, denote that the decline in tracks is probably associated to a decrease in the Tromelin nesting population. The decrease in population size may be due to a change in reproduction rate or adult mortality. The late sexual maturity of green turtles signifies that the drop in population number over a relatively short term period is predominately due to a decrease in the adult population, rather than a change in reproduction efficiency. This is corroborated by the reproduction parameters for C. mydas nesters in Tromelin falling within ranges set for other green turtle populations (Mortimer & Carr, 1987; Bjorndal & Carr, 1989; Mortimer & Portier, 1989; Godley et al., 2001; Glen et al., 2005). Poaching does not occur on the island and human disturbance is extremely low with no sections of the nesting beach destroyed, transformed or inhabited. This indicates that the drop in nesters is caused by adult green turtle mortality at sea during migration or at foraging sites. Mortality of adult turtles by fishing practices has been highly documented with low worldwide population levels attributed to unsustainable amounts of bycatch (NRC, 1990). A review of marine turtle bycatch in the SWIO found that gillnetting, prawn/shrimp trawling and long line fishing have the greatest effect on marine turtles (FAO, 2006; Bourjea et al., 2008), emphasised by their late sexual maturity (Davenport, 1997). Although protection of nesting beaches has increased, the probable decline in Tromelin s population denotes that conservation of C. mydas at sea should be further enforced. The reduction in the number of females frequenting Tromelin has directly impacted the production of juveniles from the island (Table 2), even when an increase in emergence success is seen and maximum clutch number ( ) is used in calculations. As the difference in 23

32 laying success was assumed as negligible over seasons and years, the extent of decrease in juvenile production (Table 2) should be regarded as indicative of trend, not an accurate value. This reduction in juveniles produced on Tromelin Island will further enhance the declining trend in the number of C. mydas nesters over the long-term. The decreased juvenile production, teamed with an unnatural adult mortality rate, affects the ability of green turtles to evolve under new pressures such as global climate change. It is today undeniable that global climate change is occurring at a more rapid rate than at any other time in history as a result of human induced greenhouse emissions (IPCC, 2007). This report focused on the impact of ambient temperatures on green turtle nests, and what this may mean for this species that displays TSD, as climate change takes hold. With an already endangered global population (NRC, 1990) and a decreasing amount of nesters on Tromelin beach, climate change may reduce C. mydas ability to survive or recover to sustainable levels. 4.2 Incubation Temperatures: Climatic Influences and Internest Variability Temperature profiles in C. mydas nests and sand at nest depth were similar during the initial stages of incubation. Metabolic heating within the nest caused rising temperatures within the second and last stage of incubation with a drop in incubation temperatures at the end of the incubation due to juvenile pipping (Figure 3). This pattern is in accordance with previous studies on C. mydas nest temperatures (Broderick et al., 2001b). To investigate climate impacts on nest temperatures, the relationship between climatic factors and sand temperature at nest depth were firstly analysed to exclude any clutch influences. Control nest temperatures did not follow diurnal fluctuations of ambient temperatures, thus smoothing of the latter was applied to reduce temperature extremes over daily periods. Air temperature was also delayed due to thermal diffusion through interstitial sand. Control nest temperatures followed fluctuations of smoothed and delayed air temperatures over the study period, with large drops in control nest graphically comparable to rainfall incidences (Figure 4b). Smoothed and delayed air temperature residuals significantly explained 67% of the variability in residual sand temperatures at nest depth. This combined with the shared decreasing trend outline, the impact of air temperatures on sand temperatures at nest depth on Tromelin. Rainfall is incorporated into ambient temperature impacts on nest temperatures as the two are naturally associated, however, a further influence of rainfall may occur due to changes of moisture levels within the sand. As moisture content of the soil was not monitored, and data was not collected on permeability of soils, rainfall was not independently analysed against control nest temperatures. Previous studies also showed a graphical impact of rainfall on sand temperatures (Booth & Astill, 2001a; Houghton et al., 2007). Further, Godfrey et al. (1996) correlated rainfall to sex ratio, emphasising the importance of rain on sand temperatures. Control nest temperature was used in statistical analyses against nest temperatures, as it is reflective of all climatic factors impacts on sand temperatures and is not confounded by the presence of eggs. Residuals for these analyses were gained using Loess smoothing to find those most representative of climatic impacts from raw data. Results showed that control nest temperature residuals were highly significantly correlated to all nest temperature residuals during the first third (average: p<0.001; r 2 =0.89) and the second third (average: p<0.001; r 2 =0.84) of 24

33 incubation. The higher r 2 value within the first third was due to a classic lack of biological activity in the nests (Booth & Astill, 2001a; Broderick et al., 2001a; Booth & Freeman, 2006). Results on the second third of incubation (sex determinate stage) still denote a large influence of climatic factors on nest temperatures, even when metabolic heating occurs. Climatic variables, chiefly ambient temperature and rainfall, are therefore a major influence on sex ratios of C. mydas clutches in Tromelin. In the context of climate change, as nests are impacted by climatic factors, it is important to study internest temperature variability. The nest specific parameters of hours of sun, bottom nest depth and site position were firstly tested against average first third nest temperatures, to exclude metabolic heating, by means of linear regression. Temperature within the nest during the initial stage of incubation was proven to be negatively influenced by the local factor of site position, with these groupings determined by a nests longitudinal position along the beach (Figure 11). The same factors were then tested against average second third temperatures, inclusive of egg weight and clutch size which could influence metabolic heating. The only significant effect was site position. The effect of site on this third resulted in the same r 2 value as analysis on first third temperatures (r 2 =0.63), indicating that metabolic heating does not reduce the impact of site position. Therefore the local factor of site position influences sex ratios of Tromelin green turtle. Sand samples from studied nests were all classified as medium coarse sand to very coarse sand (Blott & Pye, 2001; Appendix I). Despite low variability in grain size, it was highly negatively correlated to site position (p= 9.056e-10, r 2 =0.80), implying that average first and second third nest temperatures are positively influenced by grain size. As grains of larger sizes have a greater surface area in contact with nest temperatures, heat is more easily retained. Sand on the beach is most likely sorted by local northward alongshore currents, created by South West Alizé winds. These transport finer sands on Tromelin (i.e. medium coarse sand) northerly along the beach. Presence of medium coarse sand in the South is due to short term deposits from southern austral waves during summer (Emmanuel Cordier, personal communication, Appendix P). The range in average nest temperatures over the study (2.36 C) was similar to that between nests during sexdeterminate stages (2.17 C) (Appendix H). The extent of this range may not be accurate, as thermometer placement in the middle of the clutch cannot be verified. As nests on Tromelin endure the same climatic pressures, it is thus implied that the local factor of granulometry can alter temperatures between green turtle nests. As sex ratios are determined over a TRT range of 3 C (Mrosovsky, 1994), local factors can ultimately change the sex ratio of green turtle clutches in Tromelin. The intrabeach variability of sex ratios was demonstrated within this study where percentages of females produced from studied nests ranged between 49 and 100% (Table 5). Other local factors could also be incorporated into site effect however, as no further data was collected in the field, this could not be further explored. Sand characteristics (e.g. albedo, permeability) may play a role in temperature differences between sites. Nests in Site 3 may be placed in shallower soils with greater amounts of boulders. These would retain the cold affecting surrounding sand temperatures. Compaction levels from permeability studies were also not taken which would affect rain percolation and air diffusion into the beach sands. 25

34 4.3 Mortality The binomial model used to test emergence success against average overall temperatures of nests did not return any significant results despite a large range between nests. ICC of resultant juveniles showed no correlation to average nest temperature over the incubation periods. The lack of significant effect of temperature on clutch success and hatchling characteristics may be due to all nests retaining temperatures within the optimal range of C for developing embryos (Bustard & Greenham; 1968). No study nests exceeded lethal temperatures during incubation (Bustard, 1970) (Appendix H). 4.4 Predictions The methodology used in this study to calculate sex ratios was adopted from Miller and Limpus (1981). Miller & Limpus (1981) assumed unisex clutches developed under and above temperatures of 26 C and 29 C with a linear shift between this Transitional Range of Temperature (TRT). The modification of these extremities in this report (TRT: 27.3 C-30.3 C) was based on more recent findings (Spotila et al., 1987; Godley et al., 2002) where histological studies reported a pivotal temperature of 28.8 C for two different C. mydas populations (Tortuguero, Costa Rica and Ascension Island, respectively). Chevalier et al. (1999) reported that pivotal temperatures change according to climatic conditions of nesting sites. As studies by Spotila et al. (1987) and Godley et al. (2002) occurred at deviations from the equator similar to Tromelin (Lat: 7.9º S, Long: 14.4 W; Lat: 10.5 N Long: 83.5 W; respectively), Chevalier et al. s (1999) finding denotes that 28.8 C would be a close representation of the pivotal temperature for Tromelin. Recently, the method of Constant Temperature Equivalent (CTE), as introduced by Georges et al. (1994), has been employed to predict sex ratios (Doody et al., 2004; Booth & Freeman, 2006) due to higher accuracy. This technique is nest specific and relies on incubation duration data. This study wished to derive long-term prediction trends as opposed to nest specific predictions, hence the mean temperature of nests during the second third of incubation was the most appropriate method to use for the predictive model. The linear regression model used as a basis for sex ratio prediction was used on data that does not meet all assumptions for linear testing. The assumption of independent samples (no autocorrelation) is not met by time series temperature data. The impact of air temperature on sand temperature has been proven within this report ( Sand Temperature at Nest Depth) taking into account autocorrelation, however, this type of analysis could not be used to derive a prediction equation. Thus, linear regression was the only tool available to the author that allowed construction of a prediction equation. Sand temperature at nest depth displayed a lower initial temperature than most nests (Figure 3b & Figure 7). As local parameters have been proved to influence temperature, sand temperatures at nest depth should be recorded over more sites on the study beach. The low temperature of control nest in comparison to initial temperatures within green turtle nests may underestimate predicted temperatures, resulting in optimistic forecasts. Verification of the prediction model by comparison of predicted and recorded sand temperatures established the reliability of results for the study period (Table 4). In addition, the model follows the fluctuations of two other thermometers placed in-ground on Tromelin Island by Meteo- France. Predicted temperatures followed the same fluctuations over seasonal (Figure 15a) and long-term periods (Figure 15b) as recorded by in-ground thermometers. These thermometers were at 50cm and 1m depth and not buried on the beach of Tromelin, but under compacted 26

35 ground inland. The difference in Meteo-France thermometer temperatures and sand temperatures at depth on the beach was attributed to placement. Deviations between predicted/actual temperatures of sand at nest depth and Meteo-France thermometers inland (Figure 15a & b) are seen to increase during the cyclone season (December, January, February). This can again be explained by the thermometers placement, through substrate conductivity and the degree of contact with rain waters. The addition of metabolic heating to the control nest temperature to establish mean nest temperature during the second third underestimated average recorded nest temperatures by 0.16 C (Table 5). The prediction model for nest temperatures in the second third could not be further verified due to a lack of data. The difference of 0.16 C is a relatively small amount considering the average temperature of nests during the sex determinate stage of incubation (29.76 C). However, this resulted in a 5.4% difference in predicted sex ratio values (Table 5). As sex differentiation occurs over a TRT range of 3 C (Mrosovsky, 1994), the prediction model for sex ratios is sensitive with resultant values to be taken as representative but not exact. Derived sex ratios predicted from air temperature are therefore slightly bias towards male production, although the trend of sex ratio evolution is reliable. The sensitivity of the model results in an underestimated second third nest temperature, developing an optimistic view (underestimated percentage of females) of green turtle sex ratios in light of climate change. According to the sex ratio prediction model, monitored clutches were skewed to produce larger percentages of females. Overall, study nests produced 4 times the amount of females than males. As turtles display a peak season of nesting activity in Tromelin during hotter months of the year (Figure 2), it is important to regard results from the study period in 2009 in context to other seasons and years. Using the prediction model for , and assuming the same pattern of beach frequentation (Figure 2), females were 34% of the produced juveniles from Tromelin Island. During , this number had already increased to 51% (Table 7). These percentages of produced females are based on percentiles of monthly turtle activity on the island. As biological parameters change between seasons (Mortimer & Carr, 1987; Mazaris, 2008), these figures should be used as representative and not actual. Historically, winter on Tromelin Island (June - October) produced unisex male clutches based on average monthly temperatures. Predictions of past to present sex ratios showed an increasing linear trend towards more skewed female clutches (Figure 19). Results for the predictions for sex ratios in the year 2080 show that male unisex clutches will no longer be produced on the island of Tromelin, with all clutches highly skewed to a female bias. Further, all clutches that will continue to produce males will be laid within the low of seasonal activity of green turtles on Tromelin Island (Figure 19b), assuming the same pattern of beach frequentation (Figure 2). This would result in 96% of juveniles produced on Tromelin being female. Nesting season may evolve under new climatic conditions to avoid extreme skews of sex ratios or lethal temperatures, changing predictions of the yearly percentage of brooded females. The peak nesting period during the warmer months of the year on Tromelin (Nov-Feb; Lauret- Stepler et al., 2007) encompasses over 50% of activity (Figure 2). This nesting peak is shared with the Europa Island rookery that has a similar tropical climate, although these two stocks of green turtles are genetically separated (Bourjea et al., 2007a). This demonstrates that nesting seasonality is behavioural and cannot be qualified by genetic characteristics. The peak in nesting on Tromelin appears as a strategy to manipulate juvenile sex ratios to skew towards female 27

36 production (Figure 17). Further, nesting peak seasons of other rookeries in the SWIO, do not occur during warmer months of the year. On the islands of Grande Glorieuses (Lauret-Stepler et al., 2007) and Mayotte (Bourjea et al., 2007b), peak nesting seasons occur during winter months. These islands retain higher and more stable temperatures over a year in comparison to Tromelin (Bourjea et al., 2007b; Lauret-Stepler et al., 2007). Hence this winter nesting pattern may be due to a combination of other climatic conditions such as lower rainfall. This may also be to avoid lethal in-nest temperatures as these rookeries experience hotter summer months than Tromelin (Lauret-Stepler et al., 2007). The majority of green turtle rookeries around the world show female bias (Limpus et al., 1983; Spotila et al., 1987; Broderick et al., 2000; Booth & Astill, 2001a; Broderick et al., 2001; Godley et al., 2002). As male green turtles never return to land and studies in situ are difficult, male s roles and importance remains relatively unexplored. Tromelin Island, as a historically male bias producing rookery, may therefore play a vital role in C. mydas sexual demographics in the SWIO. This potential importance of Tromelin Island will be reduced as ambient temperatures rise and clutches produce larger proportions of females. 4.5 Climate Change Effects and Adaptation Possibility Fossils of ancestral turtles date back to 150 million years ago (Pritchard, 1979), outlining that marine turtles have adapted to changes in global weather conditions before. Population size evaluations however, estimate green turtles today number just 1% of population sizes in some areas of the world before human exploitation (Jackson, 1997), with several historical observations supporting this scientific result (cited in Lewis, 1940). Low populations reduces C. mydas ability to adapt to new pressures due to bottleneck genetic effects (Lynch, 1995) and, in the case of homing turtles, the lack of natural recruitment to nesting sites by non-indigenous females over short-term periods (Bowen & Avise, 1995). The destruction and disturbance of nesting habitats has also negatively impacted population sizes of C. mydas and their ability to evolve under new climatic pressures through a reduction in niche variability. The increasing ambient temperatures as a product of climate change will result in increasing incubation temperatures of C. mydas on Tromelin. Temperature during development for turtle embryos have been shown to affect several clutch success and juvenile characteristics. Higher temperatures produce smaller hatchlings with better locomotion abilities, higher fitness levels and larger yolk reserves (Booth & Astill, 2001b; Glen et al., 2003; Booth et al., 2004; Burgess et al., 2006) These hatchlings may have a better chance of survival during initial life history stages where swimming performance and predator evasion are vital for survival. Furthermore, emergence and hatching success rates have also been positively linked to temperature (Mazaris, 2008). The increased skews in sex ratios may also be beneficiary. In endangered populations, especially when monogamy is not practiced, this will increase nesting turtles and thus the amount of produced juveniles. The combined effect of an increase in reproduction and its success rates, as well as larger chances of survival of resultant hatchlings, will benefit Tromelin C. mydas populations as air temperature increases. This potential positive influence on green turtle populations will only be short-lived as the optimal range for these effects are surpassed. Climate change may increase nest temperatures past those of optimum development levels. C. mydas eggs develop only between temperatures of 25 C and 34 C (Bustard & Greenham, 1968; 28

37 Bustard, 1970). The prediction model developed in this report used only metabolic heating values during the second third of incubation when predicting nest temperatures for However, metabolic heating increases nest temperatures to a greater extent in the last third of development and can account for a 2.61 C rise in temperature (Broderick et al., 2001a). This would result in the summer periods (December to February), which incorporate the peak season of turtle activity and juvenile production (Figure 2), exceeding lethal C. mydas nest temperatures on Tromelin. This predicted increase in mortality rates are a severe threat to the reproduction capacity and survival of green turtles. Rainfall patterns predicted for southern Africa remain stable in the number of rain days per year (Hudson & Jones, 2002) but with an increase in intensity during incidences of rain. This will pinpoint the apparent negative effect of rain on nest temperatures to clutches laid within certain periods of the year. Although these clutches can be expected to produce larger percentages of males than predicted by rising air temperatures, the stable amount of rain days per year indicates that increased intensity may not be an escape pod from increasing sand temperatures. With the small beach of Tromelin undergoing the same climatic conditions, any ability to adapt to global climate change for this nesting population lies in peak season periods and intrabeach variability due to local factors. Nests temperatures during second third of incubation differed during the study by a range of 2.17 C, occurring between average nest temperatures throughout duration periods. It is this internest variability due to site position that may allow some clutches in new temperature regimes to still produce viable, male bias clutches. Global warming will not only affect reproductive stages of the C. mydas life cycle, but as the marine environment changes, adult feeding and breeding ground parameters will also change. Although green turtles have evolved though periods of climate change before (Pritchard, 1979), the rapid onset of global warming caused by human emissions (IPCC, 2007) reduces their ability to adapt. Destruction and disruption of several feeding and nesting habitats will be further enhanced by sea level rise, where the size of low beaches, such as on Tromelin Island (altitude 7m) will most likely be reduced. The relationship between green turtles and temperature ensures that this species will be greatly affected in all life aspects by global warming. 29

38 5 CONCLUSION AND FUTURE RESEARCH The results from this study supported the three hypotheses initially outlined. Climatic factors influenced incubation temperatures, thus impacting sex ratios of the Tromelin C. mydas population. Local factors also influenced nest temperatures on different areas of the beach. This resulted in intrabeach variation of sex ratios from studied clutches. The influence of air temperature on nest temperatures during sex determinate stage incubation was explored in the context of climate change. According to developed prediction models, global warming will skew sex ratios of C. mydas juveniles from Tromelin toward a more female bias. Despite supportive results, future research is needed to improve the accuracy of climatic and local factor relationships with C. mydas clutch temperatures, as well as the sex ratio and nest temperature prediction models. Results were unable to quantify the decline in the Tromelin green turtle population. Although the trend could be established, biological parameters must be explored in the long term to gauge a better understanding of population size. Moreover, all prediction results were based on the assumption of stability in biological parameters and peak nesting activity over yearly spans. Research into biological parameters in different seasons and over years would increase accuracy in predictions. Research programs on population assessment have been initiated by Ifremer and Kelonia on several islands within the SWIO to gain an overall understanding of regional C. mydas nesting population health. More research is required on nest temperatures to assess the extent of climate change impact. The one year nest temperature monitoring program on Tromelin, instigated by Ifremer and Kelonia, will gain insight into different temperature regimes over different seasons and weather events. To improve understanding and modelling of climatic variable impacts on C. mydas clutches, moisture levels within the nest should be investigated and study periods extended. Other temperature monitoring programs throughout the SWIO, already initiated by Ifremer and Kelonia, will help improve understanding of climatic impacts on a regional scale. Accuracy of climatic impacts, as well as the nest temperature prediction model, could be improved by more precise equipment. Thermometers used in the study had a resolution of ±0.2 C. To improve the precision, thermometers with lower resolutions should be employed. As TRT is assumed as 3 C (Mrosovsky, 1994), a change in temperature of 0.2 C can decrease predicted sex ratio accuracy by ±6.6%. Moreover, the nest temperature prediction model needs further validation by comparison with recorded second third nest temperature data over long term periods. Forecast sex ratios hinge upon pivotal temperatures discovered in other green turtle populations, however this may be different for C. mydas in Tromelin (Chevalier, 1999). Histological studies could be used to assess the pivotal temperature on Tromelin Island, increasing the accuracy of the sex ratio prediction model. Due to the ethical issues, these would only be done on dead juveniles or those unable to emerge naturally. The effects of climate change have altered weather regimes worldwide, with temperature predicted to increase (IPCC, 2007). The nesting site of Tromelin remains relatively undisturbed due to the site s remoteness, but also under researched despite a presumed declining population. This important nesting site in the SWIO (Lauret-Stepler et al., 2007) requires greater understanding of its population before the effects of climate change can be fully assessed. 30

39 6 REFERENCES Aragones, L.V. (2000). A review of the role of the green turtle in tropical seagrass ecosystems. pp In: Pilcher, N., Ismail, G. (Eds.), Sea Turtles of the Indo-Pacific: Research Management and Conservation. ASEAN Academic Press Ltd, London, UK. Balazs, G.H. & Ross, E. (1974). Observations on the pre-emergence behaviour of the green turtle. Copeia, 4: Balazs, G.H. & Chaloupka, M. (2004). Thirty-year recovery trend in the once depleted Hawaiian green sea turtle stock. Biological Conservation, 117: Bjorndal, K.A. & Carr, A. (1989). Variation in clutch size and egg size in the green turtle nesting population at Tortuguero, Costa Rica. Herpetologica, 45(2): Bjorndal, K.A., Wetherall, J.A., Bolten, A.B., Mortimer, J.A. (1999). Twenty-six years of green turtle nesting at Tortuguero, Costa Rica: an encouraging trend. Conservation Biology, 13: Bjorndal, K.A., Bolten, A.B., Chaloupka, M.Y. (2000). Green turtle somatic growth model: evidence for density dependence. Ecological Applications, 10: Blott, S.J. & Pye, K. (2001). Gradistat: a grain size distribution and statistics package for the analysis of unconsolidated sediments. Earth Surface Processes and Landforms, 26: Booth, D.T. & Astill, K. (2001a). Temperature variation within and between nests of the green sea turtle, Chelonia mydas (Chelonia: Cheloniidae) on Heron Island, Great Barrier Reef. Australian Journal of Zoology, 49: Booth, D.T., & Astill, K. (2001b). Incubation temperature, energy expenditure and hatchling size in the green turtle (Chelonia mydas), a species with temperature-sensitive sex determination. Australian Journal of Zoology, 49: Booth, D.T., Burgess, E., McCosker, J., Lanyon, J.M. (2004). The influence of incubation temperature on post-hatching fitness characteristics of turtles. International Congress Series, 1275: Booth, D.T. & Freeman, C. (2006). Sand and nest temperatures and an estimate of hatchling sex ratio from the Heron Island green turtle (Chelonia mydas) rookery, Southern Great Barrier Reef. Coral Reefs, 25: Bouchard, S.S. & Bjorndal, K.A. (2000). Sea turtles as biological transporters of nutrients and energy from marine to terrestrial ecosystems. Ecology, 81(8): Bourjea, J., Lapegue, S., Gagnevin, L., Broderick, D., Mortimer, J.A., Ciccione, S., Roos, D., Taquet, C., Grizel, H. (2007a). Phylogeography of the green turtle, Chelonia mydas, in the Southwest Indian Ocean. Molecular Ecology, 16:

40 Bourjea, J., Frappier, J., Quilllard, M., Ciccione, S., Roos, D., Hughes, G., Grizel, H. (2007b). Mayotte Island : another important green turtle nesting site in the Southwest Indian Ocean. Endangered Species Research, 3: Bourjea, J., Nel, R., Jiddawi, N.S., Koonjul, M.S., Bianchi, G. (2008). Sea turtle bycatch in the West Indian Ocean: review, recommendations and research priorities. Western Indian Ocean Journal of Marine Science, 7(2): Bourjea, J., Marmoex, C., Lauret-Stepler, M., Ciccione, S. (2009). Updated trend of green turtle tracks in the Eparses Islands, SW Indian Ocean: clear and distinct population status. Unpublished. Bowen, B.W. & Avise, J.C. (1995). Conservation genetics of marine turtles. pp In: Avise, J.C, Hamrick, J.L. (Eds), Conservation Genetics: Case Histories from Nature. Chapman and Hall, New York, USA Broderick, A.C., Godley, B.J., Hays, G.C. (2001a). Metabolic heating and the prediction of sex ratios for green turtles (Chelonia mydas). Physiological and Biochemical Zoology, 74(2): Broderick, A.C., Godley, B.J., Reece, S., Downie, J.R. (2001b). Incubation periods and sex ratios of green turtles: highly female biased hatchling production in the eastern Mediterranean. Marine Ecology Progress Series, 202: Burgess, E.A., Booth, D.T., Lanyon, J.M. (2006). Swimming performance of hatchling green turtles is affected by incubation temperature. Coral Reefs, 25: Bustard, H.R. (1970). Temperature and water tolerances of incubating sea turtle eggs. Journal of British Herpetology, 1970: Bustard, H.R., Greenham, P. (1968). Physical and chemical factors affecting hatching in the green sea turtle, Chelonia mydas (L.). Ecology, 49: Carr, A. & Hirth, H. (1961). Social facilitation in green turtle siblings. Animal Behaviour, 9: Chaloupka, M. (2001). Historical trends, seasonality and spatial synchrony in green turtle egg production. Biological Conservation, 101: Chevalier, J., Godfrey, M.H., Girondot, M. (1999). Significant difference of temperaturedependent sex determination between French Guiana (Atlantic) and Playe Grande (Costa Rica, Pacific) leatherbacks (Dermochelys coriacea). Annales des Sciences Naturelles, 20: Davenport, J. (1997). Temperature and the life-history strategies of sea turtles. Journal of Thermal Biology, 22(6):

41 Doody, J.S., Georges, A., Young, J.E. (2004). Determinants of reproductive success and offspring sex in a turtle with environmental sex determination. Biological Journal of Linnean Society, 81: 1-16 FAO (2006). Report of the workshop on assessing the relative importance of sea turtle mortality due to fisheries. 268 Zanzibar, United Republic of Tanzania, April Meeting report No. 1 GCP/INT/919/JPN. 17pp Frazer, N.B. & Erhart, L.M. (1985). Preliminary growth models for green, Chelonia mydas, and loggerhead, Caretta caretta, turtles in the wild. Copeia, 1: Frazier, J. (1975). Marine turtles of the western Indian Ocean. The International Journal of Conservation, 13: Frazier, J. (2003). Prehistoric and ancient historic interactions between humans and marine turtles. Lutz, P.L., Musick, J.A., Wyneken, J. (Eds.), The Biology of Sea Turtles Vol 2. CRC Press, Boca Raton, pp Georges, A., Limpus, C.J., Stoutjesdijk, R. (1994). Hatchling sex in the marine turtle Caretta caretta is determined by proportion of development at a temperature, not daily duration of exposure. Journal of Experimental Zoology, 270: Glen, F., Broderick, A.C., Godley, B.J., Hays, G.C. (2003). Incubation environment affects phenotype of naturally incubated green turtle hatchlings. Journal of the Marine Biological Association of the UK, 83: Glen, F., Broderick, A.C., Godley, B.J., Hays, G.C. (2005). Patterns in the emergence of green (Chelonia mydas) and loggerhead (Caretta caretta) turtle hatchlings from their nests. Marine Biology, 146: Godfrey, M.H., Barreto, R., Mrosovsky, N. (1996). Estimating past and present sex ratios of sea turtle in Suriname. Canadian Journal of Zoology, 74: Godfrey, M.H., Barreto, R., Mrosovsky, N. (1997). Metabolically-generated heat of developing eggs and its potential effect on sex ratio of sea turtle hatchlings. Journal of Herpetology, 31(4): Godley, B.J., Broderick, A.C., Glen, G., Hays, G.C. (2002). Temperature-dependent sex determination of Ascension Island green turtles. Marine Ecology Progress Series, 226: Godley, B.J., Broderick, A.C., Downie, J.R., Glen, F., Houghton, J.D., Kirkwood, I., Reece, S., Hays, G.C. (2001). Thermal conditions in nests of loggerhead turtles: further evidence suggesting female skewed sex ratios of hatchling production in the Mediterranean. Journal of Experimental Marine Biology and Ecology, 263: Hays, G.C., Adams, C.R., Speakman, J.R. (1993). Reproductive investment of green turtles nesting on Ascension Island. Canadian Journal of Zoology, 7:

42 Hays, G.C., Broderick, A.C., Glen, F., Godley, B.J. (2003). Climate change and sea turtles: a 150-year reconstruction of incubation temperatures at a major marine turtle rookery. Global Change Biology, 2003(9): Hawkes, L.A., Broderick, A.C., Godfrey, M.H., Godley, B.J. (2007). Investigating the potential impacts of climate change on a marine turtle population. Global Change Biology, 2007(13): Houghton, J.D.R., Myers, A.E., Lloyd, C., King, R.S., Isaacs, C., Hays, G.C. (2007). Protracted rainfall decreases temperature within leatherback turtle (Dermochelys coriacea) clutches in Grenada, West Indies: ecological implications for a species displaying temperature dependent sex determination. Journal of Experimental Marine Biology and Ecology, 345: Hudson, D.A. & Jones, R.G. (2002). Regional climate model simulations of present-day and future climates of southern Africa. Hadley Centre Technical Note 39, Hadley Centre for Climate Prediction and Research, Met Office, Bracknell, U.K. IPCC (Intergovernmental Panel on Climate Change) (2007). Summary for Policymakers. Cambridge University Press, Cambridge, UK. IUCN (2001). IUCN Red List Categories and criteria: Version 3.1 (9 February 2000). IUCN The World Conservation Union, Gland, Switzerland. Jackson, J.B.C. (1997). Reefs since Columbus. Coral Reefs 16, Supplement: S23-S32 Janzen, F.J. (1994). Vegetational cover predicts the sex ratio of hatchling turtles in natural nests. Ecology, 75(6): Lauret-Stepler, M., Bourjea, J., Roos, D., Pelletier, D., Ryan, P.G., Ciccione, S., Grizel, H. (2007). Reproductive seasonality and trend of Chelonia mydas in the SW Indian Ocean: a 20yr study based on track counts. Endangered Species Research, 3: Le Gall, J.Y. (1988). Biologie et evaluation des populations de tortue verte Chelonia mydas des atolls Tromelin et Europa (Océan Indien S.O.). Mésogée, 48: Le Gall, J.Y., Chateau, D., Bosc, P. (1985). Evaluation de la production de tortues vertes Chelonia mydas nouveaux-nées sur les sites de ponte Tromelin et Europa. Oceanographie Tropicale, 20(2): Lewis, C. B. (1940). Appendix on the Cayman Islands and marine turtles, In: The Herpetology of the Cayman Islands by Chapman Grant. Bull. Inst. Jam. Sci., 2: 56-6 Lynch, M. (1995). A quantitative-genetic perspective on conservation issues. In: J.C. Avise & J.L. Hamrick (Eds.). Conservation Genetics: Case Histories from Nature. Chapman and Hall, NY, USA. pp Mazaris, A.D., Kallimanis, A.S., Shardelis, S.P., Pantis, J.D. (2008). Do long-term changes in seas surface temperature at the breeding areas affect the breeding dates and reproduction 34

43 performance of Mediterranean loggerhead turtles? Implications for climate change. Journal of Experimental Marine Biology and Ecology, 367: Van de Merwe, J., Ibrahim, K., Whittier, J. (2005). Effects of nest depth, shading and metabolic heating on nest temperatures in sea turtle hatcheries. Chelonian Conservation and Biology, 5(2): Meylan, A.B. (1982). Estimation of population size in sea turtles, pp In: K. A. Bjorndal (Eds.), Biology and Conservation of Sea Turtles, Revised Edition. Smithsonian Institution Press. Washington, D.C., USA. Miller, J. D. & Limpus, C. J. (1981). Incubation period and sexual differentiation in the green turtle Chelonia mydas. In: Banks, C.B, Martin, A. (Eds.) Proceedings of the Melbourne Herpetolgical Symposium. The Royal Melbourne Zoological Gardens: Melbourne, Australia. Mortimer, J.A. & Carr, A. (1987). Reproduction and migrations of the Ascension Island green turtle (Chelonia mydas). Copeia, 1987: Mortimer, J.A. & Portier, K.M. (1989). Reproductive homing and interesting behaviour of the green turtle (Chelonia mydas) at Ascension Island, South Atlantic Ocean. Copeia, 4: Mrosovsky, N. (1994). Sex ratio of sea turtles. Journal of Experimental Zoology, 270: Mrosovsky, N. & Pieau, C. (1991). Transitional range of temperature, pivotal temperatures and thermosensitive stages for sex determination in reptiles. Amphib-Reptilia 12: Nakicenovic, N. et al. (2000). Special Report on Emissions Scenarios: A Special Report of Working Group III of the Intergovernmental Panel on Climate Change, Cambridge University Press, Cambridge, U.K., 599 pp. NRC (National Research Council) (1990). Decline of the sea turtles: causes and prevention. National Academy Press, Washington, D.C. Pritchard, P.C.H. (1979). Encyclopedia of Turtles. T.F.H. Publications, Neptune, NJ, USA. Pyper, B.J. & Peterman, R.M. (1998). Comparison of methods to account for autocorrelation in correlation analyses of data. Canadian Journal of Fisheries and Aquatic Science, 55: RITMO (2008). Réseau d Information sur les Tortues Marines d Outremer. (16/08/09) Spotila, J.R., Standora, E.A., Morreale, S.J., Ruiz, G.J. (1987). Temperature dependent sex determination in the green turtle (Chelonia mydas): effects on the sex ratio on a natural nesting beach. Herpetologica, 43(1): Yntema, C.L. & Mrosovsky, N. (1980). Sexual differentiation in hatchling loggerheads (Caretta caretta) incubated at different controlled temperatures. Herpetologica, 36:

44 36

45 APPENDICES 37

46 38

47 APPENDIX A Distribution and Life Cycle of C. mydas (Source: RITMO 2008) (Source: Author) 39

48 APPENDIX B Map and Aerial Picture of Tromelin (Lauret-Stepler et al., 2007) Nesting Beach (Date unknown, Author unknown) 40

49 APPENDIX C Vegetation Line on Tromelin beach and Tournefortia Argentea Tournefortia Argentea (Photo: T. Jacob, 2009) Vegetation Line Monitored Nests (Photo: T. Jacob, 2009) 41

50 APPENDIX D CCL Measurement and Tagging CCL measurement: (Photo: T. Jacob, 2009) Tagging: (Photo: T. Jacob, 2009) 42

51 APPENDIX E Sex Ratio Prediction Model, adapted from Miller & Limpus, 1981 Temperature (C ) y = 0.03x Pivotal temperature*= y = 0.03x % of females = (Average Temp. of Second Third-27.3)/ % Female *: Godley et al., 2002; Spotila et al.,

52 APPENDIX F Results* of Pyper and Peterman Tests on First Third Nest by Nest Residuals Vs. Corresponding Control Nest Residuals Nests p1 r 2 p0 N Nstar e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e *: All residuals gained from Loess Smoothing, alpha=1.5 p0: p value not accounting for autocorrelation N: degrees of freedom (effective sample size) not accounting for autocorrelation p1: p value accounting for autocorrelation Nstar: degrees of freedom (effective sample size) accounting for autocorrelation r 2 : correlation coefficient 44

53 APPENDIX G Results* of Pyper and Peterman Tests on Second Third Nest by Nest Residuals Vs. Corresponding Control Nest Residuals Nests p1 r 2 p0 N Nstar e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e *: All residuals gained from Loess Smoothing, alpha=0.75 p0: p value not accounting for autocorrelation N: degrees of freedom (effective sample size) not accounting for autocorrelation p1: p value accounting for autocorrelation Nstar: degrees of freedom (effective sample size) accounting for autocorrelation r 2 : correlation coefficient 45

54 1st Third Incubation Average Temperature APPENDIX H Nests Temperatures Summary 2 nd Third Incubation Average Temperature ± ± ± Nests S.D. Ranges S.D. Ranges Overall S.D. Ranges All Nests

55 NESTS SIEVING ERROR: -0.2% -0.3% 0.1% -0.1% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% MEAN (µ) SORTING (µm) MEAN: Coarse Sand Coarse Sand Coarse Sand Coarse Sand Coarse Sand Coarse Sand Coarse Sand Coarse Sand Very Coarse Sand Coarse Sand Coarse Sand Coarse Sand Very Coarse Sand Coarse Sand Coarse Sand SORTING: Mod. Sorted Mod. Sorted Mod. Sorted Mod. Sorted Mod. Sorted Mod. Sorted Mod. Sorted Mod. Sorted Mod. Well Sorted Mod. Sorted Mod. Sorted Mod. Sorted Mod. Well Sorted Mod. Well Sorted Moderately Sorted % SAND: 99.6% 99.8% 99.3% 100.0% 99.9% 99.8% 99.9% 99.9% 100.0% 99.9% 99.9% 100.0% 99.9% 100.0% 99.9% % MUD: 0.4% 0.2% 0.7% 0.0% 0.1% 0.2% 0.1% 0.1% 0.0% 0.1% 0.1% 0.0% 0.1% 0.0% 0.1% % V C. SAND: 30.8% 44.6% 32.7% 38.0% 13.4% 10.9% 15.1% 9.1% 52.2% 16.8% 16.9% 29.0% 57.0% 56.7% 23.3% % COARSE SAND: 45.1% 43.2% 42.2% 36.8% 44.2% 37.9% 37.7% 39.7% 40.1% 45.6% 38.6% 60.4% 35.2% 30.5% 42.5% % MEDIUM SAND: 22.0% 11.1% 22.5% 24.5% 41.1% 49.4% 45.6% 49.8% 7.4% 36.3% 42.9% 10.4% 7.2% 12.5% 32.8% % FINE SAND: 1.1% 0.6% 1.3% 0.5% 1.1% 1.3% 1.3% 1.1% 0.2% 1.1% 1.3% 0.2% 0.3% 0.2% 1.1% % V F. SAND: 0.6% 0.2% 0.6% 0.1% 0.1% 0.2% 0.2% 0.2% 0.0% 0.1% 0.2% 0.0% 0.1% 0.1% 0.2% % V C. SILT: 0.1% 0.0% 0.1% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% % COARSE SILT: 0.1% 0.0% 0.1% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% % MEDIUM SILT: 0.1% 0.0% 0.1% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% % FINE SILT: 0.1% 0.0% 0.1% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% % V F. SILT: 0.1% 0.0% 0.1% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% % CLAY: 0.1% 0.0% 0.1% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% Granulometry Results from Gradistat APPENDIX I 47

56 NESTS SIEVING ERROR: 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.5% 0.0% 0.0% 0.0% -0.2% 0.0% MEAN (µ) SORTING (µm) MEAN: Very Coarse Coarse Very Coarse Coarse Very Coarse Coarse Very Coarse Coarse Coarse Coarse Coarse Coarse Coarse Coarse Sand Sand Sand Sand Sand Sand Sand Sand Sand Sand Sand Sand Sand Sand SORTING: Mod. Mod. Well Mod. Mod. Well Mod. Mod. Well Mod. Mod. Mod. Mod. Mod. Mod. Mod. Mod. Sorted Sorted Sorted Sorted Sorted Sorted Sorted Sorted Sorted Sorted Sorted Sorted Sorted Sorted % SAND: 99.8% 99.7% 99.8% 99.8% 99.9% 99.8% 99.9% 99.9% 99.8% 99.8% 99.8% 99.9% 100.0% 99.9% % MUD: 0.2% 0.3% 0.2% 0.2% 0.1% 0.2% 0.1% 0.1% 0.2% 0.2% 0.2% 0.1% 0.0% 0.1% % V C. SAND: 55.5% 46.3% 61.7% 47.0% 65.3% 42.3% 59.9% 16.8% 17.0% 13.5% 42.9% 28.8% 13.0% 30.1% % COARSE SAND: 31.8% 42.2% 28.8% 41.0% 29.3% 44.5% 33.7% 43.6% 40.1% 41.6% 39.8% 41.0% 41.2% 54.4% % MEDIUM SAND: 11.6% 10.3% 8.4% 11.0% 4.8% 12.1% 6.0% 38.1% 41.3% 43.3% 15.9% 29.0% 44.4% 14.6% % FINE SAND: 0.7% 0.7% 0.6% 0.6% 0.3% 0.7% 0.3% 1.1% 1.2% 1.2% 0.9% 0.9% 1.4% 0.6% % V F. SAND: 0.3% 0.3% 0.2% 0.2% 0.1% 0.2% 0.1% 0.2% 0.2% 0.2% 0.4% 0.2% 0.1% 0.2% % V C. SILT: 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% % COARSE SILT: 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% % MEDIUM SILT: 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% % FINE SILT: 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% % V F. SILT: 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% % CLAY: 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% * All Sample Types are Unimodal, Moderately Sorted. *All samples are of the Sand Textural Group. *Mod. = Moderately 48 2

57 APPENDIX J Results of Linear Regression from Grain Sizes ~ Site and Residuals Distribution for Model Validation Linear Regression Results: Analysis of Variance Table: Response: Grain Size Df Sum Sq Mean Sq F value Pr(>F) Site e-10 *** Residuals Coefficients: Estimate Std. Error t value Pr(> t ) (Intercept) < 2e-16 *** Site Site e-09 *** Signif. codes: 0 '***' '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1 r 2 : 0.80 Residual Distribution: Residuals vs Fitted Histogram of resid(grainsite) Residuals Density Fitted density resid(grainsite) Residuals QQplot Sample Quantiles Theoretical Quantiles 49

58 APPENDIX K Results of Linear Regression from Average First Third Incubation Temperature ~ Site and Residuals Distribution for Model Validation Linear Regression Results: Analysis of Variance Table: Response: Temp1st Df Sum Sq Mean Sq F value Pr(>F) Site e-06 *** Residuals Coefficients: Estimate Std. Error t value Pr(> t ) (Intercept) < 2e-16 *** Site ** Site e-06 *** Signif. codes: 0 '***' '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1 r 2 = 0.63 Residuals Distribution: Residuals vs Fitted Histogram of resid(fmlm) Residuals Density Fitted density resid(fmlm) Residuals QQplot Sample Quantiles Theoretical Quantiles 50

59 APPENDIX L Results of Linear Regression from Average Second Third Incubation Temperature ~ Site and Residuals Distribution for Model Validation Linear Regression Results: Analysis of Variance Table: Response: Temp2nd Df Sum Sq Mean Sq F value Pr(>F) Site e-06 *** Residuals Coefficients: Estimate Std. Error t value Pr(> t ) (Intercept) < 2e-16 *** Site ** Site e-07 *** Signif. codes: 0 '***' '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1 r 2 = 0.63 Residuals Distribution: Residuals vs Fitted Histogram of resid(sm_lm) Residuals Density Fitted density resid(sm_lm) Residuals QQplot Sample Quantiles Theoretical Quantiles 51

60 APPENDIX M Linear Regression Results from Metabolic Heating ~ Clutch Number and Eggs Weight and Residuals Distribution for Model Validation Linear Regression Results: Analysis of Variance Table: Response: Met. Heating Df Sum Sq Mean Sq F value Pr(>F) Clutches Number Egg Weights Residuals Residuals Distribution: Residuals vs Fitted Histogram of resid(icc_lm) Residuals Density Fitted density resid(icc_lm) Residuals QQplot Sample Quantiles Theoretical Quantiles 52

61 APPENDIX N GLM Binomial Results from Emerged/Not Emerged ~ Average Incubation Temperature Analysis of Deviance Table Model: binomial, link: logit Response: cbind(emerged, Not_Emerged) Terms added sequentially (first to last) Df Deviance Resid. Df Resid. Dev P(> Chi ) NULL AvTemp

62 APPENDIX O Results of Linear Regression from ICC ~ Average Incubation Temperature and Residuals Distribution for Model Validation Linear Regression Results: Analysis of Variance Table: Response: ICC Df Sum Sq Mean Sq F value Pr(>F) AvTemp Residuals Residuals Distribution: Residuals vs Fitted Histogram of resid(icc_lm) Residuals Density Fitted density resid(icc_lm) Residuals QQplot Sample Quantiles Theoretical Quantiles 54

63 APPENDIX P Grain Size Distribution Dynamic Long term agglomeration of very coarse sand Granulometry gradient Short term deposit of medium coarse sand Alongshore northerly currents: transport medium coarse sand Accumulation of medium coarse sand 55

64 Pôle Halieutique : Dominante Ressources et Ecosystèmes Aquatiques Enseignant responsable : Richard SABATIE Auteur : Théa JACOB Nb pages : 35 Annexes : 16 Année de soutenance : 2009 Organisme d'accueil : Ifremer Cadre réservé à la bibliothèque centrale Adresse : rue Jean Bertho, BP 60, Le Port Cedex Maître de stage : Jérôme BOURJEA Titre : Tromelin Island: Influences on Chelonia mydas Incubation Temperature and Reproductive Traits in Light of Climate Change Résumé : Les augmentations récentes de températures, conséquences du changement climatique, font aujourd hui l objet de nombreuses études quant aux capacités d adaptation et de survies des espèces. La tortue verte Chelonia mydas, considérée aujourd hui en danger d extinction (IUCN, 2001), est une espèce dont le sexe est déterminé par la température d incubation des œufs durant le second tiers d incubation (Yntema & Mrosovsky, 1980). C. mydas est donc sensible aux variations de températures, les températures d incubations élevées produisant plus de femelles que de males. Afin d étudier l impact du réchauffement climatique sur cette population, la température de nids de C. mydas a été relevée de Mars à Mai 2009 sur l île de Tromelin. La température du sable à profondeur d un nid, la température de l air ainsi que la pluviométrie ont aussi été suivies durant cette période. Les résultats de l étude montrent que la température de l air influe sur la température du second tiers d incubation, donc sur le sexe ratio de C. mydas. L impact de facteurs locaux tels que la granulométrie sur la température d incubation des nids indique une variabilité intraplage sur le sexe ratio de C. mydas à Tromelin, variabilité positive dans un contexte d adaptation de l espèce aux changements climatiques. Un modèle de prédiction des températures d incubation des nids et du sexe ratio de C. mydas à Tromelin été construit au cours de cette étude. Cela permet d évaluer les sexe ratios passés et à venir de cette espèce en fonction de l évolution de la température ambiante. Il apparaît que la production de juvéniles à Tromelin comprenait 34% de femelles au début des années 1970, 51% au début des années 2000 et atteindra 96% en Ce modèle prédictif comprend des hypothèses fortes de stabilités des paramètres biologiques et du pic de ponte saisonnier à Tromelin. Des recherches supplémentaires sur le long terme concernant la population de tortues vertes de Tromelin, site de ponte majeur de l océan Indien (Lauret-Stepler et al., 2007) sont nécessaires à une validation rigoureuse des résultats obtenus au cours de cette étude. Abstract : Recent temperature increases caused by climate change need to be investigated to evaluate its impact on penology and survival of plant and animal species. The characteristic of Temperature-dependent Sex Determination in the endangered Chelonia. mydas indicates that global warming impacts may be emphasised in sex ratio outcomes of brooding clutches. High nest temperatures during middle thirds of incubation produce a greater proportion of females. To investigate if climate change would impact sex ratios, nest temperatures of C. mydas clutches were recorded in 2009 on Tromelin Island (South West Indian Ocean). Sand temperatures at 75cm deep on the nesting beach, ambient temperatures and rainfall were recorded over the study period. Air temperatures were found to impact C. mydas nest temperatures during the middle third of incubation, thus affecting sex ratios. Mean size of sand grains surrounding nest chambers were related to average nest temperature during the second third of incubation, indicating this local factor influence on sex ratio. The extent of the relationship between air and sand temperature was evaluated to derive a prediction model of nests temperatures and sex ratios. This model is based an assumptions of stability in both biological parameters and nesting peak season on Tromelin. Results predicted an average production of females of 51% for , against 34% in the early 1970 s, where Tromelin was a male bias rookery. Regional predicted air temperature rises were used to project future sex ratios of Tromelin clutches. Global warming will result in Tromelin producing 96% female juveniles from C. mydas clutches in Intrabeach differences caused by local factors may lessen this result. As little research has been conducted on biological parameters and nest temperatures on Tromelin C. mydas population, further data must be collected to increase confidence in these results. Mots-clés : Chelonia mydas, changement climatique, température d incubation, sexe ratio, prédictions, Tromelin. Key-words : Chelonia mydas, climate change, Temperature-dependant Sex Determination, sex ratio, predictions, Tromelin. Diffusion : Non limitée Limitée (préciser au verso) Je soussigné propriétaire des droits de reproduction du résumé du présent document, autorise toutes les sources bibliographiques à signaler et publier ce résumé.