New Technologies Reveal Insights on the Behaviour of Marine Vertebrates

New technologies reveal insights on the behaviour of marine vertebrates

Antoine Marie Dujon BSc Biology and Ecology of Populations, MSc Marine Biology and Biology of Populations

Submitted in fulfilment of the requirements for the degree of

Doctor of Philosophy

Deakin University

17 March 2017

Acknowledgments I started my PhD on migrating animals by myself migrating over 17 000 km from France to Australia. The last three years are among the most insightful of my life.

I’d like to give my sincere thanks to my main supervisor, Professor Graeme Hays, who accepted me as his PhD student and for his continued support and guidance throughout the last three years. I would also like to thank Gail Schofield who offered me so much advice, patiently supervising me, and always guiding me in the right direction. I’ve learnt a lot from her. I am also extremely grateful to Rebecca Lester for her guidance, especially during the last year of my PhD, her advice was always welcome and insightful. Many thanks to all my co-authors for their input on my work and my manuscripts

I am thankful to everyone at Deakin University who was involved with my work in any minor or major way, but importantly to Deakin University itself for providing me a scholarship, and the Centre of Integrative Ecology for the funding opportunities. Thank you to all the staff that assisted me during my PhD.

Thanks for the moments spent with the many PhD students I met during these three years: Anna Miltadious, Christian van Rijn, Jared Tromp, Matt Jago, Matthew Linn, Michael Lewis, Tom Mock, Helena Stokes, Maud Berlincourt, Johanne Martens, Jessica Hodgson, Adélaïde Sibeaux and David Mitchell. A special thanks to: Jacque-Olivier Laloë - for the many discussions about sea turtle biology and coding; Alexandre Schimel - it was a real pleasure to discuss so many science topics with you; and Fredrick Christiansen - it was a real pleasure to collaborate with you during my PhD.

A special thanks to Thomas Stieglitz who introduced me to movement ecology while I was doing my masters back in France. He showed me how fascinating quantitative ecology and movement ecology was.

I would like to thanks Romain Morvezen and Nicolas Pédron, my French friends who followed my PhD progresses from the other side of the planet. They both know that completing a PhD is challenging and always provided good advice.

iii

I am also thankful to my mother and my brother for always encouraging me to do what I love. All my thoughts to my late grand-mother who would have loved to know what I achieved here.

Publications and chapters arising from this thesis

Chapter 2 published in Methods in Ecology and Evolution

Dujon AM, Lindstrom RT, Hays GC (2014) The accuracy of Fastloc-GPS locations and implications for animal tracking. Methods in Ecology and Evolution 5:1162– 1169.

Chapter 3 is submitted to Marine Biology

Dujon AM, Schofield G, Lester RE, Hays GC (2017) Fastloc-GPS reveals daytime departure and arrival during long-distance migration and the use of different resting strategies in sea turtles. Marine Biology: in review

Chapter 4 is published in Marine Biology

Christiansen F, Esteban N, Mortimer JA, Dujon AM, Hays GC (2017) Diel and seasonal patterns in activity and home range size of green turtles on their foraging grounds revealed by extended Fastloc-GPS tracking. Marine Biology 77:285–299

Chapter 5 is unpublished

Dujon AM, Complex movement patterns by loggerhead sea turtles at foraging and wintering sites revealed by Argos-Linked Fastloc-GPS.

Chapter 6 is published in Ecosphere

Christiansen F, Dujon AM, Sprogis KR, Arnould JPY, Bejder L. (2016) Noninvasive unmanned aerial vehicle provides estimates of the energetic cost of reproduction in humpback whales. Ecosphere. doi: 10.1002/ecs2.1468

AUTHORSHIP STATEMENT Chapter 2 1. Details of publication and executive author

Title of Publication Publication details Dujon AM, Lindstrom RT, Hays GC (2014) The accuracy of The main aim of this Fastloc-GPS locations and implications for animal tracking. publication was to investigate Methods in Ecology and Evolution 5:1162–1169. the accuracy of Argos-Linked Fastloc-GPS location and to determine a method to calculate accurate, high- resolution time series of commonly used metrics in animal movement.

Name of executive author School/Institute/Division if Email or phone based at Deakin; Organisation and address if non-Deakin

Graeme C. Hays Centre for Integrative [email protected] Ecology, School of Life and Environmental Sciences 2. Inclusion of publication in a thesis

Is it intended to include this publication in a Yes If Yes, please complete higher degree by research (HDR) thesis? Section 3 If No, go straight to Section 4. 3. HDR thesis author’s declaration

Name of HDR thesis author School/Institute/Division if Thesis title if different from above. (If based at Deakin the same, write “as above”) Antoine M. Dujon Centre for Integrative New technologies reveal Ecology, School of Life and insights on the behaviour of Environmental Sciences marine vertebrates If there are multiple authors, give a full description of HDR thesis author’s contribution to the publication (for example, how much did you contribute to the conception of the project, the design of methodology or experimental protocol, data collection, analysis, drafting the manuscript, revising it critically for important intellectual content, etc.) The data have been previously collected as part of previous projects. GCH and TL provided the data, and suggested the concept of the study. All authors then discussed ways of organising the data for the analysis. I prepared and filtered the Argos-linked Fastloc-GPS tracks from the two different sea turtle datasets to retain accurate and comparable locations that could subsequently be used in the analysis. I organised the data into a database and explored the data via a series of analyses, which generated the final outputs for the publication. Initially, my analyses involved quantifying the accuracy of locations by using a function of the number of satellites and the residual values. I subsequently performed a series of simulations to determine the minimum time interval needed to compute metrics such as heading and speed of travel. I then validated this approach using the track of a green turtle migrating in the Indian ocean. I wrote all the R coding required to run the analyses for this chapter. I drafted the manuscript with assistance from GCH and TL. I produced all figures, and made subsequent improvements to the figures based on suggestions from GCH. Following submission to the journal of Methods in Ecology and Evolution, I received feedback from two reviewers, who provided around 25 comments in total. I compiled responses to all suggestions, with assistance from GCH. This chapter was completed over a period of six months.

6. Data storage

The original data for this project are stored in the following locations. (The locations must be within an appropriate institutional setting. If the executive author is a Deakin staff member and data are stored outside Deakin University, permission for this must be given by the Head of Academic Unit within which the executive author is based.)

Data format Storage Location Date Name of lodged custodian if other than the executive author Excel file http://onlinelibrary.wiley.com/doi/10.1111/2041- 10/11/2014 containing all 210X.12286/abstract locations calculated during the fixed trials

viii

Chapter 3 1. Details of publication and executive author

Title of Publication Publication details

Dujon AM, Schofield G, Lester RE, Esteban N, Hays GC The aim of this publication was (2017) Fastloc-GPS reveals daytime departure and arrival to determine the departure and during long-distance migration and the use of different arrival pattern along the diel resting strategies in sea turtles. Marine Biology: in review pattern and resting strategies for two species of migrating sea turtles.

Name of executive author School/Institute/Division if Email or phone based at Deakin; Organisation and address if non-Deakin

Antoine M. Dujon Centre for Integrative [email protected] Gail Schofield Ecology, School of Life and [email protected] Environmental Sciences, 2. Inclusion of publication in a thesis

Is it intended to include this publication in a Yes If Yes, please complete higher degree by research (HDR) thesis? Section 3 If No, go straight to Section 4. 3. HDR thesis author’s declaration

Name of HDR thesis author School/Institute/Division if Thesis title if different from above. (If based at Deakin the same, write “as above”) Antoine M. Dujon Centre for Integrative New technologies reveal Ecology, School of Life and insights on the behaviour of Environmental Sciences. marine vertebrates If there are multiple authors, give a full description of HDR thesis author’s contribution to the publication (for example, how much did you contribute to the conception of the project, the design of methodology or experimental protocol, data collection, analysis, drafting the manuscript, revising it critically for important intellectual content, etc.) This chapter was conceived by GCH and GS. Through various discussions with GCH and GS, I determined the key areas of the tracking data on which to analyse. The tracking data were originally collected by NE and GS. I prepared and filtered the Argos-linked Fastloc-GPS tracks from the two different sea turtle datasets to retain accurate and comparable locations that could subsequently be used in the analysis. I first analysed the data to determine the exact time of departure time of turtles from breeding sites and the arrival time at foraging grounds. I designed a methodology to accurately determine the time of day at which turtles departed/arrived. I ran a series of sensitivity analyses to validate the viability of this methodology, with assistance from GS. I also extended these analyses to stopover sites. I then attempted to take in account the contribution of oceanic currents in the speed of travel of sea turtles during migration using the oceanic circulation model HYCOM. However, the model was too coarse for the small spatial scales that I was using; so, this approach was abandoned. I then tested using comparisons of proximate day and night travel speeds to evaluate whether turtles travelled faster by day than by night. I initially analysed these differences using generalised linear models. However, based on the advice of REL, I applied a bootstrapping approach to determine whether the turtles were exhibiting differences in day/night travel speeds. I wrote all of the R coding required to run the analyses for this chapter. I drafted the manuscript with assistance from GCH, GS, REL and NE. I produced all the figures, which were amended based on suggestions from GCH. I submitted the manuscript to Marine Biology. I received feedback from 3 reviewers, with a total of 70 comments. I compiled all responses to the reviewer’s comments, with suggestions from my co-authors. This

4. Description of all author contributions

Name and affiliation of Contribution(s) (for example, conception of the project, design author of methodology or experimental protocol, data collection, analysis, drafting the manuscript, revising it critically for important intellectual content, etc.) Fredrik Christiansen FC developed the study with GCH as part of a larger project initially conceived by GCH, NE and JAM with input on the Centre for Integrative manuscript Ecology, School of Life and Environmental Sciences, Deakin University, Warrnambool, Vic., 3280, Australia

Cetacean Research Unit, School of Veterinary and Life Sciences, Murdoch University, Murdoch, WA 6150, Australia Nicole Esteban NE developed the study as part of a larger project initially conceived by GCH, NE and JAM with input on the manuscript Department of Biosciences, Swansea University, Singleton Park, Swansea SA2 8PP, UK Jeanne A. Mortimer JAM developed the study as part of a larger project initially conceived by GCH, NE and JAM with input on the manuscript Seychelles Islands Foundation (SIF), Victoria, Mahe, Seychelles

Department of Biology, University of Florida, Gainesville, FL 32611, USA Graeme C. Hays GCH with FC as part of a larger project initially conceived by GCH, NE and JAM. GCH provided input on the manuscript Centre for Integrative Ecology, School of Life and Environmental Sciences, Deakin University, Warrnambool, Vic., 3280, Australia 5. Author Declarations I agree to be named as one of the authors of this work, and confirm: i. that I have met the authorship criteria set out in the Deakin University Research Conduct Policy, ii. that there are no other authors according to these criteria, iii. that the description in Section 4 of my contribution(s) to this publication is accurate, iv. that the data on which these findings are based are stored as set out in Section 7 below. If this work is to form part of an HDR thesis as described in Sections 2 and 3, I further v. consent to the incorporation of the publication into the candidate’s HDR thesis submitted to Deakin University and, if the higher degree is awarded, the subsequent publication of the thesis by the university (subject to relevant Copyright provisions).

xiii

4.Description of all author contributions Name and affiliation of Contribution(s) (for example, conception of the project, design author of methodology or experimental protocol, data collection, analysis, drafting the manuscript, revising it critically for important intellectual content, etc.) Fredrik Christiansen FC conceived the study. FC and LB coordinated the fieldwork. FC, LB, and KS performed the fieldwork. FC led the data Cetacean Research Unit, processing, analyses, and interpretation of the data, with help School of Veterinary and from KS and AD. FC drafted the manuscript with input from Life Sciences, Murdoch KS, LB, JA, and AD. University, Murdoch, Western Australia 6150 Australia Kate R. Sprogis KS performed the fieldwork and helped with the data processing and interpretation. KS provided input on the Cetacean Research Unit, manuscript. School of Veterinary and Life Sciences, Murdoch University, Murdoch, Western Australia 6150 Australia JYPA provided input on the manuscript John Y. P. Arnould.

School of Life and Environmental Sciences, Deakin University, Burwood Campus, Geelong, Vic., Australia LB coordinated and performed the fieldwork and provided Lars Bejder input on the manuscript

Cetacean Research Unit, School of Veterinary and Life Sciences, Murdoch University, Murdoch, Western Australia 6150 Australia 5. Author Declarations I agree to be named as one of the authors of this work, and confirm: i. that I have met the authorship criteria set out in the Deakin University Research Conduct Policy, ii. that there are no other authors according to these criteria, iii. that the description in Section 4 of my contribution(s) to this publication is accurate, iv. that the data on which these findings are based are stored as set out in Section 7 below. If this work is to form part of an HDR thesis as described in Sections 2 and 3, I further v. consent to the incorporation of the publication into the candidate’s HDR thesis submitted to Deakin University and, if the higher degree is awarded, the subsequent publication of the thesis by the university (subject to relevant Copyright provisions).

xvi

Abstract

Many marine species migrate over long distances between areas that will increase their fitness. Understanding how animals manage to complete these migrations and how such behaviours have evolved is an important research topic. A commonly used approach to study the behaviour of migratory species is to record the trajectories of multiple individuals and to use their movement patterns to infer their behaviour. During the last three decades, the advances in satellite tracking made it possible to record the movement of marine species that regularly surface. Until recently, the amount of information possible to obtain from such tracks was limited to relatively large spatial and temporal scales, due to a limited accuracy and a reduced number of locations. In my thesis, I illustrated the potential of two emerging technologies to overcome these issues - the Argos-Linked Fastloc-GPS, and unmanned aerial vehicles (UAVs) - to obtain new insights on the behaviour of migratory marine vertebrates. I first showed how Argos-Linked Fastloc-GPS can be used to obtain new insights on the behaviour of two species of sea turtles, the loggerhead turtle in the Mediterranean Sea, and the green turtle in the Indian Ocean. I showed how to calculate high-resolution time series of metrics (such as the speed of travel of the heading) commonly used for multiple purposes in movement ecology. I then investigated the diel patterns of sea turtles during their migration and compare the strategy they use to complete their migration with terrestrial and avian species. Through this study, I was the first to show that both species primarily depart from their respective breeding grounds and stopover sites early in the morning and arrive at their respective foraging grounds during the daytime, suggesting they rely on visual cues to orientate doing migration. At the foraging ground, I quantified the differences between the small-scale foraging patterns of the two species and compared them to the patterns of terrestrial vertebrates. I showed that, within a foraging ground, green turtles use a relatively small number of patches with distinct feeding and resting areas. In comparison loggerhead turtles exhibited more complex movement patterns. The number and size of patches, as well as the area used during the daytime and night-time, depended of the location of the foraging grounds with clear differences between nearshore and offshore sites. In parallel, I also showed how UAVs can be used to provide insights on migratory behaviour that cannot be obtained with conventional satellite tracking. I illustrated

xviii

this by quantifying the change in body condition in southern humpback whales while they are at their breeding ground. I found that the body condition of mature individuals and lactating females decreased during the breeding season in whereas the body condition of immature and calves exhibited no significant variation. I also showed how to quantify the interaction between the body condition of lactating females and the body condition of their calves. As a result, I suggest that females with good body condition invest more energy into their calves compared to females in poor body condition. Thus, I was able to quantify the cost of reproduction for this migratory species . My thesis demonstrated the value of Fastloc-GPS and UAVs for monitoring migratory species because they allow us to obtain new insights on the movement and behaviour of migratory marine species, and facilitating comparison with terrestrial and avian species.

xix

Table of content

Chapter 1 : General introduction 1 What is movement ecology? 1 How can movement ecology be studied? 1 The challenges of understanding marine species movement 4 Migratory behaviours 5 Foraging and mating behaviour 9 Study species 11 Research aims and thesis structure 17

Chapter 2 : The accuracy of Fastloc-GPS locations and implications for animal tracking 21 Abstract 21 Introduction 22 Methods 23 Results 26 Discussion 32

Chapter 3 : Argos-Linked Fastloc-GPS reveals daytime departure and arrival during long-distance migration and the use of different resting strategies in sea turtles 36 Abstract 36 Introduction 37 Methods 40 Results 47 Discussion 56

Chapter 4 : Diel and seasonal patterns in activity and home range size of green turtles on their foraging grounds revealed by extended Argos-Linked Fastloc-GPS tracking 61 Abstract 61 Introduction 62 Methods 64 Results 69

Table of content

Discussion 76

Chapter 5 : Complex movement patterns by loggerhead sea turtles at foraging and wintering sites revealed by Argos-Linked Fastloc-GPS 81 Abstract 81 Introduction 82 Methods 84 Results 88 Discussion 97

Chapter 6 : Noninvasive unmanned aerial vehicle provides estimates of the energetic cost of reproduction in humpback whales 103 Abstract 103 Introduction 105 Methods 108 Results 117 Discussion 124

Chapter 7 : General discussion 130 Summary of the main findings 130 Behaviour of marine vertebrates during migration 133 Foraging patterns of sea turtles 137 Changes in body condition in southern humpback whales. 139 General conclusion 139

References 141

Appendix A 180

Appendix B 184

Appendix C 188

Appendix D 195

xxi

List of tables and figures

Chapter 1: General introduction

Figure 1.1: Examples of preferred departure times during a one-day period for a range of migratory species……………………………………………………. 9

Figure 1.2: Locations used in a typical life cycle by sea turtles. Migratory routes of green and loggerhead turtles………………………………………………….. 13

Figure 1.3: Locations used in a typical life cycle of the humpback whale. Location of the summer foraging grounds in the Antarctic Ocean and the winter breeding grounds (in Australia)………………………………………………………… 17

Figure 1.4: Simplified migration cycle for both sea turtles and humpback whales. Section of the migration is covered by the different chapters of this thesis…… 18

Chapter 2: The accuracy of Fastloc-GPS locations and implications for animal tracking

Figure 2.1: The deviation of Fastloc-GPS locations from the true position based on the number of GPS satellites………………………………………………….. 27

Figure 2.2: The deviation of Fastloc-GPS locations from the true position for 45,157 Fastloc-GPS locations as a function of their residual value.………………..... 27

Table 2.1: Accuracy statistics of a fix trial experiment involving 257 Fastloc-GPS tags and 45,157 locations…………………………………………………….. 28

Figure 2.3: Sensitivity analysis of the impact of location accuracy and number of GPS satellites used in the location calculation on speed of travel and heading estimation…………………………………………………………………….. 30

Figure 2.4: Track of a green turtle migrating from the Chagos Islands to coast of Somalia. Speed of travel and heading for this turtle as a function of time……. 31

Figure 2.5: Typical accuracy (95th percentile accuracy) of commonly used tracking technologies and of Fastloc-GPS locations………………………………….... 33

xxii

List of tables and figures

Chapter 3: Argos-Linked Fastloc-GPS reveals daytime departure and arrival during long-distance migration and the use of different resting strategies in sea turtles.

Figure 3.1: Migratory routes of loggerhead and green sea turtles tracked with Argos- linked Fastloc-GPS from the breeding to foraging grounds…………………... 41

Figure 3.2: Two examples showing how the time that turtles initiated migration was determined from the Argos-linked Fastloc-GPS locations……………………. 49

Figure 3.3: Two examples showing how the time that turtles arrived at foraging sites was determined from the Argos-linked Fastloc-GPS locations……………….. 50

Figure 3.4: Time of day that turtles initiated departure from the breeding grounds and arrived at the foraging grounds and stopover sites. Speed of travel of turtles during the final night of migration in relation to the time elapsed since dawn on the day of arrival. Examples of Argos-linked Fastloc-GPS tracks showing the movement of turtles on the night before and day of arrival at the final foraging ground……………………………………………………………………….... 51

Figure 3.5: Mean day/night speed of travel ratio and 95 % confidence intervals for nine loggerhead and five green migrating turtles. Example of two days and nights of oceanic crossing for green and loggerhead turtles. Final 10 days of migration by a loggerhead turtle in the Adriatic…………………………………………….. 55

Chapter 4: Diel and seasonal patterns in activity and home range size of green turtles on their foraging grounds revealed by extended Argos-Linked Fastloc- GPS tracking

Figure 4.1: he migratory movements of the eight tracked adult female green turtles (solid black lines) from their nesting beach on Diego Garcia, Chagos Archipelago, to their respective foraging grounds………………………………………….. 69

Table 4.1: Summary data of the eight satellite tracked adult female green turtles on their foraging grounds in the Indian Ocean…………………………………... 70

Figure 4.2: Back-transformed swim speed as a function of hour of day for the eight tracked green turtles in their Indian Ocean foraging grounds………………... 72

xxiii

List of tables and figures

Figure 4.3: Diurnal and nocturnal sites of the eight tagged female green turtles on their foraging grounds in the Indian Ocean…………………………………… 73

Table 4.2: Summary table of the 10 identified diurnal sites of the eight tracked green turtles on their foraging ground in the Indian Ocean………………………….. 74

Table 4.3: Summary table of the 11 identified nocturnal sites of the eight tracked green turtles on their foraging ground in the Indian Ocean…………………..... 75

Figure 4.4: Simulated proportion of time spent in transit for the eight green turtles on their foraging grounds in the Indian Ocean…………………………. 76

Chapter 5: Complex movement patterns by loggerhead sea turtles at foraging and wintering sites revealed by Fastloc-GPS Figure 5.1: Locations of the 31 benthic foraging and wintering sites retained for this study and identified from 24 turtles (18 males and 6 females) tracked with Argos- linked Fastloc-GPS in the Mediterranean Sea………………………………… 89

Figure 5.2: Example of two foraging sites located 50 km apart that were used by a single loggerhead turtle. Example of a loggerhead turtle using two focal patches in Zakynthos Bay. Variation of the overall home range of nearshore and offshore foraging sites with the seabed depth………………………………………….. 92

Figure 5.3: Examples of the Type 1, Type 2 and Type 3 patch use patterns over a period of 15 days in Zakynthos bay…………………………………………... 94

Table 5.1: Main characteristics of loggerhead turtles foraging behaviour in the three types of foraging patches……………………………………………………… 96

Figure 5.5: Photographs of a male loggerhead foraging on the sponge Chondrilla nucula and antagonistic interaction between two loggerhead males………….. 97

Chapter 6: Noninvasive unmanned aerial vehicle provides estimates of the energetic cost of reproduction in humpback whales

Figure 6.1: Map of the Exmouth Gulf study area in Western Australia……….. 109

Figure 6.2: An example of desired aerial photograph of humpback whales captured by a unmanned aerial vehicle that was used in analyses………………………. 111

Figure 6.3: Rate of body width change at different measurement sites…….…. 118

xxiv

List of tables and figures

Table 6.1: Linear model selection results based on minimization of Akaike’s information criterion for humpback whale body condition index……………. 120

Figure 6.4: Partial effect plot of body condition index as a function of day for calves, immature, mature, and lactating humpback whales…………………………... 121

Figure 6.5: Calf body condition as a function of female body condition for humpback whales…………………………………………………………….. 121

Figure 6.6: Partial effect plots of calf length as a function of maternal length and day……………………………………………………………………………. 122

Table 6.2: Linear model selection results based on minimization of Akaike’s information criterion (AIC) for humpback whale calf length………………... 122

Figure 6.7: Sensitivity analysis of within-photographs measurement errors showing the density distribution of the day parameter values and their associated standard errors for the best fitting model………………………………………………. 123

Figure 6.8: Sensitivity analysis of between-photographs measurement errors showing the density distribution of the day parameter values and their associated standard errors for the best fitting model……………………………………... 124

Chapter 7: General discussions

Figure 7.1: Schematics of the migratory cycle of sea turtles updated from Chapter 1………………………………………………………………………………. 131

xxv

Chapter 1: The accuracy of Fastloc-GPS locations

Chapter 1 : General introduction

What is movement ecology?

Movement ecology is the study of the movement of organisms. Movement, in its most general sense, is “the process by which individual organisms are displaced in space over time” (Turchin 1998, Page 3). Movement is a fundamental trait of life, influencing the spatiotemporal abundance, diversity and distribution of organisms (e.g. animals, plants, bacteria, viruses), and the transport of mass of energy over large scales; consequently, movement is under strong evolutionary pressure (Turchin 1998, Nathan et al. 2008, Milner-Gulland et al. 2011, Avgar et al. 2014). Thus, it could be argued that movement is a behaviour that strongly influences the morphology and physiology of animals (Dickinson et al. 2000). Using a multidisciplinary approach (i.e. by combining tools and methods from biology, physics, and engineering), the field of movement ecology aims to link the statistical properties of movement patterns to specific behaviours. Such detailed knowledge of behaviours has implications for: (1) the management of diseases and pest species; (2) limiting the impact of habitat fragmentation on populations; (3) understanding the consequences of climate change on species; and, more broadly, (4) effective conservation. As a result, understanding fundamental movement ecology can have far-reaching implications for the management of individual species.

How can movement ecology be studied?

In order to make inference regarding animal behaviour from its movement, a record of its trajectory is usually needed. A trajectory is the curve described by the animal when it moves (Turchin 1998, Calenge et al. 2009). Because this curve is mathematically continuous in space and time, sampling of the trajectory is required to discretise the path into steps connecting the successive relocation of the animal (Turchin 1998). Most of the early work on animal movement used terrestrial and avian species either on the laboratory or in the field (Turchin 1998, Milner-Gulland et al. 2011). Over recent decades, multiple methods have been used to investigate movement including visual observation, band tagging (e.g. birds, Wayne and 1

Chapter 1: The accuracy of Fastloc-GPS locations

Shamis 1977), videotaping (e.g. insects, Turchin 1998), camera traps (e. g. mammals and birds, O’Connell et al. 2010) and telemetry (e.g. radiotacking or global positioning system [GPS], White and Garrott 1990, Tomkiewicz et al. 2010). Using these methods has expanded our knowledge of the movement ecology of animals, revealing new migration, foraging and breeding patterns in a range of species (e.g. mammals, birds, insects, reptiles, Turchin 1998, Newton 2008, Milner- Gulland et al. 2011).

Tracking the movement of marine species has been a more of a challenge. The majority of marine species are considered to be elusive because they spend extended period of time underwater and never surface (e.g. most fish species) or only come to the surface for brief periods (e.g. air breathers such as sea turtles or whales). As for terrestrial species, a range of tags and devices have been deployed on marine animals to record their behaviour. This includes flipper tags (e.g. pinnipeds, sea turtles, Carr 1967, Anglia 1972), time-depth recorders to understand dive pattern and vertical movement (e.g. sea turtles, sharks, birds, Minamikawa et al. 1997, Boustany et al. 2002, Scheffer et al. 2016) or acoustic and radio tracking to provide information on horizontal movement (e.g. abalone, lampreys, sea turtles, Almeida et al. 2002, Coates et al. 2013, Hazel et al. 2013). Satellite telemetry is an alternative to these methods. Most satellite telemetry tags are equipped with an antenna that collect and/or transmit information when the animal is surfacing (i.e. when the antenna is above water) and therefore are especially useful to track air breathers (Hays, Åkesson, et al. 2001, Tomkiewicz et al. 2010). Over the years, different types of satellite telemetry systems have emerged: Argos, Argos-linked Fastloc-GPS and pop-up satellite archival tags. Argos tags first appeared in the 1980s and exploit the Doppler Effect (the frequency shift induced by the relative motion between the satellite and the platform) to locate an animal (Hays, Åkesson, et al. 2001). The typical accuracy of Argos locations range between 250 and 1500 m (Hays, Åkesson, et al. 2001, Vincent et al. 2002, Lopez et al. 2014). Argos tags are widely used to study the movement of animals at large spatial scales (e.g. migratory routes of sea turtles, sharks, manatees, Deutsch et al. 2003, Bonfil et al. 2005, Godley et al. 2008), but the limited accuracy of locations render the resolution

Chapter 1: The accuracy of Fastloc-GPS locations

of the movement at smaller spatial scales difficult, often requiring complex statistical methods (e.g. state-space models, Jonsen et al. 2005). Pop-up satellite archival tags rely on ambient light-level irradiance from which geolocations can be calculated. These tags relay information through the Argos satellites and are mostly used to track animals that very rarely surface or when their recapture is very unlikely (e.g. teleosts, elasmobranches, squids, Musyl et al. (2011)). The accuracy of pop- up satellite archival tags range between 200 km and 500 km which cause similar limitations as Argos tracking (Wilson et al. 2007). Argos-linked Fastloc-GPS, which was the technology that I used during this PhD, was developed to overcome these limitations and relies on the GPS network to estimate the location of an animal.

GPS is a satellite navigation system developed by the United States Department of Defence. With 32 satellites in orbit around earth, GPS broadcasts two signals in the L1 (1575.42 kHz for civilian use) and L2 (1227.60 kHz for military use) bands. A GPS receiver uses the pseudo-ranges (i.e. an indirect measurement of the distance between a satellite and a navigation satellite receiver which is calculated by multiplying the speed of light by the time the signal emitted by the satellite took to reach the receiver) of at least three satellites to estimate a location by trilateration (Blewitt 1997, Bajaj et al. 2002, Lehtinen et al. 2008). For most conventional terrestrial receivers, the locations are stored directly inside the device, meaning it must be manually recovered from the animal. Recovery is problematic for animals with large home ranges or that migrate over long distances. To overcome this issue, Argos-linked Fastloc-GPS tags rely on Argos satellites to transmit locations to a ground station (Sirtrack 2010). They convert the time signal of the satellites to pseudoranges and subsequently transmit them to the Argos system which relay them to ground stations. Pseudoranges are later recovered on a computer and converted to location estimates. Compared to terrestrial GPS receivers, which need 1-45 seconds to obtain a fix (Lehtinen et al. 2008), Argos- linked Fastloc-GPS receivers only need few milliseconds, making them suitable to track animals that only briefly surface and potentially generating thousands of locations per track (e.g. sea turtles, fish, pinnipeds, Schofield et al. 2007, Costa et al. 2010, Thys et al. 2015). Despite the improvement in the number and quality of 3

Chapter 1: The accuracy of Fastloc-GPS locations

locations, few researchers have explored the potential of Argos-linked Fastloc-GPS to answer key questions on behavioural ecology such as how animals navigate and orientate in the open ocean, how the physical environment influences movement, or how predation risk influences movement strategies (Hays et al. 2016).

In addition to satellite tracking technologies, unmanned aerial vehicles (UAVs) have become increasingly accessible to the scientific community over the past 3-4 years as a cheap alternative to aerial photography taken from planes, helicopters and microlights (Marris 2013, Crutsinger et al. 2016). The observer’s field of view is extended using an aerial vehicle carrying a camera that is remotely controlled from a boat or the shore. UAVs could potentially provide a wealth of information (e. g. animal density, age class, presence/absence of predators) for marine species that spend time near the surface (e.g. sea turtles or marine mammals, including whales, (oski et al. 2009; Gaspar et al. 2011; Hodgson et al. 2013) or that cannot easily be measured with satellite tracking. Another advantage of UAVs compared to Argos-Linked Fastloc-GPS is their relatively cheap price (<$2000 USD for an UAV, ~$4000 USD for an Argos-Linked Fastloc-GPS tag). However, UAVs are currently limited by the distance covered (~15 km) and by the battery’s autonomy (typically around 15-30 minutes, depending on the battery type, Koh and Wich 2012). Nevertheless, rapid improvement of the technology is expected in the coming years (Crutsinger et al. 2016). Therefore, UAVs are a promising technology to be used in parallel with Fastloc-GPS to obtain information on the animals and their environment that could be linked to their movement pattern.

The challenges of understanding marine species movement Decades of direct observation and tracking have provided numerous insights on the movement of terrestrial and avian species which is now well explored. A large number of studies revealed the migration patterns of a range of mammals, reptiles, birds and insect species (e.g. Dingle 1972; Sidney and Gauthreux 1980; Newton 2008; Milner-Gulland et al. 2011). Our deepening understanding of their foraging pattern have allowed the developed of important theories in ecology, such as the optimal foraging theory (Charnov 1976, Turchin 1998, Viswanathan 2011), which predicts how an animal in search of food will use its space and has been applied to 4

Chapter 1: The accuracy of Fastloc-GPS locations

many species (e.g. large herbivores, birds and insects, Searle et al. 2005, Pyke 2008, Viswanathan 2011) .

Compared to terrestrial species, the information gathered for marine species is limited. Most marine species spend extended period of time underwater, making direct observation and recording of their trajectory difficult. As for terrestrial species, many marine species revealed to undertake long-distance migrations (e.g. sea turtles, sharks, whales, fishes, Carr 1967, Clapham 1996, Dittman and Quinn 1996, Bonfil et al. 2005). Some of the main questions on the movement ecology of marine species that have yet to be addressed include determining: (1) whether simples rules govern the seemingly-complex movement patterns; (2) the costs and benefits of different movement patterns; (3) how animal orientate in the open sea; and (4) how those patterns can be incorporated into efficient conservation planning (Hays et al. 2016). Providing answers to these questions is constrained by the quantity and quality of available tracking data (which is limited to small numbers of individuals from a few select populations for many marine species), but also by the availability of environmental data (e.g. oceanic currents, bathymetry, prey distribution, Bradshaw et al. 2007; Fossette et al. 2012a; Hays et al. 2016).

Migratory behaviours Migration is a distinct, specialized, and complex behaviour characterized by a directed and persistent movement over relatively long durations and distances that are triggered by environmental cues. This behaviour involves activity patterns particular to departure and arrival and the use of specific energy pathway and internal resource allocation to support movement (Dingle 1996, Dingle & Drake 2007). In this thesis, I will refer to the specific case of migratory behaviour corresponding to the predictable seasonal movement of an animal between two areas. Such movement is assumed to be motivated by increasing species fitness (e.g. through the presence of food, lack of predators, mating patterns) and avoiding unfavourable conditions (e.g. extreme temperatures, presence of predation or intense competition) (Sidney & Gauthreux 1980, Dingle & Drake 2007, Milner- Gulland et al. 2011). This type of migration has evolved independently among many animal groups, such as birds (Hedenström 2008), fishes (e.g. eels, salmon, Friedland 5

Chapter 1: The accuracy of Fastloc-GPS locations

et al. 2001, Aarestrup et al. 2009), mammals (e.g. marine such as whales or terrestrial such as bats, Jenner et al. 2001, McGuire et al. 2012), reptiles (e.g. sea turtles, Lohmann et al. 2008), amphibians (Russell et al. 2005), insects (Dingle 1972) and marine invertebrates (Alerstam et al. 2003). Migratory species usually travel over long distances; Alerstam et al. (2003) provides a distance range of 1.5 to 19 000 km corresponding to 4000 to 141 million times the animal’s body length). Understanding how animals manage to complete these migrations and how such behaviours have evolved over the last million years is a major research topic (Alerstam et al. 2003, Newton 2008, Milner-Gulland et al. 2011). Migrations may be separated into four major steps: preparation, departure, in transit, and termination (Milner-Gulland et al. 2011).

Preparation for migration involves important morphological and physiological changes that prepare the animal for travelling over long distances. These changes include the storage of large amounts of energy (mostly fat) that is catabolised to power the migration (e.g. birds, sea turtles, bats, or whales, Klaassen 1996, Mcguire et al. 2012, Braithwaite et al. 2015, Bonnet et al. 2016), and metabolic changes for animals travelling between contrasting environments (e.g. anadromous and catadromous fishes traveling from fresh to salt waters and vice versa, Fontaine 1976; McCormick and Saunders 1987). For some species, the preparation involves the reorganisation of body structures to improve locomotory performance (e.g. development of full-size wings by aphids, or increase of the muscular mass and atrophy of the digestive system by birds, Newton 2008, Milner- Gulland et al. 2011).

Departure is determined at different temporal scales by circannual and circadian cycles. At the seasonal scale, changes in resource abundance and quality trigger the migration of large herbivores among foraging grounds (Fryxell & Sinclair 1988, Holdo et al. 2009). Seasonal changes in photoperiod influence the hormonal cycles of species (e.g. gonad development in birds and fish) and the initiation of migration (Gwinner 1996, Dawson et al. 2001, Alerstam et al. 2003, Binder et al. 2011). Similarly, seasonal variation in temperature triggers the spawning migration of fishes (Binder et al. 2011). At the temporal scale of the day

Chapter 1: The accuracy of Fastloc-GPS locations

of departure, animals usually depart under favourable environmental conditions which maximise the probability of successfully reaching the destination (Åkesson & Hedenström 2007, Alerstam 2009, Müller et al. 2016). Species depart at certain time of the day to reduce predation rate, energy expenditure (via passive transportation), water loss and water loss and to maximise orientation (Alerstam 2009, Müller et al. 2016). Example of species departing during the day time include insects or birds (e.g. desert locusts, throated hummingbirds, Kennedy 1951, Willimont et al. 1988) while many migratory songbird species depart at dusk to avoid predators (Åkesson et al. 1996, Alerstam 2009, Müller et al. 2016). A range of terrestrial and aquatic species (e.g. mammals, fishes) terminate their migration once they reached a suitable foraging or breeding ground (Newton 2008, Lohmann et al. 2008, Binder et al. 2011, Milner-Gulland et al. 2011).

On the way to their foraging or breeding grounds, migratory species face multiple challenges. They need to use their energy stores efficiently and navigate to their destination while avoiding predators and unfavourable conditions (Alerstam et al. 2003; Åkesson and Hedenström 2007). Many species, including birds (Åkesson & Hedenström 2007, Newton 2008), insects (Kennedy 1951, McCord & Davis 2012) and mammals (Sawyer & Kauffman 2011, McGuire et al. 2012) stop during their migration to rest and/or refuel (referred to as ‘stopovers’). Furthermore, the distribution of stopover sites is important, especially in extreme environments (e.g. in the Sahara desert, Biebach et al. 1986), and missing a stopover may be fatal for the animal in some instances (Alerstam et al. 2003). Yet, stopping is not always possible, such as when birds pass over open oceans, deserts or mountain ranges (Åkesson & Hedenström 2007, Vardanis et al. 2011, Bishop et al. 2014). In such cases, non-stop travel is required to reach the next safe area; thus, these animals must develop strategies to rest while actively travelling. Understanding how animals balance these factors to complete their migration successfully is another major research topic (e.g. Newton 2008; Bowlin et al. 2010, Binder et al. 2011; Milner-Gulland et al. 2011; Avgar et al. 2014). Stopovers may be important for marine capital breeders (i.e. species relying on their energy reserved during the reproduction) and the use of stopover sites was detected in two instances for sea

Chapter 1: The accuracy of Fastloc-GPS locations

turtles (Rice & Balazs 2008, Baudouin et al. 2015). Overall, our knowledge on whether and how marine species use stopover sites is limited.

The exact location where individuals terminate migration depends on the species, with some exhibiting strict fidelity to a breeding or foraging ground and others simply seeking suitable habitat (Lennox et al. 2016). Furthermore, species that migrate during the daytime tend to arrive during the daytime (Kennedy 1951, Strandberg & Alerstam 2007), whereas those that migrate at night tend to arrive before dawn (Biebach et al. 2000, McGuire et al. 2012) (Figure 1.1). Furthermore, while some species maintain the diel patterns exhibited at breeding and foraging grounds during migration (e.g. bats, ospreys, Strandberg and Alerstam 2007; Mcguire et al. 2012), others alter their circadian rhythm specifically for migration (e.g. songbirds, Alerstam 2009) (Figure 1.1). In comparison, for a range of marine species, it is far more common to be able to detect the day of departure or arrival, rather than the actual time of day, due to the limited volume and accuracy of transmitted locations (e.g. sea turtles, white sharks, whales, Blumenthal et al. 2006; Mate et al. 2011; Domeier and Nasby-Lucas 2013; Schofield et al. 2013b).

Understand how animals complete those four steps is central to obtain a comprehensive knowledge of the strategies animals use to successfully complete their migration.

Chapter 1: The accuracy of Fastloc-GPS locations

Figure 1.1: Illustration of the diversity of departure times (horizontal black arrows) during a one-day period (horizontal grey arrow) for a selected range of migratory species. Vertical grey dashed lines represent the timing of sunrise, midday, sunset and midnight. Day-migrating species like pigeons, doves or migratory desert locusts tend to depart shortly after the sunrise. Hummingbirds following inland migration routes forage at dawn and dusk when the flowers are filled with nectar and depart at midday when the nectar is less abundant and accessible. Soaring birds tend to depart around midday when the atmospheric conditions allow for the formation of warm ascending atmospheric currents. Most songbird species depart at sunset when they can calibrate their star and magnetic compasses. Nocturnal species (e.g. bats, lampreys or northern leopard frogs) tend to depart after sunset. Note, many other strategies exist. For example, different departure times can be observed within a given species or taxon (e.g. for hummingbirds, songbirds, waterfowls) and often depends of the local conditions. References: [1] Newton 2008, [2] Goymann et al. 2010, [3] Klaassen et al. 2008, [4] Kennedy 1951, [5] Milner-Gulland et al. 2011, [6] Almeida et al. 2002, and [7] Sidney and Gauthreux 1980.

Foraging and mating behaviour In contrast to migration, foraging involves shorter and much more complex patterns of movement, often within a defined area known as a home range. The home range is an area over which an animal or group of animals regularly travel in search of food or mates, and which may overlap with those of neighbouring animals or groups of the same species (Burt 1943, McLoughlin & Ferguson 2000). Animals can

Chapter 1: The accuracy of Fastloc-GPS locations

exhibit residency to specific areas (termed as range residency) where they occupy relatively small areas (compared to the geographical distribution of the population) over long periods of time (Maher et al. 2000, Mueller et al. 2011). Residency usually occurs when resources are sufficiently abundant across the entire population range or when the animal are in dormant states (e.g. hibernating) (Maher et al. 2000, Mueller et al. 2011). Food within a home range is rarely uniformly distributed but is rather aggregated in patches (i.e. vegetation patches for herbivores or prey patches for carnivores). Optimal foraging theory describes how an animal can maximize its food intake while minimizing the cost of travelling among food patches (Emlen 1966, MacArthur & Pianka 1966, Charnov 1976). Within this theory, Charnov’s (1976) marginal value theorem predicts that an animal should leave a patch when the food intake within the patch is less than the average food intake of the whole habitat. This theoretical framework has been successful at predicting the foraging behaviour of large terrestrial herbivores (e.g. Dorcas gazelles Gazella dorcas, elk Cervus elpahus, cattle Bos taurus, Searle et al. 2005). However, state-dependant behaviours (e.g. competition or predation rate) should also be considered when attempting to predict patch-use duration and departure decisions with accuracy (Nonacs 1991).

The application of the optimal foraging theory to patchy marine ecosystems is currently limited to easily-trackable marine birds and mammals. For example, Foo et al. (2016) found that the behaviour of lactating female Australian fur seals (Arctocephalus pusillus doriferus) supports the predictions of some optimal foraging models. Similarly, Mori and Boyd (2004) found that Antarctic fur seals (Arctocephalus gazella) adjusted their time budgets to maximize the mean rate of energy intake in response to varying prey abundance, as predicted by the optimal foraging theory. In contrast, Watanabe et al. (2014) found that the foraging behavior of Adélie penguins (Pygoscelis adeliae) was only in accordance with the marginal value theorem under certain conditions. With the advance of more accurate tracking technologies such as Argos-Linked Fastloc-GPS, I expect an increase in the number of studies investigated the foraging pattern of a range of marine species including whether or not their foraging behaviour can be described by optimal foraging theory. 10

Chapter 1: The accuracy of Fastloc-GPS locations

Study species

In this PhD, I focussed on three long-distance migratory species, the green turtle, Chelonia mydas, the loggerhead turtle Caretta caretta and the humpback whale Megaptera novaeangliae. All three species are considered to be capital breeders, which means that they rely on energy reserves acquired while feeding at foraging grounds to accomplish their migration and reproduction (Hamann et al. 2002, Bonnet et al. 2016). The two species of sea turtles are ectotherms and lay multiple clutch of eggs on sandy beaches while the humpback whale is an endotherm and gives birth to a calf which is then fed by the female for at least one year (Dawbin 1966, Lutz & Musick 1996). The research within this thesis provides new insights on the migration strategies of those three species, which will improve our understanding on how they manage their energy reserves during their life cycle.

Biology and movement ecology of sea turtles Phylogeny Terrestrial turtles appeared in the late Triassic about 200 million years ago (Pritchard 1997). During the Early Cretaceous, at least 110 million years ago, sea turtles returned to the marine environment (Hirayama 1998). Between 110 and 70 million years ago sea turtles radiated into at least 14 different lineages, from which two are still extant (from the families Cheloniidae and Dermochelyidae) (Kear & Lee 2006). Today, there are seven extant sea turtle species, of which six are cheloniids and one is a dermochelyid. The six species of cheloniids (green turtle Chelonia mydas, loggerhead turtle Caretta caretta, Kemp’s ridley Lepidochelys kempii, olive ridley, Lepidochelys olivacea, hawksbill turtle Eretmochelys imbricata and flatback turtle Natator depressus) are characterised by a hard shell, while the only remaining dermochelyid species, the leatherback turtle Dermochelys coriacea, is characterised by a soft shell (Lutz & Musick 1996).

Distribution The six species of hardshell sea turtles are true ectotherms (Lutz & Musick 1996) and inhabit warm tropical waters. Green, hawksbill, loggerhead, olive riley are distributed worldwide while the flatback turtle is endemic to Australia (Oceania)

Chapter 1: The accuracy of Fastloc-GPS locations

and Kemp’s ridley is endemic to Mexico and the US (Gulf of Mexico) (Lutz & Musick 1996, Lutz et al. 2002). Leatherback turtles are thermoregulating ectotherms and their distribution extends to colder waters at higher and lower latitudes (Paladino et al. 1990, Lutz & Musick 1996).

Life cycles All species of sea turtles are oviparous and nest on sandy beaches. Hatchling sex is determined in the egg by incubation temperatures, a process known as temperature- dependent sex determination (Girondot 1999). In turtles, more males are produced at temperatures <29 °C, while more females are produced at temperatures >29 °C (Girondot 1999). Incubation lasts between 6 and 13 weeks, depending on the species (Lutz & Musick 1996). When hatchlings emerge from the nest, they crawl seawards and start a period of at least 24-hour of sustained swimming (termed swimming frenzy) during which they move offshore (hatchling dispersal, Figure 1.2; Wyneken and Salmon 1992; Lohmann and Lohmann 1996). During the first years of life, juvenile turtles generally (except flatbacks) remain in pelagic habitats, feeding opportunistically and inhabiting temperate and tropical waters, before shifting to neritic demersal habitats (juvenile dispersal, Figure 1.2; Lutz and Musick 1996). Information about the distribution and ecology of juvenile turtles during the oceanic stage remains limited and is often referred to as the ‘lost years’ (Carr 1986, Mansfield et al. 2014). During the lost years, turtle movement is a mix of passive drifting and active swimming (Putman et al. 2012, Mansfield et al. 2014). Once they reached maturity, adult sea turtles accomplish long-distance migrations between their foraging and breeding grounds (Figure 1.2a). The foraging grounds are either located in pelagic areas (e.g. for leatherback, olive ridley and some loggerhead turtles), neritic areas (e.g. for green turtles and some loggerhead turtles) or cover the two types of habitat (e.g. for some populations of loggerhead turtles) (Lutz & Musick 1996). Breeding grounds are widely distributed in tropical waters, either near continental coastlines, or close to remote islands. Because satellites tags are typically attached when the turtles are at the breeding grounds, post-nesting migrations are the best described type of migrations. Seasonal movement between summer foraging grounds and distinct overwintering grounds, along with a

Chapter 1: The accuracy of Fastloc-GPS locations

reduction of the sea turtle activity is sometimes observed when the sea water temperature drops (Figure 1.2a; e.g. Broderick et al. 2007; Hochscheid et al. 2007).

Figure 1.2: (a) Locations used in a typical life cycle by green and loggerhead sea turtles using neritic foraging grounds at the adult stage. Discrete locations used are shown in black. The movement of individuals among the different types of grounds is indicated by black arrows and the type of movement indicated in red. The duration of each step (in blue) is indicated for green turtles in the Indian Ocean and loggerhead turtles in the Mediterranean Sea with differences shown for males and females where appropriate. Migratory routes of: (b) loggerhead sea turtles; and (c) green sea turtles tracked with Argos-linked Fastloc-GPS from the breeding to foraging grounds. Thirty- three loggerhead male turtles (black lines) and 13 loggerhead female turtles (red lines) were tracked from Zakynthos (with some passing via Kyparissia, both breeding sites are presented as white circles) in Greece, Mediterranean Sea. Eight female green turtles (red lines) were tracked from Diego Garcia, Chagos Archipelago, Western Indian Ocean (breeding site represented as a white circle). Modified from Schofield et al. (2013) and Hays et al. (2014b). References: [1] Wyneken and Salmon (1992) [2] Casale and Mariani (2014) [3] Reich et al. (2007) [4] Hochscheid et al. (2007) [5] Hays et al. (2010) [6] Mortimer et al. (2011).

Chapter 1: The accuracy of Fastloc-GPS locations

Loggerhead turtles from the Zakynthos rookery Zakynthos is one of the largest nesting rookeries of loggerhead sea turtles (Caretta caretta) in the Mediterranean Sea (Margaritoulis 2005) with the mostly-omnivorous turtles (Bjorndal 1985) returning to the rookery with an average curved carapace length (CCL) of 83.4 ± 6.1 cm (Schofield et al. 2013a). Tracking of loggerhead turtles in Zakynthos using Argos-Linked Fastloc-GPS started around 2007 and provided insights on their post-nesting migration to different foraging grounds in the Mediterranean Sea (Figure 1.2b, in the Ionian Sea, the Gulf of Gabes, to Kyparissia, the Aegean Sea or the Adriatic Sea; Schofield et al. 2013a). Typical migration distance for loggerheads is 920 ± 409 km (range: 189–1545 km) over a mean 25 ± 10 days (range: 7–42 days). The re-migration interval is 2-3 years for females and one year for males (Hays et al. 2010). The large dataset of loggerhead tracks accumulated over the years has been used to determine the size of the foraging grounds (Schofield, Hobson, Fossette, et al. 2010), to compare the migration patterns of males to females (Schofield, Dimadi, et al. 2013) and to estimate efficiency of the marine protected area in Zakynthos (Schofield, Scott, et al. 2013). In parallel, multiple studies investigated other aspects of the biology of loggerhead turtles such as energetic expenditure during breeding or the in-water behaviour of loggerheads turtles (Schofield et al. 2006; 2007b; Fossette et al. 2012). This body of work provides a good framework against which to interpret the insights obtained by analysing high-resolution Fastloc-GPS data.

The green turtle population from Diego Garcia Diego Garcia is part of the Chagos Archipelago in the Indian Ocean. Sea turtles nesting on the island migrate to foraging grounds located in different parts of the Indian Ocean (e.g. Maldives archipelago, Seychelles Archipelago, eastern Somalia coast, Chagos Bank, Hays et al. 2014). Data from the Seychelles suggest that female green turtles in Diego Garcia might re-migrate at an interval of 3-5 years (Mortimer et al. 2011). Compared with loggerheads in the Mediterranean Sea, green turtles are larger (mean CCL: 105.6 ± 3.45 cm, Hays et al. 2014), migrate over longer distances and durations (2639 ± 1264 km for 44 ± 19 days, Hays et al. 2014) and are herbivores feeding mostly on seagrasses (Bjorndal 1980). Diego Garcia is not as

Chapter 1: The accuracy of Fastloc-GPS locations

well-studied as Zakynthos however, the green turtle tracking data from this population has been used to estimate the efficiency of the marine protected area surrounding the island (Hays, Mortimer, et al. 2014) and to estimate the number of clutches laid by the females (Esteban et al. 2017). These data provide a point of comparison with the smaller omnivorous loggerhead turtles from Mediterranean Sea.

Biology and movement ecology of humpback whales Phylogeny Whales are cetaceans, an order of mammals that originated in the Eocene about 50 million years ago for the archaic whales (where they moved from terrestrial to marine ecosystems). Modern whales appeared about 25 million years ago in the Oligocene (when they developed mass-feeding adaptations and echolocation, Thewissen et al. 2009, Marx and Uhen 2010). Cetaceans have the same characteristic as other mammals (e.g. milk production, nursing with their young, three ear bones involved in the sound transmission, and lower jaws consisting of a single bone) but possess a streamlined body with a horizontal tail fluke, flippers for forelimbs and no external hind limbs (Thewissen et al. 2009). The humpback whale Megaptera novaeangliae) is one of approximately 78 extant species in the order Cetacea and has an average body length of 13 metres and weights up to 40 tons (Clapham 1996).

Distribution Humpbacks whales inhabit all major oceans over a wide range extending from the Antarctic ice edge to a latitude of 77 °N. The four global populations are the North Pacific, Atlantic, Southern Ocean and Indian Ocean populations. A large body size, a thick layer of blubber and large energy reserves, derived from a fish and krill diet, allow the Southern Hemisphere populations to forage in the greater food concentration (especially of zooplankton) of Antarctic waters (Dawbin 1966, Laws 1977). Breeding grounds in the Southern Hemisphere are located in warmer tropical waters above a latitude of 25 °S (Dawbin 1966, Jenner et al. 2001).

Chapter 1: The accuracy of Fastloc-GPS locations

Migratory behaviour Humpback whales migrate between high-latitude foraging grounds to low-latitude breeding grounds where they mate and calve (Figure 1.3, Dawbin 1966). Similarly to sea turtles, they are considered to be capital breeders because they finance the cost of their reproduction using the energy stores acquired at foraging grounds (Kasuya 1995, Bonnet et al. 2016). During spring, summer and autumn, whales from the Southern Ocean and Indian Ocean populations forage in Antarctic waters feeding mostly on zooplankton (krill, Euphausia superba) and small fishes for building energy stores (Clapham 1996, Pauly et al. 1998). In late autumn, the whales undertake their migration to the breeding grounds which are typically located around islands or associated with reef systems (Clapham 1996). Gestation and lactation last for approximately 12 months each and both peak at the breeding ground between August and October in the Southern Hemisphere (Kasuya 1995, Clapham 1996). Because of the length of gestation and lactation, female humpback whales give birth to a single calf every two or three years (Clapham 1996).

Chapter 1: The accuracy of Fastloc-GPS locations

Figure 1.3: (a) Locations used in a typical life cycle of the humpback whale migrating between a summer foraging ground, where the animal build energy reserves, and a winter breeding ground where females give birth to their calf. The movement of individuals among the different types of grounds is indicated by black arrows and the type of movement indicated in red. The duration of each step is indicated in blue. (b) Location of the summer foraging grounds in the Antarctic Ocean (in red) and the winter breeding grounds (in blue) in Australia for the Southern Ocean and Indian Ocean populations. The two black arrows show the identified migratory corridor for this species in the Southern Hemisphere. References: [1] Barlow and Clapham (1997), [2] Dawbin (1966), [3] Jenner et al. (2001). Research aims and thesis structure

The aim of my thesis is to use Argos-linked Fastloc-GPS and UAVs to obtain new insight on the behaviour of migratory marine species. To illustrate the value of the two technologies, I used three species of marine migratory vertebrates (the green turtle, the loggerhead turtle and the southern humpback whale) to cover the different phases of a typical migratory cycle. I first used Argos-linked Fastloc-GPS trajectories of green and loggerhead turtles to illustrate the behavioural patterns of sea turtles during their migration, while they are at their foraging ground and when they are overwintering (Chapter 2 to 4, Figure 1.4). In particular, I focused on 17

Chapter 1: The accuracy of Fastloc-GPS locations

patterns that occur at small spatiotemporal scales (e.g. within a day or a few km). Because it is currently logistically difficult to estimate energy expenditure for sea turtles, I used the humpback whales as a model species to examine the cost of reproduction for the humpback whale at their breeding grounds using UAVs (Figure 1.4).

Figure 1.4: Simplified migration cycle for both sea turtles and humpback whales. Black arrows represent the direction of movement of the animals among the different types of grounds. The grey arrows indicate how each component of the migration is covered by the different chapters of this thesis.

Chapter 2 aimed to quantify the typical accuracy of Argos-linked Fastloc- GPS locations and to investigate the impact of Fastloc-GPS on commonly-used metrics such as the speed of travel or heading. I performed a fixed trial where >45 000 locations were obtained to empirically estimate the accuracy of locations and I subsequently incorporated them into procedures for determining the speed of travel and heading. I then illustrated my methodology using green turtles migrating over 3000 km between Diego Garcia and Somalia in the Indian Ocean (Figure 1.4). This chapter provides an important basis for subsequent studies presented in this thesis. This chapter has been published in Methods in Ecology and Evolution.

Chapter 3 aimed to use the high quality of Argos-linked Fastloc-GPS locations to investigate the timing of departures from a breeding ground and arrival at a foraging ground, along with the diel travel speed patterns and the use of stopover sites during the migration for the green and loggerhead turtles (Figure 1.4).

Chapter 1: The accuracy of Fastloc-GPS locations

In addition, I used these data to test hypotheses about the type of cues used by sea turtles and whether or not they exhibit resting behaviours during their migration. This chapter has been submitted to Marine Biology.

Given that sea turtles use energy reserves collected while at foraging grounds during reproduction, investigating the movement of sea turtles at small spatial scales (i.e. the scale at which they feed) is a key step in understanding how sea turtles manage their energy reserves. Chapter 4 aimed to investigate spatiotemporal variation in the use of space by green turtles foraging in the Indian Ocean (Figure 1.4). I determined the number and the size of patches used by green turtles within their foraging ground and whether they used different areas during day and night. Similarly, I quantified the variation in the patches used across seasons. These data were used to assess the potential roles of food availability and predation in driving spatiotemporal patterns in the home range of the green turtle, providing insight on the complexity of green turtle movement. This chapter was published in Marine Biology.

Compared to green turtles, loggerhead turtles in the Mediterranean Sea are smaller, migrate over shorter distance and are omnivorous. Chapter 4 aimed to investigate how loggerhead sea turtles use benthic foraging and wintering habitats at multiple scales, from the general site-scale to fine-scale patches and day/night activity centres within patches. Similar to the aim of Chapter 3 for green turtles, I addressed a key ecological question on sea turtle movement: how do loggerhead sea turtles utilise their foraging areas in space and time and how do different movement patterns emerge at different spatial scales (Figure 1.4). This approaches allowed me to compare foraging strategies between the two species in two different oceans. This chapter has been submitted to Marine Ecology.

Changes in the body condition of sea turtles while at breeding sites is logistically challenging to measure. Therefore, the choice of a species easier to observe (e.g. a large animal) can sometimes be more appropriate to understand the relationship between energy expenditure and reproduction for capital breeders. I illustrate this in Chapter 5, which aimed to assess change in body condition for a capital breeder, the humpback whale at a breeding/resting ground from aerial 19

Chapter 1: The accuracy of Fastloc-GPS locations

photographs recorded using a UAV (Figure 1.4). Compared to sea turtles, whales are endothermic and the lactating females feed their calves for up to 12 months, which is theoretically associated with larger energy expenditure compared to ectotherms. This allowed me to measure variations in body condition, which I expected to be important over the breeding period, to build statistical models and to gain insights that could subsequently be applied to sea turtles. This chapter has been published by Ecosphere.

Chapter 1: The accuracy of Fastloc-GPS locations

Chapter 2 : The accuracy of Fastloc-GPS locations and implications for animal tracking Abstract

Over recent years, a breakthrough in marine animal tracking has occurred with the advent of Fastloc-GPS that provides highly accurate location data even for animals that only surface briefly such as sea turtles, marine mammals and penguins. I assessed the accuracy of Fastloc-GPS locations using fixed trials of tags in which >45 000 locations were obtained. Procedures for determining the speed of travel and heading were developed by simulating tracks and then adding Fastloc-GPS location errors. The levels of detail achievable for speed and heading estimates were illustrated by using empirical Fastloc-GPS data for a green turtle (Chelonia mydas, Linnaeus, 1758) travelling over 3000 km across the Indian Ocean. The accuracy of Fastloc-GPS locations varied as a function of the number of GPS satellites used in the location calculation. For example, when Fastloc-GPS locations were calculated using 4 GPS satellites, 50 % of locations were within 36 m and 95 % within 724 m of the true position. These values improved to 18 m and 70 m respectively when 6 satellites were used. Simulations indicated that for animals travelling around 2.5 km h-1 (e.g. turtles, penguins and seals) and depending on the number of satellites used in the location calculation, robust speed and heading estimates would usually be obtained for locations only 1 to 6 hours apart. Fastloc-GPS accuracy is several orders of magnitude better that conventional Argos tracking or light-based geolocation and consequently will allow new insights into small scale movement patterns of marine animals.

Chapter 1: The accuracy of Fastloc-GPS locations

Introduction

Understanding the movements of animals over a range of spatial and temporal scales lies at the heart of many ecological studies (Nathan 2008, Nathan et al. 2008, Nathan & Giuggioli 2013). Consequently, there are several well established approaches for tracking a range of species. Acoustic tracking is used to follow marine and freshwater species that do not surface, such as fish and invertebrates (Espinoza et al. 2011, Moland et al. 2011, Coates et al. 2013). Radio tracking using earth based receivers is used to follow animals over fairly short distances (typically a few kms) (White & Garrott 1990, Millspaugh & Marzluff 2001, Godfrey & Bryant 2003). Argos satellite tags and light-based geolocator tags have both been widely used to track large-scale movements, sometime many tens of thousands of km (see the Bridge et al. 2013 review). However often these techniques provide only fairly coarse quality locations: for example, Argos locations typically have an accuracy of several hundred meters to several km while positions estimates from light-based geolocation are accurate to only tens of km (Hays, Åkesson, et al. 2001, Teo et al. 2004, Witt et al. 2010).

Set against this backdrop of established technologies that have existed for a decade or more, there have been some major recent advances in animal biotelemetry (Rutz & Hays 2009). In particular for high resolution tracking, GPS tracking has emerged for terrestrial animals and flying birds, where there is near-continuous line of sight with satellites overhead; while Fastloc-GPS has emerged as an approach that provides high resolution locations for widely ranging marine species that are only briefly visible to the GPS satellites (Schofield, Bishop, et al. 2007, Rutz & Hays 2009). As such, Fastloc-GPS is a major breakthrough for marine animals that only surface briefly, such as sea turtles and marine mammals. Conventional GPS receivers take several seconds to generate a location estimate from a “warm start”, knowing current time within 20 seconds, current position within 100 km, and having valid almanac data. This has precluded their use on such marine taxa (Lehtinen et al. 2008, Tomkiewicz et al. 2010). However, Fastloc-GPS overcomes this problem and involves the rapid (typically tens of milliseconds) acquisition of

Chapter 1: The accuracy of Fastloc-GPS locations

GPS data when an animal surfaces and subsequent post-processing to derive position estimates (Tomkiewicz et al. 2010). Trials by the company that developed Fastloc-GPS (Wildtrack Telemetry Systems Ltd, Leeds, UK) have shown that locations are generally within a few tens of meters of the true position (Bryant 2007) and some limited trials by users have confirmed this general level of accuracy (Hazel 2009, Costa et al. 2010, Hoenner et al. 2012). Here I conduct a comprehensive assessment of the accuracy of Fastloc-GPS locations using fixed trials in which more than 45 000 Fastloc-GPS locations were obtained. Hence the outcomes of these trials will inform the ever-growing community of users for this technology of their reliability and accuracy.

It is widely known that the accuracy of tracking data influences the ability to distinguish biological signals from sampling noise in the analysis of movement patterns (Bradshaw et al. 2007, Hurford 2009). For example, with Argos location data that have been used for 20 or more years, I have previously shown how inaccurate speed of travel estimates are obtained if calculated for locations that are too close in time: in that case, the location inaccuracy may dominate compared to actual movement made by the animal (Hays, Åkesson, et al. 2001). So, for example, with Argos data I have previously suggested that locations need to be >100 km apart (i.e. 2 days apart if the animals is travelling at 50 km d-1) in order to calculate accurate speeds of travel (Hays, Åkesson, et al. 2001). In this same way, here I consider how the accuracy of Fastloc-GPS locations can be built into future procedures for calculating speed of travel and heading, parameters that lie at the heart of many movement studies (Sato et al. 2007, Codling et al. 2008, Bartumeus et al. 2008, Wilson et al. 2013).

Methods

Trials with Fastloc-GPS tags in a fixed location Between 14 May and 26 November 2013, 257 tags of assorted models equipped with Fastloc-GPS were deployed (at approximately 47°40' 36”N, 122° 08' 10”E) in an open space with an unobstructed view of the sky. Tags were always deployed in the same general area within 4 m of each other. Raw GPS data snapshots were

Chapter 1: The accuracy of Fastloc-GPS locations

collected and pre-processed into pseudo-ranges on-board the tags using the Fastloc system (Version 2.3, Wildtrack Telemetry Systems Ltd, Leeds, UK). The DAP Processor (Version 3.0, Wildlife Computers, Redmond, WA) obtained the relevant daily broadcast satellite ephemeris data (maintained by NASA http://cddis.gsfc.nasa.gov) and post-processed the pseudo-ranges into location estimates. For each location, the software reported; the number of satellites used, the ID numbers of those satellites, the estimated timestamp error, and the “residual” value of the solution. The total number of satellites can vary due to their availability within the view of the sky. The “residual” value is a measure of how well the solution matched the observed data. I assumed that the average of all locations obtained from an individual tag represented that tag’s true location.

Implications of Fastloc-GPS accuracy for speed of travel and heading derivation To investigate how location accuracy could impact calculations for speed of travel and heading (relative to geographic north), I started with theoretical tracks for animals travelling in straight-lines at various different speeds. Straight lines were selected simply to aid computational efficiency but will not impact the overall conclusions. I assumed speeds of travel of 1 km h-1, 2.5 km h-1or 5 km h-1 (24, 60, 120 km day-1). This range of speeds covers those typically seen in travelling marine animals such as sea turtles, marine mammals and penguins (see the Sato et al. 2007 review of swimming speed estimations for seabirds, pinnipeds and cetaceans and Luschi et al 1998 for sea turtles). To locations along the straight-line tracks, I added a location error randomly selected from the empirical errors found in the fixed trials. I selected location errors corresponding to those based on 4, 5 or 6 satellites, as they represent the largest expected errors. Speed and heading were then calculated by subsampling the simulated track at different time steps ranging from 1 to 24 hours. In this way for combinations of locations obtained with 4, 5 or 6 satellites, I estimated 125 000 speed of travel and heading estimates.

To determine whether a sampling interval between Fastloc-GPS locations would be long enough to produce an accurate speed and heading estimate, I defined a quality criterion corresponding to 95 % of estimated speeds and headings being within 10 % and 10 °, respectively of the true values (similar to Hays et al. 2001).

Chapter 1: The accuracy of Fastloc-GPS locations

Study case: derivation speed of travel and heading during migration of green turtles

To validate the procedures developed from the fixed location trials and the calculations with simulated tracks, I examined the level of temporal detail that was possible to obtain for speed of travel and heading estimates using a previously published track for a green turtle (Chelonia mydas). The turtle migrated across the Indian Ocean from breeding areas on Diego Garcia in the Chagos Archipelago (Hays, Mortimer, et al. 2014). This turtle was equipped with Fastloc-GPS Argos tag, i.e. the tag had a Fastloc-GPS receiver combined with an Argos transmitter to remotely relay the data (for full description see Hays et al. 2014b). To process these data, they were first filtered using a threshold residual value of 35. I then removed the remaining locations that looked visibly erroneous (i.e. appeared on land) when the tracks were viewed in Google Earth. An analysis of the speed of travel always confirmed these locations necessitated unrealistic speeds of travel (>200 km day-1). I designated the start of the post-nesting migration as the time at which the turtle left Diego Garcia and began it oceanic crossing, which continued until it arrived at the foraging ground. I next calculated speed and heading using the corresponding sampling interval determined from track simulations. I assumed a 2.5 km h-1 average speed of travel (Luschi et al. 1998, Hays, Mortimer, et al. 2014).

To assure the veracity of the calculated speed and heading time-series during migration, I calculated the autocorrelation of these values between each location at locaiton N and N+1 (Hays, Åkesson, et al. 2001). I expect a high autocorrelation value if the speed and heading estimations were coherent (Dray et al. 2010).

All analysis and simulations were performed using R software version 3.0.1 (R Development Core Team 2013). Autocorrelation value for heading was calculated using the package CircStats (Lund & Agostinelli 2014).

Chapter 1: The accuracy of Fastloc-GPS locations

Results

Typical locations accuracy Fixed trials produced a total of 45 157 snapshots with the number of satellites used to calculate locations ranging from 4 to 11 (Table 2.1). Locations using 4 to 8 satellites represented 93.9 % of total data. For each Fastloc-GPS location, I calculated the deviation from the assumed true position (distance in meters). Plots of the deviation of locations showed that the number of satellites used in the position calculation had a clear impact on location accuracy, with locations calculated using more satellites tending to be more accurate (Figure 2.1). For example, when Fastloc- GPS locations were calculated using 4 satellites, 50 % of locations were within 36 m and 95 % within 724 m of the true position. When they were calculated using 6 or more satellites, at least 50 % of locations were within 18 m and 95 % within 70 m of the true position, illustrating the increase of accuracy (see Table 2.1 for each number of satellites statistics). The “residual” value generated with each Fastloc- GPS location was also related to the location accuracy (Figure 2.2). However, this relationship was not simple. Rather, high residual values were more often associated with inaccurate locations, while for Fastloc-GPS locations that had low residual values, then the accuracy could be either good or poor. So selecting a residual value of 35 to filter locations, would certainly help to remove some of the most inaccurate Fastloc-GPS locations, but not all. Filtering in this way using a residual value of >35 had a particularly marked impact on removing the most inaccurate locations obtained with 4 satellites. Before filtering, 95 % of locations were within 1163 m of the true location. After filtering, this value decreased to 724 m. However, when looking at a distance within which 50 % of locations were found, these values before and after filtering were much closer, being 37 and 36 m respectively (Table 2.1).

Chapter 1: The accuracy of Fastloc-GPS locations

Figure 2.1: The deviation of Fastloc-GPS locations (in meters) from the true position based on the number of GPS satellites (black numbers) used in the position calculation. The deviation was sorted by magnitude before plotting. In this way, one can readily assess the distance from the true location that 50 % or 95 % of locations would fall within. To improve readability, the deviation from true location axis is truncated at 100 meters.

Figure 2.2: The deviation (in meters) of Fastloc-GPS locations from the true position for 45 157 Fastloc-GPS locations as a function of their residual value. The residual value is used as a quality index by the manufacturers to filter and remove a fraction of the locations with low accuracy. A residual value of 35 is typically used (vertical solid black line). More than half of locations (27 011, 59.8 %) have a residual value between 1 and 10.

Chapter 1: The accuracy of Fastloc-GPS locations

Table 2.1: Accuracy statistics of a fix trial experiment involving 257 Fastloc-GPS tags and 45 157 locations.

Number of satellites used for 4 5 6 7 8 9 10 11 location calculation

Number of locations unfiltered data 9898 9835 8882 7704 6096 2295 371 76 Number of locations filtered data 9718 9819 8869 7692 6080 2282 367 76 95th percentile unfiltered data (m) 1163 169 71 43 34 28 24 19 95th percentile filtered data (m) 724 165 70 43 33 27 22 19 75th percentile of unfiltered data (m) 118 55 30 22 19 16 13 15 75th percentile of filtered data (m) 109 55 30 22 19 16 13 15 50th percentile of unfiltered data (m) 37 29 18 14 12 10 8 7 50th percentile of filtered data (m) 36 29 18 14 12 10 8 7 25th percentile of unfiltered data (m) 17 16 10 8 7 6 5 5 25th percentile of filtered data (m) 17 16 10 8 7 6 5 5

Estimating the sampling interval for speed and heading derivation The probability that accurate speed of travel and heading estimates were calculated tended to increase as the time interval between locations lengthened. For example, in a very conservative way, across all combinations for the number of satellites used in the Fastloc-GPS calculations, accurate speeds and headings were obtained on more the 95 % of occasions when locations were more than 12 h apart, regardless of whether the animal was travelling at 1 km h-1 or 5 km h-1 (Figure 2.3). However, as the time interval between locations became shorter, then the number of satellites used in the Fastloc-GPS location became increasingly important as did the speed of travel of the animal. For example, for an animal travelling at 2.5 km.h-1, accurate (within 10 % and 10 ° respectively) speeds and headings were obtained on more the 95 % of occasions when the interval between locations was 6 h for pairs of locations determined using combinations of 4-4, 4-5 and 4-6 GPS satellites and 1 h for pairs of locations determined using combinations of 5-5, 5-6 and 5-6 GPS satellites. In short, there was a step increase in the utility of the Fastloc-GPS locations when ≥5 GPS satellites were used in the location calculation.

Chapter 1: The accuracy of Fastloc-GPS locations

Study case on green turtles tracks The tracked turtle migrated from the Chagos Islands to the Somalian coast (Figure 2.4). Most (84.6 %) of the locations were calculated using 5 or more satellites, meaning that I could expect 80.4 % of the data to have an accuracy of 169 m or less (Table 2.1). Number of daily locations ranged from 1 to 37 with a mean of 20 (SD = 9) locations per day. Time interval between two uplinks ranged from 14 minutes to 42 hours with a mean interval of 1 hour 15 minutes. Following the results of my track simulations, I calculated the speed of travel and heading with a sampling interval of 6 h for pairs of locations that included at least one location calculated using 4 GPS satellites, while a sampling interval of 1 h was used for all other possible pairs (i.e. 5 or more GPS satellites used in the location calculation). Speed of travel and heading showed marked variations during the migration with a high autocorrelation (r = 0.90 for speed of travel and r = 0.92 for heading, Figure 2.4). During the oceanic crossing the speed of travel of this turtle varied between about 0.5 km h-1 and almost 6 km h-1, a 12-fold variation. Heading tended to change monotonically for several days indicating a gradual change in course heading. For example, between 25/01/2013 and 28/01/2013 heading changed from 347 to 298 °. Then, interspersing these monotonic changes in heading, were several reversals in heading. For example, around 02/02/2013, the heading had been increasing from 273 to 320 °, but then reversed and changed back to 265 °. The heading reversals corresponded with distinct turns in the track (Figure 2.4). This case study, obtained from in the field Fastloc-GPS tracking data, confirms that high-resolution heading and speed of travel time series can be obtained from animal trajectories.

Chapter 1: The accuracy of Fastloc-GPS locations

Figure 2.3: Analysis of the sensitivity of speed of travel and heading estimations to variations of the number of satellites used to determine Fastloc-GPS locations. Three different speeds of travel where considered (1 km h-1 (a, b), 2.5 km h-1 (c, d) and 5 km h-1 (e, f)). Sampling intervals range from 1 to 24 h. Speed of travel and heading estimates tended to be more accurate when the animal was travelling faster, when more GPS satellites were used in the calculation of the Fastloc-GPS position and when the time interval between locations increased. I considered a sampling interval to be long enough when 95 % of speed of travel and heading estimates had an error less than 10 % or 10 ° respectively. A solid horizontal line indicates this threshold on the plots.

Chapter 1: The accuracy of Fastloc-GPS locations

Figure 2.4: Track of a green turtle migrating from the Chagos Islands to coast of Somalia (a). Triangles show the start and end of the migration and black arrows shows the main turns observed during the migration. Speed of travel (b) and heading (relative to the north) (c) for this turtle as a function of time. The black arrows correspond to the turns indicated in (a). Speed of travel and heading autocorrelation (d and e) between a value at time N and N+1.

Chapter 1: The accuracy of Fastloc-GPS locations

Discussion

My study provides a comprehensive assessment of the accuracy of Fastloc-GPS locations and hence will help inform efforts to extract the most biological information from the extensive Fastloc-GPS tracking data-sets that are now emerging for a diverse range of taxa. These taxa include pinnipeds (Vincent et al. 2010, Costa et al. 2010), fish (Sims et al. 2009, Evans et al. 2011), turtles (Hazel 2009, Schofield, Hobson, Fossette, et al. 2010, Hays, Mortimer, et al. 2014), whales (Mate 2012) and penguins (Bost et al. 2011). To investigate Fastloc-GPS accuracy, previous trials have been performed before by the developer of this technology (Wildtrack Telemetry Systems Ltd, Leeds, UK) using fixed location trials (Bryant 2007). Scientists in the field also used Fastloc-GPS tags as a crosscheck reference to investigate Argos locations accuracy (Hazel 2009, Costa et al. 2010, Hoenner et al. 2012). My results are broadly in accord with these previous findings. For example, I reported that locations obtained with 4 GPS satellites were within 724 m of the true position on 95 % of occasions, and 36 m on 50 % of occasions, while the respective values reported by Bryant (2007) were 810 m and 50 m.

So my results provide confirmation of the accuracy of Fastloc-GPS. Clearly Fastloc-GPS provides much more accurate locations than other approaches that have been used over recent decades for tracking marine species that range widely. For example, conventional Argos tracking typically gives locations that are a few kilometers from the true position. Witt et al. (2010) reported that 95 % of Argos locations of quality A and B (these often dominate in marine tracking studies), were within 3.5 ± 9.2 and 14.3 ± 135.6 km of the true position, respectively. Similarly light-based geolocation provides fairly crude position estimates. Fudickar et al. (2012) have reported that light-based geolocator position estimates are typically up to 200 km from the true position. Set against this backdrop, it is clear that Fastloc- GPS provides at least an order of magnitude improvement in location accuracy (i.e. if more than 5 satellites are used to calculate a location, the improvement will be approximatively 10 to 40 times better than the average Argos accuracy and 1,100 times better than light-based geolocators). This opens up the way for more detailed analysis of the spatio-temporal patterns of animal movement (Schofield, Bishop, et 32

Chapter 1: The accuracy of Fastloc-GPS locations

al. 2007, Hazel 2009, Schofield, Dimadi, et al. 2013, Hays, Mortimer, et al. 2014). For terrestrial animals and flyers (birds and bats) the necessity for fast acquisition of GPS ephemeris is not so important and regular GPS tags can be used which take several seconds to determine a position from switch on (Tomkiewicz et al. 2010).

Bajaj et al. (2002) report typical GPS accuracy ranges in wildlife tracking are between 18 and 91 m and can be reduced to less than ten meters if a differential correction is applied (Rempel & Rodgers 1997). For 8 different GPS collars models used for animal tracking, Villepique & Bleich (2008) reported than typically 50 % of locations were between 5 to 20 meters and 95 % of locations between 20 to 68 m of their true location. This is similar to the accuracy of Fastloc-GPS locations calculated using 6 or more satellites, i.e. about two thirds of the locations I recorded from a migrating turtle. In short, these considerations suggest that Fastloc-GPS is often as accurate as GPS locations provided in terrestrial animal tracking.

Figure 2.5: Typical accuracy (95th percentile accuracy) of commonly used tracking technologies (horizontal black bars) and of Fastloc-GPS locations (horizontal grey bars). For each technology, 95 % of locations have an accuracy within the range represented by the corresponding horizontal bar. (Source for light based geolocators, Fudickar et al. (2012); Argos, Witt et al. (2010); GPS collars, Villepique & Bleich (2008); Fastloc-GPS, this study).

Chapter 1: The accuracy of Fastloc-GPS locations

My results confirm that the “residual” value provides a useful first step in filtering Fastloc-GPS data as it removes some of the largest location outliers. This type of filtering is typically reported in manuscripts describing Fastloc-GPS results and often done automatically by manufacturers (Sirtrack 2010, Witt et al. 2010, Shimada et al. 2012). In particular, my results show that this filter applies importantly to locations obtained with 4 GPS satellites. Once 5 or more satellites are used in location calculations, then few locations will be removed by this filter. But overall, the residual value provides a less clear indicator of the location accuracy compared to the number of satellites used in the location calculation.

As well as describing Fastloc-GPS accuracy, I also explored the implications of location accuracy for estimates of speed of travel and heading. This problem is simple to understand conceptually, but harder to build into a rigorous analysis. Conceptually, it is widely known that, as the true distance between two locations decreases, so the inaccuracy of the location estimates will increasingly dominate the calculations for speed of travel and heading (Hays, Åkesson, et al. 2001, Bradshaw et al. 2007, Hurford 2009). Importantly, users need to know when reliable speeds and heading estimates are likely to be obtained under a range of different scenarios. For example, if an animal travels 200 m in 15 minutes (i.e. 1 km h-1), could this speed be accurately measured with Fastloc-GPS? In a previous study, Schofield et al. (2010) estimated speed of travel of loggerhead turtles (Caretta caretta, Linnaeus, 1758) tagged with Fastloc-GPS using one location per day. Assuming a speed of 2.5 km h-1, I showed that interval could be reduced to 6 hours if the locations were calculated using 4 satellites and to 1 hour if they were calculated using at least 5 satellites, increasing the number of possible speed of travel and heading estimations. So my simulations of animals moving at different speeds should provide some simple rules-of-thumb for users to correctly interpret Fastloc-GPS data. Of course, further smoothing of the tracking data and/or incorporation into models that take account of the location error structure (e.g. state space models, Jonsen et al. 2005) may refine the interpretation of Fastloc-GPS tracking data but will still need to incorporate estimates of location accuracy.

Chapter 1: The accuracy of Fastloc-GPS locations

The value of Fastloc-GPS tracking is that it will allow users to explore the details of migration routes and space use in more detail than previously. For example, many studies have tracked sea turtles (reviewed in Hays & Scott 2013). In almost all previous cases, conventional Argos tracking has been used. So while there are some descriptions of routes followed and some movement metrics (e.g. the straightness of routes, see Luschi et al. 1998 or Hays et al. 2014a), details of changes in speed and heading over time-scales of a few hours are poorly described. These same considerations apply to marine mammals, fish and penguins that are remotely tracked rather than are equipped with high-resolution loggers that can directly measure performance (e.g. speed) but need to be removed for data recovery. Similarly radio tracking and Argos tracking have been used to estimate home- ranges used by sea turtles and other marine taxa (Bjørge et al. 2002, Seminoff & Jones 2006, Tougaard et al. 2008, Frere et al. 2008), but here again the increased accuracy of Fastloc-GPS will allow more informed estimates of space use (e.g. Schofield et al. 2013; Hays et al. 2014a). This utility of Fastloc-GPS was evidenced in the track of a green turtle travelling across the Indian Ocean that showed small scale variations in speed and heading. Most likely these variations are due to changes in the animals swimming as well as the impact of variable ocean currents. Using Fastloc-GPS, the relative roles of swimming versus currents could be assessed (e.g. Fossette et al. 2012; Galli et al. 2012; Hays et al. 2014a). Such detailed assessment of the routes followed may allow more informed assessments of navigational mechanisms used in long-distance migration as well as allowing the impact of currents on migration to be more closely addressed (i.e. for turtles Lohmann & Lohmann 1996; Hays et al. 2003; Lohmann et al. 2008).

Here I have only investigated the impact of Fastloc-GPS accuracy on migrating green turtle tracks. Yet different species have already been tagged using this technology and I expect an increase of Fastloc-GPS use in next years. To process these future tracks, I advise users to use methodologies that: filter data using a threshold residual value and then take account of the number of GPS satellites used in each location calculation to derive an appropriate time-interval (i.e. for a migrating green turtle 6 hours if one location in a pair is calculated using 4 satellites, 1 hour otherwise) over which speed of travel and heading can be estimated. 35

Chapter 3: Diel pattern of migrating sea turtles

Chapter 3 : Argos-Linked Fastloc-GPS reveals daytime departure and arrival during long-distance migration and the use of different resting strategies in sea turtles

Abstract

Determining the time of day that animals initiate and end migration, as well as variation in diel movement patterns during migration, provides insights into the types of strategy used to maximise energy efficiency and ensure successful completion of migration. However, obtaining this level of detail has been difficult for long-distance migratory marine species. Thus, I investigated whether the large volume of highly accurate locations obtained by Argos-linked Fastloc-GPS transmitters could be used to identify the time of day that adult green (n = 8 turtles, 9 487 locations) and loggerhead (n = 46 turtles, 47 588 locations) sea turtles initiate and end migration, along with potential resting strategies during migration. I found that departure from and arrival at breeding, stopover and foraging sites consistently occurred during the daytime, suggesting that turtles used visual cues for orientation. Only seven turtles made stopovers (of up to six days and all located close to the start or end of migration) during migration, possibly to rest and/or refuel; however, observations of day versus night speed of travel indicated that turtles might use other mechanisms to rest. For instance, turtles travelled 31 % slower at night compared to day during their oceanic crossings. Furthermore, within the first 24 h of entering waters shallower than 100 m towards the end of migration, some individuals travelled 66 % slower at night, repeating this behaviour intermittently (each time for a one-night duration at 3–6 day intervals) until reaching the foraging grounds. Thus, access to data-rich, highly accurate Argos-linked Fastloc-GPS provided information about differences in day versus night activity at different stages in migration, allowing us, for the first time, to compare the strategies used by a marine vertebrate with terrestrial and avian species.

Chapter 3: Diel pattern of migrating sea turtles

Introduction

Migratory animals that invest in long-distance migration select different times of the diel cycle (e.g. dusk, midday or dawn) to initiate or terminate migration, which maximises the probability of successfully reaching the destination (Åkesson & Hedenström 2007, Alerstam 2009, Müller et al. 2016). Departure at certain times of the day helps a given species to reduce predation rates, energy expenditure via passive transportation, water loss and to maximise orientation (Alerstam 2009, Müller et al. 2016). For example, desert locusts depart shortly after sunrise to take advantage of the wind generated by rising air temperatures (Kennedy 1951). Ruby- throated hummingbirds depart during the mid-day period, leaving time for feeding in the morning and evening (Willimont et al. 1988). Many migratory songbird species depart at dusk to avoid predators (Åkesson et al. 1996, Alerstam 2009, Müller et al. 2016). Furthermore, species that migrate during the daytime tend to arrive at stopover or foraging sites during the daytime (Kennedy 1951, Strandberg & Alerstam 2007), whereas those that initiate migration at night tend to arrive before dawn (Biebach et al. 2000, McGuire et al. 2012). These observations have demonstrated that, while some species maintain the diel patterns exhibited at breeding and foraging grounds during migration (e.g. bats, Mcguire et al. 2012 ; ospreys, Strandberg and Alerstam 2007), others alter their circadian rhythm (e.g. songbird, Alerstam 2009). For such species, this also results in changes in the cues used for orientation, i.e. from solar cues for daytime travel during foraging/breeding to stellar and magnetic cues during night-time migration (e.g. songbird, Alerstam 2009). Furthermore, the time of day that turtles depart the breeding sites and arrive at the foraging sites may depend on the orientation cues used at these phases of migration. For example, daytime arrival/departure would support the use of the sun compass (Guilford & Taylor 2014), while night time arrival/departure would support the use of a magnetic compass (Ehrenfeld & Koch 1967, Åkesson et al. 1996, Åkesson & Hedenström 2007).

Many avian and terrestrial animals also stop to rest and refuel at regular intervals along the migratory route, or at transitory ‘stopover’ sites, depending on resource availability (e.g. insects, McCord and Davis 2012, Kennedy 1951; reptiles, 37

Chapter 3: Diel pattern of migrating sea turtles

Rice and Balazs 2008; birds, Schaub et al. 2001; mammals, Sawyer and Kauffman 2011). Yet, stopping is not always possible, such as when birds pass over open oceans, deserts or mountain ranges (Åkesson & Hedenström 2007, Vardanis et al. 2011, Bishop et al. 2014). In such cases, non-stop travel is required to reach the next safe area; thus, these animals must develop strategies to rest while actively travelling. For example, common swifts remain airborne during the whole of their migration and for more than 99 % of their 10-month non-breeding period over Africa, with data loggers suggesting possible mid-flight micro-sleeps during which they drop through the air for <40 seconds (Hedenström et al. 2016). Similarly, frigatebirds fly over the ocean for periods up to 10 days, sleeping for around 40 minutes per day, with either one brain hemisphere active at a time or both simultaneously (Rattenborg et al. 2016). Thus, detailed information on movement over the course of the day can potentially provide information on how animals rest during migration.

Despite the ecological value of the information, data on the time of day that many marine animals initiate and end migration, along with potential resting strategies, remains limited because of the difficulty of directly observing these animals in their natural environment, and the lack of quality (e.g. the number and accuracy of locations) in technology used to monitor movement patterns (e.g. satellite or acoustic tracking). Information does exist for some estuarine or shallow- water species. For instance, radio-tracked nocturnal sea lampreys have been shown to initiate their spawning migration from sea to rivers at night, retaining their typical circadian cycle (Almeida et al. 2002); but many studies only provide the day of departure or arrival based on changes in metrics such as speed and displacement distance, rather than the actual time of day due to the limited volume and accuracy of transmitted locations (e.g. sea turtles Blumenthal et al. 2006; Schofield et al. 2013b; white sharks, Domeier and Nasby-Lucas 2013 or whales, Mate et al. 2011). Furthermore, studies on orca and bottlenose dolphin have demonstrated the use of lateralized sleep behaviour during long-distance migration, with one hemisphere of the brain entering into slow-wave sleep while the second hemisphere remained active (Lyamin et al. 2008). For sea turtles, no clear picture has emerged on resting during long-distance migration from the few studies that are based on satellite 38

Chapter 3: Diel pattern of migrating sea turtles

telemetry and dive-profile data. For example, in water where turtles cannot reach the seabed to rest (e.g. >100-150 m deep), Minamikawa et al. (1997) suggested that turtles rest at night by investing in mid-water dives that involve steep active descents followed by gradual passive ascents. Two satellite tracking studies reported a 19-23 % difference between night and day-time travel speeds (Luschi et al. 1998; Jonsen et al. 2006, respectively). However, it is not known whether these observations are due to a reduction of the forward motion during deeper nocturnal dives or a reduction in the speed of travel due to a resting behaviour. Ultimately, extended periods (i.e. weeks) of continuous travel of around 1000 km or more by sea turtles are likely to cause fatigue, leading to the need for periodic resting, as documented for other species (Alerstam et al. 2003, Hein et al. 2012). Yet, just two studies over the last eight years have detected the use of stopover sites by one sea turtle species (green turtle, Chelonia mydas). There, individuals following a coastal migratory route used multiple stopovers (Baudouin et al. 2015), whereas individuals crossing an ocean basin frequented just one stopover site each during the middle of migration (Rice and Balazs 2008).

Advances in Argos-linked Fastloc-GPS over the last 15 years have resulted in 10–100 times greater location accuracy than standard Argos technology, because only a short period of time (typically tens of milliseconds) is required to obtain a fix, which is essential for animals that only surface to breathe briefly (Tomkiewicz et al. 2010; Chapter 2). To date, Argos-linked Fastloc-GPS has been used to provide a variety of new insights about marine species, including home ranges (e.g. northern fur seals, Kuhn et al. 2010; sunfish, Thys et al. 2015), predator-prey interactions and foraging behaviours (e.g. fur seals, Arnould et al. 2015; harbour seals, Berejikian et al. 2016; king penguin, Scheffer et al. 2016), navigation (e.g. sea turtles, Hays et al. 2014a; fur seals, Chevaillier et al. 2014), estimations of fecundity (e.g. sea turtles, Esteban et al. 2017) and human disturbance (e.g. whales, Mate 2012; sea turtles, Schofield et al. 2015). Yet, to date, few researchers have explored the potential of using the data-rich locational information that is generated by Argos-linked Fastloc-GPS to answer key questions on behavioural ecology such as how animals navigate and orientate in the open ocean (Hays et al. 2016).

Chapter 3: Diel pattern of migrating sea turtles

This study aimed to identify: (1) the time of day that sea turtles initiate and end migration; (2) potential resting strategies used by sea turtles during migration; and (3) whether those strategies were consistent across species and locations. I used Argos-linked Fastloc-GPS datasets for two different sea turtle species (loggerhead turtles Caretta caretta, green turtles Chelonia mydas) in two different ocean basins (Mediterranean Sea and Indian Ocean) to determine whether the same movement patterns were used. Sea turtles are generally active during the daytime (i.e. diurnal) when foraging (Ogden et al. 1983, Christiansen et al. 2017) and even when breeding (Hays et al. 2000: except for when emerging on beaches to nest at 12-25 day intervals). Thus, I hypothesised that migration would start and end during the day and that travel and would be faster during the daytime (as observed in Luschi et al. 1998; Jonsen et al. 2006). I also investigated differences in day-night travel speeds and the use of stopover sites to obtain insights about the resting strategies used by turtles during long-distance migration, based on the assumption that slower migration at night was likely to be due to turtles investing in some type of resting strategy. My results are expected to provide novel information on the diel strategies of migration by sea turtles, comparable to strategies already reported for avian and terrestrial animals.

Methods

Source data for case study Sea turtles from two breeding populations were used in this study: (1) male and female loggerhead turtles migrating from the breeding ground in Laganas Bay at the southern part of Zakynthos Island, Greece (37.80° N, 20.75° E) to foraging grounds throughout the Mediterranean Sea; and (2) female green turtles migrating from the breeding ground on the southern coast of Diego Garcia, Chagos Archipelago (7.31° S, 72.41° E) to foraging grounds in the central and western parts of the Western Indian Ocean (Figure 3.1).

Chapter 3: Diel pattern of migrating sea turtles

Figure 3.1: Migratory routes of: (a) loggerhead sea turtles; and (b) green sea turtles tracked with Argos-linked Fastloc-GPS from the breeding to foraging grounds. Thirty- three loggerhead male turtles (black lines) and 13 loggerhead female turtles (red lines) were tracked from Zakynthos (with some passing via Kyparissia, both breeding sites are presented as white circles) in Greece, Mediterranean Sea. Eight female green turtles (red lines) were tracked from Diego Garcia, Chagos Archipelago, Western Indian Ocean (breeding site represented as a white circle). Haddhunmathi Atoll is part of the Maldives Archipelago, and the Amirante Islands are part of the Seychelles. Two turtles travelled to Somalia and crossed paths when travelling along the coast, with the endpoints being shown as black dashes. White squares on the tracks show the stopover locations. Modified from Schofield et al. (2013) and Hays et al. (2014b).

Chapter 3: Diel pattern of migrating sea turtles

Only loggerhead and green sea turtles fitted with Argos-linked Fastloc-GPS satellite tags were used in this study. All tracks have been previously analysed, but with a different focus (e.g. Schofield et al. 2013; Hays et al. 2014b, Christiansen et al. 2017). Details on the attachment procedure of Argos-linked Fastloc-GPS tags are described in Schofield et al. 2013 for loggerhead turtles, and in Hays et al. 2014b for green turtles. Out of 56 loggerhead turtles tracked from Zakynthos between 2007 and 2012, I selected 33 males and 13 females (46 turtles in total), excluding 10 resident turtles that remained at Zakynthos during the whole tracking duration. Some of the males migrated from Zakynthos (n = 4) also visited the adjacent breeding ground of Kyparissia (150 km distant on the Peloponnese, mainland Greece; 37.25° N, 21.66° E) for 2 to 18 days; thus, data from this site were also included. I also included eight female green turtles tracked from Diego Garcia in 2012. The Mediterranean loggerhead turtles have both oceanic (primarily to the Gulf of Gabes) and neritic (coastal; primarily to the Adriatic) migratory routes, while all green turtles from the Chagos Archipelago were oceanic migrants (Schofield, Dimadi, et al. 2013, Hays, Mortimer, et al. 2014).

The curved carapace length (CCL) of the 46 loggerhead and eight green turtles was 83.4 ± 6.1 and 105.6 ± 3.45 cm, respectively (loggerhead turtles CCLs: Schofield et al. 2013; green turtles CCLs, Hays et al. 2014b). The mean distance travelled by the retained loggerhead turtles from the breeding grounds to the foraging grounds was 920 ± 409 km (range: 189–1545 km) over a mean 25 ± 10 days (range: 7–42 days) (Figure 3.1a; Schofield et al. 2013). Green turtles from Diego Garcia migrated a mean distance of 2639 ± 1264 km (range: 166–3979 km) for a mean duration of 44 ± 19 days (range 4–68 days) (Figure 3.1b; Hays et al. 2014b).

Data preparation I only used migratory tracking data between the breeding area and the destination foraging ground, including the full day on which turtles departed the breeding area through to the full day on which turtles arrived at the foraging grounds. I excluded any turtles that were resident at Zakynthos from this analysis (n = 10 out of 56 tracks). On Zakynthos, migratory turtles that were tracked in more

Chapter 3: Diel pattern of migrating sea turtles

than one year (n = 3 males) were included as separate records, as only one of the departures and arrivals for each turtle met the criteria for this study. In addition, because both males and females were tracked from Zakynthos, the data for each sex were initially analysed separately; however, the same trends were detected, so the data were merged across sexes.

I first assimilated the raw data for all turtles in Quantum-GIS V2.10.1 software (QGIS Development Team 2015). I only included Argos-linked Fastloc- GPS locations obtained with five or more satellites and with residual values of <35 (residual values are provided by the software converting the pseudoranges into location estimates, see Dujon et al. (2014) for additional detail). Loggerhead turtle locations were filtered by Sirtrack Company at the start of this study (only locations with five or more satellites were provided) but I removed 11 % of the green turtle locations (only retaining locations with five or more satellites). In addition, I removed any remaining locations that looked visibly erroneous in QGIS or that resulted in unrealistic speeds of travel (i.e. >200 km day-1) when analysed (<0.14 % of loggerhead turtle locations, <0.002 % of green turtle locations).

I then obtained real-time travel speeds (using R software, Version 3.2.3, R Development Core Team 2013) by calculating the speed of travel from locations that were separated in time by at least 3 h (but no longer than 24 hours) to ensure estimates of high accuracy (99 % of speed of travel estimates with an error <10 % of the true value, see Chapter 2). All distances in this study were calculated using the great circle distance method. Neritic and oceanic phases of migration were defined as areas that were within or deeper than the 100-m depth contour, respectively (Minamikawa et al. 1997, Hatase et al. 2007, Schofield, Hobson, Fossette, et al. 2010), and were also validated by this study (see Appendix A, Figure A1). This threshold was selected because the neritic foraging sites of loggerhead turtles tend to be located in areas with a seabed depth < 100 m (Schofield, Hobson, Fossette, et al. 2010, Foley et al. 2014). Day and night were distinguished based on local nautical dusk and dawn times (http://www.esrl.noaa.gov/gmd/grad/solcalc). All of the loggerhead and green turtles initiated and ended migration within a time window of six weeks and four months respectively, which represents a maximum

Chapter 3: Diel pattern of migrating sea turtles

variation in local dusk and dawn times of 20 min for the loggerhead turtles and of 30 min for the greens turtles. Therefore, I used a constant dawn and dusk time for both sites as the slight changes in the dusk or dawn time over that 6 weeks should not affect our results. All datasets were originally in Universal Coordinated Time (UTC), but were converted to local time to correspond with local dawn and dusk times. Data on seabed depth were obtained using the ETOPO1 global relief model (https://www.ngdc.noaa.gov/mgg/global/global.html) and from ARGOS CLS website (http://www.argos-system.org/) when a higher resolution was required (for example, inside a lagoon). The number of turtles with sufficient data for each analysis is shown in parentheses in each section of the results. All values are reported herein as mean ± 1 SD.

Departures and arrivals For each tracked turtle, the time at which turtles began migration from breeding and stopover sites was assessed from Argos-linked Fastloc-GPS locations that showed directional movement (i.e. turtle moving in a single direction offshore from the site with continuous increase in displacement distance) and an inflection in travel speed to >1 km h-1. The displacement distance was calculated as the distance between the nesting ground and the turtle location (Blumenthal et al. 2006; Schofield et al. 2010a). Migration either began: (1) immediately on departing the breeding or stopover site; or (2) 1–2 days later, after initially travelling from the breeding ground along the coast of the island, and subsequently travelling offshore to initiate the oceanic crossing, which is when I might expect turtles to switch from the cues used to follow the coastline to cues used during open sea navigation (Hays et al. 2002b). The actual onset of migration was obtained for turtles with one to six locations (mean: 1.8 ± 1.7 locations) in the 3 h either side of departure for loggerhead and from one to 13 locations either side of departure for green turtles (mean: 3.0 ± 3.3 locations). When a location occurred offshore, but intermediary locations were insufficient to determine the exact departure time, I measured the shortest distance back to the coast, henceforth referred as backtracking. Based on a sensitivity analysis (Appendix A, Figure A2), I only used turtles when the backtrack

Chapter 3: Diel pattern of migrating sea turtles

duration was less than 12 h (or 18 km) for loggerhead and 6 h (or 16 km) away for green turtles, based on mean travel speeds of 1.5 and 2.6 km h-1, respectively. Arrival at foraging and stopover sites was detected by a lack of directional movement (i.e. displacement distance from the breeding site remaining constant) and the inflection in travel speed decreasing to <1 km h-1 (Blumenthal et al. 2006; Schofield et al. 2010a). Stopover and foraging sites were distinguished by turtles remaining in the same area for <6 days (Rice & Balazs 2008) and >6 days, respectively (Schofield, Hobson, Fossette, et al. 2010). Turtles were assumed to be frequenting stopover sites when they remained at the same location for at least 6 h during the daytime and resumed travel within 6 days of arriving (Rice and Balazs 2008). The actual arrival time was obtained from one to nine locations (mean: 2.3 ± 2.1 locations) for loggerhead turtles and one to three locations for green turtles (mean: 1.1 ± 1.0 locations) in the 3 h either side of arrival. The arrival times of all other turtles were only inferred where it was possible to measure forward (henceforth referred as forward track) from the last location at sea to the first location at the foraging site within the thresholds delimited by the sensitivity analysis for each species (Appendix A, Figure A2). I excluded arrival at oceanic foraging sites (two loggerhead turtles) from the analysis because it was not possible to detect a specific arrival time from this type of movement pattern.

To determine whether turtles adjusted their speed of travel at end of migration to arrive at the foraging ground at night-time or daytime, I calculated the speed of travel for the final night and the final day of migration, and correlated it with the time of arrival at the foraging ground. This calculation was only completed for turtles that had Argos-linked Fastloc-GPS locations available within 3 h of nautical dawn and dusk to delineate the cut-off points as accurately as possible. I tested this relationship using a t-test on Pearson correlation coefficients.

Diel variation in the speed of travel I calculated the ratio in the speed of travel between day (numerator) and night (denominator) for turtles travelling in waters of different depths. A ratio value of one meant that a turtle swam at the same speed during both daytime and night- time. The speed of travel was calculated using the first and the last location available

Chapter 3: Diel pattern of migrating sea turtles

for each day and each night (separated by at least 3 h, Chapter 2). I calculated the ratios for adjacent days and nights in an attempt to avoid variation in sea currents confounding the inferred speed of travel during migration (see Luschi et al. 1998; Luschi et al. 2003). Only turtles with at least three day/night comparisons were included in this analysis. Before analysing the data, I validated that mean day/night speed of travel ratios were not affected by the straightness of the track. I found that 13 and 4 % of day/night comparisons for loggerhead and green turtles, respectively, had a straightness index <0.80 (indicating the turtle may have been deflected by oceanic currents), and that removing these data changed the mean ratio by a maximum of 8 % and 1 %, respectively. Thus, all sections of track were included in the calculation regardless of curvature.

Because the values of the ratios were not statistically independent, I used a non-parametric bootstrapping approach to calculate mean ratio values using R software. For each turtle, I resampled the day/night ratio time series 10 000 times with replacement and calculated an estimated mean ratio value for each iteration. I then averaged the 10 000 estimates and determined the 95 % confidence intervals of the mean ratio by calculating their 2.5 and 97.5 % percentiles. I only considered a difference in the mean ratio to be significant when the 95 % confidence interval did not include one. I only calculated a mean ratio using bootstrapping when at least three day/night comparisons were available for a given turtle. In addition, I performed a sign test to check whether the calculated mean day/night speed ratio could have occurred by chance across the 14 sampled turtles, assuming that turtles have an equal chance for a mean day/night ratio under and above a value of one.

I validated that the day/night ratio changed at a seabed depth of 100 m for turtles migrating across waters from >200 m to <50 m deep (i.e. at >200 m, 150- 200 m, 100-150 m, 50-100 m and <50 m). I then compared these results with those from turtles that remained within a depth of 100 m throughout migration (Appendix A, Figure A1). I then compared the day/night ratio in travel speed for the two species (green and loggerhead turtles) when crossing oceanic waters (>100 m) using the same bootstrap procedure as described above. I identified the days on which night-time travel speed was at least 1 km h-1 slower compared to daytime. Sea turtles

Chapter 3: Diel pattern of migrating sea turtles

forage mostly during daytime (Ogden et al. 1983, Narazaki et al. 2013, Christiansen et al. 2017), therefore such slow night-time speed of travel indicate possible resting behaviour. For these days, I calculated a theoretical maximum number of hours a turtle might have stopped swimming (e.g. to rest), assuming that individuals maintained day-time travel speeds rather than slowing at night.

퐷𝑖푠푡푎푛푐푒 푡푟푎푣푒푙푒푑 푎푡 푛𝑖𝑔ℎ푡 (푘푚) 푇ℎ푒표푟𝑖푡𝑖푐푎푙 푚푎푥𝑖푚푢푚 푠푡표푝 푑푢푟푎푡𝑖표푛 (ℎ) = 푁𝑖𝑔ℎ푡 푑푢푟푎푡𝑖표푛 (ℎ) − 퐷푎푦푡𝑖푚푒 푡푟푎푣푒푙 푠푝푒푒푑 (푘푚 ℎ−1)

To estimate the duration that turtles rested, I only used day/night combinations where Argos-linked Fastloc-GPS locations were available within 3 h of nautical dawn and dusk (Appendix A, Figure A2) to delineate the cut-off points as accurately as possible.

Results

General tracking I used 47 588 and 9 487 Argos-linked Fastloc-GPS locations from tracked loggerhead (n= 46) and green turtles (n = 8), respectively. The mean number of locations per day was 4.9 ± 5.9 and 7.6 ± 5.4 for loggerhead and green turtles, respectively. The mean travel speed was 1.5 ± 0.6 km h-1 for loggerhead turtles and 2.6 ± 1.2 km h-1 for green turtles. The mean time interval used to calculate those travel speeds was 7.1 ± 4.6 h (range: 3.0-23.9) for loggerhead and 5.8 ± 3.6 h (range = 3.0-23.8) for green turtles. Out of the 46 loggerhead turtles retained for this study, 16 migrated through oceanic waters, while the remainder (n = 30) primarily remained in neritic waters (Figure 3.1). All eight green turtles migrated through oceanic waters (Figure 3.1). A total of nine loggerhead (n = 66 day/night comparisons, mean: 7.6 ± 7.2, range: 3-24) and five green turtles (n = 167 day/night comparisons, mean: 33.4 ± 17.2, range: 11-51) were used to calculate ratios during the oceanic crossing, while six loggerhead turtles (28 day/night comparisons, mean: 4.7 ± 2.3, range: 3-9) were used to calculate ratios during the neritic crossing. All the turtles used to calculate ratios had at least three day/night comparisons.

Chapter 3: Diel pattern of migrating sea turtles

Departures and arrivals Out of the 46 loggerhead and eight green turtles, and using our criteria, I were able to assess a total of 26 departures (retaining 21 loggerhead and five green turtles) and 27 arrivals (retaining 21 loggerhead and six green turtles) with nine apparent uses of stopover sites (four for the loggerhead and five by the green turtles). Migration either began immediately following departure from the breeding site (for 17 loggerhead and two green turtles) or 1–2 days later, after the turtle travelled along the coast adjacent to the breeding site (for four loggerhead and three green turtles). As expected, the departures were detected based on a simultaneous increase in travel speed and displacement from breeding ground, (Figure 3.2). In comparison the arrivals were detected based on a decrease in the speed of travel and a displacement distance from the breeding site becoming constant (Figure 3.3). Out of the 27 arrivals, 17 had locations available within 3 h of nautical dawn and dusk (and were subsequently used in Figure 3.4c) while 12 did not meet this criterion. Overall, turtles primarily initiated migration from breeding and stopover sites early in the day (Figure 3.4a) with 69 % of departures observed between 04:00 h and 08:00 h. In comparison, turtles arrived at stopover sites and the foraging grounds relatively evenly between 06:00 h and 22:00 h (Figure 3.4b). There was no difference in the pattern of arrival of loggerhead turtles depending on whether they had primarily migrated through oceanic or neritic waters. There was also no difference in the pattern of arrival at foraging sites located close to the coast and those located further offshore, with all sites being <100 m deep.

I detected a significant relationship between the time of arrival at the foraging grounds and speed of travel during the final night of migration (Figure 3.4c; n = 17 arrivals based on 12 loggerhead and five green turtles; Pearsons r = 0.57, t = 2.70, p = 0.016). Specifically, turtles that were closer to the foraging grounds travelled slower on the final night and arrived early the next morning, whereas turtles that were further from the foraging grounds travelled faster on the final night and arrived later the next day (Figure 3.4e,f). Similarly, I found a significant relationship between the speed of travel during the last day of the migration and the distance to the foraging ground at the beginning of that same

Chapter 3: Diel pattern of migrating sea turtles

night (Pearsons r = 0.64, t = 3.20, p = 0.006) confirming that loggerhead turtles slowed down during that last time night when they are nearing the foraging ground (Figure 3.4d).

Figure 3.2: Two examples showing how the time that turtles initiated migration was determined from the Argos-linked Fastloc-GPS locations. Migration either began: (a) immediately on departing the breeding site (example of a loggerhead turtle departing Zakynthos Island, Greece): or (b) 1–2 days later, after initially travelling along the coast (example of a green turtle departing Diego Garcia, Chagos Archipelago). The final day of breeding is presented (white circles), along with day (red circles and lines) and night (grey circles and black lines) locations once the turtles initiated directional movement. The black arrows on the maps show the time at which turtles initiated migration (reflected by the dashed lines in c–d and e–f). (c–d) The timing of departure was confirmed by an inflection in swimming speed to above 1 km h-1 and (e–f) a continuous increase in displacement distance.

Chapter 3: Diel pattern of migrating sea turtles

Figure 3.3: Two examples showing how the time that turtles arrived at foraging sites was determined from the Argos-linked Fastloc-GPS locations. (a) Loggerhead turtle arriving at its foraging ground in the Adriatic Sea, Mediterranean. (b) Green turtle arriving at its foraging ground on the east coast of Africa (Somalia). The day (red circles and lines) and night (grey circles and black lines) locations of the turtles during migration are presented, along with the first day at the foraging ground (white circles). The black arrows on the maps show the time at which turtles arrived (indicated by the dashed lines in c–d and e–f). The timing of arrival was confirmed by (c–d) an inflection in swimming speed to below 1 km h-1 and (e-f) lack of change in displacement distance. The green turtle in (b,d,f) was swimming against the Somali current flowing southward along the Somalian coast, resulting in a speed of travel lower than the average 2.6 km h-1 calculated for this species in the Western Indian Ocean (Carbone & Accordi 2000). In (f), the distance from the breeding ground decreased because the turtle was deflected southward, probably by the current

Chapter 3: Diel pattern of migrating sea turtles

(extending the migration distance by 235 km), before reaching the coast and turning northwards to reach the foraging ground.

Figure 3.4: (a) Time of day that turtles initiated departure from the breeding grounds (grey bars for loggerhead turtles, n = 21; white bars for green turtles, n = 5) and stopover sites (dark grey bars for loggerhead turtles, n = 5; dashed bars for green turtles, n = 2). (b) Time of day that turtles arrived at the foraging grounds (grey bars for loggerhead turtles, n = 20; white bars for green turtles, n = 6) and stopover sites

Chapter 3: Diel pattern of migrating sea turtles

(dark grey bars for loggerhead turtles, n = 5; dashed bars for green turtles, n = 3). (c) Speed of travel of turtles during the final night of migration in relation to the time elapsed since dawn on the day of arrival (black circles for loggerhead turtles n = 12, and white circles for green turtles n = 5). The black line represents the linear relationship between the speed of travel and arrival time (Pearsons r = 0.57, t = 2.70, p = 0.016). Nautical dawn and dusk are represented by black (Mediterranean Sea) and grey (Western Indian Ocean) dashed vertical lines. (d) Speed of travel of turtles during the final night of migration in relation to the distance to the foraging ground at the beginning of the final night. The black line represents the linear relationship between the speed of travel and distance to foraging ground (r = 0.64, t = 3.20, p = 0.006). Examples of Argos-linked Fastloc-GPS tracks showing the movement of turtles on the night before and day of arrival at the final foraging ground, for: (f) a turtle arriving early in the day in the Gulf of Gabes; and (g) a turtle arriving late in the day in the Adriatic, and showing comparative night-time travel speeds. Turtle locations at nautical dawn (white circles) and dusk (white squares) are shown along with day (red circle and lines) and night Argos-linked Fastloc-GPS locations (grey circles and black lines). All times are presented as local time.

Stopover behaviour during migration

Arrival and departures from the stopover sites were determined using the same methodology as described in the methods section for the departure from the breeding grounds and the arrival to foraging grounds.

Four out of eight green turtles made stopovers during migration. Two individuals made one stopover each, lasting 12 hours (during daytime) and 3 days on the Mascarene Plateau (115 000 km2, 2000 km long with depth ranging from 8- 150 m, plunging to 4000 m to the abyssal plain at its edges). The stopovers occurred when more than 90 % of the migration was complete, at 204 and 245 km distance, respectively (which equates to 5 and 6 days) from the foraging grounds (total migration: 2515 and 2825 km and 37 and 53 days, respectively, due to differences in routing and travel speed). A third turtle also made one stopover of 3 days when more than 90 % of migration was complete, at the same area as the other two turtles (Mascarene Plateau); however, this turtle initially overshot the final destination, 52

Chapter 3: Diel pattern of migrating sea turtles

resulting in it travelling twice the distance before it reached foraging grounds (total migration: 3886 km and 65 days). The fourth turtle stopped twice in two Maldives atolls for 5 days and 1 day after completing 70 and 90 % of migration, respectively, at 160 and 50 km distance (or 12 and 2 days) from the foraging grounds (total migration: 1152 km, 44 days). For all green turtles, stopover sites had depths ranging from 10 to 70 m. Of note, of the two turtles that migrated to the coast of Africa (3230–3741 km over 45–61 days), neither made any stopovers, even though one passed along the edge of the Mascarene Plateau.

Three loggerhead turtles made stopovers during migration. One turtle migrating to the Adriatic had two stopovers of 12 h and 1 day when 84 and 89 % migration was complete, at 46 and 48 km straight-line distance (5 and 3 days) from the foraging ground, respectively (total migration: 1505 km, 38 days). The two stopovers were of similar distance to the foraging ground, but were separated by 29 kilometres in total migration length (1 day travel), because the turtle did not travel directly to the destination. One turtle migrating to the Gulf of Gables stopped for 6 days after completing 93 % of its migration at 69 km (2 days) from the foraging ground (total migration: 1158 km, 35 days). One turtle made a 2-day stopover after completing 17 % of migration from Zakynthos (total migration: 740 km, 16 days). The stopover site was on the north-east coast of Zakynthos, just 55 km away from the breeding area in Laganas Bay (equivalent of 1.5 days travel). However, after initially departing Zakynthos, this turtle first travelled to and remained at Kyparissia for 2.5 days, before travelling to the stopover site (four days after leaving Zakynthos). For all loggerhead turtles, stopover sites had depths ranging from 20 to 50 m.

Diel variation in the speed of travel

Loggerhead and green turtles swimming in waters >100 m depth (i.e. oceanic) had a mean speed of travel that was 31 % higher by day than by night with this behaviour being maintained for up to 24 days by loggerhead and 61 days by green turtles (mean day/night loggerhead turtles speed of travel ratio of 1.31 ± 0.16; range: 1.07–1.68; n = 66 day/night comparisons based on nine turtles; mean day/night green turtles speed of travel ratio of 1.27 ± 0.16 ; range: 1.12 – 1.44; n = 53

Chapter 3: Diel pattern of migrating sea turtles

167 day/night comparisons based on five turtles; Figure 3.5). While the mean day/night ratio was not significantly different at the individual level (likely due to the effect size and noise), the probability that all 14 turtles would have a mean day/night ratio greater than one by chance was very small (sign test, p < 0.001).

For six of the loggerhead turtles that entered shallow waters (<100 m) and had sufficient day/night comparisons, five had significantly higher speeds of travel during the day-time at the individual level (mean day/night speed of travel ratio of 1.72 ± 0.47; range: 1.37–2.50, n = 25 day/night comparisons; Figure 3.5a), with a mean ratio significantly greater than one. In comparison, the sixth turtle had a higher speed of travel at night compared to day (day/night speed of travel ratio of 0.78, n = 3 day/night comparisons).

For four out of those six turtles, I were able to determine the day/night travel speed ratio within 24 hours of entering water shallower than 100 m. These four turtles exhibited noticeably higher travel speeds by day compared to night (an average of 46 to 66 % faster by day) suggesting they rested on the sea bed during the first night after crossing the 100 m contour line. On subsequent days, similar noticeably higher travel speeds by day compared to night was detected at 3–6 day intervals until reaching the foraging grounds, suggesting that they rested every third to sixth night (Figure 3.5e). I calculated that these turtles might be theoretically stopping completely for a mean 5.8 ± 1.3 hours at night-time, assuming day-time speeds of travel also occurred at night.

It was not possible to calculate the mean day/night speed of travel ratio for the two green turtles travelling through neritic waters, because fewer than three day/night comparisons were available. However, preliminary speed of travel ratios suggests that neither turtle exhibited a change in day/night travel speed once in waters that were shallower than 100 m (and neither invested in stopovers), despite travelling for 14 and 8 days in waters of <100 m after oceanic crossings of 3741- km and 3230-km, respectively.

Chapter 3: Diel pattern of migrating sea turtles

Figure 3.5: (a) Mean day/night speed of travel ratio and 95 % confidence intervals for nine loggerhead (ID 1 to 9) and five green (ID A to E) migrating turtles. During the oceanic crossing, turtles swam continuously during day and night (grey circles for loggerhead turtles, n = 66 day/night comparisons; white circles for green turtles, n = 167 day/night comparisons). After passing into waters shallower than 100 m (neritic), five out of six loggerhead turtles travelled further by day than by night (ID 1 to 6, n = 28 day/night comparisons) while the remaining turtle swam further by night. Mean day/night speed of travel ratios indicated with a black star are significantly different from one. Example of two days and nights of oceanic crossing for (b) a loggerhead and (c) a green turtle, and an example of (d) two days and nights of neritic crossing for a loggerhead turtle. Daytime locations are shown in red, and night-time locations in grey. The speed of travel is given for each day and night. (e) Final 10 days of migration by a loggerhead turtle in the Adriatic. On entering waters shallower than 100 m (delimited by the black bold line on the track), the mean day/night ratio in the speed of travel became variable, with significantly slower night-time travel speeds on 55

Chapter 3: Diel pattern of migrating sea turtles

three nights (yellow circles) during migration, including the final night (blue circles) arriving at the foraging site (white square) during the day-time on the final day. On these nights, I estimated that this turtle rested from 3 to 5.8 hours, assuming day-time travel speeds also occurred overnight.

Discussion

My study is the first to show that two different sea turtle species from two different ocean basins exhibit similar movement patterns when departing breeding areas and arriving at stopovers and destination foraging areas. I validated my hypothesis that sea turtles (from both species and both sex) would start migration during the day, suggesting that they use visual cues for orientation.

This reliance on visual cues has previously been suggested in studies on juvenile and displaced sea turtles (Avens & Lohmann 2003, Mott & Salmon 2011, Shimada et al. 2016) and is also documented for birds, insects and other reptile species (Alerstam 2009, Southwood & Avens 2010, Guilford & Taylor 2014). Experiments performed in the laboratory by Avens and Lohmann (2003) and Mott and Salmon (2011) were completed in shallow tanks (< 2 m) and involved solar cues. Similarly, Shimada et al. (2016) showed that sea turtles displaced from their foraging ground correct their course early in the morning and in relatively shallow water (median depth of 8 m). Thus, sea turtles are able to perceive visual cues when they are relatively close to the surface. In addition, sea turtles are myopic outside water (Ehrenfeld and Koch 1967) which suggests that they use cues that can be perceived even if they are modified by the eye structure or by the refraction and absorption of light by sea water (e.g. using directional sunlight available around dawn to establish a course). It is not known whether sea turtles are also able to perceive such cues deeper in the water column.

Both species of sea turtles travelled continuously during migration, with higher speed of travel during daytime compare to night-time, particularly after entering waters shallower than 100 m, thereby confirming my original hypothesis regarding faster daytime travel speeds. In contrast, I were not able to confirm

Chapter 3: Diel pattern of migrating sea turtles

whether turtles exhibit a resting strategy similar to non-stop travelling birds or marine mammals during their oceanic crossing because the behaviour that led to those differences is open to interpretation (Lyamin et al. 2008; Hedenström et al. 2016). However, I recorded individuals of both species making stopovers. Those stopovers were all (except one) located at the end of the migration which suggests that they might not be as important for successful migration compared to stopovers made by birds, insects and terrestrial mammals (Sawyer & Kauffman 2011, McCord & Davis 2012, McGuire et al. 2012). Thus, my work shows how detailed locational information allows us to obtain novel insights about the key behaviours in marine migratory animals.

My findings strongly suggest that turtles rarely stop for resting during oceanic crossing. My results support previous studies using standard Argos-linked tracking (Luschi et al. 1998, Jonsen et al. 2006) and accelerometers (Enstipp et al. 2016), which also indicate that sea turtles swim continuously during migration. Slower travel speeds at night might be explained by turtles swimming in a less direct line at night or/and deeper dives reducing their forward motion (Enstipp et al. 2016). Minamikawa et al. (1997) suggested that the deep dives exhibited by turtles during migration are followed by a gradual passive ascent (e.g. Type 3 dives, with a single descent and ascent phase, or Type 4 dives, characterised by a gradual passive ascending interval from the maximum depth point; Minamikawa et al. 1997). Gradual passive ascent is an efficient way of lowering the cost of transport while travelling over long distances (a strategy well described for migrating birds, e.g. Hedenström 1993; Alerstam et al. 2003). Thus, travelling continuously might minimise the energetic cost of migration if an animal travels at a speed close to optimal cost of transport (Åkesson & Hedenström 2007, Southwood & Avens 2010, Enstipp et al. 2016). If turtles stopped swimming for extended periods at night during the oceanic phase of crossing (e.g. to rest), travel speeds would have been noticeably slower during the oceanic phase of migration than, which was not the case based on my Argos-linked Fastloc-GPS locations. Even though juvenile turtles have been shown to swim directly into strong sea currents to forage (Christiansen et al. 2016), other studies have shown that adult sea turtles only discern approximate headings rather than constantly reassessing their position in relation to their goal 57

Chapter 3: Diel pattern of migrating sea turtles

(Girard et al. 2006, Luschi et al. 2007, Hays, Christensen, et al. 2014, Shimada et al. 2016). Consequently, sea turtles might be more susceptible to deflection by currents if they stopped swimming to rest during oceanic crossing (see the example of the turtle on the eastern Somalian coast, Figure 3.3b). Continuous day/night migration by these two species of turtles implies the importance of reaching distant foraging grounds (1000-4000 km) in the shortest time possible to replenish energetic reserves (Åkesson & Hedenström 2007, Hein et al. 2012). Turtles tend to be capital breeders (i.e. not foraging during breeding, Hamann et al. 2002; Southwood and Avens 2010), meaning that they are likely to commence the migration with depleted energetic reserves, again emphasising the importance of reaching the feeding grounds as quickly as possible.

In contrast to these patterns observed during oceanic crossing, I detected two possible resting strategies towards the end of migration: (1) stopovers (for up to 6 days); and/or (2) intermittent slower night-time travel speeds of up to 66 % in waters shallower than 100 m. Interestingly, all stopover sites in the Mediterranean were located within 1–2 days travel distance of the breeding or foraging sites, and might have been sites previously visited by turtles while foraging, rather than being essential for completing the migration (Fagan et al. 2013). The Argos-linked Fastloc-GPS locations showed that, within 24 h of loggerhead turtles entering waters shallower than 100 m, night-time travel speeds significantly slowed, with this drop recurring every 3–6 days. This possible break in travel may be important to recover from fatigue after at least 1–3 weeks of non-stop travel in many cases. These findings support previous studies, which showed that the dive profile of turtles changes to resting dives once they reach this depth (flat-bottomed dives; described as Type 1 dives in Rice and Balazs 2008, see also Godley et al. 2003). I calculated that turtles could be stopping for a theoretical 8 h on these nights, assuming that day-time travel speeds were maintained. Yet, a similar pattern was not detected for the two green turtles that travelled along the coast of Africa, despite completing a 4000-km journey (i.e. 4 times longer than that of the loggerhead turtles). Thus, the reduction in travel speed at night that I detected during coastal travel for loggerhead turtles may only be beneficial under certain conditions.

Chapter 3: Diel pattern of migrating sea turtles

It has been suggested that the upper ceiling for migration by sea turtles is 2850 km without foraging, but 12 000 km with foraging (Hays & Scott 2013). Yet, in my study, stopover sites were not used by green turtles that migrated 4000 km, suggesting that the fat load (i.e. energy store) of migrating turtles may be higher than previously assumed or that their metabolic rate may be lower. In contrast, Baudouin et al. (2015) found that 12 out of 16 green turtles frequented regular foraging sites while migrating up to 5300 km along a coastal corridor and, in the Pacific, two out of three green turtles made one stopover while migrating about 1000 km (Rice and Balazs 2008). This variation in the use of stopovers might be dependent on individual requirements or might represent “known” refuges within a given proximity to primary foraging or breeding grounds (Hedenström & Alerstam 1992, Fagan et al. 2013). Loggerhead turtles have been shown to foray up to 400 km from breeding areas (Schofield, Hobson, Fossette, et al. 2010, Esteban et al. 2015, Mingozzi et al. 2016) and have benthic foraging grounds ranging from 10 to 100 km2 in size (Broderick et al. 2007, Schofield, Hobson, Fossette, et al. 2010, Mingozzi et al. 2016), indicating that they explore their environment over large areas. Foraging is possible along most of the coast of the Adriatic (demonstrated by published home range datasets for individual foraging sites throughout this area, Schofield et al. 2010a); yet, loggerhead turtles do not make regular stopovers when traversing this region to target foraging grounds. Thus, these turtles might not be aware of potential foraging grounds, only targeting known sites to which they exhibit high fidelity (Schofield, Hobson, Fossette, et al. 2010, Scott et al. 2014, Mingozzi et al. 2016). Only repeat tracking of the same individuals will help to understand the extent to which turtles exhibit fidelity to known stopover sites along their migratory routes (Broderick et al. 2007; Schofield et al. 2010a).

The high volume of Argos-linked Fastloc-GPS locations allowed me to pinpoint the time at which migration started and ended, as well as when turtles arrived at and departed from stopover sites. I found that turtles predominately arrived at and departed from the coast during the day-time. Other studies have also suggested that adult turtles refine their heading towards the target site using visual cues (Hays, Christensen, et al. 2014, Shimada et al. 2016), or a combination of visual and olfactory cues (Åkesson et al. 2003, Hays et al. 2003). Furthermore, 59

Chapter 3: Diel pattern of migrating sea turtles

laboratory studies showed that hatchling and juvenile turtles orientate using visual cues (Lohmann & Lohmann 1996, Avens & Lohmann 2003, Mott & Salmon 2011). Thus, turtles may depart the breeding ground in the early morning so as to use the sun compass for initial orientation (Avens & Lohmann 2003), as detected for other animals (Quinn 1980, Guilford & Taylor 2014). Once migration is underway, magnetic or other navigational cues are likely to be used to maintain heading (Avens & Lohmann 2003). In particular, I showed that the travel speed of turtles was strongly correlated with proximity to the foraging ground on the final night of migration. For instance, turtles that were within 12 hours’ travel distance from their foraging ground slowed or stopped moving the night before arrival, whereas those that were further away maintained their normal travel speed. This phenomenon suggests that turtles were responding to visual cues, adjusting their travel speed to avoid overshooting the target site.

In conclusion, using Argos-linked Fastloc-GPS, I provided information about differences in day and night movement pattern at different stages of migration, allowing me, for the first time, to compare the strategies used by a marine vertebrate with terrestrial and avian species. I showed that two sea turtle species from two ocean basins primarily initiate and end migration during the daytime, suggesting the importance of daytime cues for orientation. I also reported a variety of resting strategies may be utilised during migration, including slightly slower movement at night during the oceanic phase of migration, intermittent nights of very slow movement during the neritic phases of migration and the use of stopovers. These observations were only possible because of the availability of numerous highly accurate Argos-linked Fastloc-GPS tracking locations so access to detailed information allowed me to obtain novel insights about the key stages of migration (start and end of migration), along with potential strategies to reduce the risk of exhaustion during long-distance migration.

Chapter 4: Green turtles home range patterns

Chapter 4 : Diel and seasonal patterns in activity and home range size of green turtles on their foraging grounds revealed by extended Argos- Linked Fastloc-GPS tracking Abstract

An animal’s home range is driven by a range of factors including top-down (predation risk) and bottom-up (habitat quality) processes, which often vary in both space and time. I assessed the role of these processes in driving spatiotemporal patterns in the home range of the green turtle (Chelonia mydas), an important marine megaherbivore. My co-authors satellite tracked adult green turtles using Fastloc-GPS telemetry in the Chagos Archipelago and tracked their fine-scale movement in different foraging areas in the Indian Ocean. Using this extensive data set (5081 locations over 1675 tracking days for 8 individuals), I showed that green turtles exhibit both diel and seasonal patterns in activity and home range size. At night, turtles had smaller home ranges and lower activity levels, suggesting they were resting. In the daytime, home ranges were larger and activity levels higher, indicating that turtles were actively feeding. The transit distance between diurnal and nocturnal sites varied considerably among individuals. Further, some turtles changed resting and foraging sites seasonally. These structured movements indicate that turtles had a good understanding of their foraging grounds with regard to suitable areas for foraging and sheltered areas for resting. The clear diel patterns and the restricted size of nocturnal sites could be caused by spatiotemporal variations in predation risk, although other factors (e.g. depth, tides and currents) could also be important. The diurnal and seasonal pattern in home range sizes could similarly be driven by spatiotemporal variations in habitat (e.g. seagrass or algae) quality, although this could not be confirmed.

Chapter 4: Green turtles home range patterns

Introduction

An animal’s home range is the spatial expression of its movement pattern (Börger et al. 2008), which is the result of complex and dynamic interactions between top-down (Mech 1977, Kittle et al. 2008) and bottom-up processes (Heithaus et al. 2002, Fryxell et al. 2004), which can affect both individual fitness (Lima & Dill 1990, Heithaus & Dill 2006, Heithaus et al. 2007) and population dynamics (Wang & Grimm 2007). Hence, understanding what factors influence the home range of animals is important for predicting the potential consequences of human-induced top-down effects, such as fishery-induced apex predator declines, and bottom-up effects, such as global warming, at both an individual and population level (Boyce & McDonald 1999).

In the absence of predators, animals generally distribute themselves in a way that maximizes their net energy intake, and hence fitness, over time (Lima & Dill 1990, Langvatn & Hanley 1993, Storch 1993, Heithaus et al. 2002). Depending on the ability of an animal to perceive its environment, a forager should direct its foraging effort to subsets of the environment (patches) that, on average, yield higher benefits than the environment at large and move between these patches in a way that maximizes the total net energy intake (Charnov 1976, Brown 1988). Both terrestrial and marine mammalian grazers forage in spatiotemporally complex habitats characterized by patchy distributions of food (WallisDeVries et al. 1999, Robbins & Bell 2000). The spatial distribution of quality food patches has been shown to strongly influence the movement patterns and home ranges of large terrestrial mammalian grazers, which in turn impose patterns on the landscape, which further reinforce this behaviour (Fryxell 1991, Fryxell et al. 2004, Searle et al. 2005).

Under the risk of predation, animals generally alter their movement patterns, and consequently home ranges, in ways that reduce risk at the cost of reduced energy intake from having to reside in sub-optimal areas (Lima & Dill 1990, Houston et al. 1993, Brown 1999, Heithaus et al. 2002). From this comes the notion that herbivores exist in a “landscape of fear” (Laundré et al. 2001), with their home

Chapter 4: Green turtles home range patterns range being the result of a trade-off between maximizing energy and minimizing risk (Lima & Dill 1990, Houston et al. 1993, Brown & Kotler 2004), with selection favouring animals that optimally balance these two components in a way that maximizes fitness over time (Sih 1980, Illius & Fitzgibbon 1994, Lima 2002). The trade-off between predation risk and energy acquisition is a dynamic process, with both components often varying both spatially and temporally (Heithaus et al. 2002). For example, using fine-scale data from GPS radio collars, Creel et al. (2005) showed that elks (Cervus elaphus) reduced their use of preferred, but more risky, grassland foraging habitats when wolves (Canis lupus) were present in the area. Similarly, foraging Indo-Pacific bottlenose dolphins (Tursiops aduncus) matched the distribution of their prey when tiger sharks (Galeocerdo cuvier) were absent, but significantly deviated from these preferred habitats when shark density increased (Heithaus et al. 2002). Similar trade-offs have also been documented for African savannah herbivores (Riginos & Grace 2008, Valeix et al. 2009, Hopcraft et al. 2014), as well as dugongs (Dugong dugon) and green turtles (Chelonia mydas) (Heithaus et al. 2002, Wirsing et al. 2007).

Apart from habitat quality and predation risk, other variables can influence the movement patterns and home ranges of animals. Some of these variables are related to individual characteristics (e.g. age, body condition and reproductive status) as well as the state of the individual (e.g. hungry, satiated), whereas others are external, both biotic (e.g. competition, conspecific behaviour and habitat type) and abiotic (e.g. topography, temperature and precipitation) (McLoughlin & Ferguson 2000, Forester et al. 2007, Börger et al. 2008, Van Beest et al. 2011). Cederlund & Sand (1994) found that male moose (Alces alces) had larger home range sizes than females, due to sex-specific differences in body size. Differences in mating strategies are believed to drive differences in home range sizes between male and female Indo-Pacific bottlenose dolphins, with males having larger home ranges to maximize mating opportunities with multiple females (Sprogis et al. 2016). In green turtles in Shark Bay, Australia, body condition has been found to influence habitat use, and consequently home ranges, with turtles in poor condition selecting more profitable, but risky, microhabitats, during periods of high predation risk, compared to turtles in good condition (Heithaus et al. 2007). Finally, dugongs in Queensland, 63

Chapter 4: Green turtles home range patterns

Australia, forage closer to land during high tide compared to low tide, due to, at least partly, restricted access to intertidal food resources (Sheppard et al. 2009).

Megaherbivores play an important role in structuring primary producer communities in terrestrial, freshwater and marine habitats. Grazers can have positive effects on plant productivity, distribution, community structure, tissue nutrient content, as well as nutrient recycling, which in turn can influence the foraging behaviour and home range pattern of the grazers (Mcnaughton et al. 1997, Ritchie et al. 1998, Atwood et al. 2015). While considerable work has been done to understand the behaviour and home range of terrestrial megaherbivores (Bailey et al. 1996, Fryxell et al. 2004), relatively little attention has been focused on marine megaherbivores, despite these varied ecosystem roles. I therefore set out to assess the extent and drivers of spatiotemporal patterns in the home range of green turtles. This study is timely as it is now feasible to track this species with high resolution, for protracted periods and in remote locations using Fastloc-GPS tags that remotely relay data via the Argos satellite system (Chapter 2Chapter 2).

Methods

Tag deployment and data processing All fieldwork was approved by the Swansea University Ethics Committee, the British Indian Ocean Territory (BIOT) Scientific Advisory Group (SAG) of the UK Foreign and Commonwealth Office and the Commissioner for BIOT (research permit dated 2 October 2012). Research complied with all relevant local and national legislation. My co-authors attached Fastloc-GPS-Argos transmitters to eight adult female green turtles nesting at night on the island of Diego Garcia (7°25′S, 72°27′E) within the Chagos Archipelago during October 2012 [see Hays et al. (2014) for details]. The size of the tagged turtles and tracking details are shown in Table 4.1. To each turtle ID number, a suffix was assigned corresponding to the country in which the eventual foraging grounds were located (Se = Seychelles, Ch = Chagos, Ma = Maldives, So = Somalia). My co-authors used two models of satellite tags (model F4G 291A, Sirtrack, Havelock North, New Zealand, and SPLASH10-BF, Wildlife Computers, Seattle, Washington), both of which relayed

Chapter 4: Green turtles home range patterns

Fastloc-GPS data via the Argos satellite system (http://www.argos-system.org/). Satellite tags were programmed to acquire a maximum of one Fastloc-GPS location every 15 min, although the irregular surfacing pattern of the turtle and intermittent satellite overpasses for data relay resulted in fewer locations being obtained. From the Fastloc-GPS locations, the turtle’s net swim speed was calculated. Before doing so, however, the data were filtered to reduce measurement errors. First, locations with residual values above 35 were removed, in accordance with most Fastloc-GPS tracking studies (Chapter 2). I then processed the data through a speed filter where I removed all positions which would require the turtle to swim at unrealistic speeds (>2.3 m s−1) (Hays et al. 2014b; Chapter 2). I further restricted the location data to those points recorded by five or more satellites, which should result in an accuracy of 55 and 29 m for 75 and 50 % of locations, respectively (Chapter 2Chapter 2). This threshold further assured that more than 95 % of the speed estimations had less than 10 % error (Chapter 2). Hazel (2009) estimated the mean linear error of Fastloc-GPS locations to be 54 (±79), 42 (±53), 33 (±42) and 26 m (±19) for five, six, seven and eight satellites, respectively. Finally, a small number (<0.05 %) of locations were removed because they looked visibly erroneous (i.e. were on land or hundreds of km from the actual locations at the foraging grounds) when plotted spatially in R (R Core Team 2014). Visual examinations of plotted tracks were used to identify when the turtles reached their foraging grounds. At this point, the turtles stopped travelling in a consistent direction and instead started to move back and forth within a relatively restricted area. All location data prior to this time were excluded from analyses, while the remaining data were analysed until the tags stopped working (Table 1).

Diel patterns in movement To investigate diel movement patterns of the turtles, locations were first assigned as either daytime or night-time based on the time of sunrise and sunset for the specific area and season, which was obtained using the package insol in R. The net movement of sea turtles as a function of time of day was investigated using generalized additive mixed models (GAMMs) in R. To bind the fitted values above zero, and to make residuals homogenous, net speed was first log transformed. Because time of day is a circular variable, a cyclic cubic regression spline (type 65

Chapter 4: Green turtles home range patterns

“cc” in the R-package mgcv) was used, where the ends of the regression splines match up. To account for individual variation in movement, turtle ID was added as a random effect in the model. To account for temporal dependence between observations, a temporal auto-correlation structure within each turtle ID was incorporated in the model, where the residuals at any given time were modelled as a function of the residuals of the previous time point. Restricted maximum likelihood estimation was used for estimating model parameters.

Model validation tests were run to identify potential violations of the assumptions of the GAMM. Scatter plots of residuals versus fitted values were used to test the assumption of equal variances (homogeneity) in the model. Normality of residuals was interpreted from quantile–quantile plots and from residual histograms. Auto-correlation function and partial auto-correlation function plots were used to visually detect patterns of temporal auto-regressive and moving average parameters before and after adding the different correlation structures. Because of the irregular surfacing pattern of the turtles, net speeds were estimated over time periods of varying length. To investigate the sensitivity of the model output to this variation, the time periods over which net speed was estimated were artificially restricted to an upper threshold value ranging from 1 to 24 h. The model output was then examined visually (Appendix B, Figure B1).

Seasonal patterns in movement To identify the number of unique diurnal and nocturnal sites for each turtle, I used a Bayesian multivariate behavioural change point analysis (BCPA) on the time series of latitude and longitude for each animal, using the bcp package in R (Barry et al. 1993, Erdman & Emerson 2007). BCPA identifies partitions of sequences (time series) into contiguous blocks with constant means within each block, while assuming independence between observations, normal distributed errors and constant variance throughout each sequence (see Erdman and Emerson [2007] for details). Because the distance of one degree longitude varies across latitudes, both latitude and longitude were converted to Northings and Eastings, expressed in metres. Since a turtle could potentially change its diurnal site seasonally without having to necessarily change its nocturnal site, and vice versa, I ran separate BCPAs

Chapter 4: Green turtles home range patterns for the daytime and night-time positions. To fulfil the assumption of independence between locations (location data are naturally temporally auto-correlated), only a single location was used for each day and night, respectively. To make sure that the locations corresponded to actual daytime and night-time hours, I only included positions recorded within 3 h of midday and midnight, respectively. I used the default setting of the BCPA model (see Erdman and Emerson [2007], following the recommendations by Barry and Hartigan [1993]). For the Markov Chain Monte Carlo methods, 10,000 iterations were run, with a burn in period of 1000 iterations. From the resulting posterior probability, a lower threshold value of 0.95 (95 % probability that a given time point is a change point) was used to identify change points. Because I was interested in persistent changes in diurnal and/or nocturnal sites, rather than short-term deviations in diurnal and/or nocturnal sites, I ignored change points occurring within ten days of another change point. Locations that ended up in time periods between two identified blocks were allocated to the block located closest in space.

Home ranges Green turtle home range sizes were estimated using Kernel Utilization Distribution (KUD) (Worton 1989) using the adehabitatHR package in R, with the reference bandwidth as smoothing parameter. The area of each identified diurnal and nocturnal site was estimated independently for each turtle. Diurnal and nocturnal activity centres were identified using 50 % KUD (Worton 1989). As for the BCPA, temporal auto-correlation was accounted for by using only a single location for each day and each night, respectively. To investigate how spatiotemporal patterns in the movement of turtles influence the home range size estimates, the 95 % (overall home range) and 50 % KUD (core area) were estimated for each individual at decreasing level of spatiotemporal complexity: High, where KUD was estimated for each diurnal and nocturnal site separately and summed together for each individual to take into account both diel and seasonal patterns in home range; Medium, where KUD was estimated for daytime and night-time positions separately and then summed together for each individual, to account for diel patterns in home range; Low, where a single KUD was estimated for each individual, using one daytime and one night-time location for every 24-h period to 67

Chapter 4: Green turtles home range patterns account for temporal auto-correlation between locations; and None, where KUD was estimated directly from the filtered raw data.

Home range influence on activity budget The size and shape of a turtle’s home range are likely to influence the proportion of time that it spends foraging, resting and in transit, which constitute its activity budget. In particular, the distance between the diurnal and nocturnal sites is likely to influence the proportion of time that the turtle spend in transit between sites. The longer a turtle spends in transit, the less time it will have available for foraging and/or resting which, over time, could have consequences for the animals bioenergetic budget and ultimately fitness (New et al. 2014, Christiansen & Lusseau 2015). To better understand the potential fitness consequences of variations in the turtle’s home ranges, I developed an individual-based model for each of my eight turtles where I simulated the daily movement for each turtle over a year. For each day in the simulation, a diurnal and nocturnal site was allocated based on the number of unique sites for that individual identified by the BCPA. For animals with multiple diurnal and/or nocturnal sites, the number of simulated days spent in each site was set to be proportional to the relative amount of time spent in each site during the actual study period. After having allocated a diurnal and nocturnal site to each day, one daytime and one night-time location was drawn at random from the corresponding KUDs for those sites for each day. The transit time between the two sites was then estimated based on the distance between the two locations and the swim speed of the turtle during transit. I set the swim speed during transit to be 0.6 m s−1, based on Watanabe et al. (2011). I further assumed that the speed of travel did not differ between individuals, as cost of transport for similar-sized turtles should be similar. At the end of the simulation, the mean proportion of time spent in transit over the year and the 95 % highest posterior density intervals were estimated using bootstrapping resampling methods (1000 iterations).

Chapter 4: Green turtles home range patterns

Results

Foraging ground locations and sample size After being tagged, the turtles remained for varying lengths in the Chagos Archipelago breeding ground before starting their migrations back to their different foraging grounds across the Indian Ocean. Two turtles travelled west to the coast of Somalia, four to the Amirantes Islands, Seychelles, one travelled north to the Maldives, while the last turtle migrated to the Great Chagos Bank (Figure 4.1). A detailed description of the migration of the eight tagged turtles can be found in Hays et al. (2014).

Figure 4.1: Top-left subfigure shows the migratory movements of the eight tracked adult female green turtles (solid black lines) from their nesting beach on Diego Garcia, Chagos Archipelago, to their respective foraging grounds (red triangles) in the Indian Ocean. The smaller subfigures show the foraging grounds of each turtle (see ID number at the top of each subfigure), with blue and red dots indicating daytime and night-time locations, respectively (the sample size is shown in the lower- right corner of each subfigure). The light grey lines show the movement tracks of turtles within their foraging grounds. Grey areas indicate land.

Chapter 4: Green turtles home range patterns

After the turtles had reached their foraging grounds, the tags kept transmitting for two to 18 months, resulting in a total of 1675 tracking days (Table 4.1). After data filtering, 5081 Fastloc-GPS locations remained, ranging between 103 and 1637 per individual (Table 4.1). The average number of locations obtained per day per individual ranged between one and five. On their foraging ground, all eight turtles stayed within relatively small areas (Figure 4.1; Table 4.1). The only exception was turtle 61811-So which, after spending 152 days on its foraging ground off the coast of Somalia, made a short excursion (circa 64 km) southwest along the coast before returning back to its foraging ground after 10 days. The accumulated distance travelled during this excursion was about 64 km. To simplify my analyses, this part of the track (35 locations) was excluded from the data set. For all individuals, the locations within the foraging grounds were distributed heterogeneously in space, with clusters of positions occurring in specific areas within each foraging ground (Figure. 4.1).

Table 4.1. Summary data of the eight satellite tracked adult female green turtles on their foraging grounds in the Indian Ocean. To each turtle ID number, a suffix was assigned corresponding to the country in which the eventual foraging grounds were located (Se = Seychelles, Ch = Chagos, Ma = Maldives, So = Somalia).

Track CCL Nb Locations Lat.Dist Long.dist Turtle ID durations Start date End Date (cm) locations day-1 (km) (km) (days)

21923-Se 110 96 28/02/2013 4/06/2013 146 1.52 5.5 4.67 117568-Ch 104 538 8/11/2012 30/04/2014 1637 3.04 5.65 5.65 117569-Se 101.5 381 3/01/2013 19/01/2014 1178 3.09 20.8 5.57 117570-Ma 103 128 13/03/2013 19/07/2013 103 0.8 5.77 4.29 4394-Se 104 66 27/11/2012 1/02/2013 154 2.33 6.58 6.08 21914-Se 105 153 23/12/2012 25/05/2013 662 4.33 11.72 7.6

61811-So 111.5 223 21/12/2012 1/08/2013 1050a 4.71a 1.99a 2.89a 61813-So 106 90 7/03/2013 5/06/2013 151 1.68 1.06 3.66

Diel patterns in movement

Time of day had a significant effect on the net swim speed of turtles (F7.8,2374.2 = 118.8, p < 0.001, based on swim speeds estimated over time periods of <3 h). Individual variation accounted for 6.7 % of the total variation in the data. Adding a temporal auto-correlation structure, an auto-regression structure of lag one, 70

Chapter 4: Green turtles home range patterns improved the model significantly (Log-likelihood ratio test: L = 176.9, df = 1, p < 0.0001) and also removed any pattern of auto-correlation from the residuals. The full model explained 28.9 % (adjusted R2) of the variance in net speed.

There was a curvilinear relationship between net speed and hour of day for green turtles (Figure 4.2). The activity level (i.e. net swim speed) during night was lower (~0.2 m s−1) than during daytime hours (~0.4 m s−1). Just before sunrise, the activity of the turtles started to increase rapidly, with the turtles reaching a peak in activity between 6 and 8 am. This peak was followed by a lower level of activity (~0.4 m s−1) throughout most of the daylight hours, although significantly higher than during night. Shortly before sunset, there was a second peak in activity, between 4 and 6 pm, before the activity level dropped again for the night (Figure 4.2). While the second peak in activity was slightly lower than the first, this could be an artefact of fixing the time of sunrise to 6 am in the analyses, while sunset was allowed to vary seasonally over the year. This was done to facilitate comparison between turtles located at different time zones and latitudes. Although the magnitude of both activity peaks varied depending on the upper threshold chosen for including net speed estimates, the general pattern was consistent across threshold (Appendix B, Figure B1).

Chapter 4: Green turtles home range patterns

Figure 4.2: Back-transformed swim speed as a function of hour of day for the eight tracked green turtles in their Indian Ocean foraging grounds. The solid black line represents the fitted values of the best fitting GAMM. The white and dark grey background colours indicate daytime and night-time hours, respectively. The time of sunrise was fixed to 6 am for all turtles and the strip of light grey background colour represents dusk, which varied seasonally over the year. The dashed lines represent 95 % confidence interval. Swim speeds were estimated over time periods of 3 h and less. n = 2383 speed estimates.

Seasonal patterns in movement The BCPA identified 10 and 11 unique diurnal (Table 4.2) and nocturnal sites (Table 4.3) for my eight turtles, respectively. While most turtles were shuttling daily between a single diurnal and a single nocturnal site throughout the study period, three animals changed their diurnal and/or nocturnal site seasonally (Appendix B, Figure B2 and B3). Turtle 21923-Se spent its daytime and night-time hours in adjacent areas (D1 and N1) for the first 50 days, before abruptly changing both its diurnal and nocturnal sites to a new area (D2 and N2) located approximately 4 km north, where it remained for the last 47 days of the track (Figure 4.3, Appendix B, Figure B2 and B3). Turtle 117569-Se revisited the same diurnal and nocturnal sites over the course of the tag deployment. It spent the first 11 days in a restricted area located in the northern part of its home range (D4 and N4), before relocating to another area approximately 11 km south, where it spent 129 days (D5 and N5) (Figure 4.3, Appendix B, Figure B2 and B3). The turtle then returned to its initial site (D4 and N4), where it stayed for 135 days, before again relocating to the second site (D5 and N5), where it spent the remaining 100 days of the track. Turtle 61811-

Chapter 4: Green turtles home range patterns

So stayed in the same diurnal site over the duration of the study, but changed its nocturnal site (N9) after 187 days to a new site (N10) located about 2 km west, where it stayed at night for the remaining 16 days of the track (Figure 4.3, Appendix B, Figure B2 and B3).

Figure 4.3: Diurnal (D; blue contour lines) and nocturnal (N; red contour lines) sites of the eight tagged female green turtles on their foraging grounds in the Indian Ocean, estimated using 50 % Kernel Utilization Distributions. The numbers next to the letters indicate the ID number of the specific site, whereas a and b represent sites that had two centres of activity, but were not temporally segregated (the turtle moved back and forth between these two sites on a day-to-day basis). The ID number of each turtle can be seen on top of each sub-figure. The daytime and night-time location data that were used to estimate the home ranges are shown as blue and red dots, respectively. Only one daytime and one night-time location for every 24-h period was used to account for temporal auto-correlation between locations. No locations during transit were used. Grey areas indicate land.

Home ranges Both during day and night, the turtles restricted their movement to relatively small areas, identified from 50 % KUD (Figure 4.3). Although diurnal sites were generally larger in size (95 % KUD: mean = 20.0 km2, SD = 14.4; 50 % KUD: mean = 3.6 km2, SD = 3.1) compared to nocturnal sites (95 % KUD: mean = 10.2 km2, SD = 16.5; 50 % KUD: mean = 1.6 km2, SD = 2.1), there were two exceptions

Chapter 4: Green turtles home range patterns

(see ID 21923-Se and 61813-So, Figure 4.3; Tables 4.2, 4.3). The degree of overlap between diurnal and nocturnal sites differed markedly between individuals, as did the distance between sites (Figure 4.3). While most diurnal and nocturnal sites had a single centre of activity, some sites had two centres which the turtle regularly moved between (D7a and D7b for Turtle ID:4394-Se, D9a and D9b for Turtle ID:61811-So, D10a and D10b and N11a and N11b for Turtle ID:61813-So, Figure 4.3). There were large differences in the size of both diurnal and nocturnal sites, both within and between individuals (Tables 4.2, 4.3).

Table 4.2: Summary table of the 10 identified diurnal sites of the eight tracked green turtles on their foraging ground in the Indian Ocean.

95 % 50 % Diurnal Duration Turtle ID N KUD KUD site ID (days) area area D1 21923-Se 15 51 1.18 0.27 D3 117568-Ch 268 537 8.51 0.93 D4 117569-Se 71 145 26.16 2.60 D5 117569-Se 127 228 10.08 0.97 D6 117570-Ma 25 127 20.94 4.91 D7 4394-Se 28 61 44.14 10.56 D8 21914-Se 109 154 25.06 2.91 D9 61811-So 127 222 3.78 0.89 D10 61813-So 21 55 12.04 2.97

Chapter 4: Green turtles home range patterns

Table 4.3: Summary table of the 11 identified nocturnal sites of the eight tracked green turtles on their foraging ground in the Indian Ocean.

95 % 50 % Nocturnal Duration Turtle ID N KUD KUD site ID (days) area area N1 21923-Se 17 50 6.13 1.18 N3 117568-Ch 183 532 0.09 0.00 N4 117569-Se 74 186 22.75 3.53 N5 117569-Se 75 178 27.00 2.64 N6 117570-Ma 13 119 3.42 0.74 N7 4394-Se 19 66 4.42 0.94 N8 21914-Se 84 152 2.54 0.38 N9 61811-So 84 187 0.73 0.11 N10 61811-So 13 16 0.44 0.11 N11 61813-So 21 89 13.43 3.20

Accounting for diel and seasonal patterns in movement had large effects on the estimated home range sizes of the turtles (Appendix B, Table B1). Accounting for temporal auto-correlation between locations (low complexity) resulted in larger estimated home range sizes compared to the raw location data (no complexity) (Appendix B, Table B1). Adding diel patterns into the home range estimation (medium complexity) had a large effect on the resulting size; however, the direction and magnitude of this effect varied between individuals (Appendix B). Finally, for individuals that had multiple diurnal and/or nocturnal sites, incorporating both seasonal and diel patterns in movement (high complexity) lead to a significant reduction in home range sizes, sometimes even below that of the raw data (no complexity) (Appendix B, Table B1).

Home range influence on activity budget

My simulations showed that the eight turtles varied significantly in the proportion of time they spent in transit on their foraging grounds (Figure 4.4). While the size of the home ranges affected the daily variation in transit within individuals (the size of the error bars in Figure 4.4), individual differences in the distance between diurnal and nocturnal sites were the main cause for the large variation in transit time between individuals (Figure 4.4).

Chapter 4: Green turtles home range patterns

Figure 4.4: Simulated proportion of time spent in transit for the eight green turtles on their foraging grounds in the Indian Ocean. Error bars represent 95 % highest posterior density intervals. The means and density intervals are based on 1000 model simulations, where the daily movement for each turtle was simulated over a year. For each day in the simulation, a diurnal and nocturnal site was allocated based on the 50 % Kernel Utility Distributions for the specific turtle (Figure 4.3).

Discussion

The aim of this study was to investigate spatiotemporal patterns in the home range of green turtles to better understand the relative importance of top-down and bottom-up processes affecting this marine megaherbivore. Fastloc-GPS tags allowed me to track the fine-scale movement of green turtles for up to two years on their foraging grounds with the high quantity and quality of the locations giving me an unprecedented insight into the fine-scale movement patterns of green turtles compared to studies using conventional Argos tracking (Hays et al. 1999, Godley et al. 2002). Hence, in concurrence with Börger et al. (2008), I stress the importance of incorporating spatiotemporal patterns in animal in animal movement when estimating home range sizes.

The low level of activity during night, coupled with restricted nocturnal home range sizes, suggests that turtles were resting at night. During daytime, the activity levels were higher and the home range sizes larger, inferring that turtles were foraging within their diurnal sites during day-time. This diel movement between distinct foraging and resting sites, also observed in several other studies 76

Chapter 4: Green turtles home range patterns

(e.g. Makowski et al. 2006, MacDonald et al. 2013, Gredzens et al. 2014), could be the result of top-down effects from predation risk resulting in turtles seeking sheltered habitats during night to avoid predation from large sharks. Turtles rely on vision to detect sharks and might therefore avoid foraging at night to reduce predation risk (Heithaus et al. 2002, Makowski et al. 2006). Turtles generally rest close to reef structures, where they can find shelter under reef ledges, in small caves and crevices in the sides of the reefs (Makowski et al. 2006, Hazel 2009). Preference for safer habitats during resting has also been observed in other species, including desert baboons (Papiocynocephalus ursinus) (Cowlishaw 1997), dugongs (Dugong dugon) (Sheppard et al. 2009), spinner dolphins (Stenella longirostris) (Tyne et al. 2015) and bottlenose dolphins (Heithaus et al. 2002). Although the bottom substrate was unknown, nocturnal sites were generally smaller in size and often located closer to land presumably in habitats with more structure (e.g. caves) for shelter, although high-resolution habitat maps for these areas were not available. That the turtles showed such high fidelity to these specific sites suggests they must offer some level of protection for the turtles that makes it worthwhile to return to them. Predation risk could therefore help explain why the turtles sought out specific resting sites at night that were sometimes even spatially segregated from their daytime foraging sites.

Other possible explanations for why turtles selected specific resting sites at night also need mentioning. Resting turtles might prefer certain depths where they can stay neutrally buoyant with greater oxygen stores (more inflated lungs) and remain submerged for longer periods of time before having to breathe (Hays et al. 2000, Minamikawa et al. 2000). Unfortunately, detailed bathymetry maps of my study areas were not available to test this hypothesis. Tides and ocean currents can also influence turtle movement and habitat use, with turtles in some foraging grounds showing strong circatidal movement patterns (Brooks et al. 2009) or restricted home ranges during low tide (Limpus & Limpus 2000). While some turtles in this study showed a clear diel, rather than circatidal, pattern in activity and home range size, ocean currents still might influence habitat choice at night, with turtles selecting nocturnal sites that are protected from currents. The large variation in movement and home range patterns of green turtles recorded around the world 77

Chapter 4: Green turtles home range patterns

(Bjorndal 1980, Seminoff et al. 2002, Makowski et al. 2006, Taquet et al. 2006, Hazel 2009, Senko et al. 2010, MacDonald et al. 2013) indicate that green turtles have a high degree of plasticity in their behaviour and that their movement and home range patterns are influenced strongly by local environmental features.

I found large differences in diurnal home range sizes of turtles in this study. Further, three of the eight tracked turtles changed their home range pattern seasonally. Seasonal movement between foraging patches is a common behaviour observed in terrestrial grazers (Fryxell et al. 2004, 2008, Hopcraft et al. 2014), with animals moving between dense prey patches in a manner which maximizes energy intake over time (Charnov 1976, Brown 1988). Similarly other species of sea turtles (e.g. loggerhead turtles) are also using different foraging sites during the winter period (Mingozzi et al. 2016). Rather than being distributed homogenously over the sea floor, seagrass is generally found in well-defined patches (Robbins & Bell 2000), similar to terrestrial grass systems (WallisDeVries et al. 1999). Green turtles are known to re-graze seagrass patches within a foraging site (Bjorndal 1980, Zieman et al. 1984). Repeated grazing of seagrass patches may increase seagrass food quality by enhancing the production of new leaves that are higher in nutrient content and therefore more easily digested by the turtles (Bjorndal 1980, Zieman et al. 1984, Aragones et al. 2006). The timing of re-grazing will depend on the recovery time of the seagrass (which can vary substantially from a couple of weeks up to a year depending on the location of the seagrass bed), the timing and the intensity of the grazing (including turtle density), the seagrass species composition, depth and the location of grazing within the beds (Zieman et al. 1984, Rasheed 1999, Aragones & Marsh 2000, Rasheed et al. 2014). While this study has provided insights into the movement pattern of foraging sea turtles, the lack of information about resource (i.e. seagrass and algae) quantity and quality prevented me from testing any further hypotheses in relation to optimal foraging behaviour in this species. Nevertheless, the measured individual variation in diurnal home range sizes and the structured seasonal movement of turtles between foraging sites suggest that bottom-up processes relating to resource (i.e. seagrass and/or algae) quantity and quality could be shaping these behavioural patterns.

Chapter 4: Green turtles home range patterns

The structured and predictable nature of the movement and home range patterns in this study suggest that the turtles had a good spatial understanding of their foraging grounds, which allowed them to make informed decisions on where and when to move to find suitable foraging and resting areas. This stands in stark contrast to the random walk foraging movement of pelagic marine predators where the knowledge of the prey field is generally poor (Sims et al. 2008, Humphries et al. 2010). However, while the tracked turtles showed some similarities in movement and home range patterns, there were also some considerable differences between individuals. The transit distance between foraging and resting sites varied considerable between individuals, which resulted in differences in activity budgets between turtles, with animals transiting further having less time available for foraging compared to turtles foraging closer to their resting sites. With all of the turtles being mature females of similar size (within 10 % carapace length), it is unlikely that this difference is due to size-specific variations in food requirements and physiology, as observed by Ballorain et al. (2010). Instead, it is possible that the observed individual variation in home range sizes and transit distance reflect variation in habitat quality (food quantity and quality) between the different foraging grounds (Festa-Bianchet 1988). Turtles might be willing to travel further from their resting sites in order to reach more profitable seagrass beds, even if this means that they will have less time available per day to forage there, as long as it maximizes net energy intake over time (Charnov 1976, Brown 1988). Hence, the estimated activity budgets in this study might not necessarily reflect the turtles’ energetic budgets. In addition, other factors such as body condition and competition might also influence the movement and home range sizes of green turtles (Fretwell 1969, Heithaus et al. 2007). A direct assessment of the seagrass quality and quantity of the foraging sites in combination with direct observations of sea turtle behaviour and condition will help answer these questions. Seagrass ecosystems have been poorly studied in the western Indian Ocean and need to be given higher priority in regional habitat studies.

In summary, I highlight the value of new generation Fastloc-GPS Argos tags for resolving the details of sea turtle movements at small scales. The complexity of movements over different spatial scales points to animals that have a good 79

Chapter 4: Green turtles home range patterns knowledge of their environment, commuting between suitable foraging and resting sites and changing these sites over time in a way that likely allows patch recovery and maximize energy intake. These complexities of shifts in foraging habitat patch use over time and the associated commuting to night-time refuges, likely occur broadly across marine and terrestrial systems although resolving these complexities and generalities remains a key question (Hays et al. 2016).

Chapter 5: Loggerhead turtles home range patterns

Chapter 5 : Complex movement patterns by loggerhead sea turtles at foraging and wintering sites revealed by Argos-Linked Fastloc-GPS Abstract

How animals select and use habitats informs us about the spatial heterogeneity of resources and the level of competition for those resources. Here, I used Argos- linked Fastloc-GPS to investigate how loggerhead sea turtles use benthic foraging and wintering habitats at multiple scales, from the general site scale to fine-scale patches and day/night activity centres within patches. Turtles used up to four sites each, remaining for a mean of 150 days and returning to the same site a minimum of 52 days later, possibly reflecting the regeneration time required for benthic organisms on which they forage (e.g. sponges and molluscs). However, the area within sites was not used uniformly, with turtles exhibiting complex movement patterns within and among up to five focal patches. Fidelity to individual patches was low or variable, even for the same turtle, and might be driven by differences in resource availability (i.e. food and/or night-time refuges), competition and/or exploratory movement to investigate or locate alternative patches. My co-authors confirmed competition from direct visual observations of aggressive interactions between individuals at one focal foraging patch. During winter, most turtles showed no reduction in activity levels compared with other seasons; however, one turtle moved to alternative sites at a lower latitude and two clearly reduced activity (i.e. entered a wintering/dormant state), possibly triggered by changes in environmental conditions (e.g. sea temperature) and/or resource availability or productivity. In conclusion, my study identified three distinct movement patterns at the patch level within sites, which was linked to differences in day/night activity centre use, highlighting the need to determine what factors drive fine-scale movement patterns during foraging, which would help to ensure a suite of appropriate habitats is selected for protection by management programs.

Chapter 5: Loggerhead turtles home range patterns

Introduction

Animals must select where to forage and how long to remain in resource-rich patches to maximise individual fitness (Searle et al. 2005). Various ecological models, such as the marginal value theorem (Charnov 1976, WallisDeVries et al. 1999, Thompson & Fedak 2001), assume that patch quality (i.e. prey density, quality and distribution) is the primary constraint on foraging. For instance, small scale patches (in both space and time) of varying quality are assumed to be nested within larger scale habitats (or sites), with individuals moving when the quality of a given small patch falls to the average overall quality of the habitat (Charnov 1976). However, there is a lack of consensus on the scale at which foraging decisions are made, potentially confounded by differences in the size and mobility of different organisms, leading to variable support for this theory in studies of wildlife inhabiting terrestrial and marine systems (for an overview see: Searle et al. 2005, Foo et al. 2016).

Other factors are also likely to contribute to decisions made by individuals to remain at a single patch or move to alternative patches. For instance, the presence of competitors influences patch quality and causes additional energetic demands associated with vigilance and agonistic interactions (Stephens & Krebs 1986, Houston et al. 1993, Searle et al. 2005). The risk of predation also leads to different areas being used for specific activities, such as resting and foraging (Holt 1977, Lima & Dill 1990, Sinclair et al. 1998, Heithaus et al. 2007). Thus, animals make complex decisions regarding their movement patterns within and across patches, and hence adjust their overall home range, in response to multiple variables (Robbins & Bell 2000, Johnson et al. 2001, Seidel & Boyce 2016).

Marine turtles are primarily capital breeders (i.e. store energy while foraging for reproduction, tending not to feed during breeding; Bonnet et al. 2016) that generally forage at sites located up to thousands of kilometres from their breeding grounds. Thus, they must acquire enough energy reserves during periods of intensive foraging to sustain themselves for around four to six months, encompassing the period of migration and breeding activity away from the foraging

Chapter 5: Loggerhead turtles home range patterns sites (Hays, Mortimer, et al. 2014, Scott et al. 2014). Using satellite transmitters, my research group and others have shown that sea turtles may optimise foraging effort by exhibiting high fidelity to known high quality sites (Broderick et al. 2007, Schofield, Hobson, Fossette, et al. 2010), but also frequent additional sites, possibly in response to changes to climatic or environmental conditions (Schofield, Hobson, Fossette, et al. 2010, Schofield, Dimadi, et al. 2013). Various telemetry studies (e.g. radio, satellite, GPS) have also shown that home ranges can extend from <10 km2 to >1000 km2 for individuals from any of the seven sea turtle species (Seminoff et al. 2002, Hatase et al. 2007, Schofield, Hobson, Lilley, et al. 2010), depending on their proximity to shore and habitat type, which probably reflects resource quality (Seminoff et al. 2002, Hatase et al. 2007, Schofield, Hobson, Lilley, et al. 2010). At finer resolutions, radio and acoustic telemetry studies of green turtles have suggested that daytime foraging is concentrated in defined areas of high resource quality, while night-time resting centres are located nearshore reef areas (Brill et al. 1995, Seminoff et al. 2002, Makowski et al. 2006, Fujisaki et al. 2016), supported by direct observations (Mortimer & Portier 1989). In contrast, hawksbills appear to exhibit more complex movements, with some turtles primarily using one area, others using several areas and others exhibiting no clear pattern (Chevis et al. 2016). While loggerheads are mostly carnivorous (Bjorndal 1985), adults primarily forage on sessile benthic organisms attached to rocks (e.g. sponges) or in the seabed (e.g. molluscs) (Houghton et al. 2000, Schofield et al. 2006, Casale et al. 2008, Lazar et al. 2011), so their forage resources are distributed similarly to those of large terrestrial herbivores (Senft et al. 1987).

Here, I used Argos-linked Fastloc-GPS to investigate the movement patterns of loggerhead sea turtles in benthic (nearshore and offshore) habitats on the continental shelf of the Mediterranean Sea. Argos-linked Fastloc-GPS generates more uplinks with 10–100 greater accuracy than standard Argos technology (Tomkiewicz et al. 2010; Lopez et al. 2014; Chapter 2); thus, here, I aimed to quantify how the pattern of movement changed at multiple scales: (1) the site level; (2) the patch level within sites; and (3) day/night activity centres within patches. By defining the pattern of movement at these different scales, I will be addressing a key ecological question on sea turtle movement (Hays et al. 2016): how do sea 83

Chapter 5: Loggerhead turtles home range patterns turtles utilise their foraging areas in space and time and how different movement patterns emerge at different spatial scales. I intend to use this information to infer whether similar or different factors drive and/or limit resource use at each scale.

Methods

Study animals This study used loggerhead sea turtles equipped with high-resolution Argos-linked Fastloc-GPS units at a single breeding population (Zakynthos Island, Greece; 37.80 °N, 20.75 °E) which migrated to foraging sites up to 1545 km distant throughout the Mediterranean Sea (Figure 5.1). A total of 40 male (including 10 residents) and 17 female loggerhead turtles were tracked from Zakynthos between 2007 and 2012, of which some were excluded from the current study (see Appendix C for selection process). All tracking datasets have been previously analysed, but with a different focus (e.g. Schofield et al. 2010b, Hays et al. 2014). The curved carapace length of both male and female loggerhead turtles combined was 83.4 ± 6.1 cm (mean ± SD; Schofield et al. 2013b).

Data preparation I first assimilated the raw data for all turtles using Quantum GIS V2.10.1 software (QGIS Development Team 2015). I then removed all GPS locations obtained with four or fewer satellites and with residual values of >35 (residual values are provided by the software converting pseudoranges into location estimates; see Chapter 2Chapter 2 for full details). I then obtained real time travel speeds by calculating the speed of travel from location intervals separated by 3 h, to ensure high quality estimates (99 % of speed of travel estimates with an error <10 % of the true value; see Chapter 2Chapter 2). Finally, I removed all locations with unrealistic speeds of travel (>200 km.day-1) or that were visibly erroneous in QGIS (i.e. were on land or hundreds of km from the actual locations at the foraging grounds, representing <0.1 % of locations). Day and night were distinguished based on local nautical dusk and dawn times (http://www.esrl.noaa.gov/gmd/grad/solcalc). Data on seabed depth were obtained using the GEBCO global relief model (http://www.gebco.net/). For

Chapter 5: Loggerhead turtles home range patterns turtles that were repeat-tracked over two consecutive years, the datasets were merged.

All home ranges were estimated by determining the Kernel Utilisation Distribution (KUD; Worton 1989) using the adehabitatHR package (Calenge 2006) in R (Version 3.2.3, R Development Core Team 2013) and using the reference bandwidth as the smoothing parameter (using the 90 % KUD to infer the size of foraging areas). Because the number of GPS locations per day varied among individuals (mean: 11.7 ± 11.4; range: 2–70) and decayed the longer the turtles were tracked, I selected two locations per day (the closest to midnight and the closest to midday) per turtle to allow day/night and among individual comparisons. The adehabitatHR package requires at least five locations to compute a home range; so, all home ranges in this study were calculated using at minimum of five days of data. For each home range estimate (e.g. site, patch within site and day versus night activity centres within patches), I calculated the overall home range size and the mean size of the monthly home range (when possible) so that I could compare turtles that were tracked for different durations. I calculated the overlap between any two home ranges as the percentage of the home range of the first turtle that overlapped with that of the second turtle (Fieberg & Kochanny 2005). I used the Wilcoxon non-parametric test to compare means (reported as mean ± 1 SD throughout the manuscript). Similarly, I used Wilcoxon non-parametric paired test to compare day/night activity centres home ranges.

Site level Within this study, I defined sites as non-overlapping large-scale habitats (following Charnov 1976) used by loggerhead turtles when not breeding. I detected the date of arrival at foraging and/or wintering sites by the end of directional movement, a consistent change in travel speed to <1 km.h-1 and the displacement distance from the breeding site remaining constant for at least six consecutive days (Rice & Balazs 2008, Schofield, Hobson, Fossette, et al. 2010). I termed ‘home’ as the point that represented the first day at each site. I confirmed that this point was not a migration point by conducting a sensitivity analysis comparing how the distance from home changed when shifting the day ± 3 days from that identified (Appendix C, Figure

Chapter 5: Loggerhead turtles home range patterns

C4). Departure from one site and arrival at a second site was determined by an increase in travel speed (to >1 km h-1) on departure and decrease in travel speed (to <1 km h-1) on arrival at the next site, with sites being separated by at least 25 km (calculated from the location identified on the day of arrival at the first site, which was termed home, Figure 5.2e), and turtles remaining at the subsequent site for at least six days (Rice & Balazs 2008, Schofield, Hobson, Fossette, et al. 2010), otherwise it was considered to be a foray (Schofield, Hobson, Lilley, et al. 2010). If the home range of potential adjacent sites overlapped, they were merged as a single site (following Schofield et al. 2010b). For turtles that were resident to Zakynthos (the breeding area), I excluded all datasets from 1 March to 15 June each year within Laganas Bay, which is the mating period at this site (Schofield, Scott, et al. 2013).

In total, 31 foraging and wintering sites of 24 turtles were retained for my analyses, all of which were in neritic habitat (depth <100 m), either on the coast or in open waters (Appendix C, Figure C1, C2 and C3). For each site, I recorded the duration that each turtle spent foraging and/or wintering, and the time it spent away from the site before returning (when available), in addition to determining the home range size.

I distinguished wintering movement patterns from foraging movement patterns based on changes to the following parameters between November and March/April: (1) a steep initial drop in the number of GPS locations recorded, which plateaued throughout this period and was followed by a steep increase in the number of locations at the end; (2) a reduction in daily movement distances; (3) reduced home-range size; and/or (4) a shift to deeper water (Broderick et al. 2007). The same information was recorded for wintering sites as foraging sites.

Patch level Patches were defined as high-use areas (i.e. a high concentration of GPS locations through time) within distinct foraging and/or wintering sites, after excluding forays outside the site from the data. Patches were distinguished by fine-scale changes in movement in relation to ‘home’ (i.e. the first patch). The period spent in a patch was identified by a minimal range of movement, whereas movement between 86

Chapter 5: Loggerhead turtles home range patterns patches is characterised by a sharp change in distance from ‘home’ (patch one) or a subsequent patch. I found that turtles that remained at sites for >2 months used most (83 %) of the patches within a given site (Appendix C, Figure C5); thus, I only included turtles that remained >2 months at sites in the patch-use analysis to ensure that I captured most of patches that could potentially be used in each site. As a result, I retained 22 sites from 15 turtles (removing nine sites and nine turtles out of the 31 sites and 24 turtles previously retained at the site level).

I recorded the duration that turtles spent in each patch and time away from it before returning, and determined the home range using the 90 % KUD (following the same methods used for the site home ranges). For all turtles, I determined variation in day and night home ranges (termed activity centres) within patches using the 90 % KUD. I performed a series of Principal Component Analyses (PCA) of the quantitative variables (e.g. overall home range size, day/night activity centres overlap, day/night distance from shore) to identify the variables that explained most of the variance in the dataset at the patch level.

Direct observations in a foraging patch During 2015, a foraging patch was surveyed at intervals of up to five days during the daytime from 21 July to 10 August. In each survey, my co-authors photographed all sighted turtles, determined the sex where possible (males have tails >5 cm length with no external tags, confirmed females have tails <5 cm length with external metal or plastic flipper tag, possible females or subadults/juveniles have tails <5 cm length with no external tags; Casale et al. 2005, Rees et al. 2013) and recorded all behaviours and interactions with other turtles following Schofield et al. (2006, 2007) over a 2-hour period per survey. All turtles were identified based on a long term photo-identification database (Schofield et al. 2008) and the behaviours observed during interactions were documented (e.g. circling, chasing, biting at flippers, head-on confrontations; Schofield et al. 2006, 2007). The photo- identification database was used to validate that all (except one) individuals classified as male were in fact males based on previous documentation of their mating activity between 2000 and 2014.

Chapter 5: Loggerhead turtles home range patterns

Results

Twenty-four turtles were tracked for a mean 6.0 ± 4.5 months (range 0.3–18.8 months) following their arrival at the first benthic foraging site. Eighteen turtles migrated away from Zakynthos (to other sites in the Gulf of Gabes, the Ionian Sea or the Adriatic), while two departed after remaining on Zakynthos for several months, and four turtles remained resident to Zakynthos.

Site level A total of 31 foraging and wintering sites were detected for the 24 turtles, of which 23 sites were nearshore (all located in the Ionian Sea) and eight sites were offshore (all located in the Adriatic Sea and Gulf of Gabes; Figure 5.1). Of the 31 sites, one was a distinct wintering site, while wintering patches were detected within a further two sites. A typical example of a turtle using two sites is given in Figure 5.2ab.

Number of sites For 11 turtles, data on benthic foraging site use for the entire period between two consecutive breeding seasons (i.e. approximately one year of data or more) were obtained. Of these turtles, six used a single site and five used two to four sites during this period. For 13 turtles, the transmitters stopped at some point during the first year, while turtles were at the foraging/wintering sites. For these turtles, 11 (five males and six females) used one site and two used two sites (one male and one female).

Chapter 5: Loggerhead turtles home range patterns

Figure 5.1: Locations of the 31 benthic foraging and wintering sites retained for this study and identified from 24 turtles (18 males and 6 females) tracked with Argos- linked Fastloc-GPS in the Mediterranean Sea. The numbers in the white squares represent the number of foraging and wintering sites in each area with 23 nearshore sites (red numbers) in the Ionian Sea region (one in Corfu, seven in Amvrakikos Bay, eight on Zakynthos and seven on the Peloponnese), seven offshore (black numbers) sites in the Gulf of Gabes and one offshore site in the Adriatic Sea. Areas with a seabed depth <100 m are represented in blue. Foraging site exact locations are provided in Appendix C, Figure C2.

Duration of site use For all male turtles where the arrival and departure from sites could be determined, the first site frequented after departing the breeding area was used the longest (mean: 167.8 ± 95.0 days, range 19.7–317.9 days; n = 9 sites of 9 turtles) and subsequent sites were frequented for shorter periods (mean: 101.9 ± 53.9 days, range: 17.7–164.8; n = 7 sites of 5 turtles). For the single female where full data were available, the first site was used for just 17 days and the second site for 62 days; however, the female arrived at the foraging site in September, whereas the males arrived in June-July. These durations noticeably differed to those recorded for turtles where the transmitters stopped while frequenting a site (Appendix C).

Chapter 5: Loggerhead turtles home range patterns

Repeated use of sites Six turtles revisited foraging sites that had been previously used; two after foraging/wintering at another site and four after breeding first (Figure 5.3a). The mean interval between visits for all turtles combined was 162.4 ± 124.9 days (range: 52.3–361.2 days). Two male turtles revisited foraging sites at Amvrakikos 52.2 days later (migrating to and from the breeding ground in between) and 255.4 days later (migrating to another foraging site first, then breeding and then returning). One male from Tunisia and one male from Zakynthos also revisited the same foraging site 168.8 days later (migrating to and from the breeding ground in between) and 361.2 days later (migrating to another foraging site first, then breeding and then returning). One female turtle returned to a previously-used site off Tunisia after 72.0 days away (at another foraging site), and one male turtle returned to a previously- used site on Zakynthos after 64.9 days away (at another foraging site).

Home ranges Out of the 31 retained sites, 24 sites contained sufficient GPS locations for the home range analysis. The overall mean 90 % KUD home ranges of the nearshore and offshore sites were 24.4 ± 22.5 km2 (range: 1.4–93.1 km2, n = 18 sites) and 105.0 ± 57.9 km2 (range: 15.0-170.9 km2, n = 6 sites), respectively. The home ranges of offshore sites were significantly larger and deeper compared to nearshore sites (Figure 5.2f, Wilcoxon, p = 0.003 for the home ranges and p = 0.007 for the depth). Similar results were obtained when the monthly mean home range of each site were used (data not shown). On Zakynthos, up to five turtles used the same foraging site, with a mean 52 ± 43 % overlap in home range area (range: 0.2– 100 %); however, sites were not necessarily used at the same time, due to individuals being tracked across multiple years.

Out-of-site forays A total of 11 forays by seven turtles were documented from seven of the 34 sites. Forays lasted a mean 4.0 ± 3.0 days (range: 1.0–9.9 days) including travel time with turtles moving up to 80 km (mean: 35.7 ± 27.3 km) from the foraging site. Six forays by four turtles were recorded in the Ionian Sea, while five forays by three turtles were recorded in the Gulf of Gabes. Eight of the forays occurred 90

Chapter 5: Loggerhead turtles home range patterns between June and October (mean 3.0 ± 1.4 days, range: 1.4–4.8 days). One turtle from the Gulf of Gabes forayed for 9.2 days in December. Another turtle from the Zakynthos (Ionian) forayed once in mid-late December (1.3 days) and once in mid- late January (9.9 days).

Patch level For turtles with >2 months data, a total of 37 patches (26 nearshore and 11 offshore) were identified across 22 sites, with turtles using 1–4 patches per site (Figure 5.2cd; mean: 1.7 ± 0.9 patches per site). The number of patches used within a given site was similar in both the nearshore and offshore benthic sites, with a mean 1.6 ± 0.7 patches (range: 1–3 patches) and 1.8 ± 1.2 (range: 1-4), respectively (Wilcoxon, p = 0.4908). For both nearshore and offshore benthic sites, about 70 % of turtles used patches for <20 days and, of those, more than 40 % used patches for <3 days (even though some turtles used a single patch for up to 243 days). Out of those 37 patches, 32 contained sufficient locations for the home range analysis.

Chapter 5: Loggerhead turtles home range patterns

Figure 5.2: (a) Example of two foraging sites located 50 km apart that were used by a single loggerhead turtle (one in Zakynthos for 128 days and one in the Peloponnese for 165 days) before it moving back to Zakynthos for breeding. (b) The movement between the two foraging sites is characterised by a sharp increase in the distance (from 0.5 to 50 km) to the first location recorded at the Zakynthos foraging site. (c) Example of a loggerhead turtle using two focal patches in Zakynthos Bay (90 % KUD in blue). For comparison, the 90 % KUD home range of the site is displayed in red. (d) The turtle alternated between the two patches for durations of 19 and 30 days at Patch 1 and 20 days and 78 days at Patch 2 before leaving the site. The movement between patches is characterised by a sharp increase (from 0.8 to 7 km) of the distance to the first location recorded in Patch 1. (e) Variation of the overall home range of nearshore (n = 18 in white squares) and offshore foraging sites (n = 6, white circles) and nearshore (n = 26 in grey squares) and offshore patches (n = 6, grey circles) with the seabed depth. The solid black line represents the logarithmic 92

Chapter 5: Loggerhead turtles home range patterns

regression (r = 0.75, p < 0.001) between the overall size of home ranges of the foraging sites and depth and the solid grey black line between the size of the home range of patches and depth (r = 0.52, p < 0.001). The size of the patches increases with the depth at a slower rate compared to the overall site home ranges. This reflects the increasing contribution of exploratory movement to delineate the site limits.

Nearshore movement patterns I identified three patterns in patch use by nearshore turtles (Figure 5.4, Appendix C, Table C1): (1) ‘Type 1’, in which turtles exhibited strongly overlapping day/night activity centres, with very small home ranges (Figure 5.4abef, Table 5.1); (2) ‘Type 2’, in which turtles used day and night activity centres with low overlap, shuttling between the two, and with intermediate-sized patch home ranges (Figure 5.4acef, Table 5.1); and (3) ‘Type 3’, in which one turtle used multiple day and night activity centres with low overlap over a wide area, and with large-sized patch home ranges (Figure 5.4adef, Table 5.1).

All three patch types were used in all months of the year (Figure 5.3b). Type 1 and 3 patches extended further offshore (max. 3.5 km offshore) compared to Type 2 patches (max. 1 km offshore); however, the Type 2 and 3 night-time activity centres occurred closer to shore (<0.5 km), whereas those at Amvrakikos occurred further offshore (0.5 km and deeper than the daytime sub-patch). Furthermore, the home ranges of the night-time activity centres for Type 2 and 3 patches were about half the size of the daytime activity centres (Wilcoxon paired test, p = 0.003). Types 1, 2 and 3 patches were used for up to 243, 120 and 63 days, respectively. However, Type 2 patches seemed to be used and returned to after much shorter intervals compared to the other two patch types (Table 1). When turtles were using a single Type 1 or Type 2 patch (but not Type 3), some moved beyond the patch on occasion (termed exploratory movement here), but remained within the site limits and did not stop at another patch before returning. This exploratory movement was twice as likely in Type 1 patches versus Type 2 patches. In patches where they explored, turtles moved beyond the patch twice as frequently, travelled twice as far and for

Chapter 5: Loggerhead turtles home range patterns longer durations from Type 2 patches compared to Type 1 patches (Table 1). Additionally, the patches home range of five turtles overlapped in Zakynthos.

Figure 5.3: (a) Examples of the Type 1, Type 2 and Type 3 patch use patterns over a period of 15 days in Zakynthos Bay. The home ranges of daytime and night-time activity centres were calculated over a five-day window to clarify the differences between the three patterns. The turtles used mainly Type 1 patches (b) characterised by strongly overlapping day/night activity centre use with very tight home ranges. In Type 2 patches (c) the turtles used different day and night-time activity centres shuttling between the two, with intermediate home ranges. In Type 3 patches (d), one turtle used multiple day and night-time activity centres over a wide area, resulting in the largest home

Chapter 5: Loggerhead turtles home range patterns ranges. (e) Distance from home for the three patch-use patterns. The variations in the distance from home for Type 1 patches (black dotted line) are small compared to Type 2 (solid grey line) and Type 3 (solid black line). The turtles in Type 2 and 3 patches are exhibiting a clear diel pattern and are resting close to the shore at night and foraging further offshore during daytime. (f) Grouping of the three patch types using a principal component analysis. Type 1 patches are represented by black circles (n = 16), Type 2 by grey circles (n = 7), Type 3 by white circles (n = 2) and Offshore sites by white squares (n = 7). To produce the PCA plot, the data were reduced to three variables (overall patch home range, distance from shore, and day/night overlap between the centres of activity), which explained most (76 %) of the total variance. The offshore patches are clearly discriminated from the nearshore patches. Type 1, 2 and 3 patches clusters were mostly discriminated by the overall patch home range and the overlap in day/night activity centres. One Type 3 patch appeared as an outlier because of its large home range (37 km2) compared to other nearshore patches.

Offshore benthic movement patterns

These patches were similar in size to Type 2 and 3 nearshore patches (Wilcoxon, p = 0.126), but had strongly overlapping day and night-time activity centres similar to Type 1 nearshore patches. Patch-use duration and revisitation intervals were slightly longer than for nearshore patches (Figure 5.3b, Table 1). The turtles also made exploratory movement more often, over longer distances and for longer duration compared to nearshore Type 1 and 2 patches (Table 5.1).

Direct observations Over the 20-day observation period, 13 turtles were observed at the foraging patch on nine surveys feeding on a sponge on each occasion, Chondrilla nucula (Phylum Porifera, Class Demospongiae). Of these turtles, five were confirmed males, one was a confirmed female, and the rest were unconfirmed sex/age class. All turtles were documented foraging on sponges at the patch; however, just two of the turtles 95

Chapter 5: Loggerhead turtles home range patterns

(Males 48 and 110 in the photo-id database) were recorded using the site on more than two occasions, i.e. eight and six times respectively (Figure 5.5b). A total of seven interactions between individuals were observed (all of which involved at least one confirmed male), six of which involved Male 48, Male 110, or both males over a 14-day period. Male 48 displaced two males (Males 110 and 216) in two interactions, but also departed following an interaction with Male 110 on another occasion, with all of these interactions escalating from circling to biting the carapace (Figure 5.5b). Male 110 displaced Male 225 on one occasion, with no escalation. The other three interactions were between Male 48 and possible females (n = 2 interactions) or Male 216 and one possible immature turtle.

Table 5.1: Main characteristics of loggerhead turtles foraging behaviour in the three types of foraging patches. Nearshore and offshore Type 1 foraging patches were analysed separately. A dash in a cell means that no value was calculated for the corresponding metric (e.g. when no foray was observed in a patch).

Nearshore Nearshore Nearshore Foraging patch type Offshore Type 1 Type 2 Type 3

Number of turtles 9 3 1 5 Number of patches recorded 16 8 2 11 Mean patch home range (km2) 2.3 ± 1.9 4.3 ± 3.2 24 ± 11 8.8 ± 6.2 Mean overlap of day/night centres of activity (%) 86 ± 11 46 ± 11 36 ± 5 73 ± 22 Maximum patch-use duration (days) 243 120 63 165 Percentage of patch-use duration <3 days (%) 22 66 50 - Percentage of revisitation intervals <3 days (%) 29 59 0 - Maximum revisitation interval (days) 204 47 64 167 Percentage of patch-use durations <10 days (%) - - - 44 Percentage of revisitation intervals <4 days (%) - - - 50 Percentage of patches with exploratory movement (%) 62 32 0 50 Mean number of exploratory movement per month 3 ± 4 6 ± 2 - 7 ± 6 Extent of exploratory movement (km) 6 3 - 11 Mean exploratory movement duration (days) 2 ± 2 2 ± 2 - 3 ± 3

Chapter 5: Loggerhead turtles home range patterns

Figure 5.4: Photographs of: (a) a male loggerhead (Male 216) foraging on the sponge Chondrilla nucula (Phylum Porifera, Class Demospongiae); a common prey item for the loggerheads in Laganas Bay, Zakynthos; and (b) antagonistic interaction between two loggerhead males (Males 48 and 216) at the foraging patch. The male on the right (Male 48) is biting the hind flipper the male on the left, displacing it. (Photo credit: Kostas Papafitsoros).

Discussion

My results show that loggerhead sea turtles exhibit complex movement within and across foraging and wintering sites, showing a variety of movement patterns at the site, patch and activity centre level. Even when using a single patch, turtles appeared to actively explore the site about four times a month in about 50 % of patches, leaving for two or more days before returning to the same patch. These movement patterns are likely driven by differences in forage resource richness, competition and preference for certain types of night-time refuges, influencing the duration and frequency of patch use and, hence, site dimensions. Direct observations at one patch indicated that multiple turtles frequent the same patch, resulting in aggressive

Chapter 5: Loggerhead turtles home range patterns competition. Furthermore, I detected only a few instances of wintering activity, with this behaviour possibly being driven by a combination of environmental conditions (e.g. sea temperature) and changes to resource availability or productivity. My findings demonstrate the complexity of movement by loggerhead turtles, emphasising the importance of identifying the factors that drive variation in patch use within foraging sites so as to improve how home range estimates are used to inform management decisions when delineating protective zoning.

As detected for large terrestrial herbivores and marine fishes, turtles tended to occupy sites containing one or more patches, which are likely due to aggregation of forage resources (Senft et al. 1987, WallisDeVries et al. 1999, Searle et al. 2005). The offshore benthic sites were much larger than the nearshore benthic sites, but patch number and size were comparable, supporting the notion that resources are more broadly dispersed and have lower biomass in offshore sites (Duggins et al. 1989, Bowyer & Kie 2006, Cardona et al. 2007, Schofield, Dimadi, et al. 2013). The greater size of offshore sites could be attributed to exploratory movement beyond the patch being more frequent, involving greater distances and longer durations that for nearshore sites, but remaining close enough to the patch so as to be delineated within the site limits (and, thus, influencing home range size) (Bartumeus et al. 2016). Furthermore, movement beyond patches, but remaining within the site, might be an important strategy to actively sample the environment and monitor other potential patches in which resources might be depleted or replenishing (Bartumeus & Catalan 2009, Benhamou 2014). Turtles also made forays of up to 10 days and 80 km beyond the site before returning, which may have been driven by low resource availability in patches within the site. Furthermore, turtles have been shown to use multiple sites when shifting to lower latitudes in winter (Schofield, Dimadi, et al. 2013), but also at other times of the year (Pilcher et al. 2014), which might be linked to changes in resource quality and availability in response to changes in environmental conditions (Charnov 1976, Searle et al. 2005). For instance, individuals did not return to sites for at least 52 days, which might be sufficient for targeted benthic species to replenish, supporting the marginal value theorem (Charnov 1976). Sponges and bivalves are key forage resources of loggerheads in the Mediterranean (Houghton et al. 2000, Schofield et al. 2006, 98

Chapter 5: Loggerhead turtles home range patterns

Casale et al. 2008, Lazar et al. 2011) and require a minimum of 2-5 months to replenish (Fromont and Bergquist 1994, Okniwa et al. 2010, López-Legentil et al. 2013), aligning with the approximate re-visitation duration observed. Additionally, mobile benthic prey items (e.g. echinoderms, decapods or gastropods) can contribute significantly to the turtles’ diet (Hochscheid et al. 2013). However the density of mobile prey in the areas surrounding the patches is likely to be low, otherwise the sea turtles should have exploited those areas, which was not supported by my data.

At the patch level, I observed high plasticity in terms of duration, revisitation and day-night activity centre-use patterns. Similarly complex behaviour has been detected using acoustic tracking for hawksbills (Chevis et al. 2016), but my results contrasted with observations of green turtles, which appear to have clearly defined day and night activity centres within patches (Christiansen et al. 2017). I suggest that the observed patterns were driven by competition for resources, in addition to resource type and quality in the nearshore patches (Stephens & Krebs 1986, Houston et al. 1993, Searle et al. 2005). For instance, as the quality of the patch declines or competition increases, turtles might be more likely to depart. This suggestion of the importance of competitive interactions was supported by the fact the home ranges overlapped for up to five turtles at each of the site and patch levels on Zakynthos, along with direct underwater observations (with supporting video) of agonistic interactions among individuals during foraging. As comparison, in Florida, Fujisaki et al. (2016) observed multiple green turtles using the same patch within a foraging site while Bjorndal (1980) reported an absence of aggressive behaviour and no indication of a hierarchy among the turtles.

At some locations, I observed loggerheads using distinct night-time activity centres within patches, as has been detected for green turtles (Seminoff et al. 2002, Makowski et al. 2006, Christiansen et al. 2017), but also more broadly for other marine species (e.g. some fish species, Gladfelter 1979, Lowry and Suthers 1998). In contrast, at other patches, loggerheads used highly overlapping day and night- time activity centres. This variability in behaviour among patches has also been

Chapter 5: Loggerhead turtles home range patterns detected for individual fish within a population (Jadot et al. 2006). Use of the same patch during day and night has been linked to an optimum feeding strategy, involving high familiarity of fixed sites so as to increase feeding efficiency (Jadot et al. 2006), and minimise energy and time expenditure incurred by moving between the resting and feeding areas. Interestingly, of turtles that exhibited distinct day and night activity centres, those in Amvrakikos Gulf moved further offshore at night, similar to that observed for juvenile green turtles in Florida, USA (Hart et al. 2016) although turtles on Zakynthos moved closer to shore at night. The use of the same or different day and night-time activity centres might be due to differences in habitat substrate and the bathymetry of the immediate area, reflecting an innate mechanism to seek shelter and avoid predators, as shown for other turtles (Seminoff et al. 2002, Makowski et al. 2006, Christiansen et al. 2017) and wildlife (e.g. daily movement of copepods, Hays et al. 2001; Roe deer Padié et al. 2015). Thus, within each patch, a turtle might adjust it decision to seek nearshore night-time shelters depending on the proximity of such shelters (Heithaus et al. 2007). Loggerheads forage on a mixture of benthic taxa, and so might frequent different patches within a site or different patches across sites to access different forage matter (Houghton et al. 2000, Schofield et al. 2006, Casale et al. 2008, Lazar et al. 2011). This phenomenon might explain why I detected three different foraging movement patterns across patches, with individuals using different movement patterns in different patches. For instance, foraging on sponges might be restricted to certain reefs, resulting in tight Type 1 patch movement patterns (e.g. Figure 5.2ab), while foraging on bivalves in extensive submerged sandbanks might result in broad Type 2 or 3 movement patterns (e.g. Figure 5.2ac).

Through breaking down the patterns and scale of movement by turtles in foraging habitat, I delineate site, patch and day-night activity centres, and suggest how foraging decisions link these different scales of movement (Searle et al. 2005, Foo et al. 2016). For instance, patch size was regulated by day-night activity centres, which probably varied with respect to the type of forage matter available in a patch and access to, or importance of, night-time refuges. However, patch size did not increase at the same rate as site size: rather, even though site size increased with seabed depth, patch size remained comparatively small between nearshore and 100

Chapter 5: Loggerhead turtles home range patterns offshore benthic foraging sites, with variability in the number of patches contained in a given site. In fact, site size appeared to be driven by within-site exploratory movement back and forth from a given patch or among patches, with turtles regularly exploring the same areas outside the patches. These observations implied that sites potentially hold multiple patches, but that the actual number of viable patches within sites changes over time. Thus, regular exploration of the site informs the turtle of the status of different patches (i.e. resource type, quality and availability, and competition for this resource) and reduces uncertainty, allowing the individual to adapt to potential changes in food quality.

Exploration beyond the site was only recorded on 11 instances, thus, the decision to move from a site may be driven by changes in patch viability or the absence of alternative patches within a site (Searle et al. 2005). This observation demonstrates the need for detailed information about how a given site is used so as to determine why turtles use just one versus multiple sites. For example, quantifying the forage intake of loggerheads in individual patches would help clarify whether departure decisions at the patch scale invokes the marginal value theorem (Charnov 1976). This theorem has been successful to predict the patch departure decisions of multiple species of large terrestrial herbivorous (Searle et al. 2005) where animals leave a patch when its foraging rate drops to the overall average intake for the entire habitat (Charnov 1976). However, for several other species, the theorem failed to predict the departure and decisions were better predicted when predation rate, behaviour states (e.g. antagonistic interactions in our study) and regulation of body mass where also included in the model (Nonacs 1991). My results show that the pattern of site use by loggerheads is complex, with patch characteristics likely to be the primary driver. Thus, a turtle occupying a site with low patch viability is likely to change sites, whereas a turtle occupying a site with multiple patches of high viability is likely to remain at the site for an extended period.

My study revealed several factors that might limit, or even confound, results relating to foraging area use. Firstly, even at the site level, I found that the duration of site use was shorter, and thus likely underestimated, when the transmitter stopped before the turtle departed. Thus, to quantify home range size and delineate patch

101

Chapter 5: Loggerhead turtles home range patterns use with a high degree of accuracy, a complete record of movement patterns within a given site is required. I also showed that I had to track turtles for at least two months to identify most of the patches that it frequented within a site. This suggests that omnivorous/carnivorous turtles like loggerheads and hawksbills (Chevis et al. 2016) might exhibit far more complicated patch use compared to primarily herbivorous green turtles (Seminoff et al. 2002, Makowski et al. 2006, Christiansen et al. 2017), including much more exploratory behaviour within the site. Thus omnivorous/carnivorous turtles are likely to require longer tracking durations to adequately describe their behaviour. Finally, I showed that, for some types of patches, distinct day and night activity centres were used by loggerheads, similar to that detected for green turtles (Seminoff et al. 2002, Makowski et al. 2006, Christiansen et al. 2017). Thus, when standardising locations across individuals to allow evaluation and comparison in home range use, at least two locations should be used per day, and preferably more (as in Seminoff et al. 2002, Makowski et al. 2006), rather than the single location that is currently selected by most studies (Hawkes et al. 2011). Failing to adequately consider these factors could result in the erroneous interpretation of data leading to repercussions such as inappropriate maritime zoning when data are used by management agencies. For instance, zones could be unintentionally biased to include only areas of high daytime or night-time activity, rather than encompassing the suite of habitats used by turtles for different activities within a patch or site (Gaston et al. 2008, Grüss et al. 2011).

In conclusion, this study shows that loggerheads exhibit complex movement patterns at foraging sites, possibly as the result of complex decisions in response to multiple variables, including resource quality, refuge availability and level of competition. My findings build on existing terrestrial and marine fauna studies investigating what drives individuals to remain or depart from patches. My findings have important implications for management decisions related to marine zoning, demonstrating the importance of identifying the factors that drive decisions in each patch to ensure that the entire suite of habitats used by turtles in a given site are encompassed.

102

Chapter 6: Energetic cost of reproduction in whales

Chapter 6 : Noninvasive unmanned aerial vehicle provides estimates of the energetic cost of reproduction in humpback whales Abstract

An animal’s body condition will affect its survival and reproductive success, which influences population dynamics. Despite its importance, relatively little is known about the body condition of large whales and its relationship to reproduction. I assessed the body condition of humpback whales (Megaptera novaeangliae) at a breeding/resting ground from aerial photographs recorded using an unmanned aerial vehicle (UAV). Photogrammetry methods were used to measure the projected surface area of individual whales, which was used as an index for body condition. Repeated measurements of the same individuals were not possible; hence, this study represents a cross- sectional sample of the population. Intraseasonal changes in the body condition of four reproductive classes (calves, immature, mature, and lactating) were investigated to infer the relative energetic cost that each class faces during the breeding season. To better understand the costs of reproduction, I investigated the relationship between female body condition (FBC) and the linear growth and body condition of their dependent calves (CBC). I documented a linear decline in the body condition of mature whales (0.027 m2 d-1; n = 20) and lactating females (0.032 m2 d-1; n = 31) throughout the breeding season, while there was no change in body condition of immature whales (n = 51) and calves (n = 32). The significant decline in mature and lactating female’s body condition implies substantial energetic costs for these reproductive classes. In support of this, I found a positive linear relationship between FBC and CBC. This suggests that females in poorer body condition may not have sufficient energy stores to invest as much energy into their offspring as better conditioned females without jeopardizing their own body condition and survival probability. Measurement precision was investigated from repeated measurements of the same animals both from the same and different photographs, and by looking at residual errors in relation to the

103

Chapter 6: Energetic cost of reproduction in whales positioning of the whales in the photographs. The resulting errors were included in a sensitivity analysis to demonstrate that model parameters were robust to measurement errors. Our findings provide strong support for the use of UAVs as a noninvasive tool to measure the body condition of whales and other mammals.

104

Chapter 6: Energetic cost of reproduction in whales

Introduction

The body condition of animals in a population has important implications for individual survival and reproductive success, and consequently population dynamics. Body condition can be expressed through any physiological index that represents an individual’s energy reserves (Hanks 1981, Millar & Hickling 1990). Animals in good body condition generally have larger energy stores and therefore display more resilience and higher survival than individuals in poorer condition (Gaillard et al. 2000, Cook et al. 2004, Clutton-Brock & Sheldon 2010). Body condition is also related to the reproductive success in female mammals, and it influences factors such as the timing of reproduction (Cameron et al. 1993), the probability of conception (Loudon et al. 1983), fertility rates (Albon et al. 1983), fetal growth (Skogland 1984), offspring mass (Atkinson & Ramsay 1995), and calf survival (Festa-Bianchet 1998). The body condition of individuals constituting a population will strongly influence the population dynamics and, in turn, a population’s conservation status (Dobson 1992, Wauters & Dhondt 1995, Sæther 1997).

For capital breeding animals, where the costs of reproduction are met by endogenous energy stores during a period of fasting (Stephens et al. 2009) the link between body condition and reproduction is particularly strong (Festa-Bianchet 1998, Bonnet et al. 2002). Baleen whales (mysticetes) are considered capital breeders because they finance the costs of reproduction in low-latitude breeding grounds with stored energy acquired at high-latitude feeding grounds (Lockyer 1987a, Kasuya 1995). A large body energy storage is a necessity for baleen whales as energy stores constitute the primary source of energy during the breeding season, when whales migrate to less productive equatorial waters (Lockyer 2007). Maternal body condition in baleen whales influences fecundity (Lockyer 2007, Williams et al. 2013) and fetal development (Lockyer 2007, Christiansen et al. 2014). Although not documented for baleen whales, maternal body condition in other marine mammals, such as pinnipeds, influences offspring birth and weaning mass (Boltnev

105

Chapter 6: Energetic cost of reproduction in whales

& York 2001, Bowen et al. 2001), which, in turn, influences offspring survival (Mcmahon et al. 2000).

Much of the stored energy in baleen whales can be found in the muscle and adipose tissue (blubber and visceral fat), although a considerable amount is also stored in internal organs, bones, and other tissues (Lockyer 1986, 1987b, Vikingsson 1995, Naess et al. 1998, Miller et al. 2011, Christiansen et al. 2013). Several of these body tissues have been used as proxies for body condition in studies on baleen whale bioenergetics, including lipid concentration (Aguilar & Borrell 1990, Naess et al. 1998), blubber thickness and volume (Vikingsson 1990, Miller et al. 2011, Christiansen et al. 2013) and weight of different organs and tissues (Lockyer 1987b, Vikingsson 1995). However, a more commonly used metric for body condition in whales is body girth (Lockyer 1987b, Vikingsson 1990, Haug et al. 2002) or width (Perryman & Lynn 2002, Miller et al. 2012), which encompasses several of these tissues and, hence, provides a more holistic measure of body condition. Variation in girth and width in relation to prey abundance, seasonal fasting, feeding, and reproductive status has been measured in a number of baleen whale species, including minke (Balaenoptera acutorostrata), fin (Balaenoptera physalus), gray (Eschrichtius robustus), and right whales (Eubalaena sp.; Lockyer 1987a, b, Ichii et al. 1998, Naess et al. 1998, Haug et al. 2002, Perryman & Lynn 2002, Miller et al. 2012). These studies highlight that body girth and width measurements provide a reliable and good measure of body condition in baleen whales.

For migratory baleen whales, the temporal segregation between feeding and breeding is often reflected as seasonal changes in body condition, with an increase during the summer feeding season and a decrease during the winter breeding season (Lockyer 1987b, Vikingsson 1990, 1995, Naess et al. 1998, Konishi et al. 2008, Christiansen et al. 2013). Comparison of intraseasonal trends in body condition between different reproductive classes can provide valuable insights into the relative rates of energy acquisition and expenditure at different time periods. This includes variations in energy expenditure at different stages in their reproductive cycle (e.g., early and late gestation, and lactation, Lockyer 1986, Pettis et al. 2004,

106

Chapter 6: Energetic cost of reproduction in whales

Bost et al. 2011, Miller et al. 2012). The strong relationship between energy storage and reproduction in baleen whales (Lockyer 1987a, Williams et al. 2013, Christiansen et al. 2014) makes it possible to estimate the energetic costs of reproduction by investigating the rate of decline in the body condition of a population throughout the breeding season, when whales are fasting. The cost of reproduction is one of the main drivers of the life history of species (Stearns 1992) and is a key component in any study attempting to understand the bioenergetics of baleen whales (Lockyer 2007).

While several studies have shown that pregnant and lactating females both acquire and expend more energy than most other reproductive classes (Lockyer 1987a, b, Vikingsson 1990, Miller et al. 2012, Christiansen et al. 2013), relatively little is known about the direct relationship between female body condition and reproductive success in baleen whales. Lockyer (1987a) and Williams et al. (2013) reported that fecundity in fin whales was related to yearly variations in maternal body condition. In addition, Christiansen et al. (2014) documented that fetal growth was significantly affected by the relative body condition of females in minke whales. However, little is known about the relationship between maternal body condition and the growth, condition, and survival of their offspring postparturitions. Studies on pinnipeds highlight that the size of an offspring at birth and at weaning will strongly influence its survival probability (Boltnev et al. 1998, Mcmahon et al. 2000). Understanding how female body condition influences calf development and survival in baleen whales is therefore fundamental for our understanding of reproductive physiology and life history.

During every austral winter and spring, between May and November, humpback whales (Megaptera novaeangliae) from the breeding stock D population in the Southern Hemisphere migrate north from their summer feeding grounds in Antarctica (~56° S), along the Western Australian coastline to their tropical breeding grounds around Camden Sound (~15° S) in the Kimberley region (Chittleborough 1965, Jenner et al. 2001, Gales et al. 2010). Like most populations of large whales, breeding stock D was severely depleted through unsustainable whaling practices in the last century (Chittleborough 1965, Gibbs 2012). Following

107

Chapter 6: Energetic cost of reproduction in whales the cessation of whaling of humpback whales in the Southern Hemisphere in 1963, breeding stock D experienced a remarkable recovery of >10% increase per year (Hedley et al. 2011, Kent et al. 2012) and has almost rebounded to preexploitation numbers (>30,000 animals, Hedley et al. 2011, Bejder et al. 2016). As the population size approaches its carrying capacity (Braithwaite et al. 2012), density- dependent effect on body condition and reproduction is expected (Albon et al. 1983, Sæther 1997, Stewart et al. 2005). With the humpback whale being an important top predator, a better understanding of their body condition and its link to reproduction is fundamental to better understand the potential effects that these recovering populations will have on ecosystems in the coming decades.

The aims of this study therefore were to: (1) investigate intraseasonal variation in the body condition of four reproductive classes of humpback whales off northwestern Australia during the breeding season in order to determine the relative energetic costs that the different classes face; and (2) understand the relationship between the body condition of lactating females and the growth and condition of their dependent calves, to improve our understanding of the link between body condition and reproduction in baleen whales. To answer these questions, I used novel unmanned aerial vehicle (UAV) technology and photogrammetry methods to measure the body condition of individual whales. This is the first time that noninvasive methods have been applied to assess body condition within the family balaenopteridae, with this study also providing a feasibility study for this approach on free-ranging humpback whales.

Methods

Study area Exmouth Gulf is located at the southern end of the breeding ground for stock D on the northwest shelf of Australia, between 21°45′ S–22°33′ S and 114°08′ E–114°40′ E (Figure 6.1). The Gulf is approximately 3000 km2 in size, with a mean depth of <20 m. Exmouth Gulf serves as an important breeding and resting area for breeding stock D humpback whales on their southwards migration returning from Camden Sound to their Antarctic feeding grounds (Jenner et al. 2001). While whales start to

108

Chapter 6: Energetic cost of reproduction in whales arrive in the Gulf in July during the northern migration, most of the whales enter the Gulf during the southern migration between late August and late October (Chittleborough 1953, Jenner et al. 2001). This time period corresponds to when the majority of calves are born, with the Gulf believed to serve as an important resting and nursing area for the whales. While females with calves can stay inside the Gulf for up to two weeks before continuing their southern migration, mature males can remain for almost a month in search for females to mate with (Jenner & Jenner 2005). While in the Gulf, females with calves generally maintain a low level of activity at or near the surface, while mature males actively search and compete for females. This, together with calm weather conditions, allows the Gulf to act as an ideal site for UAV-based fieldwork on humpback whales.

Figure 6.1: Map of the Exmouth Gulf study area in Western Australia, displaying the survey track lines (solid lines) during the study period (3 August–16 September 2015) and the positions of the sightings (black circles) containing sampled humpback whales (n = 200 whales).

Sampling protocol All research was carried out under a research permit from the Department of Parks and Wildlife, Western Australia (SF010372), and an animal ethics permit from Murdoch University (R2736/15). The former permit allowed for a total of 200 whales to be sampled. The UAV was operated under a UAV Operator Certificate (CASA. UOC.0136) and a Remotely Piloted Aircraft System Licence in accordance with regulations by the Australian Civil Aviation Safety Authority.

109

Chapter 6: Energetic cost of reproduction in whales

Vertical aerial photographs of humpback whales were taken in Exmouth Gulf in August and September 2015. A 5.5-m research vessel surveyed the Gulf haphazardly in search of whales (Figure 6.1). The boat was launched from either the town of Exmouth or the Learmonth Jetty, next to Learmonth airport (Figure 6.1). Once a single or a group of whales were sighted, the boat approached to a distance of 100–300 m at idle speed. A small (50 cm diameter, 1.2 kg) water-proof “Splashdrone” quadcopter (www.swellpro.com) was deployed from a custom- made helipad at the bow of the boat and flown above the whales. Calibration of the gyro sensors of the UAV was made on land prior to launching the boat, and a dual battery system was used to keep the volatile memory of the gyro offsets throughout the day by not having to power down the UAV when switching batteries between consecutive flights. The UAV had a flight time of 8–10 min and could operate up to 1 km from the launch site. The distance between the UAV and the boat was usually kept to less than 300 m to provide a clear line of sight of the UAV and to facilitate positioning over the whales (also with a live video link—see below). The UAV was equipped with a Canon PowerShot D30 (Canon Inc., Tokyo, Japan) still camera, which took photographs at 2-s intervals, and a GoPro Hero4 (GoPro Inc., San Mateo, California, USA) continuously recording video. Both cameras were mounted vertically under the UAV. Typically, the UAV remained about 5 min above a whale resulting in, on average, 150 still images per individual. The UAV was initially flown at an altitude of 30–50 m, to obtain close-up photographs of the whale’s body shape. A live video link, providing the UAV operator with direct feed from the GoPro camera, was used to correct the position of the UAV above the whale and also confirm that photographs of adequate quality had been obtained. Desired photographs were of a whale lying flat at the surface, dorsal side facing up, with a straight body axis and peduncle (that was nonarching; Figure. 6.2). The UAV was then flown to an altitude of 80–120 m, while the research vessel moved closer to the whale until both the whale and the boat were visible in the same photograph. The size of the research vessel was then used to scale the photograph (similar to Whitehead & Payne 1978). Once the scale photograph had been obtained, the UAV returned to the vicinity of the research boat and landed safely in the water before being recovered.

110

Chapter 6: Energetic cost of reproduction in whales

Figure 6.2: (A) An example of desired aerial photograph of humpback whales captured by a unmanned aerial vehicle that was used in analyses. The whale is lying flat at the surface, dorsal side facing up, with a straight body axis and peduncle (nonarching). (B) Position of measurement sites of humpback whales recorded in this study. For clarity, only width (W) measurement sites located at 10% increments along the body axis are shown.

Morphometric measurements and scaling Photogrammetric methods were used to extract several morphometric measurements from the vertical close-up photographs of the whales (Figure 6.2; Best & Rüther 1992, Perryman & Lynn 2002, Cosens & Blouw 2003, Miller et al. 2012) using a custom-written script in R (R Core Team 2014; Appendix D). Length measurements (in pixels) included distance from the tip of the rostrum to: the notch of the tail fluke; the end of the dorsal fin; the position of the eyes (measured along the body axis of the whale); the beginning of the tail fluke (Figure 6.2). The width of the whale (in pixels), measured perpendicular to the body axis, was measured at 5% intervals along the entire body of the whale, not including 0 and 100% of the body length (19 measurements in total), and also between the eyes. The relative

111

Chapter 6: Energetic cost of reproduction in whales measurements of the whale (in pixels) were converted to absolute measurements (in cm), using the scale photographs obtained for each animal.

Identification and classification of whales Individual humpback whales were identified from the shape of their dorsal fin (Katona & Whitehead 1981), which were photographed from the research boat using a Nikon D300 camera with a 400-mm lens (Nikon Inc., Tokyo, Japan). Individual whales were classified into one of four reproductive classes: calves, immature, mature, and lactating whales. Calves and lactating females were classified based on their close and consistent association with each other. Immature and mature whales were separated based on their length, as sexual maturity in baleen whales can be determined by length rather than age (Sigurjónsson et al. 1990). From histological examination of testes from humpback whales from breeding stock D, Chittleborough (1955a) reported that the length of males at puberty ranged from 10.15 to 12.44 m with a mean length of 11.20 m. From examinations of females captured during their first estrous cycle, Chittleborough (1955b) demonstrated that females at puberty ranged in size from 10.73 to 13.26 m, with a mean length of 11.73 m. Therefore, based on these estimates, a threshold length value of 11.2 m was used to separate immature whales from mature animals. In regard to sex determination, apart from lactating females the sex of individual whales was unknown.

Body site-specific changes in width Intraseasonal changes in the body condition of baleen whales are not exhibited homogenously across the body of the animal, and the pattern of variability appears to be species specific (Vikingsson 1990, Folkow & Blix 1992, Naess et al. 1998, Miller et al. 2012, Christiansen et al. 2013). To assess which width measurements best capture intraseasonal changes in the body condition of humpback whales, I developed linear models (LMs) in R to test the effect of day (measured as day of year) and length on each width measurement. Both linear and polynomial nonlinear models were developed to test the relationship between the dependent and independent variables. Comparison of model fit was made using Akaike’s information criterion (AIC). Separate models were developed for each reproductive

112

Chapter 6: Energetic cost of reproduction in whales class to explore whether variation in body condition varied between reproductive classes.

Body condition index With the aim to develop a body condition index (BCI) that would capture the variation in width across the body of the humpback whales, I calculated the flat dorsal projected surface area for each individual (the area as seen from above) from the body morphometric data. The flat projected surface area of the body of the whales was modeled as a series of trapezoids connected to each other at each width 2 measurement site. The projected surface area (m ) for each trapezoid segment, As, was calculated:

ℎ 퐴 = × (푎 + 푏) 푠 2

Where a is the width (m) of the base (i.e., the anterior width measurement) and b is the width (m) of the top (i.e., the posterior width measurement) and h is the distance (i.e., length, m) between the two width measurements along the body axis of the animal. Because the head, fins, and tail fluke of cetaceans contain relatively little energy reserves (Brodie 1975, Koopman 1998), only segments (s) between the position of the eyes down to 80 % of the body length from the rostrum were included in the BCI (i.e., projected surface area) (m2) estimate:

푆

퐵퐶퐼 = ∑ 퐴푠 푆=1

Where S is the total number of segments (s) between the position of the eyes and 80 % of the body length from the rostrum. This quantitative index of body condition accounted for variation in width across the body of the whales and how it changed throughout the breeding season. A generalized additive model (GAM) with a thin plate regression spline smoother was used to investigate whether the size of the head in relation to total body length changed with respect to the length of the whale.

113

Chapter 6: Energetic cost of reproduction in whales

Intraseasonal changes in body condition Linear models were developed to investigate intraseasonal changes in BCI of humpback whales for the four reproductive classes. In addition to day (of year and reproductive class, length was included as a covariate to account for difference in the structural size between individuals, which will influence BCI (i.e., projected surface area). In the model selection process, covariates and interactions between covariates were added sequentially to the null model and different polynomial nonlinear factors were used to test the relationship between the dependent and independent variables. As above, model selection was made using AIC.

Effects of female body condition on calf condition and growth From the best fitting model of BCI, the relative body condition of calves (CBC) was estimated as (Christiansen et al. 2014):

퐵퐶퐼퐶.푂푏푠,푖 − 퐵퐶퐼퐶.퐸푥푝,푖 휀퐶.푖 퐶퐵퐶푖 = = 퐵퐶퐼퐶.퐸푥푝,푖 퐵퐶퐼퐶.푒푥푝,푖

Where 퐵퐶퐼퐶.푂푏푠,푖 is the observed BCI (i.e., projected surface area) of calf i 3 (m ) and 퐵퐶퐼퐶.퐸푥푝,푖 is the expected (or predicted) BCI of a calf of the same length and at the same day in the season. A positive value of CBC suggests that the calf was considered to be in a relatively better condition than an average calf, while a negative CBC value suggests that it was considered to be in a relatively poorer condition. The CBCs were compared to the relative body condition of their mothers (female body condition, FBC; Christiansen et al. 2014):

퐵퐶퐼퐹.푂푏푠,푖 − 퐵퐶퐼퐹.퐸푥푝,푖 휀퐹.푖 퐹퐵퐶푖 = = 퐵퐶퐼퐹.퐸푥푝,푖 퐵퐶퐼퐹.푒푥푝,푖

Where 퐵퐶퐼퐹.푂푏푠,푖 and 퐵퐶퐼퐹.퐸푥푝,푖 are the observed BCI and expected BCI (i.e., projected surface area) of the mother of calf i (m3) at the same day in the season, respectively. Again, a positive value of FBC means that the mother was considered to be in a relatively better condition than the average mother, while a negative FBC value suggests that the mother was considered to be in a relatively poorer condition. LMs were then developed to investigate the effect of FBC on CBC (similar to Christiansen et al. 2014).

114

Chapter 6: Energetic cost of reproduction in whales

The effect of FBC on calf growth (i.e., length) was investigated using LMs. To account for the growth of calves through the breeding season, day of year was included as a covariate in the model. The effect of maternal size (i.e., length) was also assessed, as maternal length is often positively correlated to offspring size in mammals, including whales (Kovacs & Lavigne 1986, Best & Rüther 1992, Boltnev & York 2001, Perryman & Lynn 2002). Again, both linear and polynomial nonlinear relationships between calf length and maternal length were investigated and the best fitting model was selected using AIC.

Model validation For all models, model validation tests were run to identify potential violations of the assumptions of the models. Homogeneity of variances was investigated from scatter plots of residuals vs. fitted values and residuals against each explanatory variable in the model. Normality of residuals was interpreted from quantile–quantile plots and from residual histograms. Influential points and outliers were identified using leverage and Cook’s distance. All model assumptions were fulfilled.

Sensitivity analysis When measuring body morphometrics from aerial photographs, a number of potential measurement measurement errors related to picture quality, measurement precision, and image distortion need to be investigated and, if necessary, accounted for.

To obtain accurate measurements of the body condition of the whales, the contour of the animal’s body in the water needs to be clearly visible in the photographs. Waves, water spray, and turbidity can distort the body contour in a photograph, resulting in a reduction in the accuracy of body morphometric measurements. Measurement precision within photographs was assessed by having three independent researchers measure the body morphometrics of the same whale from the same photograph. From these measurements, the coefficient of variation (COV) in “relative” body condition (the projected surface area in pixels) of each animal was calculated. To obtain a measure of accuracy of the “absolute” body condition (the projected surface area in m2), measurement errors relating to the scaling of the photographs (the conversion from pixels to m2) were also quantified. 115

Chapter 6: Energetic cost of reproduction in whales

Again, three independent researchers scaled the same photograph for each whale, and the COV in “absolute” body condition was calculated. To minimize the risk of measurement errors affecting the body condition analyses, an upper threshold value of 0.05 was chosen for both COV values, with animals above this threshold being excluded from the main analyses.

To further investigate the effect of measurement errors, a sensitivity analysis was carried out to quantify the effect of measurement errors on the day parameter values (the intraseasonal effect) of the best fitting model. One thousand bootstrap iterations were run. For each of the iterations and for each individual, a random body condition value was drawn from a Gaussian distribution with the mean equal to the body condition of the individual and the standard deviation being estimated from the COV in “absolute” body condition (the projected surface area in m2). The resulting density distributions of the day parameters and their associated standard errors (SEs) were visually examined to investigate the robustness of the model results to within-photographs measurement errors.

Depending on how the whale was positioned in the water (the degree of arching or curving of the body), and the parallax, BCI (i.e., projected surface area) can be either positively or negatively biased. Similar to Perryman & Lynn (2002), measurement precision (differences between repeated measurements of the same individual from different photographs) was assessed by calculating the COV for five whales that had each been photographed at least three times during the same encounter. From each photograph, the body condition of the whale was estimated, resulting in three independent body condition measurements of the same whale from different photographs. Subsequently, the COV in body condition for each of the five whales was then calculated. Based on this estimate, a sensitivity analysis was carried out to test the effect of between-photographs measurement errors on the day parameter estimates of the best fitting model. One thousand bootstrap iterations were run. Again, for each model iteration and for each individual, a random body condition value was drawn from a Gaussian distribution with the mean being equal to the body condition of the individual and the standard deviation being calculated from the COV in body condition resulting from the between-photographs

116

Chapter 6: Energetic cost of reproduction in whales measurement errors. The effect on the day parameter values of the best fitting model was visually examined from density distribution plots.

Potential measurement errors resulting from the positioning of the whale in the photograph (i.e., the distance from the midpoint of the photograph frame) was investigated by plotting the residuals of the best fitting model against its spatial coordinates (X and Y pixels) in the photograph.

Results

A total of 200 humpback whales were photo-graphed vertically by the UAV in Exmouth Gulf between 3 August and 16 September 2015 (Figure 6.1). Over this 44-d period, I spent 26 d on the water, equaling 186 h of research effort (Figure 6.1). After initial filtering of photographs based on body position (removing laterally curved, arched, and animals rolling on their side) and picture quality, 134 animals remained for analyses’ purposes. This included 32 calves (body length [m]: mean = 5.64, SD = 0.81, min = 4.14, max = 7.76), 51 immature (body length [m]: mean = 9.74, SD = 0.85, min = 7.96, max = 10.95), 20 mature (body length [m]: mean = 11.86, SD = 0.81, min = 11.20, max = 14.42), and 31 lactating females (body length [m]: mean = 12.20, SD = 0.98, min = 10.77, max = 14.73). Photograph identification records indicated that no individual was measured more than once during the study period. Hence, the body condition data represent a cross-sectional sample of the population and not repeated measurements of the same individuals. I did not observe any visual behavioral responses of the whales toward the UAV.

117

Chapter 6: Energetic cost of reproduction in whales

Figure 6.3: Rate of body width change at different measurement sites (Figure 6.2) for calves (n = 32), immature (n = 51), mature (n = 20), and lactating (n = 31) humpback whales. Error bars represent 95% confidence intervals. The dashed lines represent the level where width remains constant through the breeding season.

Body site-specific changes in width The change in body width of humpback whales through the study period varied between measurement sites and also between reproductive classes. While calves and immature whales showed no intraseasonal change in body width, mature and lactating whales showed a decrease in width at a number of measurement sites (Figure 6.3). For mature whales, the decrease was highest around 50–65 % of the body length from the rostrum, while lactating females showed a decrease in width over a larger portion of their body, between 35 % and 80 % of the body length from the rostrum (Figure 6.3). The proportional size of the head of humpback whales increased nonlinearly with the length of the animals (F5.51,127.49 = 37.75, P < 0.001, R2 = 66.8 %, n = 134), from 20 % as calves (4–6 m body length) to 25 % as late immatures/ matures (>10 m body length). As expected, the head of the whales showed no intraseasonal variation in width for any of the reproductive classes (Figure 6.3). Similarly, the lower section of the peduncle (>80 %) showed no significant change in width over the season.

118

Chapter 6: Energetic cost of reproduction in whales

Intraseasonal changes in body condition Whale length, reproductive class, and the interaction between reproductive class and day explained 98.7 % (R2) of the variance in BCI (Model 9 in Table 6.1). Most of this variance was explained by length (98.1 %), while the remaining 0.6 % was explained by reproductive class (52.4 %) and the interaction term (47.6 %). The effect of length on BCI was best explained by a quadratic polynomial relationship

(F2,124 = 4712.1, P < 0.001). The overall body condition of humpback whales varied between reproductive classes (F3,124 = 10.6, P < 0.001), with calves having the smallest BCI (mean = 3.05 m2, SD = 0.97, min = 1.58, max = 5.77), followed by immature (mean = 7.98 m2, SD = 1.36, min = 5.62, max = 10.31), mature (mean = 12.03 m2, SD = 1.69, min = 9.90, max = 17.25), and finally lactating whales (mean = 12.64 m2, SD = 1.76, min = 9.62, max = 16.74). The effect of day (intraseasonal effect) on BCI varied between reproductive classes (F4,124 = 7.2, P < 0.001) (Figure 6.4). Similar to the site-specific changes in width, only mature and lactating whales showed a change in their BCI over the study period, while calves and immature whales showed no intraseasonal variation in BCI. Mature whales showed a decrease in BCI at a rate of 0.027 m2 d-1 (SE = 0.0083), while lactating females showed a decrease in BCI at a rate of 0.032 m2 d-1 (SE = 0.0076) (Figure 6.4). The rate of change in BCI varied significantly between mature whales and lactating females (t = −4.2, P < 0.001).

119

Chapter 6: Energetic cost of reproduction in whales

Table 6.1: Linear model selection results based on minimization of Akaike’s information criterion (AIC) for humpback whale body condition index

df df

Model Variables F (among) (within) P R2 AIC Δ A I C 1 BCI ~ Length 3621.6 1 132 < 0.0001 0.96 300.9 1 1 2 BCI ~ (Length + Length2) 3345.2 2 131 < 0.0001 0.98 221.8 3 89 3 BCI ~ Day 2.2 1 132 0.1431 0.02 747.3 5.. 6 2 33 4 BCI ~ (Day + Day ) 1.1 2 131 0.3375 0.02 749.2 45 .6 5 BCI ~ Reproductive class 281.8 3 130 < 0.0001 0.87 483.4 763 0 2 . 6 BCI ~ (Length + Length ) + 1584.1 5 128 < 0.0001 0.98 202.5 702 0 Reproductive class . 9. 0 7 BCI ~ (Length + Length2) × 751.3 11 122 < 0.0001 0.99 202.6 2 0 Reproductive class . 8 BCI ~ (Length + Length2) + 1492.0 6 127 < 0.0001 0.99 187.3 14 . Reproductive class + Day 8 9 BCI ~ (Length + Length2) + 1053.9 9 124 < 0.0001 0.99 182.5 0 Reproductive class × Day† . 0 10 BC I ~ (Length + Length2) × 416.1 23 110 < 0.0001 0.99 193.5 1 1 Reproductive class × Day . 0 Notes: Variable abbreviations: body condition index (BCI; relating to projected surface area), Day of year, calf, immature, mature, or lactating female (reproductive class). † the most parsimonious model (Model 9).

Effects of female body condition on calf condition and growth

There was a positive linear relationship between FBC and CBC (F1,22 = 4.5, P = 0.044), with calves increasing in body condition at a rate of 0.594 m2 (SE = 0.2786) per m2 increase in FBC (Figure 6.5). FBC explained 17.1% (R2) of the variance in

CBC. The lengths of the calves were affected by maternal length (F1,21 = 22.0, P <

0.001) and day (F1,21 = 11.4, P = 0.003) (Model 4 in Table 6.2). The model explained 61.4 % of the variance in calf body length, with maternal length explaining 40.4 % and day 21.0 %. Calves showed an increase in length at a rate of 0.033 m d-1 (SE = 0.0097) through the study period and it was positively related to the size of their mothers, with calves increasing in length at a rate of 0.456 m (SE = 0.1005) per m

120

Chapter 6: Energetic cost of reproduction in whales increase in maternal length (Figure 6.6). FBC had no effect on calf length (Models 3, 6, 7, 8, 9 in Table 6.2).

Figure 6.4: Partial effect plot of body condition index (i.e., projected surface area) as a function of day of year for calves (n = 32), immature (n = 51), mature (n = 20), and lactating (n = 31) humpback whales. The dashed lines represent 95% confidence intervals. Length has been fixed at the mean length for each reproductive class (calf = 5.64 m, immature = 9.74 m, mature = 11.86 m, lactating = 12.20 m). Rug plots show the distribution of the data points.

Figure 6.5: Calf body condition (CBC) as a function of female body condition (FBC) for humpback whales (CBC = 0.594 × FBC). The dashed lines represent 95% confidence intervals. The dotted vertical line crossing the x-axis at zero represents the average FBC. n = 24.

121

Chapter 6: Energetic cost of reproduction in whales

Figure 6.6: Partial effect plots of calf length as a function of (A) maternal length and (B) day. The dashed lines represent 95% confidence intervals. In the left subfigure (A), day has been fixed at 240 (August 28), and in the right subfigure (B), length has been fixed at the mean length of lactating females (12.20 m). n = 24.

Table 6.2. Linear model selection results based on minimization of Akaike’s information criterion (AIC) for humpback whale calf length

df df

Model Variables F (among) (within) P R2 AIC ΔAIC

1 CL ~ Day 6.7 1 22 0.0165 0.23 51.2 14.4

2 CL ~ Maternal length 14.9 1 22 0.0008 0.40 45.2 8.4

3 CL ~ FBC 0.0 1 22 0.9361 0.00 57.6 20.8

4 CL ~ Maternal length + Day† 16.7 2 21 < 0.0001 0.61 36.8 0.0

5 CL ~ Maternal length × Day 11.1 3 20 0.0002 0.62 38.0 1.3

6 CL ~ Maternal length + Day + FBC 10.6 3 20 0.0002 0.61 38.7 2.0

7 CL ~ Maternal length × Day + FBC 7.9 4 19 0.0006 0.63 40.0 3.3

8 CL ~ Maternal length + Day × FBC 8.1 4 19 0.0006 0.63 39.7 3.0

9 CL ~ Maternal length × Day × FBC 5.9 7 16 0.0016 0.72 38.9 2.2

Notes: Variable abbreviations: calf length (CL), Day of year (day), female body condition (FBC). † The most parsimonious model (Model 4)

Sensitivity analysis The day parameter values estimated from the bootstrapping procedure were relatively narrow in their distribution both within (Figure 6.7) and between (Figure 6.8) photographs. The residuals of the best fitting model (Model 9 in Table 6.1)

122

Chapter 6: Energetic cost of reproduction in whales showed no signs of spatial autocorrelation (spatial dependence) between data points. Finally, altering the body length threshold (to separate immature whales from mature whales) between 10.2 and 12.2 m (1 m below and above the chosen threshold value of 11.2 m, respectively) did not significantly affect the day parameter estimates.

Figure 6.7:Sensitivity analysis of within-photographs measurement errors showing the density distribution of the day parameter values (slope parameter) and their associated standard errors (SEs) for the best fitting model (Model 9 in Table 6.1) based on 1000 bootstrapping iterations. For each iteration and individual, a random body condition value was drawn from a distribution of values with the mean equivalent to the mean body condition of the individual and the standard deviation resulting from three independent body condition measurements (from the same photograph) for each whale. n = 134

123

Chapter 6: Energetic cost of reproduction in whales

Figure 6.8: Sensitivity analysis of between-photographs measurement errors showing the density distribution of the day parameter values (slope parameter) and their associated standard errors (SEs) for the best fitting model (Model 9 in Table 6.1) based on 1000 bootstrapping iterations. For each iteration and individual, a random body condition value was drawn from a distribution of values with the mean equivalent to the mean body condition of the individual and the standard deviation resulting from three independent body condition measurements (from three different photographs) from five whales. n = 5 Discussion

Intraseasonal changes in body condition The aim of this study was to investigate intraseasonal variation in the body condition of different reproductive classes of humpback whales. Because I were unable to make repeated measurements of the same individuals, this study represents a cross-sectional sample of the body condition of the population. I documented that the body condition of mature and lactating whales decreased significantly through the breeding season, while there was no change in the body condition of immature whales and calves. Similar differences in energy storage between reproductive classes have been documented in minke whales, fin whales, and right whales and are likely to reflect differences in energetic costs during the breeding season (Lockyer 1986, 1987b, Vikingsson 1990, 1995, Miller et al. 2011, Christiansen et al. 2013). Mature and lactating whales carry the costs of reproduction (i.e., lactation and finding a mate) and thus, their body condition

124

Chapter 6: Energetic cost of reproduction in whales reduced significantly (Lockyer 1987a, b, Vikingsson 1990, 1995, Miller et al. 2012, Christiansen et al. 2013). In contrast, immature whales do not carry the added energetic costs of reproduction; however, they still rely on stored energy to support the costs of migration, daily field metabolic rates, growth, and body maintenance. As suggested by Miller et al. (2011), intra-seasonal variation in the body condition of immature whales may not be detectable as they might rely on stored lipids in their blubber and other tissues to support the energetic costs during the breeding season. A decrease in lipid concentration would not necessarily be visible in the body shape (i.e., width) of the whales (Christiansen et al. 2013) and could therefore be a limitation when using photogrammetry methods to assess body condition.

The rate of decline in body condition was greater for lactating females than for other mature whales. Lactation is considered the energetically most costly part of the reproductive cycle in mammals (Gittleman & Thompson 1988), including large whales (Lockyer 1981), and to cover these high costs, pregnant females build up larger energy stores than any other reproductive class during the summer feeding season (Christiansen et al. 2013). As a consequence, pregnant whales early in the breeding season generally have the greatest body condition, while lactating females late in the breeding season have the poorest body condition (Lockyer 1987a, Aguilar & Borrell 1990, Perryman & Lynn 2002, Miller et al. 2011). The high rate of decline in the body condition of lactating females in this study suggests that this is also the case for humpback whales.

That mature whales decreased significantly in BCI through the breeding season suggest that the energetic costs of reproduction were relatively high for this reproductive class. Catch data from Western Australia suggest that the majority of migrating mature (nonlactating) whales are males (Chittleborough 1965). Consequently, the cost of reproduction for male humpback whales is relatively high. High reproductive costs for mature males have been documented in mammalian species exhibiting female defense or scramble competition mating systems (Forsyth et al. 2005, Lane et al. 2010). On breeding grounds, male humpback whales actively compete for females where they physically fight and display aggressive behaviors toward each other in order to gain access to receptive females (Baker & Herman

125

Chapter 6: Energetic cost of reproduction in whales

1984). Such behaviors are likely to be energetically costly and could explain the observed difference in intraseasonal trends in body condition between mature and immature whales.

Some of the observed differences in absolute body condition (the intercepts in Figure 6.4) between reproductive classes are likely due to differences in the timing of migration. During the northern migration from Antarctica to the Kimberley region, immature whales and lactating females with calves born the previous year are the first to arrive, followed by anestrous females, mature males, and finally pregnant females (Chittleborough 1965). During the southern migration, the order is more or less the same, with immature whales arriving first, followed by mature whales and finally lactating females whom have recently given birth. At the time whales enter Exmouth Gulf, immature and mature whales should therefore have a relatively lower body condition compared with lactating females. This difference in timing of migration must be accounted for when calculating the absolute costs of migration and reproduction for different reproductive classes of whales (Christiansen et al. 2013). Measuring the body condition of humpback whales at different locations along their migratory route would help to further distinguish intraseasonal changes in body condition from potential differences caused by variation in the timing of migration between individuals and reproductive classes. The decline in body condition for mature and lactating humpback whales was highest around the mid- and caudal regions of the body. In balaenopterids, the posterior region of the body plays an important role in energy storage (Lockyer 1987b, Naess et al. 1998, Christiansen et al. 2013). Naess et al. (1998) showed that during the feeding season the blubber thickness and lipid content of minke whales increased the most at the caudal region of the body, just behind the dorsal fin (the posterior end of the dorsal fin of humpback whales in this study was located at ~70% of the body length from the rostrum). Christiansen et al. (2013) further confirmed that blubber deposition for both mature and pregnant minke whales was highest toward the posterior end of the body. Similarly in fin and sei whales (Balaenoptera borealis), (Lockyer 1987b) reported that the caudal region posterior to the dorsal fin serves as the main area of lipid storage in both the blubber and muscle. While our findings demonstrate a significant decrease in body width in the 126

Chapter 6: Energetic cost of reproduction in whales caudal region for mature and lactating whales, the lower tail region of the whales (>80 % of body length from the rostrum) showed no pronounced variation in width, which suggests that this region might play a more structural role in humpback whales, by providing aid during locomotion and streamlining the caudal body (Koopman et al. 1996, 2002, Struntz et al. 2004). Lactating females showed a decrease in width along a larger portion of their body (35–80 %) compared with mature whales (50– 65 %). This is similar to lactating southern right whales, which displayed the highest decrease in width between 40 % and 80 % of the body length from the rostrum (Miller et al. 2012). Similarly, Perryman & Lynn (2002) found that the widest part of the body of pregnant and early lactating gray whales was located further back than that of other reproductive classes. I found that by modeling BCI as the projected surface area of humpback whales, this across body variation in width could be captured by a single metric of body condition. This, in turn, made it relatively easy to test the effect of different covariates on body condition, using standard statistical methods in ecology. I therefore highly recommend this single metric approach when studying the body condition of baleen whales.

Effects of female body condition on calf condition and growth Newly born calves on the breeding grounds need to grow in size and build up a sufficiently thick blubber layer to survive the migration back to the feeding grounds in cold polar waters (Corkeron & Connor 1999). I documented a significant positive relationship between FBC and the condition of their calves. The iteroparous nature of baleen whales and the high costs of lactation (Lockyer 1981) suggests that female humpback whales with insufficient energy reserves should reduce their energetic investment into their offspring, by producing smaller (i.e., shorter) or poorer conditioned calves, to maintain their survival (Peacock 1991, Lockyer 2007, Pontier et al. 2012, Christiansen et al. 2014). Our results support this hypothesis and are further strengthened by the findings of Christiansen et al. (2014) who found that pregnant minke whales in poorer body condition reduced their energetic investment in their fetus proportionately to their own body condition. Thus, it would appear that female baleen whales throughout both gestation and lactation will prioritize their own body condition and survival, above that of their offspring, which is consistent with a K-strategist life history (MacArthur & Wilson 1967). 127

Chapter 6: Energetic cost of reproduction in whales

While the length of humpback whale calves increased throughout the breeding season, as expected, their BCI showed no intraseasonal variation. This was surprising, given that calves are born with relatively low fat reserves and need to accumulate these as quickly as possible to reduce heat loss. The BCI of calves increased with body length, meaning that the surface to volume area and therefore heat loss should decrease as the calf grows bigger. The best strategy for calves to reduce heat loss might therefore be to invest their excess energy into growth (i.e., length) rather than fat reserves, so that they can become larger overall and reduce their surface to volume area. While I found a positive relationship between maternal length and the length of the calves, which has also been found in other mammals (Skogland 1984, Kovacs & Lavigne 1986, Boltnev & York 2001, Lockyer 2007), including baleen whales (Best & Rüther 1992, Perryman & Lynn 2002), I found no significant relationship between FBC and calf length. Instead, females in better condition produced calves in better condition. An explanation for this could be that the growth rate of calves is always kept at its physiological maximum, irrespective of FBC (within the range of values observed in this study), to prepare the calf for the migration back to the cooler feeding grounds. While exceptionally good conditioned females can also afford to fatten their calves, to provide extra insulation and an energetic buffer, poorer conditioned females will prioritize the growth of their calves in size (i.e., length) at the expense of calf condition, to yield a higher survival probability overall.

Using unmanned aerial vehicles to assess body condition in baleen whales This study is the first to apply UAV technology and photogrammetry methods to assess body condition in a balaenopterid. Compared with conventional aircrafts, UAVs are less expensive and safer and can also be operated in more remote regions. Here, I demonstrate that even relatively inexpensive (<$1,000 USD) UAVs can be used to successfully measure body condition in baleen whales over relatively short time periods (i.e., 44 d). I show that even relatively small changes in condition of humpback whales can be reliably detected in measurements from vertical aerial photographs. I further show how measurement errors can be quantified and incorporated into analyses of body condition. Our sensitivity analysis demonstrated that our findings were robust to measurement errors both within and between 128

Chapter 6: Energetic cost of reproduction in whales photographs of the same whale and that there was no measurement bias associated with where in the photograph the whale was positioned. Apart from humpback whales, photogrammetry has so far been used to successfully measure the body condition of right whales (Miller et al. 2012) and gray whales (Perryman & Lynn 2002). I strongly encourage this approach to be extended further to other baleen whales, in an overall attempt to improve our understanding of large whale bioenergetics and reproductive biology.

Management implications This study demonstrates how photogrammetry can be used to assess the body condition of humpback whales from aerial photographs recorded using UAV technology. This noninvasive approach provides a valuable tool to monitor the health of baleen whale populations globally. With most humpback whale populations recovering at an impressive rate, this approach can be used to measure density-dependent effects on body condition and reproduction (Fowler 1990, Stewart et al. 2005). Prey availability is likely to be a key determinant of body condition in baleen whales, and interannual variation in prey availability can be linked to changes in body condition and reproduction in baleen whales (Lockyer 1986, Ichii et al. 1998). Finally, developing a global health index for baleen whale populations will allow for comparison with other populations, which will provide a more holistic understanding of the status of baleen whale species and to aid in conservation.

129

Chapter 7: General discussion

Chapter 7 : General discussion

Summary of the main findings

The five chapters of this thesis illustrate how Argos-linked Fastloc-GPS and UAVs are valuable tools to obtain new insights on the movement of elusive marine species. I have presented detailed examples for three long-distance migratory species (green and loggerhead turtles, humpback whales) and for a range of behaviours (e.g. foraging, migration, and breeding) (see Figure 7.1 for overview). I found that Argos-linked Fastloc-GPS provided a large amount of data on sea turtle movement (e.g. departure, arrival and diel pattern during migration, but also small-scale foraging movement patterns) that cannot be easily obtained using other tracking methods (e.g. Argos satellite tracking). As a result, I was able to evaluate the movement patterns of sea turtles (e.g. migration and foraging strategies) at a similar detail as that achieved in the past for avian and terrestrial species. I also showed that UAVs are useful to obtain information that cannot be easily obtained using tracking data (e.g. animal size, or changes in body condition) and when the animal is too large to be captured (e.g. whales). Overall, information collected using Argos- linked Fastloc-GPS and UAVs provided novel insights on the evolution and energy management strategies of capital breeders during their life cycle.

130

Chapter 7: General discussion

Figure 7.1: Schematics of (a) the migratory cycle of green and loggerhead sea turtles updated from Chapter 1 (Introduction) by adding stopover sites to the cycle, (b) the migratory cycle of humpback whales. The main findings of the thesis are reported in (c) with the numbers in parentheses referring to those presented in (a) and (b).

In chapter 6, I illustrated how UAVs could be used to determine important life- history parameters for long-distance migratory species with high accuracy. I measured the size and body condition of humpback whales at the breeding ground, to evaluate the interaction between the body condition of lactating females and their calves. Both body size and body condition can be used to quantify the cost of 131

Chapter 7: General discussion reproduction for species relying on fat stores (e.g. polar bears, Atkinson and Ramsay 1995; northern elephant seals, Crocker et al. 2001). Such information cannot be obtained from tracking datasets on their own (e.g. sea turtle tracking dataset). Nonetheless, this study was limited to a measurement of the body condition for a cross-sectional sample of the population. The results could be refined by obtaining multiple measurements for the same individuals over a breeding season to calculate more accurate energy budgets. As UAVs technology improves and with improvement of their autonomy such repeated measures become easier (Crutsinger et al. 2016). In addition, UAVs could also be deployed when the whales or sea turtles are at their foraging grounds, in parallel to Argos-Linked Fastloc-GPS tracking devices. In the case of whales, UAVs could be used to investigate how body condition changes by using photogrammetry, while short term Fastloc-GPS tracking data could be used to investigate food patch patch use when in Antarctica (Dalla Rosa et al. 2008). For sea turtles, UAVs and Fastloc-GPS could be used conjointly to investigate the relationship between foraging behaviour and predator avoidance (extending existing work on species like green turtles and dugongs, e.g. Heithaus et al. 2007; Wirsing et al. 2007). More specifically, UAVs could provide data on predator abundance and distribution that are lacking in Chapter 4 and 5 (e.g. on foraging sites located nearshore) while Fastloc-GPS would allow to resolve sea turtles small scale movement patterns. Additionally, departure from breeding grounds and arrival at foraging grounds could be determined using Argos-Linked Fastloc-GPS for both species.

My work clearly showed that Fastloc-GPS tracking and UAVs are valuable additions to the scientist’s toolbox, providing the opportunity to obtain novel information, such as the pattern of departure/arrival at breeding/foraging sites by migratory sea turtles as well as fine-scale patch use during foraging, in addition to obtaining accurate measurements of the body size and condition of whales. Considering the potential of both technologies, I anticipate that the scientific community will increasingly use them to explore aspects of the life-history of elusive species that were not previously possible.

132

Chapter 7: General discussion

Behaviour of marine vertebrates during migration

Departure/arrival of sea turtles

Using high-resolution Fastloc-GPS locations, I found that loggerhead and green sea turtles from two oceanic basins (Mediterranean and Pacific) consistently arrived and departed to and from stopover/foraging sites during the daytime (Chapter 3). In addition, turtles that were within 12 hours travel distance from their foraging ground, slowed or stopped moving the night before arrival, whereas those that were further away maintained their normal travel speed. This phenomenon suggests that turtles were responding to visual cues, and might adjust their travel speed to avoid overshooting the target site. This suggestion is consistent with previous experiments investigating the use of visual cues by sea turtles. On nesting beaches, blindfolded turtles cannot locate the sea, except by chance (Ehrenfeld & Carr 1967). The results of arena tests on hatchling turtles underline the importance of the horizon in sea- finding orientation. When the horizon is obstructed, sea turtles are less efficient at finding the sea (Ehrenfeld & Carr 1967). In my study, adult sea turtles departed in mornings, which is time of day that the azimuth of the rising sun indicates east all year-round reasonably reliably, and is consistent with laboratory findings (Guilford & Taylor 2014). In contrast, some species that orient themselves during migration use solar cues to calibrate their sun compass at sunset (e.g. warblers, Phillips and Moore 1992; homing pigeon, Phillips and Waldvogel 1988). In a clock-shifting experiment, sea turtles exposed to a laboratory photocycle advanced by seven hours modified their orientation westward in a way that was consistent with the use of solar cues (Mott & Salmon 2011). In addition, Avens and Lohmann (2003) found that juvenile turtles deprived of visual cues that are placed in a distorted magnetic field, have significantly different orientations to turtles that have access to one or both cues. Thus, both visual cues and magnetic fields are probably needed for successful orientation.

133

Chapter 7: General discussion

Resting patterns during migration

I found that both loggerhead and green sea turtles from the two ocean basins swam continuously during the oceanic crossing, with faster travel speed during the day compared to at night. My results are consistent with the findings of previously published Argos-tracking studies (e.g. Luschi et al. 1998; Jonsen et al. 2006). Enstipp et al. (2016) found that a displaced green turtle equipped with accelerometers and a time-depth recorder could swim continuously for ~3.5 days to migrate back to its breeding ground which is consistent with the patterns observed over longer distances in my study. Therefore the day/night difference in travel speed could be due to a reduction of the forward motion during deep nocturnal dives (Minamikawa et al. 1997; Hays et al. 2001a; Rice and Balazs 2008). Travelling continuously might minimise the energetic cost of migration if an animal travels at speeds close to the speed of the optimal cost of transport (Åkesson & Hedenström 2007, Hein et al. 2012). However, Fastloc-GPS, accelerometer or time-depth recorder data could not reveal whether sea turtles exhibit any en-route sleeping or resting strategies, as has been documented for marine mammals and birds. For example, the study of electroencephalograms (EEG) of long-distance migratory cetaceans shows that they perform unihemispheric sleeps, with one hemisphere of the brain sleeping, while the other hemisphere remains active (Lyamin et al. 2008). Furthermore, frigate birds travel continuously for up to 10 days, with their EEGs showing that they rest for less than 40 minutes a day (Rattenborg et al. 2016). In Africa, common swifts remain airborne throughout their entire migration and for most of their non-breeding period (>99 %). Basically, they rest while airborne by passively gliding for periods of <40 seconds (as revealed by accelerometry, Hedenström et al. 2016). Sea turtle might use similar strategies, possibly resting one hemisphere of the brain at the time; however, the sleep patterns of reptiles remain poorly understood. Terrestrial mammals typically exhibit rapid eye movement sleep (REM) characterized by a pattern that closely resembles that of waking in most brain regions and non-REM sleep characterized by greatly reduced activity in brainstem systems (Siegel 2008). Marine mammals, like cetaceans, only exhibit unihemispheric slow waves, and their brain activity is indistinguishable from that of quiet waking (Siegel 2008). A variety of resting strategies have been documented 134

Chapter 7: General discussion in reptiles. Periods of inactivity have been documented for all chelonians (which includes sea turtles), and described as a state of prolonged immobility, different from basking, with the eyes closed, the plastron resting on the ground, and the head fully relaxed (Houghton et al. 2008, Libourel & Herrel 2015). Houghton et al. (2008) found evidence of sleeping during long benthic dives using an inter- mandibular angle sensor measuring the gulping movement of a hawksbill turtle. Libourel and Herrel (2015) reported that, out of 10 studies of terrestrial and marine turtles EEGs, five exhibited no change in EEGs between states of wakefulness and resting, including one conducted on the loggerhead turtle, and five reported a diminution in the EEG amplitude and/or frequency. One recent study has reported REM sleep for the Australian dragon Pogona vitticeps (Shein-Idelson et al. 2016). Clearly more studies are needed on the sleeping patterns of sea turtles to address how they rest during long-distance migration over extended periods of time.

Stopover pattern of sea turtles

I found that some sea turtles made stopovers shortly after leaving the breeding grounds or just before arriving at the foraging grounds. The stopover duration was relatively short (<6 days), suggesting the turtles did not spend extensive periods of time foraging for refuelling. Only one stopover was recorded at the beginning of the migration; thus further investigation is required to determine whether if this event is representative of sea turtles behaviour or not. Overall the role of these stopovers remains unclear. Terrestrial mammals, insects and avian species use stopover sites to rest and refuel during migration (Åkesson & Hedenström 2007, Sawyer & Kauffman 2011, McCord & Davis 2012, McGuire et al. 2012). The morphology of birds constrains the amount (i.e. the weight) of fuel-load they can carry; therefore, stopovers are essential for many migrating bird species, which alternate between periods of fast-travelling and intense refuelling (Goymann et al. 2010). It is unlikely that such a weight limitation applies to sea turtles, due to the relatively high density of water compared to air and thus the turtle’s buoyancy. Therefore, these sites might be primarily used for resting. In order to determine whether sea turtles exhibit fidelity to stopover sites in the same way that exhibit fidelity to their primary foraging sites (Broderick et al. 2007, Schofield, Hobson,

135

Chapter 7: General discussion

Fossette, et al. 2010), the same individuals must be tracked repeatedly over multiple years.

Sea turtle also greatly reduced their travel speed at night-time (>1 km h-1 difference between day and night travel speeds) within 24 hours of entering shallow water during migration (<100 m), and repeated this pattern at a 3-6 days interval until they reached their foraging ground, which suggested that they might rest on the seabed for several hours. Equipping sea turtles with an Argos-linked Fastloc- GPS, a time depth-recorder and an accelerometer (e. g. a setup similar to the Enstipp et al. 2016 study) would provide insights on whether sea turtles rest on the seabed during stopovers or at night during certain parts of migration (Chapter 3).

As with sea turtles, tracking enough migrating whales for sufficient periods of time to test hypotheses related to migration behaviours is a challenge; consequently, limited information is available on whether they use stopover sites. During their migrations between the high latitudes and the tropics, Atlantic fin and blue whales stop for a mean of four days in the Azores, possibly to build energy stores (Silva et al. 2013). Jervis Bay in Australia is a resting stopover site for the mothers and calves of humpback whales migrating to Antarctica (Bruce et al. 2014). Understanding the stopover patterns of marine vertebrates could provide new information on how these capital breeders manage their energy reserves during migration. This would allow us to determine whether these species use similar strategies to those described for avian and terrestrial species (e.g. refuelling or resting, Åkesson and Hedenström 2007; Sawyer and Kauffman 2011).

In summary, this thesis revealed the utility of Argos-Linked Fastloc-GPS to show the fine-scale movement patterns related to the time of day that turtles initiate and terminate migration, as well as potential resting strategies during migration. Thus, Argos-Linked Fastloc-GPS could be used to gain further insights on other marine migrating species such as whales, sharks, penguin and pinnipeds, allowing for an in-depth understanding of their movement patterns. Overall, this thesis showed that sea turtles infrequently rest during migration, probably to conserve energy. This places strong importance on the way that turtles use foraging grounds

136

Chapter 7: General discussion to build energy reserves, and the importance of investigating fine-scale habitat use at these grounds.

Foraging patterns of sea turtles

Diel pattern of migratory sea turtles

In the next two chapters (Chapter 4 and Chapter 5), I again used high-resolution Argos- linked Fastloc-GPS but here applied it to demonstrate its utility in determining the foraging patterns of loggerhead and green sea turtles in two different ocean basins. Interestingly, these two species exhibited very different movement patterns, possibly due to the fact that loggerheads are omnivores while green turtles are primarily herbivores.

Most green turtles shuttled between daytime foraging patches to night-time resting areas. The use of different activity centres during day (foraging) and night (resting) has been linked to a preference for safer habitat at night for a variety of species, including desert baboons Papio cynocephalus ursinus (Cowlishaw 1997), dugongs Dugong dugon (Sheppard et al. 2009), spinner dolphins Stenella longirostris (Tyne et al. 2015) and bottlenose dolphins Tursiops aduncus (Heithaus & Dill 2002). Sea turtles rely on vision to detect predators like sharks; thus, green turtles in the Indian Ocean might avoid foraging at night to reduce predation risk (Heithaus et al. 2002, Makowski et al. 2006).

In comparison to green turtles, loggerheads exhibited more complex movement patterns: (1) shuttling between separate but consistent day and night activity centres, as recorded for green turtles; (2) using the same overlapping day and night activity centres; and (3) shuttling between multiple day and night activity centres. The use of the same overlapping day and night activity centres has been linked to an optimum feeding strategy, involving high familiarity of fixed sites to increase feeding efficiency (Jadot et al. 2006). The reasons for variation in activity centre use may be attributed to loggerhead frequenting different patches within a site, mostly likely because the prey items they forage on might be restricted to certain habitats (e.g. reefs for sponges, extensive submerged sandbanks for bivalves) that are not suitable resting habitat (Chapter 5).

137

Chapter 7: General discussion

Shaping of sea turtle home ranges

The main factor shaping the home ranges of the foraging sites of green turtles was the shuttling movement between daytime foraging patches and night-time resting areas. Green turtles made relatively few forays outside their foraging sites and also made few exploratory movements around feeding patches (Chapter 4). This suggested that green turtles have good knowledge of their environment and have access to a stable source of food. In contrast, the home range of loggerhead turtles was shaped by frequent exploratory movements around focal patches (Chapters 5). Movement beyond patches, but remaining within the site, might be an important strategy to actively sample the environment and monitor other potential patches that might be depleted or regenerating (Bartumeus et al. 2016). The direct observation of loggerhead sea turtles at one patch confirmed the presence of antagonistic interactions, which are not commonly observed among green turtles at other study areas (Chapter 5, Bjorndal 1985). Therefore, loggerhead turtles might modify their space use and hence, their home range in response to those interactions. However, for the two species of sea turtles studied here, the actual distribution of food is not known. Knowledge on the distribution of seagrass beds on which green turtles feed in the Indian Ocean is limited and often confined to coastal areas (e.g. Carbone and Accordi 2000), which makes it more difficult to interpret foraging movement. Similarly, it is difficult to quantify the distribution of prey on the continental shelf in the Mediterranean Sea, which limits the ability to analyse how the distribution of food correlates with fine-scale animal movement. Such information is important to test a range of models and theories developed on terrestrial animals (e.g. Charnov marginal value theorem and optimal forging theory; Charnov 1976; Bartumeus and Catalan 2009). Like sea turtles, the home range of humpback whales feeding on pelagic krill is shaped by the use of multiple patches, but at a larger spatial scale (hundreds of kilometres) and with a high inter-individual variability (Heide- Jørgensen & Laidre 2007, Dalla Rosa et al. 2008). However, due to the low accuracy of transmitters (typically Argos tags), the small-scale movement pattern of whales is as of yet poorly resolved (Heide-Jørgensen & Laidre 2007, Dalla Rosa et al. 2008) so this remains an open question in ecology.

138

Chapter 7: General discussion

Changes in body condition in southern humpback whales.

In Chapter 6, I described how the body condition of southern humpback whales change while they are at the breeding ground. I, with my co-authors, clearly illustrated the cost of reproduction for lactating females by showing their body condition significantly decreased over the breeding season while the body condition of immature and calves remained constant. Additionally, lactating females showed a decrease in width along a larger portion of their body (35–80 %) compared with mature whales (50–65 %) illustrating the extra energetic cost of lactation. While it was not possible to link changes in whale shape with an absolute value of weight loss, I could compare these observations with finding on other marine species. For example, green turtles in Ascension island and hawksbill turtles in Brazil loose around 19 % and 15 % of their body weight during the breeding season respectivly (Hays, Broderick, Glen, et al. 2002, Santos et al. 2010). Female elephant seals (Mirounga angustirostris) can lose around 42 % of their body mass while Antarctic fur seal (Arctocephalus gazella) and Galapagos fur seal (Arctocephalus galapagoensis) can lose 26 % and 22 % of their fat mass, respectively (Costa et al. 1986, Costa & Trillmich 1988). This chapter covers the final phase of the migratory cycle that was not investigated for sea turtles because measurements of the changes in their body condition are logistically challenging and cannot be achieved using UAVs due to the shell (carapace) covering the body. By using humpback whales, our findings illustrate how costly reproduction can be for marine migratory species (including sea turtles), and that be able to accurately estimate energy expenditure is important to understand the biology of those species.

General conclusion

In this thesis, I demonstrated how recently developed technologies, such as Argos- linked Fastloc-GPS and UAVs, could be used to provide detailed information on the movement patterns and life-history characteristics of elusive marine migratory species. I showed that these two techniques could be used to elucidate key behaviours, such as the time of day that these animals start and end migration or even rest, as well as fine-scale movement patterns at foraging grounds. I also

139

Chapter 7: General discussion showed that UAVs are a useful addition to the scientist’s toolbox, suggesting how they could be used to complement the use of Argos-linked Fastloc-GPS to capture difficult-to-obtain key information. These two technologies provide new ways of obtaining in-depth information on elusive animals, opening a whole new range of applications that will provide the opportunity to determine how the strategies of marine vertebrates compare to those of avian and terrestrial species.

140

References

Aarestrup K, Økland F, Hansen M, Righton D, Gargan P, Castonguay M, Bernatchez L, Howey P, Sparholt H, Pedersen MI, McKinley RS (2009) Oceanic spawning migration of the European eel (Anguilla anguilla). Science 325:1000 Aguilar A, Borrell A (1990) Patterns of lipid content and stratification in the blubber of Fin Whales (Balaenoptera physalus). J Mammal 71:544–554 Åkesson S, Alerstam T, Hedenström A (1996) Flight initiation of nocturnal passerine migrants in relation to celestial orientation conditions at twilight. J Avian Biol 27:95–102 Åkesson S, Broderick AC, Glen F, Godley BJ, Luschi P, Papi F, Hays GC (2003) Navigation by green turtles: which strategy do displaced adults use to find Ascension Island? Oikos 103:363–372 Åkesson S, Hedenström A (2007) How migrants get there: migratory performance and orientation. Bioscience 57:123–133 Albon SD, Mitchell B, Staines BW (1983) Fertility and body weight in female red deer: a density-dependent relationship. J Anim Ecol 52:969–980 Alerstam T (2009) Flight by night or day? Optimal daily timing of bird migration. J Theor Biol 258:530–536 Alerstam T, Hedenström A, Åkesson S (2003) Long-distance migration : evolution and determinants. Oikos 103:247–260 Almeida PR, Quintella BR, Dias NM (2002) Movement of radio-tagged anadromous sea lamprey during the spawning migration in the River Mondego (Portugal). Hydrobiologia 483:1–8 Anglia E (1972) Dispersal of common seals (Phoca vitulina), tagged in the Wash, East Anglia. Notes from Mammal Soc 29:715–716 Aragones LV, Lawler IR, Foley WJ, Marsh H (2006) Dugong grazing and turtle cropping: grazing optimization in tropical seagrass systems? Oecologia 149:635–647

141

References

Aragones L, Marsh H (2000) Impact of dugong grazing and turtle cropping on tropical seagrass communities. Pacific Conserv Biol 5:277–288 Arnould JPY, Monk J, Ierodiaconou D, Hindell MA, Semmens J, Hoskins AJ, Costa DP, Abernathy K, Marshall GJ (2015) Use of anthropogenic sea floor structures by Australian fur seals: potential positive ecological impacts of marine industrial development? PLoS One 10:e0130581 Atkinson SN, Ramsay MA (1995) The effects of prolonged fasting of the body composition and reproductive success of female polar bears (Ursus maritimus). Funct Ecol 9:559–567 Atwood TB, Connolly RM, Ritchie EG, Lovelock CE, Heithaus MR, Hays GC, Fourqurean JW, Macreadie PI (2015) Predators help protect carbon stocks in blue carbon ecosystems. Nat Clim Chang 5:1038–1045 Avens L, Lohmann KJ (2003) Use of multiple orientation cues by juvenile loggerhead sea turtles Caretta caretta. J Exp Biol 206:4317–4325 Avgar T, Street G, Fryxell JM (2014) On the adaptive benefits of mammal migration. Can J Zool 92:481–490 Bailey DW, Gross JE, Laca EA, Rittenhouse LR, Coughenour MB, Swift DM, Sims PL, Bailey DW (1996) Mechanisms that result in large herbivore grazing distribution patterns. J Range Manag 49:386–400 Bajaj R, Ranaweera SL, Agrawal DP (2002) GPS: location-tracking technology. Computer (Long Beach Calif) 35:92–94 Baker CS, Herman LM (1984) Aggressive behavior between humpback whales (Megaptera novaeangliae) wintering in Hawaiian waters. Can J Zool 62:1922– 1937 Ballorain K, Ciccione S, Bourjea J, Grizel H, Enstipp M, Georges J-Y (2010) Habitat use of a multispecific seagrass meadow by green turtles Chelonia mydas at Mayotte Island. Mar Biol 157:2581–2590 Barlow J, Clapham PJ (1997) A new birth interval approach to estimating demographic parameters of humpback whales. Ecology 78:535–546

142

References

Barry D, Hartigan JA, Journal S, Statistical A, Mar N, Barry D, Hartigan JA (1993) A bayesian analysis for change point problems a bayesian analysis for change point problems. J Am Stat Assoc 88:309–319 Bartumeus F, Campos D, Ryu WS, Lloret-Cabot R, Méndez V, Catalan J (2016) Foraging success under uncertainty: search tradeoffs and optimal space use. Ecol Lett 19:1–15 Bartumeus F, Catalan J (2009) Optimal search behavior and classic foraging theory. J Phys A Math Theor 42:434002 Bartumeus F, Catalan J, Viswanathan GM, Raposo EP, Luz MGE Da (2008) The influence of turning angles on the success of non-oriented animal searches. J Theor Biol 252:43–55 Baudouin M, Thoisy B de, Chambault P, Berzins R, Entraygues M, Kelle L, Turny A, Maho Y Le, Chevallier D (2015) Identification of key marine areas for conservation based on satellite tracking of post-nesting migrating green turtles (Chelonia mydas). Biol Conserv 184:36–41 Beest FM Van, Rivrud IM, Loe LE, Milner JM, Mysterud A (2011) What determines variation in home range size across spatiotemporal scales in a large browsing herbivore? J Anim Ecol 80:771–785 Bejder M, Johnston DW, Smith J, Friedlaender A, Bejder L (2016) Embracing conservation success of recovering humpback whale populations: evaluating the case for downlisting their conservation status in Australia. Mar Policy 66:137–141 Benhamou S (2014) Of scales and stationarity in animal movements. Ecol Lett 17:261–272 Berejikian BA, Moore ME, Jeffries SJ (2016) Predator-prey interactions between harbor seals and migrating steelhead trout smolts revealed by acoustic telemetry. Mar Ecol Prog Ser 543:21–35 Best P, Rüther H (1992) Aerial photogrammetry of southern right whales, Eubalaena australis. J Zool London 228:595–614

143

References

Biebach H, Biebach I, Friedrich W, Heine G, Partecke J, Schmidl D (2000) Strategics of passerine migration across the Mediterranean Sea and the Sahara Desert: a radar study. Ibis (Lond 1859) 142:623–634 Biebach H, Friedrich W, Heine G (1986) Oecologia in trans-sahara migrating passerine birds. Oecologia 69:370–379 Binder TR, Cooke SJ, Hinch SG (2011) The biology of fish migration. In: Farrell AP (ed) Encyclopedia of fish physiology: from genome to environment. Academic Press, San Diego, California, USA, p 1921–1927 Bishop CM, Hawkes LA, Chua B, Frappell PB, Milsom WK, Natsagdorj T, Newman SH, Scott GR, Takekawa JY, Wikelski M, Butler PJ (2014) The roller coaster flight strategy of bar-headed geese conserves energy during Himalayan migrations. Science 347:250–254 Bjørge A, Bekkby T, Bryant EB (2002) Summer home range and habitat selection of harbour seal (Phoca vitulina) pups. Mar Mammal Sci 18:438–454 Bjorndal KA (1980) Nutrition and grazing behavior of the green turtle Chelonia mydas. Mar Biol 56:147–154 Bjorndal KA (1985) Nutritional ecology of sea turtles. Copeia 3:736–751 Blewitt G (1997) Basics of the GPS technique : observation equations (B Johnson, Ed.). University of Newcaste, Sweden Blumenthal JM, Solomon JL, Bell CD, Austin TJ, Ebanks-Petrie G, Coyne MS, Broderick AC, Godley BJ (2006) Satellite tracking highlights the need for international cooperation in marine turtle management. Endanger Species Res 2:51–61 Boltnev AI, York AE (2001) Maternal investment in Northern fur seals (Callorhinus ursinus): interrelationships among mother’s age, size, parturition date, offspring size and sex ratios. J Zool 254:219–228 Boltnev AI, York AE, Antonelis GA, Boltnev AI, York AE, Antonelis GA (1998) Northern fur seal young: interrelationships among birth size, growth, and survival. Can J Zool 76:843–854

144

References

Bonfil R, Meÿer M, Scholl MC, Johnson R, O’Brien S, Oosthuizen H, Swanson S, Kotze D, Paterson M (2005) Transoceanic migration, spatial dynamics, and population linkages of white sharks. Science 310:100–103 Bonnet X, Bradshaw D, Shine R (2016) Capital versus income breeding : an ectothermic perspective. Oikos 83:333–342 Bonnet XB, Lourdais OL, Shine RS, Maulleau G (2002) Reproduction in a typical capital breeder : costs, currencies, and complications in the Aspic Viper. Ecology 83:2124–2135 Börger L, Dalziel BD, Fryxell JM (2008) Are there general mechanisms of animal home range behaviour? A review and prospects for future research. Ecol Lett 11:637–650 Bost C-A, Goarant A, Scheffer A, Koubbi P, Duhamel G, Charrassin J-B (2011) Foraging habitat and performances of King penguins Aptenodytes patagonicus Miller, 1778 at Kerguelen islands in relation to climatic variability. In: Duhamel G. & Welsford D. (ed) The Kerguelen Plateau: Marine Ecosystem and Fisheries. Société d’Ichtyologie, p 199–202 Boustany AM, Davis SF, Pyle P, Anderson SD, Boeuf BJ Le, Block BA (2002) Expanded niche for white sharks. Nature 415:35–36 Bowen WD, Iverson SJ, Boness DJ, Oftedal OT (2001) Foraging effort, food intake and lactation performance depend on maternal mass in a small phocid seal. Funct Ecol 15:325–334 Bowlin MS, Bisson I-A, Shamoun-Baranes J, Reichard JD, Sapir N, Marra PP, Kunz TH, Wilcove DS, Hedenström A, Guglielmo CG, Åkesson S, Ramenofsky M, Wikelski M (2010) Grand challenges in migration biology. Integr Comp Biol 50:261–279 Bowyer RT, Kie JG (2006) Effects of scale on interpreting life-history characteristics of ungulates and carnivores. Divers Distrib 12:244–257 Boyce MS, McDonald LL (1999) Relating populations to habitats using resource selection functions. Trends Ecol Evol 14:268–272

145

References

Bradshaw CJA, Sims DW, Hays GC (2007) Measurement error causes scale- dependent threshold erosion of biological signals in animal movement data. Ecol Appl 17:628–638 Braithwaite JE, Meeuwig JJ, Hipsey MR (2015) Optimal migration energetics of humpback whales and the implications of disturbance. Conserv Physiol 3:1– 15 Braithwaite JE, Meeuwig JJ, Jenner KCS (2012) Estimating cetacean carrying capacity based on spacing behaviour. PLoS One 7:e51347 Bridge ES, Kelly JF, Contina A, Gabrielson RM, MacCurdy RB, Winkler DW (2013) Advances in tracking small migratory birds: a technical review of light- level geolocation. J F Ornithol 84:121–137 Brill RW, Balazs GH, Holland KN, Chang RK, Sullivan S, George JC (1995) Daily movements, habitat use, and submergence intervals of normal and tumor- bearing juvenile green turtles (Chelonia mydas L.) within a foraging area in the Hawaiian Islands. J Exp Mar Bio Ecol 185:203–218 Broderick AC, Coyne MS, Fuller WJ, Glen F, Godley BJ (2007) Fidelity and overwintering of sea turtles. Proc R Soc B Biol Sci 274:1533–1539 Brodie PF (1975) Cetacean energetics, an overview of intraspecific size variation. Ecology 56:152–161 Brooks LB, Harvey JT, Nichols WJ (2009) Tidal movements of East Pacific green turtle Chelonia mydas at a foraging area in Baja California Sur, México. Mar Ecol Prog Ser 386:263–274 Brown JS (1988) Patch use as an indicator of habitat preference, predation risk, and competition. Behav Ecol Sociobiol 22:37–47 Brown JS (1999) Vigilance, patch use and habitat selection: foraging under predation risk. Evol Ecol Res 1:49–71 Brown JS, Kotler BP (2004) Hazardous duty pay and the foraging cost of predation. Ecol Lett 7:999–1014

146

References

Bruce E, Albright L, Sheehan S, Blewitt M (2014) Distribution patterns of migrating humpback whales (Megaptera novaeangliae) in Jervis Bay, Australia: a spatial analysis using geographical citizen science data. Appl Geogr 54:83–95 Bryant E (2007) 2D Location Accuracy Statistics for Fastloc® Cores Running Firmware Versions 2.2 & 2.3. Wildtrack Telemetry Systems Ltd Burt WH (1943) Territoriality and home range concepts as applied to mamals. J Mammal 24:346–352 Calenge C (2006) The package “adehabitat” for the R software: a tool for the analysis of space and habitat use by animals. Ecol Modell 197:516–519 Calenge C, Dray S, Royer-Carenzi M (2009) The concept of animals’ trajectories from a data analysis perspective. Ecol Inform 4:34–41 Cameron RD, Smith WT, Fancy SG, Gerhart KL, White RG (1993) Calving success of female caribou in relation to body weight. Can J Zool 71:480–486 Carbone F, Accordi G (2000) The Indian Ocean coast of Somalia. Mar Pollut Bull 41:141–159 Cardona L, Revelles M, Sales M, Aguilar A, Borrell A (2007) Meadows of the seagrass Posidonia oceanica are a significant source of organic matter for adjoining ecosystems. Mar Ecol Prog Ser 335:123–131 Carr A (1967) So excellent a fishe. A natural history of sea turtles (A Carr, Ed.). Natural History Press, New York, USA Carr A (1986) Rips, FADS, and little loggerheads. Bioscience 36:92–100 Casale P, Abbate G, Freggi D, Conte N, Oliverio M, Argano R (2008) Foraging ecology of loggerhead sea turtles Caretta caretta in the central Mediterranean Sea: evidence for a relaxed life history model. Mar Ecol Prog Ser 372:265– 276 Casale P, Freggi D, Basso R, Argano R (2005) Size at male maturity, sexing methods and adult sex ratio in loggerhead turtles (Caretta caretta) from italian waters investigated through tail measurementse. Herpetol J 15:145–148

147

References

Casale P, Mariani P (2014) The first “lost year” of Mediterranean sea turtles: Dispersal patterns indicate subregional management units for conservation. Mar Ecol Prog Ser 498:263–274 Cederlund G, Sand H (1994) Home-range size in relation to age and sex in moose. J Mammal 75:1005–1012 Charnov EL (1976) Optimal foraging, the marginal value theorem. Theor Popul Biol 9:129–136 Chevaillier D, Karpytchev M, Mcconnell BJ, Moss S, Vincent C (2014) Can gray seals maintain heading within areas of high tidal current? Preliminary results from numerical modeling and GPS observations. Mar Mammal Sci 30:374– 380 Chevis MG, Godley BJ, Lewis JP, Lewis JJ, Scales KL, Graham RT (2016) Movement patterns of juvenile hawksbill turtles (Eretmochelys imbricata) at a Caribbean coral atoll: long-term tracking using passive acoustic telemetry. Mar Ecol Prog Ser 32:309–319 Chittleborough RG (1953) Aerial observation on the humpback whale, Megaptera nodosa (Bonnaterre), with notes on other species. Aust J Mar Freshw Res 4:219–226 Chittleborough RG (1955a) Aspects of reproduction in the male humpback whale, Megaptera nodosa (Bonnaterre). Aust J Mar Freshw Res 6:1–29 Chittleborough RG (1955b) Puberty, physical maturity, and relative growth of the female humpback whale, Megaptera nodosa (Bonnaterre), on the Western Australian coast. Aust J Mar Freshw Res 6:315–327 Chittleborough RG (1965) Dynamics of two populations of the humpback whale, Megaptera novaeangliae (Borowski). Aust J Mar Freshw Res 16:33–128 Christiansen F, Esteban N, Mortimer JA, Dujon AM, Hays GC (2017) Diel and seasonal patterns in activity and home range size of green turtles on their foraging grounds revealed by extended Fastloc-GPS tracking. Mar Biol 77:285–299 Christiansen F, Lusseau D (2015) Linking behavior to vital rates to measure the effects of non-lethal disturbance on wildlife. Conserv Lett 8:424–431

148

References

Christiansen F, Putman NF, Farman R, Parker DM, Rice MR, Polovina JJ, Balazs GH, Hays GC (2016) Spatial variation in directional swimming enables juvenile sea turtles to reach and remain in productive waters. Mar Ecol Prog Ser 557:247–250 Christiansen F, Vikingsson GA, Rasmussen MH, Lusseau D (2013) Minke whales maximise energy storage on their feeding grounds. J Exp Biol 216:427–436 Christiansen F, Víkingsson GA, Rasmussen MH, Lusseau D (2014) Female body condition affects foetal growth in a capital breeding mysticete. Funct Ecol 28:579–588 Clapham PJ (1996) The social and reproductive biology of humpback whales: an ecological perspective. Mamm Rev 26:27–49 Clutton-Brock T, Sheldon BC (2010) Individuals and populations: the role of long- term, individual-based studies of animals in ecology and evolutionary biology. Trends Ecol Evol 25:562–573 Coates JH, Hovel KA, Butler JL, Klimley AP, Morgan SG (2013) Movement and home range of pink abalone Haliotis corrugata: implications for restoration and population recovery. Mar Ecol Prog Ser 486:189–201 Codling EA, Plank MJ, Benhamou S (2008) Random walk models in biology. J R Soc Interface 5:813–834 Cook JG, Johnson BK, Cook RC, Riggs RA, Delcurto TIM, Bryant LD, Irwin LL (2004) Effects of summer-autumn nutrition and parturition date on reproduction and survival of Elk. Wildl Monogr 155:1–61 Corkeron PJ, Connor RC (1999) Why do baleen whales migrate? Mar Mammal Sci 15:1228–1245 Cosens S, Blouw A (2003) Size- and age-class segregation of bowhead whales summering in Northern Foxe Basin: a photogrammetric analysis. Mar Mammal Sci 19:284–296 Costa DP, Boeuf BJ Le, Huntley AC, Ortix CL (1986) The energetics of lactation in the northern elephant seal, Mirounga angustirostris. J Zool 209:21–33

149

References

Costa DP, Robinson PW, Arnould JPY, Harrison A-L, Simmons SE, Hassrick JL, Hoskins AJ, Kirkman SP, Oosthuizen H, Villegas-Amtmann S, Crocker DE (2010) Accuracy of ARGOS locations of pinnipeds at-sea estimated using Fastloc GPS. PLoS One 5:e8677 Costa DP, Trillmich F (1988) Mass changes and metabolism during the perinatal fast: a comparison between Antarctic (Arctocephalus gazella) and Galápagos fur seals (Arctocephalus galapagoensis). Physiol Zool 61:160–169 Cowlishaw G (1997) Trade-offs between foraging and predation risk determine habitat use in a desert baboon population. Anim Behav 53:667–686 Creel S, Winnie J, Maxwell B, Hamlin K, Creel M (2005) Elk alter habitat selection as an antipredator response to wolves. Ecology 86:3387–3397 Crocker DE, Williams JD, Costa DP, Boeuf BJ Le (2001) Maternal traits and reproductive effort in northern elephant sea. Ecology 82:3541–3555 Crutsinger GM, Short J, Sollenberger R (2016) The future of UAVs in ecology : an insider perspective from the Silicon Valley drone industry. J Unmanned Veh Syst 4:161–168 Dalla Rosa L, Secchi ER, Maia YG, Zerbini AN, Heide-Jørgensen MP (2008) Movements of satellite-monitored humpback whales on their feeding ground along the Antarctic Peninsula. Polar Biol 31:771–781 Dawbin WH (1966) The seasonal migratory cycle of humpback whales. In: Norris Kenneth S (ed) Whales, dolphins and porpoises. University of California Press, Berkeley, p 145–170 Dawson A, King VM, Bentley GE, Ball GF (2001) Photoperiodic control of seasonality in birds. J Biol Rhythms 16:365–380 Deutsch CJ, Reid JP, Bonde RK, Easton DE, Kochman HI, O’Shea TJ (2003) Seasonal movements, migratory behavior, and site fidelity of West Indian manatees along the atlantic coast of the United States. Wildl Monogr 151:1– 77 Dickinson MH, Robert TF, Koehl MAR, Kram R, Lehman S (2000) How animals move: An integrative view. Science 288:100–106 Dingle H (1972) Migration strategies of insects. Science 175:1327–1335

150

References

Dingle H (1996) Migration: the Biology of life on the move. Oxford University Press, New York Dingle H, Drake VA (2007) What is migration? Science 57:113–121 Dittman AH., Quinn TP (1996) Homing in Pacific salmon: mechanisms and ecological basis. J Exp Biol 199:83–91 Dobson FS (1992) Body mass, structural size, and life-history patterns of the columbian ground squirrel. Am Nat 140:109–125 Domeier ML, Nasby-Lucas N (2013) Two-year migration of adult female white sharks (Carcharodon carcharias) reveals widely separated nursery areas and conservation concerns. Anim Biotelemetry 1:1–9 Dray S, Royer-Carenzi M, Calenge C (2010) The exploratory analysis of autocorrelation in animal-movement studies. Ecol Res 25:673–681 Duggins DO, Simenstad CA, Estes JA (1989) Magnification of secondary production by kelp detritus in coastal marine ecosystems. Science 245:170– 174 Dujon AM, Lindstrom RT, Hays GC (2014) The accuracy of Fastloc-GPS locations and implications for animal tracking. Methods Ecol Evol 5:1162–1169 Ehrenfeld DW, Carr A (1967) The role of vision in the sea-finding orientation of the green turtle (Chelonia mydas). Anim Behav 15:25–36 Ehrenfeld DW, Koch AL (1967) Visual accommodation in the green turtle. Science 155:827–828 Emlen JM (1966) Role of time and energy in food preference. Am Nat 100:611– 617 Enstipp MR, Ballorain K, Ciccione S, Narazaki T, Sato K, Georges J-Y (2016) Energy expenditure of adult green turtles at their foraging grounds and during simulated oceanic migration. Funct Ecol 30:1810–1825 Erdman C, Emerson JW (2007) bcp: an R package for performing a Bayesian analysis of change point problems. J Stat Softw 23:1–13 Espinoza M, Farrugia TJ, Webber DM, Smith F, Lowe CG (2011) Testing a new acoustic telemetry technique to quantify long-term, fine-scale movements of aquatic animals. Fish Res 108:364–371

151

References

Esteban N, Dam RP van, Harrison E, Herrera A, Berkel J (2015) Green and hawksbill turtles in the Lesser Antilles demonstrate behavioural plasticity in inter-nesting behaviour and post-nesting migration. Mar Biol 162:1153–1163 Esteban N, Mortimer JA, Hays GC (2017) How numbers of nesting sea turtles can be overestimated by nearly a factor of two. Proc R Soc B Biol Sci 284:20162581 Evans K, Baer H, Bryant E, Holland M, Rupley T, Wilcox C (2011) Resolving estimation of movement in a vertically migrating pelagic fish: does GPS provide a solution? J Exp Mar Bio Ecol 398:9–17 Fagan WF, Lewis MA, Auger-Méthé M, Avgar T, Benhamou S, Breed G, LaDage L, Schlägel UE, Tang W-W, Papastamatiou YP, Forester J, Mueller T (2013) Spatial memory and animal movement. Ecol Lett 16:1316–29 Festa-Bianchet M (1988) Seasonal range selection in bighorn sheep: conflicts between forage quality, forage quantity, and predator avoidance. Oecologia 75:580–586 Festa-Bianchet M (1998) Condition-dependent reproductive success in bighorn ewes. Ecol Lett 1:91–94 Fieberg J, Kochanny CO (2005) Quantifying home-range overlap : the importance of the Utilization Distribution. J Wildl Manage 69:1346–1359 Foley AM, Schroeder BA, Hardy R, MacPherson SL, Nicholas M (2014) Long- term behavior at foraging sites of adult female loggerhead sea turtles (Caretta caretta) from three Florida rookeries. Mar Biol 161:1251–1262 Folkow LP, Blix AS (1992) Metabolic rates of minke whales (Balaenoptera acutorostrata) in cold water. Acta Physiol Scand 146:141–150 Fontaine M (1976) Physiological mechanisms in the migration of marine and amphihaline fish. Adv Mar Biol 13:241–355 Foo D, Semmens JM, Arnould JPY, Dorville N, Hoskins AJ, Abernathy K, Marshall GJ, Hindell MA (2016) Testing optimal foraging theory models on benthic divers. Anim Behav 112:127–138

152

References

Forester JD, Ives AR, Turner MG, Anderson DP, Fortin D, Beyer HL, Smith DW, Boyce MS (2007) State-space models link elk movement patterns to landscape characteristics in Yellowstone National Park. Ecol Monogr 77:285–299 Forsyth DM, Duncan RP, Tustin KG, Gaillard J-M (2005) Male reproduction in a sexually dimorphic ungulate. Ecology 86:2154–2163 Fossette S, Putman NF, Lohmann KJ, Marsh R, Hays GC (2012) A biologist’s guide to assessing ocean currents : a review. Mar Ecol Prog Ser 457:285–301 Fossette S, Schofield G, Lilley MKS, Gleiss AC, Hays GC (2012) Acceleration data reveal the energy management strategy of a marine ectotherm during reproduction. Funct Ecol 26:324–333 Fowler CW (1990) Density dependance in northern fur seal (Callorhinus ursinus). Mar Mammal Sci 6:171–195 Frere E, Quintana F, Gandini P, Wilson RP (2008) Foraging behaviour and habitat partitioning of two sympatric cormorants in Patagonia, Argentina. Ibis (Lond 1859) 150:558–564 Fretwell SD (1969) On territorial behavior and other factors influencing habitat distribution in birds. Part 3. Breeding success in a local population of field sparrows. Acta Biotheor 19:45–52 Friedland KD, Walker R V., Davis ND, Myers KW, Boehlert GW, Urawa S, Ueno Y (2001) Open-ocean orientation and return migration routes of chum salmon based on temperature data from data storage tags. Mar Ecol Prog Ser 216:235– 252 Fromont J, Bergquist PR (1994) Reproductive biology of three sponge species of the genus Xestospongia (Porifera: Demospongiae: Petrosida) from the Great Barrier Reef. Coral Reefs 13:119–126 Fryxell JM (1991) Forage quality and aggregation by large herbivores. Am Nat 138:478–498 Fryxell JM, Hazell M, Börger L, Dalziel BD, Haydon DT, Morales JM, McIntosh T, Rosatte RC (2008) Multiple movement modes by large herbivores at multiple spatiotemporal scales. Proc Natl Acad Sci 105:19114–19119

153

References

Fryxell JM, Sinclair ARE (1988) Causes and consequences of migration by large herbivores. Trends Ecol Evol 3:237–241 Fryxell JM, Wilmshurst JF, Sinclair ARE (2004) Predictive models of movement by Serengeti grazers. Ecology 85:2429–2435 Fudickar AM, Wikelski M, Partecke J (2012) Tracking migratory songbirds: accuracy of light-level loggers (geolocators) in forest habitats. Methods Ecol Evol 3:47–52 Fujisaki I, Hart KM, Sartain-Iverson AR (2016) Habitat selection by green turtles in a spatially heterogeneous benthic landscape in dry Tortugas National Park, Florida. Aquat Biol 24:185–199 Gaillard J-M, Festa-Bianchet M, Delorme D, Jorgenson J (2000) Body mass and individual fitness in female ungulates: bigger is not always better. Proc R Soc London, Ser B 267:471–477 Gales N, Double MC, Robinson S, Jenner C, Jenner M, Gedamke J, Childerhouse S, Paton D (2010) Satellite tracking of Australian humpback (Megaptera novaeangliae) and pygmy blue whales (Balaenoptera musculus brevicauda). International Whaling Commission, Agadir, 9 pages Galli S, Gaspar P, Fossette S, Calmettes B, Hays GC, Lutjeharms JRE, Luschi P (2012) Orientation of migrating leatherback turtles in relation to ocean currents. Anim Behav 84:1491–1500 Gaspar T, Oliveira P, Silvestre C (2011) UAV-based marine mammals positioning and tracking system. In: Proceedings of the 2011 IEEE International Conference on Mechatronics and Automation. Beijing, China, p 1050–1055 Gaston KJ, Jackson SF, Cantú-Salazar L, Cruz-Piñón G (2008) The ecological performance of protected areas. Annu Rev Ecol Evol Syst 39:93–113 Gibbs M (2012) Whale catches from 19th century shore stations in Western Australia. J Cetacean Res Manag 12:129–135 Girard C, Sudre J, Benhamou S, Roos D, Luschi P (2006) Homing in green turtles Chelonia mydas: oceanic currents act as a constraint rather than as an information source. Mar Ecol Prog Ser 322:281–289

154

References

Girondot M (1999) Statistical description of temperature-dependent sex determination using maximum likelihood. Evol Ecol 1:479–486 Gittleman JL, Thompson SD (1988) Energy allocation in mammalian reproduction. Integr Comp Biol 28:863–875 Gladfelter WB (1979) Twilight migrations and foraging activities of the copper sweeper Pempheris schomburgki (Teleostei: Pempheridae). Mar Biol 50:109– 119 Godfrey JD, Bryant DM (2003) Effects of radio transmitters: review of recent radio- tracking studies. In: Murray W (ed) Conservation applications of measuring energy expenditure of New Zealand birds: assessing habitat quality and costs of carrying radio transmitters. Departement of Conservation, Wellington, New Zealand, p 83–95 Godley BJ, Blumenthal JM, Broderick AC, Coyne MS, Godfrey MH, Hawkes LA, Witt MJ (2008) Satellite tracking of sea turtles: where have we been and where do we go next? Endanger Species Res 4:3–22 Godley BJ, Broderick AC, Glen F, Hays GC (2003) Post-nesting movements and submergence patterns of loggerhead marine turtles in the Mediterranean assessed by satellite tracking. J Exp Mar Bio Ecol 287:119–134 Godley BJ, Richardson S, Broderick AC, Coyne MS, Glen F, Hays GC (2002) Long-term satellite telemetry of the movements and habitat utilisation by green turtles in the Mediterranean. Ecography (Cop) 25:352–362 Goymann W, Spina F, Ferri A, Fusani L (2010) Body fat influences departure from stopover sites in migratory birds: evidence from whole-island telemetry. Biol Lett 6:478–481 Gredzens C, Marsh H, Fuentes MMPB, Limpus CJ, Shimada T, Hamann M (2014) Satellite tracking of sympatric marine megafauna can inform the biological basis for species co-management. PLoS One 9:e98944 Grüss A, Kaplan DM, Guénette S, Roberts CM, Botsford LW (2011) Consequences of adult and juvenile movement for marine protected areas. Biol Conserv 144:692–702

155

References

Guilford T, Taylor GK (2014) The sun compass revisited. Anim Behav 97:135–143 Gwinner E (1996) Circadian and circannual programmes in avian migration. J Exp Biol 199:39–48 Hamann M, Limpus CJ, Owens DW (2002) Reproductive cycles of males and females. In: Wyneken J (ed) The biology of sea turtles, Vol. II. CRC Press, Boca Raton, Florida, USA, p 135–161 Hanks J (1981) Characterization of population condition. In: Smith CW, Fowler TD (eds) Dynamics of large mammal populations. John Wiley and Sons, New York, New York State, USA, p 47–73 Hart K, White C, Iverson A, Whitney N (2016) Trading shallow safety for deep sleep: juvenile green turtles select deeper resting sites as they grow. Endanger Species Res 31:61–73 Hatase H, Omuta K, Tsukamoto K (2007) Bottom or midwater: alternative foraging behaviours in adult female loggerhead sea turtles. J Zool 273:46–55 Haug T, Lindstrøm U, Nilssen KT (2002) Variations in Minke Whale (Balaenoptera acutorostrata) diet and body condition in response to ecosystem changes in the Barents Sea. Sarsia 87:409–422 Hawkes LA, Witt MJ, Broderick AC, Coker JW, Coyne MS, Dodd M, Frick MG, Godfrey MH, Griffin DB, Murphy SR, Murphy TM, Williams KL, Godley BJ (2011) Home on the range: spatial ecology of loggerhead turtles in Atlantic waters of the USA. Divers Distrib 17:624–640 Hays GC, Adams C, Broderick AC, Godley BJ, Lucas DJ, Metcalfe JD, Prior AA (2000) The diving behaviour of green turtles at Ascension Island. Anim Behav 59:577–586 Hays GC, Åkesson S, Broderick AC, Glen F, Godley BJ, Papi F, Luschi P (2003) Island-finding ability of marine turtles. Proc R Soc London B, Biol Sci 270:S5–S7 Hays GC, Åkesson S, Godley BJ, Luschi P, Santidrian P (2001) The implications of location accuracy for the interpretation of satellite-tracking data. Anim Behav 61:1035–1040

156

References

Hays GC, Broderick AC, Glen F, Godley BJ (2002) Change in body mass associated with long-term fasting in a marine reptile: the case of green turtles (Chelonia mydas) at Ascension Island. Can J Zool 80:1299–1302 Hays GC, Broderick AC, Godley BJ, Lovell P, Martin C, McConnell BJ, Richardson S (2002) Biphasal long-distance migration in green turtles. Anim Behav 64:895–898 Hays GC, Christensen A, Fossette S, Schofield G, Talbot J, Mariani P (2014) Route optimisation and solving Zermelo’s navigation problem during long distance migration in cross flows. Ecol Lett 17:137–143 Hays GC, Ferreira LC, Sequeira AMM, Meekan MG, Duarte CM, Bailey H, Bailleul F, Bowen WD, Caley MJ, Costa DP, Eguíluz VM, Fossette S, Friedlaender AS, Gales N, Gleiss AC, Gunn J, Harcourt R, Hazen EL, Heithaus MR, Heupel M, Holland K, Horning M, Jonsen I, Kooyman GL, Lowe CG, Madsen PT, Marsh H, Phillips RA, Righton D, Ropert-Coudert Y, Sato K, Shaffer SA, Simpfendorfer CA, Sims DW, Skomal G, Takahashi A, Trathan PN, Wikelski M, Womble JN, Thums M (2016) Key questions in marine megafauna movement ecology. Trends Ecol Evol 31:463–475 Hays GC, Fossette S, Katselidis KA, Schofield G, Gravenor MB (2010) Breeding periodicity for male sea turtles, operational sex ratios, and implications in the face of climate change. Conserv Biol 24:1636–1643 Hays GC, Kennedy H, Frost BW (2001) Individual variability in diel vertical migration of marine copepod: why some individuals remain at deph when others migrate. Limnol Oceanogr 46:2050–2054 Hays GC, Luschi P, Papi F, Seppia C Del, Marsh R (1999) Changes in behaviour during the inter-nesting period and post-nesting migration for Ascension Island green turtles. Mar Ecol Prog Ser 189:263–273 Hays GC, Mortimer JA, Ierodiaconou D, Esteban N (2014) Use of long-distance migration patterns of an endangered species to inform conservation planning for the world’s largest marine protected area. Conserv Biol 28:1636–1644

157

References

Hays GC, Scott R (2013) Global patterns for upper ceilings on migration distance in sea turtles and comparisons with fish, birds and mammals. Funct Ecol 27:748–756 Hazel J (2009) Evaluation of fast-acquisition GPS in stationary tests and fine-scale tracking of green turtles. J Exp Mar Bio Ecol 374:58–68 Hazel J, Hamann M, Lawler IR (2013) Home range of immature green turtles tracked at an offshore tropical reef using automated passive acoustic technology. Mar Biol 160:617–627 Hedenström A (1993) Migration by soaring or flapping flight in birds: The relative importance of energy cost and speed. Proc R Soc B Biol Sci 342:353–361 Hedenström A (2008) Adaptations to migration in birds: behavioural strategies, morphology and scaling effects. Philos Trans R Soc Lond B Biol Sci 363:287– 99 Hedenström A, Alerstam T (1992) Climbing performance of migrating birds as a basis for estimating limits for fuel-carrying capacity and muscle work. J Exp Biol 164:19–38 Hedenström A, Norevik G, Warfvinge K, Andersson A, Bäckman J, Åkesson S (2016) Annual 10-month aerial life phase in the common swift Apus apus. Curr Biol 26:3066–3070 Hedley SL, Bannister JL, Dunlop RA (2011) Abundance estimates of Southern Hemisphere breeding stock “D” humpback whales from aerial and land-based surveys off Shark Bay, Western Australia, 2008. J Cetacean Res Manag 3:209– 221 Heide-Jørgensen MP, Laidre KL (2007) Autumn space-use patterns of humpback whales (Megaptera novaeangliae) in West Greenland. J Cetacean Res Manag 9:121–126 Hein AM, Hou C, Gillooly JF (2012) Energetic and biomechanical constraints on animal migration distance. Ecol Lett 15:104–110 Heithaus MR, Dill LM (2002) Food availability and tiger shark predation risk influence bottlenose dolphin habitat use. Ecology 83:480–491

158

References

Heithaus MR, Dill LM (2006) Does tiger shark predation risk in uence foraging habitat use by bottlenose dolphins at multiple spatial scales. Oikos 114:257– 264 Heithaus MR, Frid A, Dill LM (2002) Shark-inflicted injury frequencies, escape ability, and habitat use of green and loggerhead turtles. Mar Biol 140:229–236 Heithaus MR, Frid A, Wirsing AJ, Dill LM, Fourqurean JW, Burkholder D, Thomson J, Bejder L (2007) State-dependent risk-taking by green sea turtles mediates top-down effects of tiger shark intimidation in a marine ecosystem. J Anim Ecol 76:837–844 Hirayama R (1998) Oldest known sea turtle. Nature 392:705–708 Hochscheid S, Bentivegna F, Bradai MN, Hays GC (2007) Overwintering behaviour in sea turtles: dormancy is optional. Mar Ecol Prog Ser 340:287– 298 Hochscheid S, Travaglini A, Maffucci F, Hays GC, Bentivegna F (2013) Since turtles cannot talk: What beak movement sensors can tell us about the feeding ecology of neritic loggerhead turtles, Caretta caretta. Mar Ecol 34:321–333 Hodgson A, Kelly N, Peel D (2013) Unmanned aerial vehicles (UAVs) for surveying marine fauna: a dugong case study. PLoS One 8:1–15 Hoenner X, Whiting SD, Hindell MA, McMahon CR (2012) Enhancing the use of Argos satellite data for home range and long distance migration studies of marine animals. PLoS One 7:e40713 Holdo RM, Holt RD, Fryxell JM (2009) Opposing rainfall and plant nutritional gradients best explain the Wildebeest migration in the Serengeti. Am Nat 173:431–445 Holt D (1977) Predation, apparent competiton, and the structure of prey communities. Theor Popul Biol 12:197–229 Hopcraft JGC, Morales JM, Beyer HL, Borner M, Mwangomo E, Sinclair ARE, Olff H, Haydon DT (2014) Competition, predation, and migration: individual choice patterns of Serengeti migrants captured by hierarchical models. Ecol Monogr 84:355–372

159

References

Houghton JDR, Cedras A, Myers AE, Liebsch N, Metcalfe JD, Mortimer JA, Hays GC (2008) Measuring the state of consciousness in a free-living diving sea turtle. J Exp Mar Bio Ecol 356:115–120 Houghton JDR, Woolmer A, Hays GC (2000) Sea turtle diving and foraging behaviour around the Greek Island of Kefalonia. J Mar Biol Assoc United Kingdom 80:761–762 Houston AI, McNamara JM, Hutchinson JMC (1993) General results concerning the trade-off between gaining energy and avoiding predation. Philos Trans R Soc B Biol Sci 341:375–397 Humphries NE, Queiroz N, Dyer JRM (2010) Environmental context explains Lévy and Brownian movement patterns of marine predators. Nature 465:1066–1069 Hurford A (2009) GPS measurement error gives rise to spurious 180° turning angles and strong directional biases in animal movement data. PLoS One 4:e5632 Ichii T, Shinohara N, Fujise Y, Nishiwaki S, Matsuoka K (1998) Interannual changes in body fat condition index of minke whales in the Antarctic. Mar Ecol Prog Ser 175:1–12 Illius AW, Fitzgibbon C (1994) Costs of vigilance in foraging ungulates. Anim Behav 47:481–484 Jadot C, Donnay A, Acolas ML, Cornet Y, Bégout Anras ML (2006) Activity patterns, home-range size, and habitat utilization of Sarpa salpa (Teleostei: Sparidae) in the Mediterranean Sea. ICES J Mar Sci 63:128–139 Jenner C, Jenner M (2005) Distribution and abundance of humpback whales and other mega-fauna in Exmouth Gulf, Western Australia, during 2004/2005. Unpublished report to Straits Resources Jenner KCS, Jenner MNM, Mccabe KA (2001) Geographical and temporal movements of humpback whales in western australian waters. Appea J 38:692–707 Johnson CJ, Parker KL, Heard DC (2001) Foraging across a variable landscape: behavioral decisions made by woodland caribou at multiple spatial scales. Oecologia 127:590–602

160

References

Jonsen ID, Flemming JM, Myers RA (2005) Robust state-space modeling of animal movement data. Ecology 86:2874–2880 Jonsen ID, Myers RA, James MC (2006) Robust hierarchical state-space models reveal diel variation in travel rates of migrating leatherback turtles. J Anim Ecol 75:1046–57 Kasuya T (1995) Overview of cetacean life histories: an essay in their evolution. Dev Mar Biol 4:481–497 Katona SK, Whitehead HP (1981) Identifying humpback whales using their natural markings. Polar Rec (Gr Brit) 20:439–444 Kear BP, Lee MSY (2006) A primitive protostegid from Australia and early sea turtle evolution. Biol Lett 2:116–9 Kennedy JS (1951) The migration of the Desert Locust (Schistocerca gregaria Forsk.). I. The Behaviour of swarms. II. A Theory of long-range migrations. Philos Trans R Soc B 235:163–290 Kent CS, Jenner C, Jenner M, Bouchet P, Rexstad E (2012) Southern hemisphere breeding stock D humpback whale population estimates from North West Cape, Western Australia. J Cetacean Res Manag 12:29–39 Kittle AM, Fryxell JM, Desy GE, Hamr J (2008) The scale-dependent impact of wolf predation risk on resource selection by three sympatric ungulates. Oecologia 157:163–175 Klaassen M (1996) Metabolic constraints on long-distance migration in birds. J Exp Biol 199:57–64 Klaassen RHG, Strandberg R, Hake M, Alerstam T (2008) Flexibility in daily travel routines causes regional variation in bird migration speed. Behav Ecol Sociobiol 62:1427–1432 Koh LP, Wich SA (2012) Dawn of drone ecology: low-cost autonomous aerial vehicles for conservation. Trop Conserv Sci 5:121–132 Konishi K, Tamura T, Zenitani R, Bando T, Kato H, Walløe L (2008) Decline in energy storage in the Antarctic minke whale (Balaenoptera bonaerensis) in the Southern Ocean. Polar Biol 31:1509–1520

161

References

Koopman HN (1998) Topographical distribution of the blubber of harbor porpoises (Phocoena phocoena). J Mammal 79:260–270 Koopman HN, Iverson SJ, Gaskin DE (1996) Stratification and age-related differences in blubber fatty acids of the male harbour porpoise (Phocoena phocoena). J Comp Physiol B 165:628–639 Koopman HN, Pabst DA, McLellan WA, Dillaman RM, Read AJ (2002) Changes in blubber distribution and morphology associated with starvation in the harbor porpoise (Phocoena phocoena): evidence for regional differences in blubber structure and function. Physiol Biochem Zool 75:498–512 Koski WR, Allen T, Ireland D, Buck G, Smith PR, Macrender AM, Halick MA, Rushing C, Sliwa DJ, McDonald TL (2009) Evaluation of an unmanned airborne system for monitoring marine mammals. Aquat Mamm 35:347–357 Kovacs KM, Lavigne DM (1986) Maternal investment and neonatal growth in phocid seals. J Anim Ecol 55:1035–1051 Kuhn C, Tremblay Y, Ream R, Gelatt T (2010) Coupling GPS tracking with dive behavior to examine the relationship between foraging strategy and fine-scale movements of northern fur seals. Endanger Species Res 12:125–139 Lane JE, Boutin S, Speakman JR, Humphries MM (2010) Energetic costs of male reproduction in a scramble competition mating system. J Anim Ecol 79:27–34 Langvatn R, Hanley TA (1993) Feeding-patch choice by red deer in relation to foraging efficiency - An experiment. Oecologia 95:164–170 Laundré JW, Hernández L, Altendorf KB (2001) Wolves, elk, and bison: reestablishing the “landscape of fear” in Yellowstone National Park, U.S.A. Can J Zool 79:1401–1409 Laws RM (1977) Seal and whales of the Southern Ocean. Philos Trans R Soc London B 279:81–89 Lazar B, Gracan R, Katic J, Zavodnik D, Jaklin A, Tvrtkovic N (2011) Loggerhead sea turtles (Caretta caretta) as bioturbators in neritic habitats: an insight through the analysis of benthic molluscs in the diet. Mar Ecol 32:65–74

162

References

Lehtinen M, Happonen A, Ikonen J (2008) Accuracy and time to first fix using consumer-grade GPS receivers. In: 2008 16th International Conference on Software, Telecommunications and Computer Networks. IEEE, p 334–340 Lennox RJ, Chapman JM, Souliere CM, Tudorache C, Wikelski M, Metcalfe JD, Cooke SJ (2016) Conservation physiology of animal migration. Conserv Physiol 4:1–15 Libourel P-S, Herrel A (2015) Sleep in amphibians and reptiles : a review and a preliminary analysis of evolutionary patterns. Biol Rev 91:833–866 Lima SL (2002) Putting predators back into behavioral predator–prey interactions. Trends Ecol Evol 17:70–75 Lima SL, Dill LM (1990) Behavioral decisions made under the risk of predation: a review and prospectus. Can J Zool 68:619–640 Limpus CJ, Limpus DJ (2000) Mangroves in the diet of Chelonia mydas in Queensland, Australia. Mar Turt Newsl 89:13–15 Lockyer C (1981) Estimation of the energy costs of growth, maintenance and reproduction in the female minke whale, (Balaenoptera acutorostrata), from the Southern hemisphere. Cambridge, UK Lockyer C (1986) Body fat condition in northeast Atlantic fin whales, Balaenoptera physalus, and its relationship with reproduction and food resource. Can J Fish Aquat Sci 43:142–147 Lockyer C (1987a) The relationship between body fat, food resource and reproductive energy costs in north Atlantic fin whales (Balaenoptera physalus). Symp Zool Soc London 57:343–361 Lockyer C (1987b) Evaluation of the role of fat reserves in relation to the ecology of North Atlantic fin and sei whales. In: Huntley AC, Costa DP, Worth GAJ, Castellini MA (eds) Approaches to Marine Mammal Energetics. Society for Marine Mammalogy - Special Publication No. 1, p 183–203 Lockyer C (2007) All creatures great and smaller: a study in cetacean life history energetics. J Mar Biol Assoc UK 87:1035 Lohmann KJ, Lohmann CMF (1996) Orientation and open-sea navigation in sea turtles. J Exp Biol 199:73–81

163

References

Lohmann KJ, Luschi P, Hays GC (2008) Goal navigation and island-finding in sea turtles. J Exp Mar Bio Ecol 356:83–95 López-Legentil S, Erwin PM, Velasco M, Turon X (2013) Growing or reproducing in a temperate sea: optimization of resource allocation in a colonial ascidian. Invertebr Biol 132:69–80 Lopez R, Malardé J-P, Royer F, Gaspar P (2014) Improving Argos doppler location using multiple-model Kalman filtering. IEEE Trans Geosci Remote Sens 52:4744–4755 Loudon ASI, McNeilly AS, Milne JA (1983) Nutrition and lactational control of fertility in red deer. Nature 302:145–147 Lowry MB, Suthers IM (1998) Home range, activity and distribution patterns of a temperate rocky-reef fish, Cheilodactylus fuscus. Mar Biol 132:569–578 Lund U, Agostinelli C (2014) Circstats: Circular statistics R package version 0.2– 4. Luschi P, Benhamou S, Girard C, Ciccione S, Roos D, Sudre J, Benvenuti S (2007) Marine turtles use geomagnetic cues during open-sea homing. Curr Biol 17:126–133 Luschi P, Hays GC, Papi F (2003) A review of long-distance movements by marine turtles, and the possible role of ocean currents. Oikos 103:293–302 Luschi P, Hays GC, Seppia C Del, Marsh R, Papi F (1998) The navigational feats of green sea turtles migrating from Ascension Island investigated by satellite telemetry. Proc R Soc B, Biol Sci 265:2279–2284 Lutz PL, Musick JA (1996) The biology of sea turtles, Volume I (PL Lutz and JA Musick, Eds.). CRC Press, New York, New York State, USA Lutz PL, Musick JA, Wyneken J (2002) The biology of sea turtles, Volume II (WJW Lutz, P. L., Musick J. A., Ed.). CRC Press, New York, New York State, USA Lyamin OI, Manger PR, Ridgway SH, Mukhametov LM, Siegel JM (2008) Cetacean sleep: an unusual form of mammalian sleep. Neurosci Biobehav Rev 32:1451–1484 MacArthur RH, Pianka ER (1966) On optimal use of a patchy environement. Am Nat 100:603–609

164

References

MacArthur RH, Wilson EO (1967) The theory of island biogeography (Princeton University Press, Ed.). Princeton University Press, Princeton, New Jersey, USA MacDonald BD, Madrak S V., Lewison RL, Seminoff JA, Eguchi T (2013) Fine scale diel movement of the east Pacific green turtle, Chelonia mydas, in a highly urbanized foraging environment. J Exp Mar Bio Ecol 443:56–64 Maher CR, Lott DF, Lott DF (2000) A review of ecological determinants of territoriality within vertebrate species. Am Midl Nat 143:1–29 Makowski C, Seminoff JA, Salmon M (2006) Home range and habitat use of juvenile Atlantic green turtles (Chelonia mydas L.) on shallow reef habitats in Palm Beach, Florida, USA. Mar Biol 148:1167–1179 Mansfield KL, Wyneken J, Porter WP, Luo J (2014) First satellite tracks of neonate sea turtles redefine the “lost years” oceanic niche. Proc R Soc B Biol Sci 281:20133039 Margaritoulis D (2005) Nesting activity and reproductive output of Loggerhead Sea Turtles, Caretta caretta, over 19 seasons (1984-2002) at Laganas Bay, Zakynthos, Greece: the largest rookery in the Mediterranean. Chelonian Conserv Biol 4:916–929 Marris E (2013) Drones in science: fly, and bring me data. Nature 498:156–158 Marx FG, Uhen MD (2010) Climate, critters, and cetaceans: Cenozoic drivers of the evolution of modern whales. Science 327:993–996 Mate BR (2012) Implementation of acoustic dosimeters with recoverable month- long GPS/TDR tags to interpret controlled-exposure experiments for large whales. Adv Exp Med Biol 730:203–205 Mate BR, Best PB, Lagerquist BA, Winsor MH (2011) Coastal, offshore, and migratory movements of South African right whales revealed by satellite telemetry. Mar Mammal Sci 27:455–476 McCord J, Davis A (2012) Characteristics of monarch butterflies (Danaus plexippus) that stopover at a site in coastal South Carolina during fall migration. J Res Lepid 45:1–8

165

References

McCormick SD, Saunders RL (1987) Preparatory physiological adaptations for marine life in salmonids: osmoregulation, growth and metabolism. Am Fish Soc Symp 1:211–229 McGuire LP, Guglielmo CG, Mackenzie SA, Taylor PD (2012) Migratory stopover in the long-distance migrant silver-haired bat, Lasionycteris noctivagans. J Anim Ecol 81:377–385 McLoughlin PD, Ferguson SH (2000) A hierarchical pattern of limiting factors helps explain variation in home range size. Écoscience 7:123–130 Mcmahon CR, Burton HR, Bester MN (2000) Weaning mass and the future survival of juvenile southern elephant seals, Mirounga leonina, at Macquarie Island. Antarct Sci 72:749–753 Mcnaughton SJ, Banyikwa FF, Mcnaughton MM (1997) Promotion of the cycling of diet-enhancing nutrients by African grazers. Science 278:1798–1800 Mech LD (1977) Productivity, mortality, and population trends of wolves in northeastern Minnesota. J Mammal 58:559–574 Millar JS, Hickling GJ (1990) Fasting endurance and the evolution of mammalian body size. Funct Ecol 4:5–12 Miller CA, Best PB, Perryman WL, Baumgartner MF, Moore MJ (2012) Body shape changes associated with reproductive status, nutritive condition and growth in right whales Eubalaena glacialis and E. australis. Mar Ecol Prog Ser 459:135–156 Miller CA, Reeb D, Best PB, Knowlton AR, Brown MW, Moore MJ (2011) Blubber thickness in right whales Eubalaena glacialis and Eubalaena australis related with reproduction, life history status and prey abundance. Mar Ecol Prog Ser 438:267–283 Millspaugh JJ, Marzluff JM (2001) Radio tracking and animal populations (JJ Millspaugh and JM Marzluff, Eds.). Academic Press, San Diego, California Milner-Gulland EJ, Fryxell JM, Sinclair ARE (2011) Animal migration: a synthesis (EJ Milner-Gulland, JM Fryxell, and ARE Sinclair, Eds.). Oxford University Press, Oxford

166

References

Minamikawa S, Naito Y, Itaru U (1997) Buoyancy control in diving behavior of the loggerhead turtle, Caretta caretta. J Ethol 15:109–118 Minamikawa S, Naito Y, Sato K, Matsuzawa Y, Bando T, Sakamoto W (2000) Maintenance of neutral buoyancy by depth selection in the loggerhead turtle Caretta caretta. J Exp Biol 203:2967–2975 Mingozzi T, Mencacci R, Cerritelli G, Giunchi D, Luschi P (2016) Living between widely separated areas: Long-term monitoring of Mediterranean loggerhead turtles sheds light on cryptic aspects of females spatial ecology. J Exp Mar Bio Ecol 485:8–17 Moland E, Olsen EM, Andvord K, Knutsen JA, Stenseth NC, Sainte-Marie B (2011) Home range of European lobster (Homarus gammarus) in a marine reserve: implications for future reserve design. Can J Fish Aquat Sci 68:1197–1210 Mori Y, Boyd IL (2004) The behavioral basis for nonlinear functional responses and optimal foraging in antarctic fur seals. Ecology 85:398–410 Mortimer JA, Brandis RG von, Liljevik A, Chapman R, Collie J (2011) Fall and rise of nesting green turtles (Chelonia mydas) at Aldabra Atoll, Seychelles: positive response to four decades of protection (1968–2008). Chelonian Conserv Biol 10:165–176 Mortimer JA, Portier KM (1989) Reproductive homing and internesting behavior of the green turtle (Chelonia mydas) at Ascension Island , South Atlantic Ocean. Copeia 1989:962–977 Mott CR, Salmon M (2011) Sun compass orientation by juvenile green sea turtles (Chelonia mydas). Chelonian Conserv Biol 10:73–81 Mueller T, Olson KA, Dressler G, Leimgruber P, Fuller TK, Nicolson C, Novaro AJ, Bolgeri MJ, Wattles D, Destefano S, Calabrese JM, Fagan WF (2011) How landscape dynamics link individual-to population-level movement patterns: A multispecies comparison of ungulate relocation data. Glob Ecol Biogeogr 20:683–694

167

References

Müller F, Taylor PD, Sjöberg S, Muheim R, Tsvey A, Mackenzie SA, Schmaljohann H (2016) Towards a conceptual framework for explaining variation in nocturnal departure time of songbird migrants. Mov Ecol 4:DOI:10.1186/s40462-016-0089-2 Musyl MK, Domeier ML, Nasby-Lucas N, Brill RW, McNaughton LM, Swimmer JY, Lutcavage MS, Wilson SG, Galuardi B, Liddle JB (2011) Performance of pop-up satellite archival tags. Mar Ecol Prog Ser 433:1–28 Naess A, Haug T, Nilssen EM (1998) Seasonal variation in body condition and muscular lipid contents in Northeast Atlantic minke whale Balaenoptera acutorostrata. Sarsia 83:211–218 Narazaki T, Sato K, Abernathy KJ, Marshall GJ, Miyazaki N (2013) Loggerhead turtles (Caretta caretta) use vision to forage on gelatinous prey in mid-water. PLoS One 8:e66043 Nathan R (2008) An emerging movement ecology paradigm. Proc Natl Acad Sci U S A 105:19050–19051 Nathan R, Getz WM, Revilla E, Holyoak M, Kadmon R, Saltz D, Smouse PE (2008) A movement ecology paradigm for unifying organismal movement research. Proc Natl Acad Sci U S A 105:19052–19059 Nathan R, Giuggioli L (2013) A milestone for movement ecology research. Mov Ecol 1:1 New LF, Clark JS, Costa DP, Fleishman E, Hindell MA, Klanjšček T, Lusseau D, Kraus S, McMahon CR, Robinson PW, Schick RS, Schwarz LK, Simmons SE, Thomas L, Tyack P, Harwood J (2014) Using short-term measures of behaviour to estimate long-term fitness of southern elephant seals. Mar Ecol Prog Ser 496:99–108 Newton I (2008) The migration ecology of birds (I Newton, Ed.). Academic Press, San Diego, California, USA Nonacs P (1991) State dependent behavior and the Marginal Value Theorem. Behav Ecol 12:71–83 O’Connell AF, Nichols JD, Karanth KU (2010) Camera traps in animal ecology: methods and analyses. Springer, Tokyo

168

References

Ogden JC, Robinson L, Whitlock K, Daganhardt H, Cebula R (1983) Diel foraging patterns in juvenile green turtles (Chelonia mydas L.) in St. Croix United States virgin islands. J Exp Mar Bio Ecol 66:199–205 Okniwa N, Miyaji T, Sasaki T, Tanabe K (2010) Shell growth and reproductive cycle of the Mediterranean mussel Mytilus galloprovincialis in Tokyo Bay, Japan: relationship with environmental conditions. Plankt Benthos Res 5:214– 220 Padié S, Morellet N, Hewison AJM, Martin JL, Bonnot N, Cargnelutti B, Chamaillé-Jammes S (2015) Roe deer at risk: teasing apart habitat selection and landscape constraints in risk exposure at multiple scales. Oikos 124:1536– 1546 Paladino F V., O’Connor MP, Spotila JR (1990) Metabolism of leatherback turtles, gigantothermy, and thermoregulation of dinosaurs. Nature 344:858–860 Pauly D, Trites AW, Capuli E, Christensen V (1998) Diet composition and trophic levels of marine mammals. ICES J Mar Sci 55:467–481 Peacock N (1991) An evolutionary perspective on the patterning of maternal investment in pregnancy. Hum Nat 2:351–385 Perryman WL, Lynn MS (2002) Evaluation of nutritive condition and reproductive status of migrating gray whales (Eschrichtius robustus) based on analysis of photogrammetric data. J Cetacean Res Manag 4:155–164 Pettis HM, Rolland RM, Hamilton PK, Brault S, Knowlton AR, Kraus SD (2004) Visual health assessment of North Atlantic right whales (Eubalaena glacialis) using photographs. Can J Zool 82:8–19 Phillips JB, Moore FR (1992) Calibration of the sun-compass of a migratory bird by sunset polarized light patterns. Behav Ecol Sociobiol 31:189–194 Phillips JB, Waldvogel JA (1988) Celestial polarized light patterns as a calibration reference for sun compass. J Theor Biol:55–67

169

References

Pilcher NJ, Perry L, Antonopoulou M, Abdel-Moati MA, Abdessalaam TZ Al, Albeldawi M, Ansi M Al, Al-Mohannadi SF, Baldwin R, Chikhi A, Das HS, Hamza S, Kerr OJ, Kiyumi A Al, Mobaraki A, Suwaidi HS Al, Suweidi AS Al, Sawaf M, Tourenq C, Williams J, Willson A (2014) Short-term behavioural responses to thermal stress by hawksbill turtles in the Arabian region. J Exp Mar Bio Ecol 457:190–198 Pontier D, Gaillard J-M, Allainé D (2012) Maternal investment and demographic tactics in placental mammals. Oikos 66:424–430 Pritchard PHC (1997) Evolution, phylogeny, and current status (C Press, Ed.). Boca Raton, Florida, USA Putman NF, Verley P, Shay TJ, Lohmann KJ (2012) Simulating transoceanic migrations of young loggerhead sea turtles: merging magnetic navigation behavior with an ocean circulation model. J Exp Biol 215:1863–70 Pyke GH (2008) Optimal foraging theory. Annu Rev Ecol Syst 15:523–575 QGIS Development Team (2015) QGIS Geographic Information System. Quinn TP (1980) Evidence for celestial and magnetic compass orientation in lake migrating sockeye salmon fry. J Comp Physiol 137:243–248 R Development Core Team (2013) R: a language and environment for statistical computing. Rasheed MA (1999) Recovery of experimentally created gaps within a tropical Zostera capricorni (Aschers.) seagrass meadow, Queensland Australia. J Exp Mar Bio Ecol 235:183–200 Rasheed MA, McKenna SA, Carter AB, Coles RG (2014) Contrasting recovery of shallow and deep water seagrass communities following climate associated losses in tropical north Queensland, Australia. Mar Pollut Bull 83:491–499 Rattenborg NC, Voirin B, Cruz SM, Tisdale R, Dell’Omo G, Lipp H-P, Wikelski M, Vyssotski AL (2016) Evidence that birds sleep in mid-flight. Nat Commun 7:12468

170

References

Rees AF, Margaritoulis D, Newman R, Riggall TE, Tsaros P, Zbinden JA, Godley BJ (2013) Ecology of loggerhead marine turtles Caretta caretta in a neritic foraging habitat: movements, sex ratios and growth rates. Mar Biol 160:519– 529 Reich KJ, Bjorndal KA, Bolten AB (2007) The “lost years” of green turtles: using stable isotopes to study cryptic lifestages. Biol Lett 3:712–4 Rempel RS, Rodgers AR (1997) Effects of differential correction on accuracy of a GPS animal location system. J Wildl Manage 61:525–530 Rice MR, Balazs GH (2008) Diving behavior of the Hawaiian green turtle (Chelonia mydas) during oceanic migrations. J Exp Mar Bio Ecol 356:121– 127 Riginos C, Grace JB (2008) Savanna tree density, herbivores, and the herbaceous community: Bottom-up vs. top-down effects. Ecology 89:2228–2238 Ritchie ME, Tilman D, Knops JMH (1998) Herbivore effects on plant and nitrogen dynamics in oak savanna. Ecology 79:165–177 Robbins BD, Bell SS (2000) Dynamics of a subtidal seagrass landscape : seasonal and annual change in relation to water depth. Ecology 81:1193–1205 Russell AP, Bauer AM, Johnson MK (2005) Migration in amphibians and reptiles: An overview of patterns and orientation mechanisms in relation to life history strategies. In: Elewa D, Ashraf MT (eds) Migration of organisms: climate geography ecology. Springer, Berlin Heidelberg, Germany, p 151–203 Rutz C, Hays GC (2009) New frontiers in biologging science. Biol Lett 5:289–292 Sæther B-E (1997) Environemental stochasticity and population dynamics of large herbivores: a search for mechanisms. Trends Ecol Evol 12:143–149 Santos AJB, Freire EMX, Bellini C, Corso G, Antos ARJBS, Reire ELMXF, Ellini CLB (2010) Body mass and the energy budget of gravid hawksbill turtles (Eretmochelys imbricata) during the nesting deason. J Herpetol 44:352–359

171

References

Sato K, Watanuki Y, Takahashi A, Miller PJO, Tanaka H, Kawabe R, Ponganis PJ, Handrich Y, Akamatsu T, Watanabe Y, Mitani Y, Costa DP, Bost C-A, Aoki K, Amano M, Trathan P, Shapiro A, Naito Y (2007) Stroke frequency, but not swimming speed, is related to body size in free-ranging seabirds, pinnipeds and cetaceans. Proc R Soc London B, Biol Sci 274:471–477 Sawyer H, Kauffman MJ (2011) Stopover ecology of a migratory ungulate. J Anim Ecol 80:1078–1087 Schaub M, Pradel R, Jenni L, Lebreton JD (2001) Migrating birds stop over longer than usually thought: An improved capture-recapture analysis. Ecology 82:852–859 Scheffer A, Trathan PN, Edmonston JG, Bost CA (2016) Combined influence of meso-scale circulation and bathymetry on the foraging behaviour of a diving predator, the king penguin (Aptenodytes patagonicus). Prog Oceanogr 141:1– 16 Schofield G, Bishop CM, MacLean G, Brown P, Baker M, Katselidis KA, Dimopoulos P, Pantis JD, Hays GC (2007) Novel GPS tracking of sea turtles as a tool for conservation management. J Exp Mar Bio Ecol 347:58–68 Schofield G, Dimadi A, Fossette S, Katselidis KA, Koutsoubas D, Lilley MKS, Luckman A, Pantis JD, Karagouni AD, Hays GC (2013) Satellite tracking large numbers of individuals to infer population level dispersal and core areas for the protection of an endangered species. Divers Distrib 19:834–844 Schofield G, Hobson VJ, Fossette S, Lilley MKS, Katselidis KA, Hays GC (2010) Fidelity to foraging sites, consistency of migration routes and habitat modulation of home range by sea turtles. Divers Distrib 16:840–853 Schofield G, Hobson VJ, Lilley MKS, Katselidis KA, Bishop CM, Brown P, Hays GC (2010) Inter-annual variability in the home range of breeding turtles : implications for current and future conservation management. Biol Conserv 143:722–730 Schofield G, Katselidis KA, Dimopoulos P, Pantis JD (2008) Investigating the viability of photo-identification as an objective tool to study endangered sea turtle populations. J Exp Mar Bio Ecol 360:103–108

172

References

Schofield G, Katselidis KA, Dimopoulos P, Pantis JD, Hays GC (2006) Behaviour analysis of the loggerhead sea turtle Caretta caretta from direct in-water observation. Endanger Species Res 2:71–79 Schofield G, Katselidis KA, Pantis JD, Dimopoulos P, Hays GC (2007) Female- female aggression: structure of interaction and outcome in loggerhead sea turtles. Mar Ecol Prog Ser 336:267–274 Schofield G, Scott R, Dimadi A, Fossette S, Katselidis KA, Koutsoubas D, Lilley MKS, Pantis JD, Karagouni AD, Hays GC (2013) Evidence-based marine protected area planning for a highly mobile endangered marine vertebrate. Biol Conserv 161:101–109 Schofield G, Scott R, Katselidis KA, Mazaris AD, Hays GC (2015) Quantifying wildlife-watching ecotourism intensity on an endangered marine vertebrate. Anim Conserv 18:517–528 Scott R, Marsh R, Hays GC (2014) Ontogeny of long distance migration. Ecology 95:2840–2850 Searle KR, Hobbs NT, Shipley LA (2005) Should I stay or should I go? Patch departure decisions by herbivores at multiple scales. Oikos 111:417–424 Seidel DP, Boyce MS (2016) Varied tastes: home range implications of foraging- patch selection. Oikos 125:39–49 Seminoff JA, Jones TT (2006) Diel movements and activity ranges of Green turtle (Chelonia mydas) at a temperate foraging area in the Gulf of California, Mexico. Herpetol Conserv Biol 1:81–86 Seminoff JA, Resendiz A, Nichols WJ (2002) Home range of green turtles Chelonia mydas at a coastal foraging area in the Gulf of California, Mexico. Mar Ecol Prog Ser 242:253–265 Senft RL, Coughenour MB, Bailey DW, Rittenhouse LR, Sala OE, Swift DM (1987) Large herbivore foraging and ecological hierarchies. Bioscience 37:789–799

173

References

Senko J, Koch V, Megill WM, Carthy RR, Templeton RP, Nichols WJ (2010) Fine scale daily movements and habitat use of East Pacific green turtles at a shallow coastal lagoon in Baja California Sur, Mexico. J Exp Mar Bio Ecol 391:92– 100 Shein-Idelson M, Ondracek JM, Liaw H-P, Reiter S, Laurent G (2016) Slow waves, sharp waves, ripples, and REM in sleeping dragons. Science 352:590–595 Sheppard JK, Jones RE, Marsh H, Lawler IR (2009) Effects of tidal and diel cycles on dugong habitat use. J Wildl Manage 73:45–59 Shimada T, Jones R, Colin L, Mark H (2012) Improving data retention and home range estimates by data-driven screening. Mar Ecol Prog Ser 457:171–180 Shimada T, Jones R, Limpus C, Hamann M (2016) Time-restricted orientation of green turtles. J Exp Mar Bio Ecol 484:31–38 Sidney A, Gauthreux J (1980) Animal migration, orientation, and navigation (A Sidney and J Gauthreux, Eds.). Academic Press, New York, USA Siegel JM (2008) Do all animals sleep? Trends Neurosci 31:208–213 Sigurjónsson J, Halldórsson SD, Konráðsson A (1990) New information on age and reproduction in minke whales (Balaenoptera acutorostrata) in Icelandic waters. Page Document SC/42/NHMi27. Noordwijkerhout, The Netherlands Sih A (1980) Optimal behaviour: can foragers balance two conflict demands? Science 210:1041–1043 Silva MA, Prieto R, Jonsen I, Baumgartner MF, Santos RS (2013) North Atlantic blue and fin whales suspend their spring migration to forage in middle latitudes: building up energy reserves for the journey? PLoS One 8:e76507 Sims DW, Queiroz N, Humphries NE, Lima FP, Hays GC (2009) Long-term GPS tracking of ocean sunfish Mola mola offers a new direction in fish monitoring. PLoS One 4:e7351 Sims DW, Southall EJ, Humphries NE, Hays GC, Bradshaw CJA, Pitchford JW, James A, Ahmed MZ, Brierley AS, Hindell MA, Morritt D, Musyl MK, Righton D, Shepard ELC, Wearmouth VJ, Wilson RP, Witt MJ, Metcalfe JD (2008) Scaling laws of marine predator search behaviour. Nature 451:1098– 102

174

References

Sinclair ARE, Pech RP, Dickman CR, Hik D, Mahon P, Newsome AE (1998) Predicting effects of predation on conservation of endangered prey. Conserv Biol 12:564–575 Sirtrack (2010) FastlocTM user guide version: 1.3. Sirtrack Ltd, Havelock North Skogland T (1984) The effects of food maternal conditions on fetal growth and size in wild reindee. Ramgifer 4:39–46 Southwood A, Avens L (2010) Physiological, behavioral, and ecological aspects of migration in reptiles. J Comp Physiol B Biochem Syst Environ Physiol 180:1– 23 Sprogis KR, Raudino HC, Rankin R, MacLeod CD, Bejder L (2016) Home range size of adult Indo-Pacific bottlenose dolphins (Tursiops aduncus) in a coastal and estuarine system is habitat and sex-specific. Mar Mammal Sci 32:287–308 Stearns SC (1992) The evolution of life histories. Oxford University Press, Oxford, UK Stephens PA, Boyd IL, McNamara JM, Houston AI (2009) Capital breeding and income breeding: their meaning, measurement, and worth. Ecology 90:2057– 2067 Stephens DW, Krebs JR (1986) Foraging theory. Princeton University Press, Princeton Stewart KM, Bowyer RT, Dick BL, Johnson BK, Kie JG (2005) Density-dependent effects on physical condition and reproduction in North American elk : an experimental test. Oecologiaecologia 143:85–93 Storch I (1993) Habitat selection by capercaillie in summer and autumn: Is bilberry important? Oecologia 95:257–265 Strandberg R, Alerstam T (2007) The strategy of fly-and-forage migration, illustrated for the osprey (Pandion haliaetus). Behav Ecol Sociobiol 61:1865– 1875 Struntz DJ, Mclellan WA, Dillaman RM, Blum JE, Kucklick JR, Pabst DA (2004) Blubber development in Bottlenose Dolphins (Tursiops truncatus). J Morphol 20:7–20

175

References

Taquet C, Taquet M, Dempster T, Soria M, Ciccione S, Roos D, Dagorn L (2006) Foraging of the green sea turtle Chelonia mydas on seagrass beds at Mayotte Island (Indian Ocean), determined by acoustic transmitters. Mar Ecol Prog Ser 306:295–302 Teo SLH, Boustany A, Blackwell S, Walli A, Weng KC, Block BA (2004) Validation of geolocation estimates based on light level and sea surface temperature from electronic tags. Mar Ecol Prog Ser 283:81–98 Thewissen JGM, Cooper LN, George JC, Bajpai S (2009) From land to water: the origin of whales, dolphins and porpoises. Evol Educ Outreach 2:272–288 Thompson D, Fedak MA (2001) How long should a dive last? A simple model of foraging decisions by breath-hold divers in a patchy environment. Anim Behav 61:287–296 Thys TM, Ryan JP, Dewar H, Perle CR, Lyons K, Sullivan JO, Farwell C, Howard MJ, Weng KC, Lavaniegos BE, Gaxiola-castro G, Erasmo L, Bojorquez M, Hazen EL, Bograd SJ (2015) Ecology of the Ocean Sunfish, Mola mola, in the southern California Current System. J Exp Mar Bio Ecol 471:64–76 Tomkiewicz SM, Fuller MR, Kie JG, Bates KK (2010) Global Positioning System and associated technologies in animal behaviour and ecological research. Philos Trans R Soc B, Biol Sci 365:2163–2176 Tougaard J, Teilmann J, Tougaard S (2008) Harbour seal spatial distribution estimated from Argos satellite telemetry: overcoming positioning errors. Endanger Species Res 4:113–122 Turchin P (1998) Quantitative analysis of movement: measuring and modeling population redistribution in animals and plants. Sinauer Associates Incorporated, Sunderland, Massachusetts Tyne JA, Johnston DW, Rankin R, Loneragan NR, Bejder L (2015) The importance of spinner dolphin (Stenella longirostris) resting habitat: Implications for management. J Appl Ecol 52:621–630

176

References

Valeix M, Loveridge AJ, Chamaillé-Jammes S, Davidson Z, Murindagomo F, Fritz H, MacDonald DW (2009) Behavioral adjustments of African herbivores to predation risk by lions: Spatiotemporal variations influence habitat use. Ecology 90:23–30 Vardanis Y, Klaassen RHG, Strandberg R, Alerstam T (2011) Individuality in bird migration: routes and timing. Biol Lett 7:502–505 Vikingsson GA (1990) Energetic studies on fin and sei whales caught off Iceland. Rep Int Whal Comm:365–373 Vikingsson GA (1995) Body condition of fin whales during summer off Iceland. In: Schytte Blix A, Walløe L, Ulltang Ø (eds) Whales, Seals, Fish and Man. Elsevier Science, Amsterdam, Pays-Bas, p 361–369 Villepique JT, Bleich VC, Pierce BM, Stephenson TR, Botta RA, Bowyer RT (2008) Evaluating GPS collar error: a critical evaluation of Televilt POSREC- ScienceTM collars and a method for screening location data. Calif Fish Game 94:155–168 Vincent C, McConnell BJ, Delayat S, Elder J-F, Gautier G, Ridoux V (2010) Winter habitat use of harbour seals (Phoca vitulina) fitted with FastlocTM GPS/GSM tags in two tidal bays in France. NAMMCO Sci Publ 8:285–302 Vincent C, Mcconnell BJ, Ridoux V, Fedak MA (2002) Assessment of Argos location accuracy from satellite tags deployed on captive Gray Seals. Mar Mammal Sci 18:156–166 Viswanathan G (2011) The physics of foraging : an introduction to random searches and biological encounters. Cambridge University Press, New York, USA WallisDeVries MF, Laca EA, Demment MW (1999) The importance of scale of patchiness for selectivity in grazing herbivores. Oecologia 121:355–363 Wang M, Grimm V (2007) Home range dynamics and population regulation: An individual-based model of the common shrew Sorex araneus. Ecol Modell 205:397–409 Watanabe YY, Ito M, Takahashi A (2014) Testing optimal foraging theory in a penguin–krill system. Proc R Soc B 281:20132376

177

References

Watanabe YY, Sato K, Watanuki Y, Takahashi A, Mitani Y, Amano M, Aoki K, Narazaki T, Iwata T, Minamikawa S, Miyazaki N (2011) Scaling of swim speed in breath-hold divers. J Anim Ecol 80:57–68 Wauters LA, Dhondt AA (1995) Lifetime reproductive success and its correlates in female Eurasian red squirrels. Oikos 72:402–410 Wayne RM, Shamis JD (1977) An annotated bibliography of bird marking techniques. Bird-Banding 48:42–61 White GC, Garrott RA (1990) Analysis of wildlife radio-tracking data. Academic Press, San Diego, California Whitehead H, Payne R (1978) New techniques for assessing populations of right whales without killing them. FAO Fish Ser 5:189 Williams R, Vikingsson GA, Gislason A, Lockyer C, New L, Thomas L, Hammond PS (2013) Evidence for density-dependent changes in body condition and pregnancy rate of North Atlantic fin whales over four decades of varying environmental conditions. ICES J Mar Sci 70:1273–1280 Willimont LA, Senner SE, Goodrich LJ (1988) Fall migration of ruby-throated hummingbirds in the Northeastern United States. Wilson Bull 100:203–206 Wilson RP, Griffiths IW, Legg PA, Friswell MI, Bidder OR, Halsey LG, Lambertucci SA, Shepard ELC (2013) Turn costs change the value of animal search paths. Ecol Lett 16:1145–1150 Wilson SG, Stewart BS, Polovina JJ, Meekan MG, Stevens JD, Galuardi B (2007) Accuracy and precision of archival tag data: a multiple-tagging study conducted on a whale shark (Rhincodon typus) in the Indian Ocean. Fish Oceanogr 16:547–554 Wirsing AJ, Heithaus MR, Dill LM (2007) Living on the edge: dugongs prefer to forage in microhabitats that allow escape from rather than avoidance of predators. Anim Behav 74:93–101 Witt MJ, Åkesson S, Broderick AC, Coyne MS, Ellick J, Formia A, Hays GC, Luschi P, Stroud S, Godley BJ (2010) Assessing accuracy and utility of satellite-tracking data using Argos-linked Fastloc-GPS. Anim Behav 80:571– 581

178

References

Worton BT (1989) Kernel methods for estimating the Utilization Distribution in home-range studies stable. Ecology 70:164–168 Wyneken J, Salmon M (1992) Frenzy and postfrenzy swimming activity in loggerhead, green and leatherback hatchling sea turtles. Copeia 2:478–484 Zieman JC, Iverson RL, Ogden JC (1984) Herbivory effects on Thalassia testudinum leaf growth and nitrogen content. Mar Ecol Prog Ser 15:151–158

179

Appendix A

Appendix A Supplementary information for chapter 3

Validation of the use of the 100-m depth contour as the transition point between oceanic and neritic turtle travel speeds

We validated that the 100-m contour was the transition point between oceanic and neritic speeds of travel based on the mean day/night speed of travel ratios. We first calculated the day/night ratio for turtles that only migrated through waters shallower than 100 m. We then grouped the day/night ratios for oceanic turtles into five depth categories as they travelled from waters deeper than 200 m to waters shallower than 50 m when migrating to the foraging grounds (0–50 m, 50– 100 m, 100–150 m, 150–200 m, >200 m). Subsequently we calculated a mean day/night speed of travel ratio for each category using the bootstrapping procedure described in the method section.

Loggerhead turtles that only travelled in neritic waters (<100 m deep; n = 4 turtles and 14 day/night comparisons) always exceeded a mean day/night ratio of one, with turtles travelling significantly further by day (Figure A2). Day/night ratios of both loggerhead and green turtles travelling in water >100 m deep showed no significant difference in speed of travel ratio for either species (n = 66 and 167 day/night comparisons for nine loggerhead green and five green turtles, respectively). Day/night ratio for loggerhead turtles that travelled from >200 m to <50 m showed that the mean ratio only exceeded one when they travelled within water shallower than 100 m (0–50 m, n = 17 day/night comparisons; 50–100 m, n = 10; 100–150 m, n = 8; 150 – 200 m, n = 5, >200 m, n = 53).

180

Appendix A

Figure A1. Mean day/night speed of travel ratio and 95 % confidence intervals for loggerhead and green turtles at different depths. Day/night ratio for loggerhead turtles that only travelled in neritic waters (<100 m) is shown as a grey square (n = 14 day/night comparisons using tracks from four turtles). Day/night ratios for loggerhead and green turtles that migrated through oceanic waters (>100 m) are shown as grey and white diamonds, respectively (n = 66 and 167 day/night comparisons using nine loggerhead and five green turtles, respectively), with no significant difference between day and night being detected for either of these two groups. The day/night ratio for loggerhead turtles (grey circles) that travelled from >200 m to <50 m showed that the mean ratio significantly exceeded one when they travelled in waters shallower than 100 m. Mean day/night speed of travel ratios with a black star are significantly different from one.

Sensitivity analysis for determining the maximum backtrack duration

We calculated the number of hours for which it was possible to backtrack turtles while maintaining sufficient accuracy to infer departure time. We selected three loggerhead and three green turtle tracking datasets with a large volume of accurate tracks over the first 48 h of departure from the breeding ground and for 181

Appendix A which the exact time of departure was apparent. Then, for each location beyond this initial location, we backtracked to the starting point to obtain an estimated departure time. The estimated departure time from each location was then compared against the actual departure time. During backtracking, we assumed a constant speed of travel of 1.5 km h-1 for loggerhead turtles, and 2.6 km h-1 for green turtles based on the average migratory speeds of travel calculated in the current study (excluding stopovers).

Error in the calculated departure time was less than two hours for the first 12 h of backtracking for loggerhead and the first 6 h for green turtles in 95 % of cases (Supplementary Figure A3). This difference is directly attributed to differences in mean travel speeds by the two species. Therefore, we selected threshold backtracking durations of 12 h for loggerhead turtles and 6 h for green turtles, to minimise error when calculating the departure times. The same threshold of 12 h and 6 h was used to calculate the arrival at the foraging grounds.

Figure A2: Sensitivity analysis to determine the threshold time to backtrack as a method to infer the departure time from the breeding grounds with the highest accuracy possible for: (a) three loggerhead turtles; and (b) three green turtles. Horizontal solid lines correspond to an exact departure time estimation. The dashed

182

Appendix A

vertical black lines represent the maximum backtrack time thresholds selected for calculating departure time.

183

Appendix B

Appendix B Supplementary information for chapter 4

Figure B1: Output from sensitivity analysis for turtle activity patterns, showing the net swim speed of turtles as a function of hour of day at different upper threshold values for the time period duration (hours) over which net speed was estimated. The threshold value can be seen in the top centre of each sub-figure, with the sample size (n=number of speed estimates) given just below.

184

Appendix B

Figure B2: Diurnal latitudinal (km Northings) and longitudinal (km Eastings) positions of the eight tracked green turtles in their Indian Ocean feeding grounds, as a function of time since arrival on the feeding grounds. The colour of the dots represents which diurnal site the locations have been assigned to by the Bayesian multivariate behavioural change point analysis (BCPA). The ID number of each turtle can be seen on top of each sub-figure and the sample size (n=number of locations) for each BCPA is given in the bottom centre of each sub-figure.

185

Appendix B

Figure B3: Nocturnal latitudinal (km Northings) and longitudinal (km Eastings) positions of the eight tracked green turtles in their Indian Ocean feeding grounds, as a function of time since arrival on the feeding grounds. The colour of the dots represents which nocturnal site the locations have been assigned to by the Bayesian multivariate behavioural change point analysis (BCPA). The ID number of each turtle can be seen on top of each sub-figure and the sample size (n=number of locations) for each BCPA is given in the bottom centre of each sub-figure.

186

Appendix B

Table B1: Summary table of the estimated home range sizes (km2), obtained from the Kernel Utility Distributions (KUD) of the eight tracked female green turtles on their foraging grounds in the Indian Ocean, when incorporating different levels of spatiotemporal complexity (High = Seasonal and diel movement patterns, Medium = Diel movement patterns, Low = Temporal auto-correlation only, None = Filtered raw data). All estimates, except for the raw locations, are based on one daytime and one night-time position per 24- hours.

187

Appendix C

Supplementary information for chapter 5 Selection of the loggerhead turtles for the analyses.

A total of 40 male (including 10 residents) and 17 female loggerhead turtles were tracked from Zakynthos between 2007 and 2012. We excluded 13 turtles from the analyses because the transmitter stopped before they departed the breeding sites or reached the foraging/wintering sites (n = 11) or the turtle remained in oceanic waters (n = 2). Of the 44 remaining turtles, we identified 58 foraging and wintering sites (i.e. several turtles used more than one site). We excluded sites where data were available for fewer than six days (n = 2 sites) or where <1 GPS location per day was available (n = 4 sites, excluding four turtles). In addition, one site was removed because the turtle frequented a fish farm, and another site was removed because it was primarily oceanic. This left 50 sites used by one or more of the 40 individual turtles.

For those 50 sites, when viewing the distance from home graphs, we detected two distinct trends (referred as Type A and Type B) in the magnitude of daily movement by turtles foraging and wintering in neritic habitat, i.e. <100 m seabed depth (Schofield et al. 2010) (Figure C1). The movement patterns of turtles at Type B sites significantly differed to those at Type A sites (Wilcoxon, p < 0.001). For sites with Type A patterns (n = 31 sites across 24 turtles), turtle movement was restricted to within a 5.4 ± 3.4 km distance from home (range: 1.1–17.5). For sites with type B patterns (n = 19 sites of 16 turtles), turtles moved a 26 ± 13.7 km distance from home (range: 9.5–49.4 km). Type A movement patterns encompassed all turtles that used nearshore shallow waters (i.e. <2 km from shore and to a maximum of 40 m seabed depth; Schofield et al. 2010), and so were assumed to reflect benthic foraging behaviour. In contrast, all turtles that exhibited Type B movement were in waters deeper than 40 m seabed depth and >2 km offshore, with the possibility of mid-water foraging activity (Hatase et al. 2007). Thus, we excluded all offshore sites where turtles exhibited Type B movement patterns (n = 188

Appendix C

19 sites of 16 turtles), but retained offshore sites where turtles exhibited Type A patterns (n = 8 sites). In addition, one oceanic site used by a turtle frequenting multiple sites was removed. Note that sea turtles occasionally made forays out of those Type A sites (an example is provided in Figure C2).Thus, in total, 31 foraging and wintering sites of 24 turtles were retained for our analyses (Figure C3).

Figure C1: (a) Changes in the distance to home over 60 days after arrival at the foraging site for turtles exhibiting Type A (black) and Type B (grey) foraging movement. The turtle exhibiting Type A movement remained within 5 km of its initial location. In contrast, the turtle exhibiting Type B movement moved up to 50 km from its initial location. (b) Boxplots showing the typical distance to home for Type A and Type B foraging movement patterns. The variation in the distance from home were typically smaller for Type A compared to Type B foraging sites. The boundary of the boxes indicate the 25th and 75th percentiles and the line within the boxes marks the median. The whiskers (error bars) above and below the box indicate the 90th and 10th percentiles. Black dots indicates outliers.

189

Appendix C

Figure C2: Example of a loggerhead turtle foraying out of its foraging sites located in Zakynthos (dark grey locations) up to the Peloponnese (red locations, Kotichi Lagoon). The crossing lasted two days each way and the turtle stayed at a single site on the Peloponnese for three days. Arrival and departure from the foraging site and the Peloponnese are indicated by black arrows.

Figure C3: Locations of the 31 benthic foraging sites retained for the analyses. At the retained sites, the loggerhead turtles (n = 24) were all exhibiting Type A movement patterns. Offshore sites are indicated by dark grey circles and nearshore sites by dark grey squares. Areas with a seabed depth <100 m are represented in blue.

190

Appendix C

Sensitivity analysis validating the determination of the home location.

To confirm that the first location recorded on the foraging site could be used as home, we compared how the distance from home changed when the day of arrival was artificially shifted by ±3 days. If the turtle was still migrating (Day -3 to Day - 1 before arrival), we expected the distance from home to increase rapidly until the turtle completed migration, and plateau when the turtle started foraging. Conversely, if the migration was already complete when home was defined, we would expect small variations in the distance to home (Day +1 to Day +3) corresponding to turtle foraging movement (Figure C4).

As predicted, the distance to home increased rapidly across Day -3 to Day - 1, when the turtle was assumed to be still migrating and plateaued at Day 0 when the animal was assumed to have started to forage. Furthermore, moving the home location to three days after arrival (Day 1 to Day 3) had no major impact on determining location of home, and no impact on the subsequent analyses.

Figure C4: Example of how shifting the day of arrival influences estimates of distance from home. Vertical dashed line is the actual defined day of arrival (Day 0). The

191

Appendix C

distance to home increased rapidly in the three days before Day 0, when the turtle was assumed to be still migrating and plateaued at Day 0 when the animal was assumed to have started to forage. Small variations in the distance to home (Day 0 to Day +3) correspond to turtle foraging movement.

Sensitivity analysis determining the duration required to detect the majority of patches within a foraging/wintering site.

To determine the duration of residency required to detect the majority of patches used by turtles at foraging/wintering sites, we selected 19 sites (n = 13 turtles) where the departure/arrival dates were available. Patches were identified for each site, and the cumulative percentage of patches used in function of the time elapsed since the first location at the foraging site was calculated. The 13 turtles visited 84 % (26 out of 31) of the patches in the first two months (Figure C5).

Figure C5: Cumulative percentage of patches used as a function of the time since arrival at the foraging/wintering site. Thirteen turtles visited 84 % (26 out of 31) of 192

Appendix C

the patches in the first two months (vertical dashed grey bar) after the arrival at the foraging/wintering site. Four resident turtles from Zakynthos visited a patch more than two months after the arrival at the foraging site.

Effect of incorporating the Fastloc-GPS tags cessation of transmission on calculating the site use duration.

For all male turtles where the arrival and departure from foraging/wintering sites could be determined, the first site frequented after departing the breeding area was used for a mean 167.8 ± 95.0 days (range 19.7-317.9 days; n = 9 sites of nine turtles) and subsequent sites were frequented for shorter periods (101.9 ± 51.0 days, range: 17.7-164.8; n = 7 sites of five turtles).

For males where transmissions stopped while at foraging a foraging site, mean site use duration was 138.2 ± 113.9 days for the first site (range: 18.1-369.6; n = 11 sites of 11 turtles), and 132.2 days for subsequent sites (n = 1 turtles). The cessation of transmission of turtles while at their primary sites may have slightly reduced detected site use duration, as four devices stopped during the first four months of transmission.

Slightly different results were obtained for females, with arrival and departure being recorded for just one female at the first and subsequent sites (n = 17.4 days and 62.7 days, respectively). The inclusion of females where tracking stopped while at sites raised the average use duration, but still not to the same level as that of males (first site n = 5 sites of five turtles, mean: 34.5 ± 29.0, range: 7.9- 73.8 days; subsequent sites: n = 2 sites of one turtle, mean: 59.3 ± 4.9, range: 55.8- 62.7 days). This suggests that the inclusion of the sites for which the transmitter failed may result in an underestimate of the site-use duration and thus should not be included in the analyses. Consequently, we only include turtles for which the transmitter recorded the whole period spent a given foraging site in our analyses.

193

Appendix C

Table C1: Details of the loggerhead turtles retained for the analyses. The number of sites and the number of the three types of patches (see main text for full description of the three types or patches) are reported for each turtle along with their geographic location. A dash in the three last columns indicates we were not able to determine patch use for the corresponding turtles. (Zak: Zakynthos, Tun: Tunisia, Amv: Amvrakikos, Pel: Peloponnese, Cor: Corfu) Number Number Number Turtle Nearshore Site of Type of Type of Type Year Sex Location ID / Offshore # 1 2 3 patches patches patches 2008 Male 84412 Zak Nearshore 1 - - - 2008 Male 15120 Zak Nearshore 1 0 2 0 Zak Nearshore 2 1 1 0 2008 Female 84410 Tun Offshore 1 1 0 0 Tun Offshore 2 1 0 0 2008 Male 4395 Amv Nearshore 1 1 2 0 Amv Nearshore 2 1 1 0 2009 Female 84417 Pel Nearshore 1 - - - 2009 Female Solar Pel Nearshore 1 - - - 2009 Female 93788 Adr Offshore 1 - - - 2009 Female 93790 Amv Nearshore 1 - - - 2009 Female 93791 Tun Offshore 1 - - - 2009 Male 93785 Zak Nearshore 1 1 2 0 2009 Male 93784 Zak Nearshore 1 1 0 1 Zak Nearshore 2 1 0 0 Pel Nearshore 3 0 0 1 2010 Male 95610 Pel Nearshore 1 - - - 2010 Male 95611 Pel Nearshore 1 - - - 2010 Male 4395 Tun Offshore 1 2 0 0 2010 Male 15120 Tun Offshore 1 2 0 0 2011 Male 61810 Tun Offshore 1 1 0 0 2011 Male 61809 Amv Nearshore 1 1 0 0 2011 Male 61813 Zak Nearshore 1 1 0 0 Pel Nearshore 2 2 0 0 2011 Male 21914 Cor Nearshore 1 1 0 0 2012 Male 117572 Tun Offshore 1 4 0 0 2012 Male 15120 Amv Nearshore 1 1 0 0 2012 Male 15119 Amv Nearshore 1 1 0 0 Amv Nearshore 2 1 0 0 2012 Male 117571 Pel Nearshore 2 - - - 2012 Male 117573 Zak Nearshore 1 2 0 0

194

Appendix D

Appendix D Supplementary information for chapter 6

Description of the whale.morpho function

The whale.morpho function extracts different morphological metrics from an aerial photograph of a whale. The metrics are calculated using multiple reference points placed by the user on the photograph of the body of the whale (Figure D1). The whale.morpho function’s code is available at the following URL : https://goo.gl/KomQLZ

The list of extracted metrics is:

• Rostrum.X: The distance in X between the rostrum and the bottom-left corner of the full picture (in % of total width (X) of the photo). • Rostrum.Y: The distance in Y between the rostrum and the bottom-left corner of the full picture (in % of total height (Y) of the photo). • Fluke.X: The distance in X between the notch of the tail fluke and the bottom-left corner of the full picture (in % of total width (X) of the photo). • Fluke.Y: The distance in Y between the notch of the tail fluke and the bottom-left corner of the full picture (in % of total height (Y) of the photo). • Total.length.pix: Total length of the whale (in pixels). • Length.to.blowhole.pix: The length between the rostrum and the blowhole of the whale (in pixels). • Length.to.start.of.dorsal.fin.pix: The length between the rostrum and the beginning of the dorsal fin of the whale (in pixels). • Length.to.end.of.dorsal.fin.pix: The length between the rostrum and the end of the dorsal fin of the whale (in pixels). • Length.to.start.of.fluke.pix: The length between the rostrum and the start of the tail of the whale (in pixels).

195

Appendix D

• Length.to.eyes.pix: The length between the rostrum and the line of the eyes of the whale (in pixels). • Width.at.eyes.pix: The width between the eyes of the whale (in pixels). • Width.fluke.pix: The length between the two extremities of the tail fluke (in pixels). • Width.5.proc.pix to Width.95.proc.pix: Width measurements of the body of the whale (in pixels).

Figure D1: Position of the different reference points used by the whale.morpho function to compute the morphometric measurements.

196

Appendix D

Usage whale.morpho(myimage, tiff)

Arguments myimage: Path to the picture of the whale. The picture must be JPEG or TIFF format. tiff: Set to TRUE if the picture is a TIFF. Set to FALSE if the picture is a JPEG. Examples rm(list=ls()) # clear all the files in memory ### load the packages required to run the whale.morpho function library(jpeg) library(tiff) library(raster) source("whale.morpho.R") #Run the function for a JPEG picture res <- whale.morpho(myimage = "C:/Whales/whale_picture.jpg", TIFF = FALSE) res #Run the function for a TIFF picture res <- whale.morpho(myimage = "C:/Whales/whale_picture.tif", TIFF = TRUE) res

#NOTE: When running your script, make sure not to run anything past the code line launching the whale.morpho function, otherwise the function will crash. First run your script up to the line calling the whale.morpho function, analyse your picture, and after that run the end of your script.

197