Movement ecology in pelagic seabirds Zuzana Zajková Aquesta tesi doctoral està subjecta a la llicència Reconeixement- NoComercial 4.0. Espanya de Creative Commons. Esta tesis doctoral está sujeta a la licencia Reconocimiento - NoComercial 4.0. España de Creative Commons. This doctoral thesis is licensed under the Creative Commons Attribution-NonCommercial 4.0. Spain License. Layout design by Arais Reyes Meza Original illustrations in covers of chapters 1 to 4 by Esther Charles Jordán © Cover design by Esther Charles Jordán with contribution of José Manuel Reyes-González. The copyright of illustrations, figures and text of this document rests with their authors. No quotation from this thesis and no information derived from it may be published without the author’s prior consent. This thesis was financially suported by a pre-doctoral grant (APIF/2012) from the University of Barcelona. Facultat de Biologia Departament de Biologia Evolutiva, Ecologia i Ciències Ambientals Programa de Doctorat en Biodiversitat MOVEMENT ECOLOGY IN PELAGIC SEABIRDS Memòria presentada per ZUZANA ZAJKOVÁ per optar al grau de Doctora per la Universitat de Barcelona Barcelona, 2019 Supervisor and tutor: Dr. Jacob González-Solís Bou Departament de Biologia Evolutiva, Ecologia i Ciències Ambientals Facultat de Biologia Universitat de Barcelona Supervisor: Dr. Frederic Bartumeus Ferré Centre d’Estudis Avançats de Blanes CEAB - CSIC ECOLOGÍA DEL MOVIMIENTO EN AVES MARINAS I have always liked to know birds as individuals, rather than as statistics. (…) No generalization in ecology is ever 100% valid. Somewhere, sometime, there is or will be a guillemot that plunges like a gannet and a tern that swims underwater. In an age when the computer speaks with the voice of unerring certainty, I find the unpredictable character of birds rather reassuring. They were here before us, and they will surely outlive us. Their lives have a reality and immediacy that may escape us in our increasingly secure and synthetic world. Antony J. Gaston - Seabirds. A natural history - Acknowledgements At the moment of choosing a topic for the final project of my Degree in Ecology, I choose to focus on plants and their conservation. Not only because I really like plants, the other reason which drove my decision, was that I found it pretty tough and challenging to study animals, as they are on a constant move. Who would say that much much later I would engage in a thesis to study animals, with a particular focus on an animal group with the most spectacular movements and migrations! Any of the work you are about to read, would not be possible without supervisors of this thesis, Jacob González-Solís and Fede Bartumeus. Each of them contributed in different way, related to their research background. Thanks to Jacob, I learned a lot about seabirds’ biology and ecology, his deep knowledge helped me to decipher the information coming from the small tracking de- vices. With Jacob I enjoyed my first seabird fieldwork on a remote island of Montaña Clara (Canary Islands), full of bites of Cory’s shearwaters and Bulwer petrels, camping in a stone “house” and with a private lagoon! I still remember the first time I met Fede in his office in Blanes, where I found myself staring at a white board (occupying the whole wall!) full of strange annotations. And I am glad that time later, also words like seabirds, geolocator and wet-dry started to appear in between all those strange annotations, together with a corner dedicated to a selection of visualizations of our work! I ap- preciate a lot his enthusiasm and patience and pragmatic approach to all of our discoveries. Although I need to admit that finding an intersection of the two worlds of my supervisors (and mine, too) was not always easy and straightforward. Over this journey, we discovered several dead ends, but I think that at the end we could agree that our diverse backgrounds resulted in a very fruitful collaboration. My time as PhD. student was almost equally split between the University of Barcelona and CEAB in Blanes. Starting at the University, I was lucky enough to become a part of great group of people. With all: Teresa, Vero, Laura, Fernanda and Jose we had funny times supported by regular con- sumptions of chocolate. Although the “new blood” Vir, Marta and Laura entered more less in the time I moved to Blanes, they greatly matched the mood of the group. And of course, it would have not been the same without having Francesc, Irene, Eli, Manolo, Mario, Alberto and Oriol around! i For the second part of my PhD. I moved to Blanes to work at CEAB. During almost two years I shared the office with an amazing person and polychaetes specialist, João. It was a unique op- portunity for me to get deep insights about all those “worms”, not even talking about discovering all the amazing music of the world. I miss our regularly irregular beer-times in Casino, together with Nixon and Jose! I would like to thank to the members of my PhD. evaluation committee, Maite Louzao, Robin Freeman and Raül Ramos, for their contributions and comments to my work developed especially over the first years of my doctoral studies. I am thankful to all co-authors of the work contained in this thesis. Particularly, to Prof. Peter Beck- er, for giving me the possibility to contribute to their study by adding the analysis of activity data of Common terns, becoming the first published article of this thesis. During the third year of my PhD. I had a great pleasure to spend two months at the lab at Max Planck of Ornithology in Radolfzell, lead by Dr. Kamran Safi. I received a great welcoming from the lab and I learned a lot, especially with the help of Kami, Bart and Anne, who helped me not to “re-invent the wheel” over and over again. How I enjoyed all the misty autumn days and nights, bike rides and walks around the Konstanz lake and the city. And of course the happy hours in nice company and talks and cakes! Any of the analysis and visualizations would not be possible without the community of R enthu- siasts. I am greatly doubted to all developers of plenty of R packages which I have used over the years. Here I want to thank to Joan Garriga, the man behind the R package used in the last chapter of this thesis. Thank you for your patience while teaching me how to use cluster computing to run all those immense calculations, with the objective to find out what the animals were doing. My big thanks goes to Mara and Xavier, who believed that my skills which I developed focused on studying complex seabirds’ behaviour were applicable also in non-academic data science proj- ects with totally different focus. Although I need to recognize that combining a full time job as a data scientist while finishing the PhD. was probably not the best idea ever, causing that things like “free time” somehow vanished from my vocabulary. Thank you both, together with David and Leandro, for being supportive and interested in my parallel academic journey, especially in the last moths. I am lucky enough to have around people supporting me all the time to achieve my goals. All the friends which I have met at different space-times, some of them remain spatially close, many spread all around the world. Thank you Matias, Berta, Petr for our great adventures on the walls and in the mountains. Thank you Holger and Arais for always being around and for great jazzy parties at your place. Thank you Hugo for your coaching sessions, while hiking or searching for birds. Thank you Katka and your family for our regular winter wine meetings, Zuzka and the Vin- ii klovci for being always close. And of course thank to Mazúrovci and Šutekovci for all the great time! Maťko, thank you for encouraging all my decisions and steps for all those years, for always being around when needed. Most probably, there wouldn’t be any thesis without our decision to start a new adventure in Spain. How many things we have learned and how many peaks and mountains have we climbed together here around! Looking forward for the next one. Ďakujem moc. The huge thank you belongs to my great family, mum and dad, who always supported my deci- sions, my brother and sister for always being around. And even new members of the family, Mario and small Liam! Ďakujem, mamina a ocino, Miro, Dada, za vašu podporu počas celých týchto bláznivých rokov, keď sme sa aj kvôli tejto práci videli pomenej, ale zato zas oveľa intenzívnejšie. And last, but not least, to you, Jose, thank you for all. I would probably need all the pages of this thesis to write down everything I would like to mention to express my great gratitude for all your help during the process of this thesis. But as you know, it would probably take me years, there- fore, thank you, my darling. Arenys de Mar, Barcelona, September 2019 iii ABSTRACT Movement is a fundamental component of behaviour and thus both are inextricably linked, so variation in movement patterns usually reflects different behaviours. The way individuals allo- cate time budgets to different behaviours within circadian rhythms and over the annual cycle will ultimately provide knowledge about evolutionary processes and adaptive capacity, which is also important to proper conservation actions of endangered species. Seabird movements have been studied over the last 20 years with the wide deployment of geolocator-immersion loggers, but wet-dry data seem underused according to literature published. Along 4 chapters this thesis presents novel insights about movements and behaviour of 4 little-known seabird species from the Atlantic Ocean: Boyd’s shearwater (Puffinus boydi), Common tern (Sterna hirundo), Atlantic pe- trel (Pterodroma incerta) and Cory’s shearwater (Calonectris borealis). Using wet-dry data alone or combined with positional data we uncovered the timing of major life cycle events and revealed circadian and circa-annual activity patterns of such species. In highly mobile migratory seabirds, the existence of radically different behavioural contexts linked to phenology and the need to exploit different marine environments over the year lead to different behavioural budgets. In the last chapter, we present a new analytical protocol based on state-of-the-art algorithms to decipher behaviours from wet-dry data. We reveal the hierarchical and modular nature of seabird behaviour at an unprecedented level of detail and used cutting-edge data visualization to high- light key insights. Our framework paves the way to use behavioural annotation for addressing old and new questions of interest in ecology from new perspectives using geolocator-immersion sensors. Overall, through this thesis I highlight the irreplaceable utility of wet-dry data to get unique insights in ecology and behaviour over the annual cycle of seabirds, a difficult-to-observe group of birds that remain out of the human sight most of their life. Geolocator-immersion sen- sors continue to be the most extended loggers to track year-round movements of seabirds, since they ensure the welfare of tagged individuals. Therefore, the results compiled in this thesis should encourage researchers to incorporate the use wet-dry data within hypothesis-driven frameworks, which surely would contribute to increase our knowledge of seabird ecology at sea. v RESUMEN El movimiento es un componente fundamental del comportamiento animal, de forma que vari- aciones en los patrones de movimiento reflejan cambios comportamentales. El tiempo destinado a los diferentes comportamientos según los ritmos circadianos y a lo largo del ciclo anual puede ayudar a entender los procesos evolutivos y la capacidad de adaptación, algo importante tam- bién de cara a la conservación de especies amenazadas. La ecología de las aves marinas en mar abierto ha sido ampliamente estudiada en las dos últimas décadas gracias a los geolocalizadores por niveles de luz. Muchos modelos de geolocalizador registran datos de conductividad en agua salada, pero esta información parece infrautilizada a la luz de la literatura publicada. Esta tesis aporta nuevos conocimientos sobre la ecología en mar abierto de 4 especies de aves marinas del océano Atlántico: la pardela chica de Cabo Verde (Puffinus boydi), el charrán común (Sterna hirun- do), el petrel atlántico (Pterodroma incerta) y la pardela cenicienta (Calonectris borealis). En esta tesis, usando los datos de conductividad solos o en combinación con datos de posicionamiento, desvelamos con detalle la fenología y los ritmos circadianos y anuales en el comportamiento de las aves marinas. En especies migratorias, la exposición a contextos diferentes a lo largo del año conduce a diferentes patrones comportamentales. En el último capítulo presentamos un nuevo protocolo analítico basado en datos de conductividad. Gracias al uso de algoritmos de aprendiza- je automático desgranamos el comportamiento de las aves marinas a un nivel sin precedentes, desvelando su naturaleza jerárquica y modular. En conjunto, esta tesis remarca la enorme utilidad de los datos de conductividad para estudiar los patrones comportamentales a lo largo del ciclo anual en las aves marinas, un grupo animal difícil de observar al pasar la mayor parte del año en mar abierto. Los geolocalizadores provistos de sensor de conductividad siguen siendo los únicos aparatos de seguimiento remoto que aseguran el bienestar de las aves marinas instrumentadas durante largos periodos de tiempo. Los resultados expuestos en esta tesis deberían promover un mayor uso de los datos de conductividad, lo que contribuiría a aumentar nuestro conocimiento sobre la ecología de las aves marinas. vii INDEX Acknowledgements Abstract Resumen INTRODUCTION Preamble General introduction Objectives and structure of the thesis Supervisors’ Report CHAPTER 1: Year-round movements of a small seabird and oceanic isotopic gradient in the tropical Atlantic CHAPTER 2: Common Terns on the East Atlantic Flyway: Temporal-spatial distribution during the non-breeding period CHAPTER 3: Spatial ecology, phenological variability and moulting patterns of the endangered Atlantic petrel, Pterodroma incerta CHAPTER 4: Air-water behavioural dynamics reveal complex at-sea ecology in global migratory seabirds GENERAL DISCUSSION CONCLUSIONS BIBLIOGRAPHY ANNEXES ANNEX 1: Published articles ANNEX 2: Stable isotope analysis ANNEX 3: Outreach contributions i v vii 1 2 15 17 20 48 94 130 182 200 202 217 247 255 PREAMBLE This thesis is intended to provide new knowledge about movement and behaviour of seabird species, both for a better understanding of their ecology at sea and to bring new insights that hopefully can contribute to the conservation of this avian group. This thesis encompasses several aspects of seabird ecology at sea along 4 chapters focused on different species. Each of them was written as a self-contained piece of research and thus can be read and under- stood independently. Presenting them together I wanted to highlight how a very basic source of data, the wet-dry data recorded by geolocator-immersion loggers, can provide unique in- sights in behavioural and ecological studies. This data can reveal important aspects of move- ment and behaviour within the life cycle of seabirds, a difficult-to-observe group of birds as they remain out of the human sight most of their life. 2GENERAL INTRODUCTION Animal movement represents the continuous succession of locations of an individual over time. Movement is a feature governing different biological processes, from individuals to populations and communities, since every individual moves to engage in a variety of behaviours that determines fit- ness and ultimately population dynamics (Nathan 2008, Damschen et al. 2008, Jeltsch et al. 2013). Therefore, in vagile species, movement is a fundamental component of behaviour and thus both are inextricable linked. Behaviours mediated by movement represent the way animals react to internal and external stimuli, serving as mediators between the environment and individual fitness (Nathan et al. 2008). Indeed, variation in movement patterns usually reflects different behaviours, includ- ing those most glaring, such as foraging, dispersal, migration, social interaction, mate search or escaping from predators (Sutherland et al. 2013). Behavioural strategies, i.e. fine-scale behaviours that animals usually display interrelated, are the result of evolutionary processes exhibited in a population in certain environmental conditions, since behaviours evolve to maximize the fitness of individuals (Van Buskirk 2012). Therefore, gathering information on movement and behaviour is fundamental for assessing how individuals and populations react in different environments. The way animals allocate behavioural budgets within circadian rhythm and over annual life cycles may ultimately provide knowledge about evolutionary processes and adaptive capacity to face changes in the environment (Sih et al. 2010, Wong & Candolin 2015). Studying movement and behaviour over a wide range of scales and contexts, and from a multidi- mensional perspective, is a critical step to understand behavioural strategies and their relationship with environmental conditions. This has remained challenging and for a long time mainly restricted to description throughout focal observation bouts, thus leading to different degree of subjectivity and severely time constrained (Altmann 1974). Fortunately, during the last two decades the ad- vances in remote tracking technologies have revolutionized the study of movement and behaviour of wild animals, even promoting the rise of a new research framework, the movement ecology (Nathan et al. 2008, Nathan & Giuggioli, 2013, Kays et al. 2015). The recent advent of biologging, i.e. the use of animal-borne sensors (Boyd et al. 2004, Cooke et al. 2004, Rutz & Hays 2009, Ropert- Coudert & Wilson 2005, Ropert-Coudert et al. 2012) provided the required bypass to address behavioural questions. Nowadays, we can find a full assortment of wearable devices fitted with multi-sensors, capable of recording not only location but a diverse array of ancillary data with unprecedented detail over a wide range of temporal and spatial scales (e.g. Wilmers et al. 2015, Chmura et al. 2018). The difficulty to observe and study movement of wild animals in the marine environment had precluded addressing questions about behaviour in the context of their natural history-life and con- servation. In this sense, the advances in tracking technologies have brought tremendous insights into the study of elusive marine megafauna (Block et al. 2011, McIntyre 2014, Roncon et al. 2018, Harcourt et al. 2019). In species such as sharks, tuna, cetaceans or seabirds, individuals can travel thousands of kilometers across ocean basins year round, playing a major role in the energy bal-ance and providing important goods and services across marine ecosystems (Tavares et al. 2019), and thus, biologging becomes essential for their study (Block et al. 2011, McIntyre 3INTRODUCTION 2014, Roncon et al. 2018, Harcourt et al. 2019). In the particular case of seabirds, their suitability as model species and the possibility to address many questions throughout biologging have led to a bloom in seabird research, including a proliferation of methods and analytical techniques, which has ultimately led to enhance our understanding of their ecology. SEABIRDS AS MODEL SPECIES Seabirds represent a diverse and polyphyletic avian group comprising several different families with complex natural life-history traits. Across seabird species, many ecological traits shaping their life style are shared, including extended immaturity, long-life and high adult survival, social and mostly sexual monogamy, low reproductive rates, small clutch size and an extended incubation and chick-rearing periods. Moreover, most seabirds are top-predators occupying the upper trophic level in marine and coastal food webs. But above all, what mainly represents seabird life-style is their high dependence on the marine environment for most part of their annual life cycle. Indeed, the most pelagic species spend most of their life at open sea and only come to land for breeding, as they need solid ground to lay the egg (Gaston 2004). Living in the oceans imposes various constraints on seabird life style. Seasonal patterns and an- nual variations in climatic events shape highly dynamic environments in terms of productivity. Moreover, resources appear patchily distributed and are usually little predictable in space and time at medium or fine scale, thought they could be predictable at larger scales as also depend on static oceanographic features (e.g. shelf slopes, coastline shape, sea mountains) (Weimerskirch 2007). A singular trait of seabirds that has evolved to cope with these constraints is an extraordinary move- ment capacity. Flight performance and wing shape of many seabirds, especially the most pelagic species, allow them to fly over vast distances in relatively short times (Hertel & Balance 1999). In fact, seasonality in marine environments leads many species to perform long migratory move- ments year round, even moving between different ocean basins and across hemispheres to live in an “endless summer” in search of more abundant resources (Shaffer et al. 2006, González-Solís et al. 2007, Egevang et al. 2010). Moreover, even during the breeding period when seabirds become central-place foragers, after every visit to land they need to fly far away off-shore to forage (Phillips et al. 2017). Altogether, seabirds represent a particular case of free-range marine top-predators, as their movement and behaviour year round are severely constrained by their own phenology and the marine environment seasonality. Lastly, it should be remarked that seabird conservation is a global concern, as they are one of the world’s most rapidly declining vertebrate groups because of the effect of human activities (Croxall et al 2012, Dias et al. 2019). On land, seabirds suffer from the introduction of invasive predators (cats, rats, mice), poaching, human disturbance, habitat loss and light pollution (Croxall et al. 2012, Rodríguez et al. 2019). At sea, seabirds are threatened by fishing activities, by-catch, habitat degra- dation, pollution and climate change (Rodríguez et al., 2019, Díaz et al. 2019). Marine ecosystems are also recognized as globally threatened (Halpern et al. 2007). In this context, seabirds are some- times considered as bio-indicators of marine ecosystem’s health (Furness & Camphuysen 1997). 4Nevertheless, their use as bio-indicators could improve with a comprehensive understanding of their behaviour at sea (Durant et al. 2009). Hence, every new insight about movement and behaviour of seabirds can greatly contribute to improve conservation and management actions of these species and their habitats (Croxall et al 2012, Lascelles et al. 2016, Dias et al. 2019). STUDYING SEABIRD MOVEMENT AND BEHAVIOUR A diverse array of tracking devices and tools have been used to study movement and behaviour of seabirds. The following is a brief introduction to the two main tools that have primarily contributed to the development of this thesis. I succinctly introduce them to ease understanding of their usage in the research shown in the next chapters. Geolocation-immersion loggers The study of year-round movement of pelagic seabirds at sea has been addressed over the last 20 years with the wide deployment of light-level geolocation loggers (Global Location Sensing units, GLS) (Burger & Shaffer 2008, Wilson & Vandenabeele 2012). These miniature archival data log- gers, also known as solar geolocators, measure the ambient light in a regular schedule (measuring and recording resolution depend on the models) together with a time stamp in Greenwich Mean Time (GMT). Light records allow for determining latitude and longitude on a daily basis using astronomical algorithms (Wilson et al. 1992), but the method fails to infer latitude properly around the equinoxes and provides an average accuracy of 186 ± 114 km, being the error greater towards the equator (Phillips et al. 2004). Despite this low spatio-temporal resolution and the lack of reli- ability in latitude during the equinoxes (Hill & Bran 2001), GLS accuracy has been enough to reveal spectacular migrations and non-breeding areas of many seabird species (e.g. Shaffer et al. 2006, González-Solís et al. 2007, Egevang et al. 2010). Low spatio-temporal resolution is compensated by the size and weight of these devices (currently even < 1 g) and a long battery life (usually > 1 year) that allow for long tracking periods. The increasing miniaturization of GLS currently allows researchers to track also ever smaller species (e.g. Egevang et al. 2010, Quillfeldt et al. 2013, Pollet et al. 2014, Ramos et al. 2015). Despite other kind of devices are available, GLS still are the single devices allowing the study of spatial ecology over long periods of time (Wilson & Vandenabeele 2012, López-López 2016). GPS loggers also need to be recovered to download the data but record high resolution spatio-temporal data. However, they have short battery life and are usually attached with TESA tape to back feath- ers, resulting in short-time attachment (days to few weeks), although technological improvements are rapidly cutting the distance between GLS and GPS loggers. Other devices, such as PTT, GPS- PTT or GPS-GSM, are usually equipped with solar panels and thus have no battery life limitations. Moreover, information can be retrieved remotely without the need of recapture. However, to track animals over long periods, these devices require long-lasting attachment systems such as harness, known to cause severe damage when used in some seabird species (Mallory & Gilbert, 2008). GLS, in contrast, are attached on a plastic or metallic ring placed on the tarsus of the birds, enabling even 5INTRODUCTION multi-year deployments and large sample sizes, with no detrimental effect observed on behaviour and fitness of birds (Igual et al. 2010, Vandenabeele et al. 2011, but see Brlík et al 2019). Therefore, GLS currently remain as the most cost-effective balanced tracking devices to get insights into the spatial ecology and movement of pelagic seabird species over the entire annual cycle while ensuring the welfare of tagged individuals (Vandenabeele et al. 2011, Vandenabeele et al. 2012, Kürten et al. 2019). Together with ambient light intensity and time stamp, some models of GLS, frequently called geolocation-immersion loggers, also register wet-dry data. These devices detect conductivity be- tween the anode and cathode terminals, which occurs in contact with saltwater, recording the time immersed. These data have been commonly used to study activity as a proxy of seabird behaviour (Wilson et al. 1995, Gutowsky et al. 2014). That is why in the literature wet-dry data from GLS may be also referred as saltwater immersion data or activity data. It is appropriate to mention here that high-resolution accelerometers can provide highly detailed behavioural information, particularly when combined with GPS devices (Cianchetti-Benedetti et al. 2017, Yoda 2019). However, they have similar disadvantages as those commented before for GPS, namely high energy and data-memory consumption, which precludes their use for extended periods of time (Yoda 2019). Some recent multi-sensor devices equipped with solar panels can store accelerometer plus GPS data over long periods and allow for remote downloading (Bouten et al. 2013), yet they also require harness for a long-lasting attachment, impeding their use to study at-sea behaviour of pelagic seabirds over long periods of time. Data Visualization According to the fast development in technology over the last two decades, the amount of move- ment and behaviour data collected from wildlife species have rocketed (Kays et al. 2015; Hays et al. 2016). Tracking data often provide enough information to immediately identify ecological insights, such as migratory pathways or home ranges of the species. However, as data size increases, more advanced graphics to unveil complex patterns are needed. In this sense, analysis of movement and behavioural data, as rich quantitative information, are an appropriate target to take advantage of data visualization tools. Effective data visualization can assist research bidirectionally, that is, as a knowledge-discovery tool helping to rise new hypotheses throughout exploration, or the other way around, helping to interpret results in the context of expectations and previous hypotheses (Tukey, 1977). Therefore, data visualization becomes an invaluable tool to assist research in movement ecol- ogy and animal behaviour. STRUCTURE, USAGE, AND INSIGHTS FROM WET-DRY DATA Studying how individuals allocate their time budget to different behaviours allows for better inter- preting behavioural strategies within circadian rhythm, over annual life cycles and in different en- vironments and conditions (Phillips et al. 2017). As commented above, GLS models used for seabird research also measure conductivity in saltwater, which have been used to infer activity patterns. The 6way GLS record wet-dry data varies according to models. In some of them, such as those initially produced by the British Antarctic Survey (Fox 2010), the default schedule stored data in 10-minutes blocks, where samples taken each 3s tested for contact with saltwater. This resulted in values rang- ing from 0 to 200 for each block (0 being dry the entire block, 200 being wet the entire block). Later models of GLS, such as those manufactured by Biotrack Ltd., store wet-dry data in a more continu- ous way by registering the time stamp of every change of state (wet to dry and vice versa). Lastly, most recent models (e.g. those models manufactured by Migrate Technology Ltd.) offer a variety of schedules to record wet-dry data, some of them matching schedules of previous models and thus being more frequently used by researchers. The advent of GLS completely changed the way seabirds are studied. I carried out a systematic literature review to evaluate the use of GLS in seabird research over time but specially to evaluate the use of wet-dry data so far. Details about this review are included in Box 1. Once performed this review, the first evidence was that wet-dry data and activity patterns can provide useful insights on a variety of dimensions of seabird ecology, hard to obtain otherwise for elusive species. For example, in combination with positional information, wet-dry data have been used to define major important events over the breeding period. As a case in point, Militão et al. (2017) inferred the first visit to the colony and the duration of incubation stints in the endangered Cape Verde petrel (Pterodroma feae) by this means. In many seabird species, the arrival to breeding areas usually coincides with the equinox period, when positional data from GLS is unreliable and in such circumstances wet-dry data can help estimating the arrival date to the breeding colony. Regarding the breeding period, wet-dry data have also been used to compare at-sea activity patterns between successful and failed breeders (Catry et al. 2013, Ramos et al. 2018, Ponchon et al. 2019). Many researchers have com- monly aggregated wet-dry data to assess the proportion of total time spent on water/in flight, and then looked at the variability between stages of the annual life cycle and across different groups, such as sexes (e.g. Pinet et al. 2012, De Felipe et al. 2019), ages (Catry et al. 2011, Missagia et al. 2015, Clay et al. 2018), or natal origin (e.g. Catry et al. 2011). These approaches usually evaluated circadian and circa-annual at-sea activity rhythms based on daylight/darkness activity (Phalan et al. 2007, Dias et al 2012), some of them also considering the effect of the moonlight (e.g. Yamamoto et al. 2008, Ramos et al, 2016). The proportion of daylight and darkness activity has also been used to calculate a night-flight index (Dias et al. 2012, Ramos et al. 2015, 2016). Some authors extended the use of wet-dry to broadly quantify foraging effort, estimating the number of landings/take-offs on hourly or daily time-scales (e.g. Phalan et. 2007, Mackley et al. 2010, Dias et al. 2012, Dias et al. 2016, Rayner et al. 2012). Some other approaches assumed wet-dry states and their alternation to be representative of basic behavioural modes. For example, some authors classified wet-dry data structured in ‘0-200’ schedule into three modes based on a simple threshold: ‘sitting on water’ (as representing resting or drifting, values from 195 to 200), ‘probable foraging’ (5-195) and ‘flying/ roosting’ (0-5) (McKnight et al. 2011, Mattern et al. 2015). Guilford et al. (2009) used unsupervised clustering to infer those same three modes from 0-200 wet-dry data but estimated on a daily basis (i.e. assigning each day to a unique most probable mode). Wet-dry data from more recent GLS models, stored in a continuous way, have been used in similar way to calculate the duration and number of wet-dry changes as indicative of landing and take-off rate (Catry et al. 2004, Shaffer et 7INTRODUCTION al. 2001). Finally, few studies have actually taken advantage of transitions in wet-dry data recorded in continuous schedule to discern between foraging, flight and sitting on water (Dias et al. 2012, Gutowsky et al. 2014, Ponchon et al. 2019). More advanced approaches combining wet-dry data with information from other devices (GPS, time-depth recorders) have been used within a supervised machine learning framework to carry out behavioural annotation, although classifying again the three basic modes (‘flight’, ‘sitting on water’ and ‘foraging’, see Dean et al. 2012). As shown above, a considerable research has been undertaken to uncover seabird behaviour from several perspectives. However, the second notable conclusion from the literature review was that wet-dry data seem underused, since only 53% of papers using GLS also made use of wet-dry data (see Table B1 in Box 1). Moreover, I found some less attended topics and remarkable gaps. Within these topics, the increase in the use of wet-dry data would probably contribute to provide important insights. In the following I comment some of the topics where I believe wet-dry data is clearly un- derused in seabird ecology studies: • There are many seabird species with medium to small body size whose basic information re- garding movements and behaviour at sea are unknown. Nevertheless, the progressive minia- turization and extensive use of GLS is allowing to fill this gap gradually. Lack about this basic knowledge also exists in medium-small sized species from tropical distribution, as a clear bias towards species with boreal and Antarctic/sub-Antarctic distribution has prevailed among re- searchers. • Behaviour inferred from wet-dry data can easily enrich positional data to study the seasonal timing of life-history events (i.e. phenology). As major events over the annual cycle shape behaviour, the latter could easily inform the onset and duration of every event. For example, not only migratory schedules but also important events such as incubation shifts, duration of incubation stints, or hatching data, may be potentially inferred from a careful inspection or ap- propriate visualization of wet-dry data. This is particularly useful in species breeding in remote locations where on land recurrent nest monitoring may be difficult. However, wet-dry data have been underused for this topic, since only 15% of articles using wet-dry data evaluated aspects related to phenology at some extent. • It is likely that physiological changes shape activity and behavioural budgets. In regards with physiology but also phenology, feather moulting is probably one of the most important pro- cesses that can constrain seabird behaviour and movement. However, and surprisingly, this ef- fect has been rarely studied using wet-dry data (Cherel et al. 2016), despite once again a careful inspection or appropriate visualization of wet-dry data can greatly assist research in this regard. I found only 8% of articles using wet-dry data to relate at some extent with this topic, and many of them did not addressed the issue explicitly. • Investigating the causes and consequences of movement and behaviour should consider carry- over effects, that is, how an individual previous experience explains its following performance 8BOX 1: The use of geolocator-immersion loggers in seabird ecology research: a literature review I carried out a systematic literature review to evaluate the use of light-level geolocators to study seabirds. I also evaluated to which extend researchers have taken advantage of wet-dry data from geolocation-immersion models to explore aspects of seabird ecology in more detail. I performed a search of published research articles using the Web of Science (WoS, Thomson Reuters & Clarivate Analytics). Using WoS I searched in ISI Web of Knowledge (WoK) and Zoological Record databases, filtering to report only peer-reviewed journal articles and trun- cating the search on 31 December 2018. I searched in both databases since I detected that some published articles on the issue were not included in WoK. At first, I used the following query, which I referred to as “spatial” for defining topics: TS = (seabird* AND (*geolocat* OR GLS)) This search provided 224 published papers in which light-level geolocators were broadly used to produce positional data and investigate seabirds’ spatial distribution in different ways. Once gathered, I compiled titles and abstracts of the articles from the search and used Natural (O’Connor et al. 2014). Behavioural performance can be directly evaluated from wet-dry data. However, little research has been carried out in this sense: only 3% of articles using wet-dry data addressed this topic (Catry et al. 2013, Schultner et al. 2014, Shoji et al. 2015, Fayet et al. 2016, Ramos et al. 2018). • Wet-dry data have been mostly used to investigate foraging: 39% of the articles using wet-dry data addressed this topic. Some of them, as commented above, went beyond wet-dry states and identified three behavioural modes, namely foraging, flying and sitting on water. In the litera- ture review I only found one article using solely wet-dry data for behavioural annotation (also called behavioural classification, Guilford et al. 2009). Apart from inferring the same three behavioural modes, Guilford et al. (2009) based their analytical procedure on a predefined 24 h window upon which they aggregated the data. Therefore, virtually none of the articles pub- lished to date has intended to identify a greater array of behaviours using solely wet-dry data neither considering the natural temporal sequence of wet-dry events. Certainly, there is still room to take advantage of wet-dry data, a source of information that has been available to researchers since more than a decade ago, but still clearly underused. Therefore, along this thesis I generally aimed to contribute with new insights and tools based on wet-dry data to fill these gaps. 9INTRODUCTION Language Processing algorithms to create a text corpus. I calculated the most frequent words in the corpus as a proxy of most frequent topics and explored results with data visualization tools to evaluate the topics’ representativeness (see Fig. B1 and B2). Fig. B1: Top 10 ranking of words by their frequency in published articles about seabird research using light-level geolocators. Fig. B2: Word cloud of the 100 most frequent words computed from text included in titles and abstracts of published articles related to seabird research using light-level geolocators. The size relates to the fre- quency of each term in the whole dataset, so the visualization depicts the most frequent topics. 10 Topics as behaviour, activity or wet/dry were not included in the top 10 ranking (Fig. B1), but were present in the text content (Fig. B2). Next, I wanted to quantify the extent in the use of geolocators together with wet-dry data to investigate activity and behaviour of seabirds. To do so, I searched for published articles using the following query, which I referred to as “activity” topic: TS = (seabird* AND (*geolocat* OR GLS) AND (activit* OR wet OR dry OR “wet-dry” OR at-sea behaviour)) This second search provided 120 articles, all of them already contained in the first “spatial” query results. Last, I wanted to evaluate the number of articles published addressing some specific topics in which I thought wet-dry data could greatly contribute to foster seabird ecology research. These topics were “foraging”, “phenology”, “moult” and “carry-over”, which I used in the following queries: TS = (seabird* AND (*geolocat* OR GLS) AND (activit* OR wet OR dry OR “wet-dry” OR at-sea behaviour) AND (foraging)) TS = (seabird* AND (*geolocat* OR GLS) AND (activit* OR wet OR dry OR “wet-dry” OR at-sea behaviour) AND (seasonal* OR phenolog*)) TS = (seabird* AND (*geolocat* OR GLS) AND (activit* OR wet OR dry OR “wet-dry” OR at-sea behaviour) AND (moult)) TS = (seabird* AND (*geolocat* OR GLS) AND (activit* OR wet OR dry OR “wet-dry” OR at-sea behaviour) AND (carry-over OR carryover)) I also looked at the topic of using wet-dry as data source for behavioural annotation (also re- ferred as behavioural classification): TS = (seabird* AND (*geolocat* OR GLS) AND (activit* OR “wet” OR “dry” OR “wet-dry” OR “at-sea behaviour”) AND (behav* NEAR annotation* OR behav* NEAR classif*)) The number of papers in each topic and relative representativeness are presented in Table B1. Note that a same article may be related with various topics. The bloom in the number of seabird articles over the last two decades thanks to the use of geolocator-immersion loggers is evident; the temporal trend of each topic is shown in Fig. B1. However, the last two years the number of articles has decreased (see Fig. B3). 11INTRODUCTION Table B1. Number of articles (n) found in the systematic literature review for each topic selected. The column Proportion represent the percentage respect to the total, and this total corresponds to the topic “Spatial” (i.e. n=224). The column Proportion respect to activity represents the percentage of each topic respect to the number of articles in the topic “Activity” (i.e. n=119). Recall that a same article may be represented in more than one topic, so the values in proportion columns sum more than 100. Topic Number of articles Proportion Proportion respect to “activity” Spatial 224 - - Activity 119 53.1 - Foraging 87 38.8 73.1 Phenology 34 15.2 28.6 Moult 18 8.0 15.1 Carryover 6 2.7 5.0 Behavioural annotation 1 0.4 0.8 Fig. B3: Temporal trend of articles published related to each of the topics considered. The stacked area plot highlights that wet-dry (activity) data are underused for some of these topics. 12 Boyd’s shearwater (Puffinus boydi, Order Procellariiformes, Family Procellariidae) is part of the Little−Audubon’s shearwater complex (Puffinus assimilis−lherminieri) which encompass small- sized (140 - 290 g) dark-and-white shearwaters of pelagic habits and spread within tropical and temperate waters. Mainly due to their morphologic similarities, there has been a lot of controversy in the taxonomy and the species within this complex have been assigned to different taxa within ‘assimilis’ and ‘lherminieri’ groups (Cramp & Simmons 1977, Warham 1990, Carboneras 1992, Brooke 2004). The Boyd’s shearwater is endemic to the Cape Verde Islands, where breeding sites are thought to be located on most islands and islets (Hazevoet 1995). Population is estimated of ca. 5 000 pairs (BirdLife International 2015). The species is thought to breed on most islands and islets of the archipelago, nesting in burrows in soft soil or in rocky cavities. Birds take a long breeding sea- son (ca. 6 months), that generally starts during the boreal winter. After the females lays one single egg, both parents share breeding duties (Carboneras et al. 2016). Its diet is mostly based on squid and small pelagic and demersal fish (Neves et al. 2012, J. A. Ramos et al. 2015). Due to unsolved taxonomic status, the conservation status of taxa is ‘Least Concern’ (BirdLife International 2015, Carboneras et al. 2016). Cory’s Shearwater (Calonectris borealis, Order Procellariiformes, Family Procellariidae) is a large-sized shearwater (605-1060 g) of pelagic habits. The species breeds on islands and islets of the Macaronesian archipelago, in the North-East Atlantic Ocean. Females lay one single egg on late May-early June, in simple nests located inside burrows. Chicks fledge between late October and early November. Individuals migrate to several wintering areas in the southern Atlantic Ocean, returning early in February to the breeding colonies. Their diet is based mainly on epipelagic fish and squid (Reyes-González & González-Solís, 2016). It is considered as not globally threatened by the IUCN (‘Least Concern’, BirdLife International 2018a). Species Study colony Order Procellariiformes Boyd's shearwater (Puffinus boydi) Cory's shearwater (Calonectris borealis) Atlantic Petrel (Pterodroma incerta) Order Charadriiformes Common Tern (Sterna hirundo) Raso Is., Cima Is. (Cape Verde) Gran Canaria Is. (Canary Islands) Gough Island (Tristan da Cunha) Wilhemshaven (Germany) STUDIED SPECIES AND FIELDWORK SITES The work included in this thesis refers to seabird species from two different orders: Procellari- iformes and Charadriiformes (Table 1), with studied colonies spread over different sites within the Atlantic Ocean and ranging from tropical to temperate to sub-Antarctic water distribution ranges (Figure 1). Table 1: Studied species and studied colonies. 13INTRODUCTION Fig. 1: Map showing the location of the breeding colonies (black triangles) of the seabird populations studied in this thesis. The Atlantic petrel (Pterodroma incerta, Order Procellariiformes, Family Procellariidae) is a medium-sized (420 – 720 g) pelagic gadfly petrel (Pterodroma spp.). The species distribution is restricted to the South Atlantic Ocean (Enticott 1991, Orgeira 2001, Cuthbert 2004), with breeding sites located exclusively in the Tristan da Cunha group of volcanic islands, to which the species is endemic as breeder. Birds breed during the austral winter, females lay a single egg in a burrow in June-July, and chicks fledge in December (Richardson 1984, Cuthbert 2004). Diet is composed basically of squid (Klages & Cooper 1997). The population of Atlantic petrels is approximately 1 14 million pairs, breeding at Gough Island, but the strong population decline caused by the high rate of chick predation by introduced house mice (Mus musculus) has led to list the species as ‘Endan- gered’ by the IUCN (BirdLife International 2018b, Caravaggi et al. 2019). Common Tern (Sterna hirundo, Order Charadriiformes, Family Laridae) is a small-sized sea- bird (120–135 g). The species’ breeding distribution comprises the Palearctic and North America. It breeds in a variety of coastal and inland habitats (sand beaches, marshes, rocky islands, even nesting successfully on artificial nesting platforms), laying 2-3 eggs between April-June. Terns are considered inshore feeders since generally feed in waters not far from their colonies or roosting sites in their wintering grounds. They mostly feed on small fish, occasionally also on crustaceans and insects (Granadeiro et al. 2002, Bugoni et al. 2004). Common terns are strongly migratory, travel- ling to the southern hemisphere to winter. The species is considered as not globally threatened by the IUCN (‘Least Concern’) with a global population of 1 600 000–3 600 000 individuals (Gochfeld et al. 2019). 15 OBJECTIVES AND STRUCTURE OF THE THESIS The main aim of this thesis was to provide new insights into the factors shaping the movement and at-sea behaviour of pelagic seabirds, highlighting the utility of wet-dry data from geolocator-im- mersion sensors. My specific objectives, and accordingly my role in the development of the studies presented as chapters of this thesis, were: 1. to extend our knowledge about the year-round movements and seasonal timing of life-history events in seabirds using wet-dry data, 2. to reveal the circadian and circa-annual at-sea activity rhythms of little-known seabird species, 3. to develop a novel methodological protocol based on wet-dry data to identify and annotate complex behaviours, 4. to quantify the relative prevalence and transition probabilities of behaviours in order to evaluate complexity in seabirds’ behavioural strategies, 5. lastly, I additionally aimed to devise effective data visualizations to assist the process of re- search in animal movement and behavioural ecology. The work contained in this thesis is organized in the following 4 chapters: In Chapter 1, we used a combination of geolocation, wet-dry data and stable isotope analysis to reveal phenology, non-breeding distribution and migratory routes of a little-known tropical seabird endemic to Cape Verde Islands, the Boyd’s shearwater Puffinus boydi. Using 5-year geolocation da- taset, I described the onset and ending of major events over the annual life cycle, through combining positional with wet-dry data. This chapter contributes to objective 1 of the thesis. In Chapter 2, we studied the temporal-spatial distribution of the Common tern Sterna hirundo along the East Atlantic Flyway. We unveiled migratory routes, stopover sites and non-breeding ar- eas of the studied population. In this study, I used wet-dry data to disentangle differences in activity patterns at two hierarchical scales: on a daily basis and across stages of the annual life cycle. This chapter contributes to objectives 1 and 2 of the thesis. In Chapter 3, we extended the knowledge about the spatial ecology of an endangered species, At- lantic petrel Pterodroma incerta. We used geolocation-immersion loggers to assess in detail phenol- ogy, at-sea distribution, behaviour and habitat preferences over the entire annual cycle. I combined positional and wet-dry data to describe in detail major events over the annual life cycle. Besides, I used data visualization in exploratory analyses to assist defining plausible hypotheses, which led us to find likely carry-over effects of breeding success on individual phenology and behaviour. This chapter contributes to objectives 1, 2 and 5 of the thesis. 16 In Chapter 4, we developed a novel protocol to provide new insights into behavioural organiza- tion of seabirds based uniquely on wet-dry data. In this protocol, I used a breakpoint algorithm to segment continuous wet-dry data, to which we later applied dimensionality reduction and un- supervised clustering algorithms. Throughout our approach, we built up continuous behavioural spaces and evaluated prevalence and transition probabilities of behaviours. Moreover, I introduced a novel application of network analysis to explore behavioural strategies throughout quantifying the changes in organization and importance of behavioural modes between the different stages of the annual life cycle. As a proof of concept, we applied the protocol on data from Cory’s shearwater (Calonectris borealis), which allowed us to find a diverse array of behaviours and analyse them at various spatial and temporal scales. The study was firmly supported by data visualization, which we used to assist all steps along the research process. This chapter contributes to objectives 1, 2, 3, 4 and 5 of the thesis. 17 SUPERVISORS’ REPORT Dr. Jacob González-Solís and Dr. Frederic Bartumeus, as supervisors of the doctoral thesis entitled “Movement ecology in pelagic seabirds” certified that the dissertation presented here has been car- ried out by Zuzana Zajková and grant her the right to defend her thesis in front of a scientific panel. The dissertation work comprises two articles published and one accepted in highly ranked peer reviewed journals included in the Science Citation Index. As supervisors, we have participated in the design, guidance and correction of the work and earlier drafts of the manuscripts included in this thesis. The contribution of the doctoral candidate to each manuscript and the impact factor (Thomson Institute for the Scientific Information) is detailed be- low: CHAPTER 1 Year-round movements of a small seabird and oceanic isotopic gradient in the tropical Atlantic Zajková Z, Militão T, González-Solís J (2017) Year-round movements of a small seabird and oce- anic isotopic gradient in the tropical Atlantic. Marine Ecology Progress Series 579:169-183. Impact factor (2017): 2.292 Z. Zajková has contributed to the analysis of geolocator data, stable isotopes analysis and scientific writing. CHAPTER 2 Common Terns on the East Atlantic Flyway: temporal–spatial distribution during the non- breeding period Becker, PH, Schmaljohann, H, Riechert, J, Wagenknecht, G, Zajková, Z & González-Solís, J (2016) Common Terns on the East Atlantic Flyway: temporal–spatial distribution during the non-breeding period. Journal of Ornithology 157: 927– 940. Impact factor (2015):1.419 Z. Zajková has contributed to the analysis of wet-dry immersion data and scientific writing. CHAPTER 3 Spatial ecology, phenological variability and moulting patterns of the endangered Atlantic petrel, Pterodroma incerta Marina Pastor-Prieto, Raül Ramos, Zuzana Zajková, José Manuel Reyes-González, Manuel L. Ri- vas, Peter G. Ryan & Jacob González-Solís. Accepted in Endangered Species Research. Biologging and Conservation Special issue. 18 Impact factor (2019): 2.122 Z. Zajková has contributed to the analysis of data and scientific writing. CHAPTER 4 Air-water behavioural dynamics reveal complex at-sea ecology in global migratory seabirds Zuzana Zajková, José Manuel Reyes-González, Teresa Militão, Jacob González-Solís & Frederic Bartumeus. To be submitted to Current Biology. Impact factor (2018): 9.193 Z. Zajková has contributed to the study design, data analysis and scientific writing. We also certify that any of the co-authors in the referred papers have used any or part of the work for their own doctoral theses. Barcelona, 20th September 2019 Dr. Jacob González-Solís Bou Departament de Biologia Evolutiva, Ecologia i Ciències Ambientals Facultat de Biologia Universitat de Barcelona Dr. Frederic Bartumeus Ferré Centre d’Estudis Avançats de Blanes Chapter 1: Year-round movements of a small seabird and oceanic isotopic gradient in the tropical Atlantic Zuzana Zajková1, Teresa Militão1 & Jacob González-Solís1 1 Institut de Recerca de la Biodiversitat (IRBio) and Dept. Biologia Evolutiva, Ecologia i Ciències Ambientals, Universitat de Barcelona, Av. Diagonal 643, Barcelona 08028, Spain Published in: Zajková Z, Militão T, González-Solís J (2017) Year-round movements of a small seabird and oceanic isotopic gradient in the tropical Atlantic. Mar Ecol Prog Ser 579:169-183 ABSTRACT Despite the proliferation of seabird tracking studies, there is a relative paucity of studies on small tropical seabirds. We present for the first time the distribution and movements of the little-known Boyd’s shearwater Puffinus boydi, a Procellariiform endemic to the Cape Verde Islands. We tracked 28 birds from 2 breeding sites (Ilhéu Raso and Ilhéu de Cima) with geolocator loggers from 2007 to 2012. We also analysed stable isotopes of carbon and nitrogen in the 1st primary (P1), the 6th rectrice (R6) and the 1st (S1) and 8th (S8) secondary feathers to reveal moulting pattern and oceanic isotopic gradients. Birds migrated on average 1450 km westward, to the central Atlantic Ocean (5 to 15° N, 30 to 40° W), where they stayed on average 114 d, from May to August. Boyd’s shearwaters exploited oceanic waters year-round and showed δ13C values similar to other oceanic seabird spe- cies and δ15N values indicating the lowest known trophic level among all central Atlantic seabirds. Isotope values in flight feathers suggest most animals moult their P1 and R6 around the breeding ground, whereas all birds moult S1 and S8 at the non-breeding quarters. Correlations of δ13C and δ15N values from S8 with the longitude of the non-breeding area indicate the existence of large-scale isotopic gradients matching those known at baseline levels. Combining geolocator tracking and stable isotope analyses in feathers not only allowed us to describe in detail the annual life cycle and distribution of the species, but also the oceanic isotopic gradients in the tropical Atlantic. 22 CHAPTER 1 : INTRODUCTION Over recent decades, studies on the biology and ecology of tropical seabirds have been mainly focused on diet, foraging and perfor- mance at the breeding colonies (Ashmole & Ashmole 1967, Ballance & Pitman 1999, Spear et al. 2007). More recently, studies have been extended to include the relationship between breeding performance and environmental fea- tures (Surman et al. 2012, Catry et al. 2013). In contrast, at-sea distribution of many tropical seabirds remains poorly known and the sparse information available is mostly based on ship- board and coastal observations (Jaquemet et al. 2004, Ballance 2007). Despite the standardized approaches used in ship surveys (Tasker et al. 1984, Camphuysen & Garthe 2004), unreliable at-sea identification of some species (Ainley et al. 2012) and usually unknown origin and breeding status of observed individuals make these counts difficult to interpret. In the last 2 decades, the rise in the use of ex- trinsic and intrinsic markers has underpinned an exponential increase in studies on the pe- lagic ecology of seabirds. Regarding extrinsic markers, the light level logger (geolocator) has become an essential device for studying year- round movements in much more detail than ever before, improving our understanding on the ecological needs and constraints of seabirds at sea (e.g. González-Solís et al. 2007, Guilford et al. 2012). However, the increasing use of ge- olocators to study seabird distribution and be- haviour has been clearly biased towards species from temperate and subantarctic waters. Thus, there is still a clear lack of knowledge about the year-round at-sea ecology and distribution of tropical seabirds, with only a few species well studied (Catry et al. 2009, Pinet et al. 2011, Dias et al. 2015, Precheur 2015, Paiva et al. 2016, Ra- mos et al. 2016). Similarly, intrinsic markers, such as stable isotope analysis (SIA) of δ13C and δ15N of vari- ous tissues have been widely used to study seabird trophic ecology. Typically, δ13C val- ues have been used to determine the diet of seabirds whereas δ15N values reflect trophic level in a general manner (Hobson et al. 1994, Cherel et al. 2008). However, isotopic values of δ13C and δ15N at baseline are also known to vary geographically in the marine environment (McMahon et al. 2013a,b). Spatial maps of iso- topic landscapes, so called ‘isoscapes’, reflect- ing this variability, are just now beginning to emerge, mostly based on large-scale studies on plankton (Somes et al. 2010, McMa- hon et al. 2013a,b). Whether this spatial isotopic variabil- ity propagates up to the food chain and can pro- vide insights into the foraging movements or wintering areas of predators is still a matter of study (Quillfeldt et al. 2005, Cherel & Hobson 2007, Navarro et al. 2013). In this regard, com- bining SIA with tracking studies can help vali- date the relationship between isotope values and foraging movements (Jaeger et al. 2010). Despite increasing interest in linking isotope values of feathers to seabird movements, espe- cially during the less known non-breeding sea- son, only few studies showed a correspondence between δ13C and δ15N in feather isotope val- ues and non-breeding distribution of seabirds tracked with geolocators (Phillips et al. 2009, González-Solís et al. 2011, Hedd et al. 2012). The lack of basic knowledge regarding year- round distribution, phenology and trophic ecol- ogy becomes a matter of conservation con- cern in polytypic species difficult to identify at sea and with unclear taxonomic status. The Little−Audubon’s shearwater complex (Puffi- nus assimilis−lherminieri, Procellariiformes), small-sized seabirds spread within tropical and temperate waters, is a particularly poorly known seabird complex as shown by the vari- ous taxonomic revisions that occurred over re- cent decades (reviewed in Austin et al. 2004). After many years of controversy, Audubon’s shearwater is now suggested to include 3 sub- species, the Audubon’s shearwater P. l. lher- minieri, the Barolo shearwater P. l. baroli and the Boyd’s shearwater P. l. boydi, with a con- servation status of ‘Least Concern’ (BirdLife International 2015, Carboneras et al. 2016a), although in the present study we preferred to follow a precautionary principle and maintain the specific status of the taxon P. boydi (Haze- 23 Year-round movements of a small seabird and oceanic isotopic gradient in the tropical Atlantic voet 1995, Robb & Mullarney 2008). Indeed, the conservation status and taxonomy of sever- al closely related seabird taxa still remain con- troversial partly due to our lack of knowledge on their spatial ecology, since this is important for understanding migratory connectivity, re- productive isolation mechanisms, and therefore potential for lineage divergence (Ramos et al. 2016). Therefore, studies on the phenology and year-round distribution of species within sea- bird complexes with controversial taxonomic relationships are particularly timely. Recent geolocation and stable isotope studies on Barolo shearwater breeding on the Maca- ronesian archipelagos of the Azores and the Salvagens (Neves et al. 2012, Paiva et al. 2016) showed this subspecies to disperse in the sur- roundings of the breeding colonies outside the breeding period. However, there is very little knowledge about detailed biology of the closely related Boyd’s shearwater P. boydi Mathews, 1912, endemic to the Cape Verde Islands, es- pecially those aspects related to phenology, year-round distribution and trophic ecology. Roscales et al. (2011) revealed the distribution and trophic position of Boyd’s shearwaters only at the end of the breeding season, when animals foraged close to the colony. Away from breed- ing grounds, Boyd’s shearwater has only been seen in small numbers off the Senegal coast in October (Hazevoet 1997, Dubois et al. 2009), in 1976 1 bird was trapped on St. Helena (Bourne & Loveridge 1978) and a suspected observa- tion of 1 individual was reported from the Ca- nary Islands in December 2012 (Velasco 2013). However, the majority of the observations of individuals of this species have been reported all year round in Cape Verde and surrounding waters (Bourne 1955, Hazevoet 1995, Dubois et al. 2009), which suggests a non-migratory behaviour, even though non-breeding grounds remain unknown. To fill in this gap, we provide the first de- tailed study on the year-round movements and distribution of the Boyd’s shearwater, based on geolocation and SIA of feathers over multiple years. We aim to (1) reveal main foraging ar- eas during breeding and non-breeding seasons, the detailed phenology of their life cycle and, in particular, clarify whether Boyd’s shearwa- ter performs dispersal movements or oriented migration to a specific non-breeding area; and (2) to bring new insights into the existence of isoscapes and their potential use to study the movement of tropical top predators by linking the isotopic values of feathers with individual non-breeding areas. MATERIALS AND METHODS Study site and species We conducted fieldwork during the breeding seasons of 2007 to 2012 in the Cape Verde Is- lands, on Ilhéu Raso (16° 36’ N, 24° 35’ W) and Ilhéu de Cima (14° 58’ N, 24° 38’ W), 2 islets 180 km apart. We visited the colonies during the incubation period, from February to early April, depending on the year. Additionally, we visited Raso in November 2009. The Boyd’s shearwater is a taxon within the Little−Audubon’s shearwater complex (Puffi- nus assimilis−lherminieri, Procellariiformes) (reviewed in Austin et al. 2004). Traditionally, ‘assimilis’ and ‘lherminieri’ were recognized as 2 species groups, but with numerous taxa within each group (Cramp & Simmons 1977, Warham 1990, Carboneras 1992, Brooke 2004). In the last decade, a molecular study by Austin et al. (2004) proposed 3 geographically discrete clades of the complex identified in the North Atlantic, Southern (Australasia) and tropical Pacific and Indian oceans. A recent revision (Carboneras et al. 2016a,b) has suggested the separation of little shearwater Puffinus assimil- is, distributed in the southern hemisphere, from the Audubon’s shearwater Puffinus lherminieri, distributed in the North Atlantic Ocean and Caribbean Sea. Particularly, 2 North-Atlantic taxa, Barolo shearwater Puffinus baroli (breed- ing in the Azores, Madeira, and the Canary Islands) and Boyd’s shearwater Puffinus boydi (breeding in the Cape Verde Islands) have been switched between ‘assimilis’ and ‘lhermin- ieri’ groups by various authors over the years. 24 CHAPTER 1 : Hazevoet (1995) considered P. boydi as an in- dependent species. The Boyd’s shearwater is endemic to the Cape Verde Islands, where it is thought to breed on most islands and islets (not known from Maio and extinct on Sal) (Haze- voet 1995), with a population estimation of ca. 5000 pairs (BirdLife International 2015). Birds (body mass ≈ 160 g) nest in burrows in soft soil or in rocky cavities. Both parents share incuba- tion of a single white egg that may take 44−60 d to hatch (Carboneras et al. 2016a) and breeding lasts from January to June. Birds are thought to disperse after breeding; however, some were reported visiting the breeding colony at the end of August (Bourne 1955) and throughout the year (Hazevoet 1995). Diet is not well known; the few stomachs examined by Bourne (1955) contained fish and cephalopods up to 8 cm. A closely related species (P. baroli) from the Azores feeds mostly on cephalopods and fish (Neves et al. 2012, J. A. Ramos et al. 2015). Bird tracking and spatial data analysis During the study period (2007−2012) we de- ployed a total of 90 geolocators on 68 individu- als of Boyd’s shearwaters. We captured breed- ing birds by hand in the burrow and deployed geolocators, which we retrieved after ≥1 yr. Over the course of the study, we used 3 differ- ent types of loggers from the British Antarctic Survey (BAS): Mk9 (n = 32), Mk13 (n = 15) and Mk18-H (n = 43). Each logger was attached with a cable tie to a plastic ring, which was deployed on the tarsus of the bird; weight of equipment was approximately 2 g (1.25 % of body mass). We deployed only 1 geolocator per breeding pair. Geolocators recorded ambient light intensity, time and immersion in seawater. Light levels were measured every 60 s and, depending on the type of device, the maximum value within each 5 min (Mk18-H logger) or 10 min (Mk9 and Mk13 loggers) interval was recorded. We processed raw light data and visually super- vised each transition using TransEdit from BASTrack software (British Antarctic Survey). The sunrise and sunset times were estimated applying the light threshold value of 20. To es- timate sun elevation angle, we calibrated the loggers before deployment and after recovery on an open site without shading. The value of sun elevation angle was calculated and applied for each logger, ranging from −5.82 to −3.49 (mean −4.54). Light level data were converted into latitude derived from day length and lon- gitude derived from the time of local midday with respect to Greenwich Mean Time, us- ing BASTrack software. This process results in estimation of 2 positions of the animal per day (Delong et al. 1992, Hill 1994, Afanasyev 2004), with a mean error ± SD of 186 ± 114 km (Phillips et al. 2004). Furthermore, as the latitude estimates are highly sensitive to errors and changes in day length, positions in equato- rial regions may present lower accuracy (Hill & Braun 2001). In addition, cloudy weather at sunrise and/or sunset may lead to error es- timated to 340 km in latitude and 105 km in longitude (Nisbet et al. 2011). It is worth mentioning that interpretation of geolocation positions especially in equatorial latitudes should be accepted with caution, espe- cially the latitude estimations around equinoxes (Hill & Braun 2001, Ekstrom 2004, Lisovski et al. 2012). Detailed examination of error in lati- tude estimation is necessary to avoid the pos- sible misleading interpretation of geolocation positions. Particularly in this study, previous visual examination of positions showed a clear pattern (Figs. S1 & S2 in Supplement 1 at www. int-res.com/articles/suppl/m579p169_supp/) repeated in all individuals during the breeding period and resulting from a shift in the latitu- dinal error between interequinoctial intervals: positions before spring equinox — reflecting mostly incubation and the early chick-rearing period — were distributed northerly from the colony, whereas the positions after the spring equinox — reflecting the chick-rearing period — were distributed southerly from the colony. To avoid possible misleading interpretation that during incubation animals forage in the north and during chick rearing in the south of the col- ony, we pooled together prelaying, incubation and chick-rearing as a breeding period. 25 Year-round movements of a small seabird and oceanic isotopic gradient in the tropical Atlantic Obtained positions were filtered for each logger separately applying a 3-level filtering method, removing positions (1) 15−30 d before and af- ter equinoxes, (2) with obvious interference at dawn or dusk, and (3) when flight speeds sus- tained over a 48 h period were higher than 30 km h−1 applying iterative backward/forward speed filtering (McConnell et al. 1992). The speed threshold was defined after visual ex- amination of distributions of flight speeds. We also excluded positions from the day of deploy- ment and recovery of the logger. Overall, 66 % of original locations were retained for further analysis. Kernel density utilization distribution (UD) estimates were generated from filtered loca- tions (projection: Lambert Equal-Area Azi- muthal, centred to the centroid of all locations) during different periods of the life cycle sepa- rately for each bird and year of tracking using package adehabitatHR (Calenge 2006) in R (R Core Team 2016). Kernel contours of 50 % (‘core-area’) were calculated using a smooth- ing parameter (h) equivalent to the mean er- ror of the geolocators (Phillips et al. 2004). We examined various spatial parameters for each track: (1) the area exploited during the breed- ing and non-breeding periods (50% UD; in km2); (2) location of the centroids of breeding and non-breeding areas (50% UD), which were calculated using ‘centroid’ function from pack- age geosphere (Hijmans et al. 2012); (3) the total distance (great-circle) from the breeding colony to the centroid of the non-breeding area and (5) the accumulated distance covered with- in the non-breeding area (without migration), which were estimated using the functions ‘dis- tance’ and ‘distance-Track’ from the argosfilter package (Freitas 2012), respectively. Geolocators also recorded salt-water im- mersion data sampled every 3 s and registered summary value every 10 min (varying from 0, when the logger was dry the entire 10 min pe- riod, to 200, when the logger was permanently wet). This information was used to help define some phenological parameters (see next sub- section). Phenology Dates defining the phenology of species were identified visually from geographical positions, light and immersion data. During equinox pe- riods, when latitude estimation is not accurate (Hill & Braun 2001), we used only changes in longitude and in immersion data to detect changes in movements and estimate dates of arrival to and departure from the breeding colony. We estimated various phenological param- eters: last night spent at the colony (continu- ous dry record over prolonged period of time during darkness), departure from the breed- ing and non-breeding area (the first day that the bird’s location was outside the cluster of previous day’s positions and was followed by directed movement away from this area), dura- tion of the non-breeding period and migratory movements, arrival to the breeding and non- breeding area (the first day the bird entered the cluster of positions after a directed movement towards that area), the first day and night an individual spent in the burrow (detected by a continuous dry record over a prolonged period of time during daylight and darkness), first day of incubation (min. 2 consecutive days spent in the burrow), duration of the incubation period (from the first day of incubation until the return from the last foraging trip, including time spent outside on foraging trips between incubation shifts), and, finally, incubation shift and forag- ing trip duration. Parameters referring to incubation duration were estimated only for individuals with 2 or more continuous years of tracking data (with the same geolocator or the geolocator that was replaced during incubation and recovered the following year). For those individuals we could estimate the onset, duration and end of incuba- tion from light and immersion data. As some loggers failed to collect data for the entire deployment period or some animals did not breed, sample sizes for different phenological parameters vary somewhat between analyses. Based on these parameters, we identified and considered 4 periods of the life cycle: (1) breed- 26 CHAPTER 1 : ing, period between logger deployment and de- parture on migration and period between the arrival to the colony from migration and recov- ery of the logger, (2) postnuptial migration, (3) non-breeding, period between arrival to non- breeding area and start of prenuptial migration and (4) prenuptial migration. One individual did not migrate and spent the non-breeding sea- son in the vicinity of the Cape Verde Islands, so we considered the last night the animal spent at the colony (burrow) as the end of the breeding period. Similarly, the start of the subsequent breeding season was assigned as the first night the animal visited the burrow. We used repeated measures ANOVA with in- dividual as an error term (to account for pseu- do-replication as few individuals were tracked >1 yr) to test for differences between the dura- tion of the post- and prenuptial migration and the size of the core range areas between the breeding and non-breeding periods. Stable isotope analysis Boyd’s shearwater is expected to moult the first primary feather at the end of the breeding pe- riod, just before migration, reflecting the iso- topic composition of the breeding area (Cramp & Simmons 1977, Roscales et al. 2011). Known primary moult patterns of similar shearwa- ter species are described as descendent, i.e. from the innermost to the outermost primary feather, with a duration of 3−5 mo, while the outermost rectrice feather is among the last to be moulted (Monteiro et al. 1996, Bridge 2006, Ramos et al. 2009). Moult of secondary feath- ers of shearwaters has been previously linked with the non-breeding area (Neves et al. 2012, Paiva et al. 2016). In this study, carbon (δ13C) and nitrogen (δ15N) isotope ratios were exam- ined in different wing-feather types: 1st prima- ry (the innermost), 1st and 8th secondary and 6th rectrices (the outermost) feathers (hereafter named as P1, S1, S8 and R6, respectively). All feathers were sampled when we recovered the geolocator and they were stored in plastic bags before the analysis. For birds with the same logger recovered after ≥ 2 yr, feather sampling also occurred at the point of logger recovery, but these feathers are only related with the last year of tracking. In total, the dataset for statis- tical analysis consisted of 32 sets of 4 feathers from 28 individuals (4 individuals with feath- ers from 2 different years). To avoid any possible contamination, feather samples were washed in 0.25 M sodium hy- droxide solution, rinsed with distilled water and oven dried at 40°C for 24 h. Subsequently, we manually cut each feather to small fragments using stainless steel scissors and weighed a sample of 0.30−0.32 mg on a precision scale. Stable isotope values are expressed in delta no- tation (δ) as parts per thousand (‰) according to the following: δX = [(Rsample/Rstandard) − 1], where X is 15N or 13C and R is the correspond- ing ratio 15N/14N or 13C/12C, respectively. Rstandard values for 15N and 13C were based on atmo- spheric N2 and the Vienna Pee Dee Belemnite standard, respectively. Replicate measurements of laboratory standards (2 standards for every 12 unknowns) indicated measurement errors of approximately 0.2 and 0.1 ‰ for nitrogen and carbon, respectively. The analysis of stable isotopes was carried out at Scientific-Technical Services of the University of Barcelona. Statistical analyses of isotopic data We could not directly test for differences be- tween the 2 colonies as they were sampled in different years (Ilhéu Raso: 2007 and 2008; Ilhéu de Cima: 2009, 2010 and 2011). After visual comparison, there was no indication in systematic differences in isotopic values be- tween colonies; therefore, all the statistical analyses were performed with isotopic data of both colonies pooled together. To test for differ- ences among feathers in isotopic data, we first checked δ13C and δ15N values for normal dis- tribution using Q-Q plots and Shapiro-Wilks’ test. We used linear mixed models (LMM, R package lmerTest; Kuznetsova et al. 2015) to compare isotopic values among feathers (fixed factor) and we accounted for pseudo-replica- tion including individual and sampling year as random factors. The p-values were calculated 27 Year-round movements of a small seabird and oceanic isotopic gradient in the tropical Atlantic from Type 3 F-statistics with Satterthwaite’s approximation for degrees of freedom, while pairwise comparisons were calculated based on differences of least squares means (function ‘difflsmeans’ package lmerTest) and adjusted using Bonferroni correction. Based on the isotopic differences found among feathers (see ‘Results’), we inferred that the S1 and S8 were moulted during the non- breeding period. Because both showed similar isotopic values, and to allow a comparison with the isotopic data of previous studies on a close- ly related species (Neves et al. 2012, Paiva et al. 2016), we used S8 for subsequent analyses. To link isotopic values of S8 feathers with the non-breeding area of each individual we deter- mined the individual centroids of 50% kernel of non-breeding area. We used LMM to exam- ine whether the variation in S8 feather isotopic values could be explained by the location of their non-breeding area (latitude and/or longi- tude of centroid as fixed, individual and year as random factors; R package lmerTest; Kuznetso- va et al. 2015). The best-supported model was selected using the Akaike Information Crite- ria corrected for small sample sizes (AICc) (R package MuMIn; Bartoń 2016). To understand the possible influence of the sampling year we verified its importance based on the likelihood ratio test (function ‘rand’ from package lmerT- est) and by calculating the variance explained by the sampling year. Statistical analyses were carried out using the R software version 3.2.1 (R Core Team 2016). All values are presented as means ± SD, and we assumed a significance level p < 0.05. RESULTS Recovery of loggers We retrieved 43 loggers (recovery rate 47.8 %) from 32 unique individuals. Most loggers were recovered in the year following deploy- ment; however, 7 loggers were retrieved after 2 consecutive years. Eight other individuals (n = 6 and n = 2, respectively) were tracked over 2 and 3 consecutive years, by recovering and de- ploying a new logger each year. Eleven loggers failed or did not contain enough data for further analysis. Overall, the final dataset contained 38 year-long tracks of 28 unique individuals (9 from Ilhéu Raso, 19 from Ilhéu de Cima), including 10 individuals with 2 yr of track- ing. We calculated kernel UD density for 38 tracks for non-breeding (2007, 9 tracks; 2008, 3 tracks; 2009, 6 tracks; 2010, 12 tracks; 2011, 8 tracks) and 35 tracks for breeding season, as 3 tracks from 2010/2011 did not contain enough locations for kernel estimation. Phenology of annual cycle Boyd’s shearwaters presented some variability in their phenological parameters, especially in the timing of the first day and the first night in the burrow (Table 1) and on the duration of the non-breeding period (Table 1, Fig. 1 with individual phenologies). Furthermore, the du- ration of the pre-nuptial migration was statis- tically longer than the postnuptial migration (repeated-measures ANOVA, F 1, 45 = 7.463, p = 0.009), with birds travelling for 7.2 ± 6.0 d to reach the breeding colony on their prenuptial migration in contrast with 4.9 ± 2.6 d to reach the non-breeding area on their postnuptial mi- gration (Table 1). Seasonal changes in at-sea distribution During breeding, birds dispersed in different directions around the breeding colony and in proximity to Cape Verde Islands. With 1 excep- tion (bird ID 2007_047), which foraged in the neritic area of the African coast in November- December, the tracked birds did not forage in neritic waters but north of the breeding colo- nies, reaching up to 30° N (Fig. 2, Fig. S3 in Supplement 1 and the animation in Supple- ment 2 at www.int-res.com/articles/suppl/ m579p169_supp/). The estimated individual core range area during the breeding season (50% kernel UD) ranged from 292 000 to 764 400 km2 (470 600 ± 111 500 km2, n =35). At the beginning of May, birds started their post-nuptial migration consistently in a west- 28 CHAPTER 1 : Table 1. Year-round phenology of Boyd’s shearwaters from Ilhéu de Cima and Raso (Cape Verde), tracked with geolocators from 2007−2012; data are mean ± SD and range values over 5 yr of tracking study 2011_475 2011_473 2011_458 2011_456 2011_455 2011_446 2011_444 2011_195 2010_520 2010_519 2010_517 2010_512 2010_510 2010_458 2010_457 2010_455 2010_454 2010_452 2010_446 2010_438 2009_522 2009_519 2009_518 2009_517 2009_512 2009_510 2008_042 2008_036 2008_007 2007_056 2007_047 2007_042 2007_040 2007_036 2007_028 2007_025 2007_019 2007_007 Feb Mar Apr May June July Aug Sep Oct Nov Dec Jan Feb Mar Apr Fig. 1. Individual phenologies of Boyd’s shearwaters, ordered by year of logger deployment (year of deploy- ment_bird ID; n = 38). Each horizontal bar represents 1 yr of tracking, colours represent different stages of breeding cycle: breeding in yellow, postnuptial migration in dark red, non-breeding in blue, prenuptial migration in purple. Points refer to last night (◆), first night (●) and first day (○) the bird spent in the burrow, and onset of incubation (＊). Starts on the day of deployment, ends on the day of retrieval of the logger (or when logger stopped collecting data) Phenological parameter n Mean ± SD Range Last night colony 36 26 Apr ± 17.4 21 Mar - 28 May Colony departure 38 4 May ± 16.5 4 Apr - 7 Jun Non-breeding area arrival 38 9 May ± 16.6 7 Apr - 12 Jun Non-breeding area departure 37 31 Aug ± 18.6 1 Aug - 16 Oct Colony arrival 37 7 Sep ± 19.5 4 Aug - 22 Oct First night burrow 33 18 Sep ± 28.5 4 Aug - 2 Dec First day burrow 32 31 Oct ± 59.6 13 Aug - 11 Jan Incubation start 24 9 Feb ± 12.2 22 Jan - 6 Mar Number of incubation shifts 6 3.8 ± 0.8 3 - 5 Duration of incubation (d) 6 47.0 ± 2.8 42 - 50 Duration of incubation shift (d) 23 6.0 ± 1.9 2 - 10 Duration of incubation foraging trip (d) 23 6.3 ± 1.9 2 - 9 Duration of postnuptial migration (d) 38 4.9 ± 2.6 0 - 13 Duration of non-breeding period (d) 37 114.0 ± 18.1 50 - 140 Duration of prenuptial migration (d) 37 7.2 ± 6.0 0 - 33 29 Year-round movements of a small seabird and oceanic isotopic gradient in the tropical Atlantic ward direction along a migration corridor between 7° and 15° N (Fig. S4). The mean distance between the breeding colony and non- breeding area (to the centroid of 50 % kernel UD) was 1450 ± 398 km (range 106−2391 km, n = 38). The main non-breeding area of Boyd’s shearwaters was in the Central Atlantic Ocean, west of Cape Verde Basin, over the Mid-Atlan- tic Ocean Ridge, from 5 to 15° N and from 30 to 40° W (50% kernel UD; Fig. 3, Fig. S3 in Supplement 1, and Supplement 2). However, 1 individual migrated further west to 9° N, 43° W (bird ID 2009_510), while another went fur- ther north to 21° N, 36° W (bird ID 2007_007). The estimated individual core range area dur- ing the non-breeding season (50% kernel UD) ranged from 300 700 to 795 400 km2 (467 700 ± 120 000 km2, n = 38), which did not signifi- cantly differ from the size of core range areas during the breeding season (repeated measures ANOVA, F 1, 41 = 0.027, p = 0.870, n = 35). Dur- ing the non-breeding period, birds dispersed or steadily moved over a huge area. Total distance covered within the non-breeding area was on average 33 670 ± 5628 km (range 17 440−47 690 km, n = 38), moving on average 253.1 ± 32.8 km over approximately 24 h by a mean velocity of 10.5 ± 1.4 km h−1. From all tracked birds, only 1 individual (bird ID 2007_040) did not migrate and stayed in the vicinity of Cape Verde Island year-round. The timing of pre- nuptial migration mostly overlapped with the autumn equinox period, but data for a few indi- viduals suggest that animals use a similar route to return to breeding grounds (Fig. S4). Stable isotope analysis Boyd’s shearwaters presented a wider range of nitrogen (6.39 to 12.60 ‰) than carbon values (−17.96 to −15.24 ‰) (Table 2, Fig. 4). Signifi- cant differences were found between feathers (P1, S1, S8 and R6) in both nitrogen (LMM, F 3, 94.323 = 29.965, p < 0.001) and carbon values (LMM, F 3, 91.355 = 53.684, p < 0.001). The differ- ences in nitrogen values were between P1 and both S1 and S8 (pairwise comparison, both p < 0.001; Table S1 in Supplement 1). No differ- ence was found between P1 and R6 (pairwise comparison, p = 0.719), or between S1 and S8 (pairwise comparison, p = 1.000). Significant differences were found between all feathers for carbon values (pairwise comparison, for all p < 0.001), except for S1 and S8 (p = 0.417). Although differences found between feathers were statistically significant, the magnitude of those differences was small (Table S1 in Sup- plement 1), comparing with variation among individuals (Fig. 4). Geographic isotopic gradient (isoscapes) The best-supported models (Table 3, Table S2a) suggest that the variation of isotope values of S8 was highly related with longitude (LMM, δ15N: F 1, 29.915 = 57.945, p < 0.001; δ13C: F 1, 28.326 = 29.139, p < 0.001; Fig. 5A,B), but not with latitude (LMM, δ15N: F 1, 25.094 = 1.512, p = 0.230; δ13C: F 1, 29.200 = 0.511, p = 0.485) values of the non-breeding centroid of Boyd’s shearwaters. Indeed, the R2m values, which describe the Table 2. Isotopic values of δ15N and δ13C (‰) in the 1st primary (P1), the 1st (S1) and 8th (S8) secondary feathers and the 6th rectrice (R6) of Boyd’s shearwaters breeding in the Cape Verde Islands. P1 and R6 feathers showed similar isotopic values but distinct than S1 and S8 Feather n δ15N δ13C Mean ± SD Range Mean ± SD Range P1 32 8.75 ± 1.12 7.29 to 12.60 -16.68 ± 0.43 -17.96 to -16.10 S1 32 7.61 ± 0.59 6.39 to 9.46 -16.04 ± 0.30 -16.93 to -15.59 S8 32 7.57 ± 0.61 6.44 to 9.82 -15.92 ± 0.35 -16.83 to -15.24 R6 32 8.51 ± 1.13 6.64 to 11.01 -16.37 ± 0.44 -17.20 to -15.51 30 CHAPTER 1 : Fig. 2. Individual kernel density utilisation distributions for Boyd’s shearwaters during the breeding season, tracked with geolocators during 5 yr (2007−2012). Ellipses: individual 50 % density contours; points: individual centroids of 50 % density contours; black square: breeding colony. Bathymetry used as background 31 Year-round movements of a small seabird and oceanic isotopic gradient in the tropical Atlantic Fig. 3. Individual kernel density utilisation distributions for Boyd’s shearwaters during the non-breeding season. Other details as in Fig. 2 32 CHAPTER 1 : Table 3. Linear mixed models testing for spatial gradient in isotopic values of nitrogen and carbon of Boyd’s shearwaters breeding in the Cape Verde Islands. Results of second-order Akaike’s Information Criterion (AICc), delta AICc and Akaike weights are shown. The best supported model (in bold) includes longitude as fixed factor. All models include individual and year as random factors A 6.0 7.0 8.0 9.0 10.0 11.0 12.0 13.0 P1 S1 S8 R6 Feather δ1 5 N (‰ ) B -15.0 -16.0 -17.0 -18.0 P1 S1 S8 R6 Feather δ1 3 C (‰ ) Year 2007 2008 2009 2010 2011 Fig. 4. Stable isotopes (A) δ15N and (B) δ13C of sampled feathers (P1: 1st primary; S1 and S8: 1st and 8th secondary, respectively; R6: 6th rectrice) of tracked Boyd’s shearwaters breeding in the Cape Verde Islands (n = 32) in 2007−2012. Each line corresponds to 1 individual coloured by the year of tracking. Circle with range represents mean ± 95 % confidence interval Model Df δ15N δ13C AICc ΔAICc AICc wt AICc ΔAICc AICc wt Long 5 37.56 0.00 0.81 0.75 0.00 0.82 Long + Lat 6 40.49 2.93 0.19 3.79 3.04 0.18 Null 4 66.45 28.90 0.00 19.62 18.87 0.00 Lat 5 67.60 30.03 0.00 21.83 21.08 0.00 proportion of variance explained by the fixed factor alone, were high in the model with the fixed factor longitude (0.63 and 0.39 for δ15N and δ13C, respectively; Fig. 5A,B), but not in the ones with latitude (0.05 and 0.01 for δ15N and δ13C, respectively). Accounting for both longi- tude and latitude did not significantly improve the longitude-gradient model (LMM, δ15N: χ2 = 0.121, df = 1, p = 0.729; δ13C: χ2 = 0.012, df = 1, p = 0.913). Our best-supported models suggest that annual isotopic variability was negligible for nitrogen values of S8, accounting for 0 % of random variance (χ2 = 0, df = 1, p = 1.000) (Table S2b). In contrast, annual variation ac- 33 Year-round movements of a small seabird and oceanic isotopic gradient in the tropical Atlantic counted for almost half (53.3 %) of the random variance of the carbon values (χ2 = 7.630, df = 1, p = 0.006). DISCUSSION This is the first study on the movements and year-round distribution of the Boyd’s shearwa- ter. We showed that Boyd’s shearwaters per- form oriented migratory movements and ex- ploit oceanic habitats year-round. Furthermore, we revealed the existence of a longitudinal isotopic gradient in the tropical north Atlantic by relating the isotopic values of the feathers moulted during the non-breeding period and the location of the individual non-breeding area. Boyd’s shearwaters showed some variability in various aspects of their breeding phenology. Small species breeding in the tropics may expe- rience relatively constant environmental condi- tions, which may cause minimal synchrony in breeding (Brooke 1990). The few individuals that started the postnuptial migration relatively earlier, in the beginning of April, were pre- sumably failed breeders; however, we do not have breeding success information of each bird to confirm this hypothesis. The longer dura- tion of prenuptial migration in relation to the postnuptial one is an opposite pattern to many long-distance migrants (Nilsson et al. 2013) and may be a consequence of prevailing trade winds which advantaged shearwaters during post-nuptial migration through a tailwind but disadvantaged them during pre-nuptial migra- tion through a headwind (Liechti 2006). Birds started to arrive at the colony in early August, which confirms observations of shearwaters visiting Ilhéu de Cima at the end of August (Bourne 1955). After returning to the breeding colony, birds were asynchronous in terms of the first day spent in the burrow during day- light; these dates were spread over 4 mo. Those differences might be sex-related, with males visiting burrows earlier than females in some shearwater species (Hedd et al. 2012, Müller et al. 2014), probably due to their role in nest de- fence. However, this asynchrony was also ob- served in Barolo shearwaters, in a study where only males were tracked (Neves et al. 2012). As sex of animals tracked in this study was un- known and only 1 member of the breeding pair was tracked, we could not estimate the laying date and define the first incubation shift. How- ever, we were able to estimate the beginning of incubation, on average February 9, which is earlier than Barolo’s shearwaters in the Azores (Neves et al. 2012, but see Monteiro et al. 1996). Indeed, the incubation period (42−50 d) was slightly shorter than the periods reported for Puffinus lherminieri (44−60 d, Carboneras et al. 2016b), P. assimilis assimilis (55 d) breeding at Lord Howe Island (Priddel et al. 2003) and P. a. haurakiensis (54−57 d) on Lady Alice Island (Booth et al. 2000). Incubation shift length and duration of foraging trips during incubation were similar to Barolo shearwaters tracked in the Azores (Neves et al. 2012). However, the foraging-trip durations of Boyd’s shearwaters differed from those of Barolo shearwaters for- aging mostly within the Canary Current sys- tem (being longer than those of birds breeding in Salvagem Grande but shorter than those in Porto Santo) (Paiva et al. 2016). Since we would expect incubation behaviour and forag- ing strategies to be similar among such closely related taxa, this variability most likely reflects differences in environmental conditions across localities, such as differences in the distance to suitable foraging areas and their typically low predictability in tropical waters, which pos- sibly results in differences in egg neglect epi- sodes (and therefore duration of the incubation) and foraging trip length (and therefore duration of incubation shifts) across populations. During the breeding period, Boyd’s shear- waters mainly foraged around the Cape Verde archipelago. Individual core ranges seemed to fluctuate north and south of the archipelago, and some geolocator positions may have even reached the Canary Islands or the Azores (Fig. S3 in Supplement 1, and Supplement 2), but this is most likely due to the effect of the equinoxes on the latitudinal errors (Figs. S1 & S2). Since longitudinal errors of the geolocator methodol- 34 CHAPTER 1 : ogy are relatively small and the African coast is just east of the archipelago, our results clearly showed that birds do not visit the African shelf to forage in neritic waters. With the exception of 1 individual for a few weeks, all birds were largely oceanic during breeding and over the 5 yr of the study (Fig. 2). Similarly, the closely related Barolo shearwaters breeding in Ma- deira and other small seabird species in Cape Verde also show oceanic distribution during the breeding period (J. A. Ramos et al. 2015, R. Ramos et al. 2015, 2016, Paiva et al. 2016). The oceanic behaviour of Boyd’s shearwaters is also suggested by the low carbon values in their first primary feathers (P1), similar to those reported for other oceanic species, such as the Barolo shearwaters in several Macaro- nesian localities (Roscales et al. 2011, Neves et al. 2012, J. A. Ramos et al. 2015, Paiva et al. 2016), but also by Bulwer’s and Fea’s petrel in Cape Verde (Roscales et al. 2011). These results contrast with the importance of the continental shelf inferred for the Barolo shearwaters breed- ing on Salvagens (J. A. Ramos et al. 2015, Paiva et al. 2016) and also for Audubon’s shearwaters breeding in the Caribbean (Precheur 2015, P. Jodice unpubl. data). Further studies using more accurate loggers are needed to confirm these results as this apparent neritic behaviour may just result from the latitudinal error of the geolocation method (Fig. S1). After breeding, Boyd’s shearwaters per- formed a longitudinal-oriented migration, heading westward to the oligotrophic waters of the central North Atlantic Ocean. Despite their short migration, Boyd’s shearwaters constantly moved during the non-breeding season, cover- ing on average more than 30 000 km. These movements may be a foraging strategy to in- crease the chances of finding prey in tropical oceanic waters, which typically show lower productivity and predictability of resources than upwelling systems (Weimerskirch 2007). However, distance calculations should be treat- ed with caution as they may be overestimated due to the positional error. The longitudinal- oriented migration was noticeably consistent across years at coarse scale, wintering in the same area of the Atlantic (except 1 individual remaining around the Cape Verde Islands). All birds spending their non-breeding period in this area also showed clear oceanic habits. The lack of direct observations of Boyd’s shear- waters from their migration and non-breeding grounds may be due to the lack of observers in those areas and/or to the problematic iden- tification of the taxa at sea. To our knowledge, there are just a few sightings of individuals of the little shearwater complex of unknown provenance (RNBWS 2014), illustrating once again the enormous insights geolocation is providing into the spatial ecology of seabirds, particularly in closely related taxa with few morphological differences and unclear taxo- nomic status. In contrast with our results, pre- vious tracking studies on Barolo shearwaters in the Azores and Salvagens mainly showed a dispersive behaviour after breeding (Neves et al. 2012, Paiva et al. 2016). In addition, there is no spatio-temporal overlap in distribution of the different taxa of the complex, pointing out substantial differences in their migratory behaviour and distribution and potential for lineage divergence, which deserves some at- tention when discussing the taxonomy within the little shearwater complex. To understand year-round trophic ecology and study the existence of oceanic isoscapes through the analyses of stable isotopes in feathers it is essential to know the moulting patterns of the study model. Unfortunately, there is a lack of information about moult in Boyd’s shearwater. Shearwaters usually show simple descendent moult that takes 3 to 5 mo to complete (Bridge 2006), starting with the in- nermost primary feather (P1), which in some species may be moulted even before the bird leaves the breeding area (Cramp & Simmons 1977, Monteiro et al. 1996). According to our geolocation data, birds spent on average 114 d outside of the breeding area, which theoretical- ly should leave enough time to complete moult in the non-breeding area. Our SIA supports this hypothesis. P1 and R6 were isotopically similar, suggesting that they are moulted in the same area, probably near the breeding area at 35 Year-round movements of a small seabird and oceanic isotopic gradient in the tropical Atlantic the end of the breeding and non-breeding pe- riod, respectively. These 2 feathers differed from the S1 and S8, which we inferred were moulted during the non-breeding period in the North Central Atlantic, since the isotopic val- ues of the S8 showed a high correlation with the longitude of the centroid of the non-breeding area of each individual. We inferred P1 and R6 to be moulted in the same area (surroundings of the breeding colony), so we expected to find lower isotopic variability compared to S1 and S8, which were moulted in different non-breed- ing areas with potentially different baselines. Therefore, the larger range of isotopic values of the P1 and R6 than S1 and S8 may reflect the inter-individual variability in the phenology at the beginning and at the end of the moulting period, with some birds advancing or delaying their moulting patterns in relation to migration depending, for example, on their breeding suc- cess. Moreover, moulting pattern of rectrices is typically more asynchronic among and within individuals than the rest of the flight feathers (Ramos et al. 2009), adding variability in the timing of moult and in turn in the standard deviation and range of the isotopic values we found in R6. The inter-annual variability in stable isotope values was low for nitrogen, but relatively high for carbon values. However, the broad pattern found in longitudinal gradients was similar over the years (Fig. S5). Baselines of nitrogen and carbon values are known to vary between seasons and years due to changing environ- mental factors (temperature) and/or productiv- ity in marine environment (Goering et al. 1990, Rolff 2000, Graham et al. 2010). Inter-annual differences in stable isotopes were also found in Barolo shearwaters, but the origin is diffi- cult to determine, since these differences may result from changes in diet, foraging areas and/ or baseline conditions due to environmental factors, or a combination thereof (Neves et al. 2012, J. A. Ramos et al. 2015, Paiva et al. 2016). Many seabird species cross the equatorial area of the Atlantic Ocean during their trans- equatorial migrations, but do not forage in this area for extended periods (González-Solís et al. 2007, Guilford et al. 2009, Hedd et al. 2012). So far, the only tracked species known to use the equatorial Atlantic waters as one of their main non-breeding areas is the Bulwer’s petrel Bul- weria bulwerii (Dias et al. 2015, R. Ramos et al. 2015), although in a different period than the Boyd’s shearwater, since Bulwer’s petrels breed during the non-breeding period of the Boyd’s shearwaters. Temporal segregation in the breeding cycles of Bulwer’s petrel and Boyd’s shearwaters may suggest that this is driven by competition for food, but their segregation in trophic level, as indicated by the greater δ15N in the former than in the latter (Roscales et al. 2011), would not support this interpretation. In- stead, temporal segregation may partly result from competition for nesting sites (Fagundes et al. 2016). Indeed, previous studies on breeding seabirds of the tropical and subtropical Atlantic indicated that the trophic position of the Boyd’s shearwater is the lowest among all pelagic seabirds, together with Barolo and Audubon’s shearwaters (Roscales et al. 2011, Neves et al. 2012, Mancini et al. 2014, Paiva et al. 2016). No conventional dietary analysis of Boyd’s shearwaters has been conducted so far (but see Bourne 1955), but its low trophic level indi- cates the consumption of small juvenile squid and fish and crustaceans, as found in the diet of the Barolo shearwater (Neves et al. 2012, J. A. Ramos et al. 2015). Previous studies have sug- gested seasonal changes in the diet of the Baro- lo shearwater (Neves et al. 2012, J. A. Ramos et al. 2015, Paiva et al. 2016), as indicated by an increase in δ15N values in feathers moulted in the non-breeding season compared to those moulted in the breeding season, suggesting that shearwaters targeted prey with higher trophic level during the non-breeding season (Neves et al. 2012). We also found seasonal changes in the isotopic values quite consistent over 5 years, but changes in δ15N were the opposite, i.e. we observed a decrease in δ15N and an in- crease in δ13C values from feathers moulted in the non-breeding (S1 and S8) compared to those grown in the breeding season (P1 and R6; Fig. 4). However, this opposite trend in δ15N values and its significant correlation with 36 CHAPTER 1 : A y = 12.358 + (0.133 * long) R2m = 0.63 R2c = 0.89 p < 0.001 C B y = −17.941 + (−0.055 * long) R2m = 0.39 R2c = 0.71 p < 0.001 D 6.0 7.0 8.0 9.0 10.0 0° 10°N 20°N −18.0 −17.0 −16.0 −15.0 −14.0 0° 10°N 20°N 50°W 40°W 30°W 20°W 50°W 40°W 30°W 20°W 50°W 40°W 30°W 20°W 50°W 40°W 30°W 20°W Longitude Longitude δ1 5 N (‰ ) La tit ud e δ1 3 C (‰ ) La tit ud e 7 8 9 δ15N (‰) −16.5 −16 −15.5 δ13C (‰) Fig. 5. Relationship between (A) δ15N and (B) δ13C values of 8th secondary feather (S8) of Boyd’s shearwaters tracked with geolocators (n = 32; 2007−2012) and the longitude of the centroid, reflecting the area exploited during the non-breeding season (May−August). Points represent individual centroids of 50 % kernel utilization distribution during the non-breeding season. Equation (negative values for western longitude) and dark grey line refer to intercept and slope for fixed factors of linear mixed model (longitude as fixed factor, with year and individual as random) for all years pooled together. (C, D) Spatial distribution of individual centroids of 50 % kernel utilization distributions during the non-breeding season and their respective gradient in (C) δ15N and (D) δ13C values of S8. Black squares: breeding colonies (Ilhéu Raso and Ilhéu de Cima) longitude suggest that these changes just reflect baseline isotopic gradients in longitude (Fig. 5). Indeed, correlations between longitude of the non-breeding centroids with the isotopic values of feathers grown in this period broadly match isoscapes based on plankton samples from the same area (Somes et al. 2010, McMa- hon et al. 2013a). Spatial patterns indicating greater values of δ13C and smaller in δ15N in the central oligotrophic subtropical Atlantic Ocean were confirmed by a recent study on plankton biomass (Mompeán et al. 2013). Knowledge of baselines is also essential in any isotopic stud- ies of trophic ecology, since baseline adjust- ment allows for the comparison of species from different geographical origin (Navarro et al. 2013). The strong longitudinal gradient in val- ues of nitrogen and carbon found in this study suggests propagation of isotopic variability up to the food chain on a coarse scale. However, failure to find latitudinal gradients may be re- lated to latitudinal error inherent to geolocation methodology. Another constraint in gradient models is the limitation in modelling tech- niques to incorporate all sources of uncertainty and error of location estimations. Furthermore, care should be taken, as the high isotopic vari- ability among individuals and the reduction 37 Year-round movements of a small seabird and oceanic isotopic gradient in the tropical Atlantic of the moulting area to a centroid may hinder the potential use of this isotopic gradient to in- fer the non-breeding areas of untracked birds. A study using data with more precise spatial resolution and more detailed knowledge about timing of moult would be required to create complex isoscapes and investigate the potential geographic assignment to foraging movements or non-breeding areas of top predators in the tropical Atlantic Ocean using SIA, but our re- sults show some promising potential for this. Overall, in this study we provided detailed information about the year-round distribution, trophic ecology, phenology and moulting pat- terns of Boyd’s shearwater. The combined use of geolocators and SIA allowed us to bring new insights to the biology and ecology of a poorly known tropical species. ACKNOWLEDGEMENTS We thank all the people who helped us at sev- eral stages of fieldwork (S. Martins, R. Ramos, L. Zango). We are also grateful to the Direção Nacional do Ambiente of Cape Verde who provided logistical support and the necessary authorisations. Z.Z. was supported by a PhD grant (APIF) from the University of Barce- lona (Spain) and T.M. by a PhD grant (SFRH/ BD/47467/2008) from Fundação para a Ciência e Tecnologia (Portugal). This study was funded by the MICIN, Spain (CGL2009- 11278/BOS and CGL2013-42585-P) and Fondos FEDER. We are thankful to Sarah Saldanha for lan- guage corrections, Martin Stýblo for his en- couragement and José Manuel Reyes-González for helpful discussions and comments on pre- vious versions of the manuscript. We are also grateful to Steffen Oppel and an anonymous reviewer for their comments, which greatly im- proved the previous version of the manuscript. LITERATURE CITED Afanasyev V (2004) A miniature daylight level and activity data recorder for tracking ani- mals over long periods. Mem Natl Inst Polar Res 58:227−233 Ainley DG, Ribic CA, Woehler EJ (2012) Add- ing the ocean to the study of seabirds: a brief history of at-sea seabird research. Mar Ecol Prog Ser 451:231−243 Ashmole NP, Ashmole MJ (1967) Comparative feeding ecology of sea birds of a tropical oceanic island. Peabody Mus Nat Hist Yale Univ Bull 24 Austin JJ, Bretagnolle V, Pasquet E (2004) A global molecular phylogeny of the small puffinus shearwaters and implications for systematics of the little-audubon’s shearwa- ter complex. Auk 121:847−864 Ballance L (2007) Understanding seabirds at sea: why and how? Mar Ornithol 35:127−135 Ballance LT, Pitman RL (1999) Foraging ecol- ogy of tropical seabirds. In: Adams NJ, Slotow RH (eds) Proc 22 Int Ornithol Con- gr, Durban. BirdLife South Africa, Johan- nesburg, p 2057−2071 Bartoń K (2016) MuMIn: multi-model infer- ence. https://cran. r-project.org/web/pack- ages/MuMIn/index.html BirdLife International (2015) Puffinus lhermin- ieri. The IUCN Red List of Threatened Spe- cies 2015: eT45959182A 84678618 Booth A, Minot E, Imber M, Fordham R (2000) Aspects of the breeding ecology of the North Island little shearwater Puffinus as- similis haurakiensis. NZ J Zool 27:335−345 Bourne WRP (1955) The birds of Cape Verde Islands. Ibis 97: 508−555 Bourne WRP, Loveridge A (1978) Small shear- waters from Ascension and St. Helena, South Atlantic Ocean. Ibis 120:65−66 Bridge ES (2006) Influences of morphology and behavior on wing-molt strategies in seabirds. Mar Ornithol 34:7−19 Brooke M (1990) The Manx shearwater. Aca- demic Press, London Brooke M (2004) Albatrosses and petrels across the world. Oxford University Press, Oxford Calenge C (2006) The package ‘adehabitat’ for the R software: a tool for the analysis of space and habitat use by animals. Ecol Model 197:516−519 38 CHAPTER 1 : menting migrations of northern elephant seals using day length. Mar Mamm Sci 8:155−159 Dias MP, Alho M, Granadeiro JP, Catry P (2015) Wanderer of the deepest seas: mi- gratory behaviour and distribution of the highly pelagic Bulwer’s petrel. J Ornithol 156: 955−962 Dubois PJ, Holmström N, Verneau A (2009) La péninsule du Cap-Vert à Dakar, Sénégal, est-elle la «Mecque» du sea- watching? Or- nithos 16:220−232 Ekstrom PA (2004) An advance in geoloca- tion by light. Mem Natl Inst Polar Res 58:210−226 Fagundes AI, Ramos JA, Ramos U, Medeiros R, Paiva VH (2016) Breeding biology of a winter-breeding procellariiform in the North Atlantic, the Macaronesian shear- water Puffinus lherminieri baroli. Zoology 119:421−429 Freitas C (2012) argosfilter: Argos locations filter. R package version 0.63. https://cran.r- project.org/web/packages/ argosfilter Goering J, Alexander V, Haubenstock N (1990) Seasonal variability of stable carbon and nitrogen isotope ratios of organisms in a North Pacific Bay. Estuar Coast Shelf Sci 30:239−260 González-Solís J, Croxall JP, Oro D, Ruiz X (2007) Trans-equatorial migration and mix- ing in the wintering areas of a pelagic sea- bird. Front Ecol Environ 5:297−301 González-Solís J, Smyrli M, Militão T, Grémil- let D, Tveraa T, Phillips RA, Boulinier T (2011) Combining stable isotope analyses and geolocation to reveal kittiwake migra- tion. Mar Ecol Prog Ser 435:251−261 Graham BS, Koch PL, Newsome SD, McMa- hon CR, Aurioles-Gamboa D (2010) Us- ing isoscapes to trace the movements and foraging behavior of top predators in oce- anic ecosystems. In: West JB, Bowen GJ, Dawson T, Tu KP (eds) Isoscapes: under- standing movement, pat- tern, and process on Earth through isotope mapping. Spring- er, New York, NY, p 299−318 Guilford T, Meade J, Willis J, Phillips RA and Camphuysen KCJ, Garthe S (2004) Recording foraging seabirds at sea: standardised re- cording and coding of foraging behaviour and multi-species associations. Atl Sea- birds 6:1−32 Carboneras C (1992) Family Procellariidae (petrels and shearwaters). In: Del Hoyo J, Elliot A, Sargatal J (eds) Handbook of the birds of the world, Vol 1. Lynx Edicions, Barcelona, p 216−271 Carboneras C, Jutglar F, Kirwan GM (2016a) Audubon’s shearwater (Puffinus lhermin- ieri). In: del Hoyo J, Elliott A, Sargatal J, Christie DA, de Juana E (eds) Handbook of the birds of the world alive. Lynx Edicions, Barcelona Carboneras C, Jutglar F, Kirwan G (2016b) Lit- tle shearwater (Puffinus assimilis). In: del Hoyo J, Elliott A, Sargatal J, Christie DA, de Juana E (eds) Handbook of the birds of the world alive. Lynx Edicions, Barcelona Catry T, Ramos JA, Le Corre M, Phillips RA (2009) Movements, at-sea distribution and behaviour of a tropical pelagic seabird: the wedge-tailed shearwater in the western In- dian Ocean. Mar Ecol Prog Ser 391:231−242 Catry T, Ramos JA, Catry I, Monticelli D, Granadeiro JP (2013) Inter-annual variabil- ity in the breeding performance of six tropi- cal seabird species: influence of life-history traits and relationship with oceanographic parameters. Mar Biol 160:1189−1201 Cherel Y, Hobson KA (2007) Geographical variation in carbon stable isotope signa- tures of marine predators: a tool to inves- tigate their foraging areas in the Southern Ocean. Mar Ecol Prog Ser 329:281−287 Cherel Y, Le Corre M, Jaquemet S, Ménard F, Richard P, Weimerskirch H (2008) Re- source partitioning within a tropical seabird community: new information from stable isotopes. Mar Ecol Prog Ser 366:281−291 Cramp S, Simmons KEL (1977) Handbook of the birds of Europe, the Middle East and North Africa: the birds of the Western Pa- learctic, Vol I. Oxford University Press, London Delong RL, Stewart BS, Hill RD (1992) Docu- 39 Year-round movements of a small seabird and oceanic isotopic gradient in the tropical Atlantic others (2009) Migration and stopover in a small pelagic seabird, the Manx shearwater Puffinus puffinus: insights from machine learning. Proc Biol Sci 276:1215−1223 Guilford T, Wynn R, McMinn M, Rodríguez A and others (2012) Geolocators reveal migration and pre-breeding behaviour of the critically endangered Balearic shear- water Puffinus mauretanicus. PLOS ONE 7:e33753 Hazevoet CJ (1995) The birds of the Cape Verde Islands. An annotated check list. British Ornitologists Union, London Hazevoet CJ (1997) Notes on distribution, conservation, and taxonomy of birds from the Cape Verde Islands, including records of six species new to the archipelago. Bull Zool Mus Univ Amst 15:89−100 Hedd A, Montevecchi WA, Otley H, Phillips RA, Fifield DA (2012) Trans-equatorial migration and habitat use by sooty shear- waters Puffinus griseus from the South At- lantic during the nonbreeding season. Mar Ecol Prog Ser 449:277−290 Hijmans RJ, Williams E, Vennes C (2012) Geo- sphere: spherical trigonometry. R package version 1.2-28. http://cran. r-project.org/ web/packages/geosphere/geosphere.pdf Hill RD (1994) Theory of geolocation by light levels. In: Le Boeuf BJ, Laws RM (eds) El- ephant seals: population ecology, behavior, and physiology. University of California Press, Berkeley, CA, p 227−236 Hill RD, Braun MJ (2001) Geolocation by light level. In: Sibert JR, Nielsen J (eds) Electron- ic tagging and tracking in marine fisheries. Reviews: methods and technologies in fish biology and fisheries, Vol 1. Springer, Dor- drecht, p 315−330 Hobson KA, Piatt JF, Pitocchelli J (1994) Using stable isotopes to determine seabird trophic relationships. J Anim Ecol 63:786−798 Jaeger A, Lecomte VJ, Weimerskirch H, Rich- ard P, Cherel Y (2010) Seabird satellite tracking validates the use of latitudinal isoscapes to depict predators’ foraging ar- eas in the Southern Ocean. Rapid Commun Mass Spectrom 24: 3456−3460 Jaquemet S, Le Corre M, Weimerskirch H (2004) Seabird community structure in a coastal tropical environment: importance of natural factors and fish aggregating devices (FADs). Mar Ecol Prog Ser 268:281−292 Kuznetsova A, Brockhoff PB, Christensen RHB (2015) lmerTest: tests in linear mixed effects models. R package version 2.0-29. https://cran.r-project.org/web/packages/ lmerTest/lmerTest.pdf Liechti F (2006) Birds: blowin’ by the wind? J Ornithol 147: 202−211 Lisovski S, Hewson CM, Klaassen RHG, Ko- rner-Nievergelt F, Kristensen MW, Hahn S (2012) Geolocation by light: accuracy and precision affected by environmental fac- tors. Methods Ecol Evol 3:603−612 Mancini PL, Hobson KA, Bugoni L (2014) Role of body size in shaping the trophic structure of tropical seabird communities. Mar Ecol Prog Ser 497:243−257 McConnell BJ, Chambers C, Fedak MA (1992) Foraging ecology of southern elephant seals in relation to the bathymetry and produc- tivity of the Southern Ocean. Antarct Sci 4:393−398 McMahon KW, Hamady LL, Thorrold SR (2013a) A review of ecogeochemistry ap- proaches to estimating movements of ma- rine animals. Limnol Oceanogr 58:697−714 McMahon KW, Hamady LL, Thorrold SR (2013b) Ocean eco- geochemistry: a review. Oceanogr Mar Biol Annu Rev 51:327−374 Mompeán C, Bode A, Benitez-Barrios VM, Dominguez- Yanes JF, Escanez J, Fraile- Nuez E (2013) Spatial patterns of plankton biomass and stable isotopes reflect the in- fluence of the nitrogen-fixer Trichodesmi- um along the subtropical North Atlantic. J Plankton Res 35:513−525 Monteiro LR, Ramos JA, Furness RW, del Nevo AJ (1996) Movements, morphology, breed- ing, moult, diet and feeding of seabirds in the Azores. Colon Waterbirds 19: 82−97 Müller MS, Massa B, Phillips RA, Dell’Omo G (2014) Individual consistency and sex dif- ferences in migration strategies of Scopolis shearwaters Calonectris diomedea despite 40 CHAPTER 1 : systematic year differences. Curr Zool 60: 631−641 Navarro J, Coll M, Somes CJ, Olson RJ (2013) Trophic niche of squids: insights from iso- topic data in marine systems worldwide. Deep Sea Res II 95:93−102 Neves VC, Bried J, González-Solís J, Roscales JL, Clarke MR (2012) Feeding ecology and movements of the Barolo shearwater Puffi- nus baroli baroli in the Azores, NE Atlan- tic. Mar Ecol Prog Ser 452:269−285 Nilsson C, Klaassen RHG, Alerstam T (2013) Differences in speed and duration of bird migration between spring and autumn. Am Nat 181:837−845 Nisbet ICT, Mostello CS, Veit RR, Fox JW, Afanasyev V (2011) Migrations and winter quarters of five common terns tracked us- ing geolocators. Waterbirds 34:32−39 Paiva VH, Fagundes AI, Romão V, Gouveia C, Ramos JA (2016) Population-scale foraging segregation in an apex predator of the North Atlantic. PLOS ONE 11:e0151340 Phillips RA, Silk JRD, Croxall JP, Afanasyev V, Briggs DR (2004) Accuracy of geoloca- tion estimates for flying seabirds. Mar Ecol Prog Ser 266:265−272 Phillips RA, Bearhop S, McGill RAR, Daw- son DA (2009) Stable isotopes reveal indi- vidual variation in migration strategies and habitat preferences in a suite of seabirds during the nonbreeding period. Oecologia 160:795−806 Pinet P, Jaquemet S, Pinaud D, Weimerskirch H, Phillips RA, Le Corre M (2011) Migra- tion, wintering distribution and habitat use of an endangered tropical seabird, Barau’s petrel Pterodroma baraui. Mar Ecol Prog Ser 423:291−302 Precheur C (2015) Ecologie et conservation du puffin d’Audubon (Puffinus lherminieri lherminieri) de la Réserve Nat- urelle des Ilets de Sainte-Anne (Martinique). PhD thesis, Zoologie des vertébrés, Université des Antilles Priddel D, Hutton I, Carlile N, Bester A (2003) Little shearwaters, Puffinus assimilis assi- milis, breeding on Lord Howe Island. Emu 103:67−70 Quillfeldt P, McGill RAR, Furness RW (2005) Diet and foraging areas of Southern Ocean seabirds and their prey inferred from stable isotopes: a review and case study of the Wilson’s storm-petrel. Mar Ecol Prog Ser 295:295−304 R Core Team (2016) R: a language and envi- ronment for statistical computing. Version 3.3.0. R Foundation for Statistical Comput- ing, Vienna. www.r-project.org Ramos JA, Fagundes AI, Xavier JC, Fidalgo V, Ceia FR, Medeiros R, Paiva VH (2015) A switch in the Atlantic Oscillation correlates with inter-annual changes in foraging loca- tion and food habits of Macaronesian shear- waters (Puffinus baroli) nesting on two islands of the subtropical Atlantic Ocean. Deep Sea Res I 104:60−71 Ramos R, González-Solís J, Ruiz X (2009) Linking isotopic and migratory patterns in a pelagic seabird. Oecologia 160:97−105 Ramos R, Sanz V, Militão T, Bried J and oth- ers (2015) Leap-frog migration and habitat preferences of a small oceanic seabird, Bul- wer’s petrel (Bulweria bulwerii). J Biogeogr 42:1651−1664 Ramos R, Ramírez I, Paiva VH, Militão T and others (2016) Global spatial ecology of three closely-related gadfly petrels. Sci Rep 6:23447 RNBWS (Royal Naval Birdwatching Society) (2014) World Database. www.rnbws.org.uk/ search-2/ (accessed May 2, 2014) Robb M, Mullarney K (2008) Petrels night and day: a sound approach. The Sound Ap- proach, Dorset Rolff C (2000) Seasonal variation in δ13C and δ15N of size-fractionated plankton at a coastal station in the northern Baltic proper. Mar Ecol Prog Ser 203:47−65 Roscales JL, Gómez-Díaz E, Neves V, González-Solís J (2011) Trophic versus geographic structure in stable isotope sig- natures of pelagic seabirds breeding in the Northeast Atlantic. Mar Ecol Prog Ser 434:1−13 Somes CJ, Schmittner A, Galbraith ED, Lehm- 41 Year-round movements of a small seabird and oceanic isotopic gradient in the tropical Atlantic ann MF and others (2010) Simulating the global distribution of nitrogen isotopes in the ocean. Global Biogeochem Cycles 24: GB4019 Spear LB, Ainley DG, Walker WA (2007) For- aging dynamics of seabirds in the eastern tropical Pacific Ocean. In: Marti CD (ed) Studies in avian biology, No. 35. Cooper Ornithological Society, Camarillo, CA, p 1–99 Surman CA, Nicholson LW, Santora JA (2012) Effects of climate variability on breeding phenology and performance of tropical sea- birds in the eastern Indian Ocean. Mar Ecol Prog Ser 454:147−157 Tasker ML, Jones PH, Dixon T, Blake BF (1984) Counting seabirds at sea from ships: a review of methods employed and a sug- gestion for a standarized approach. Auk 101:567−577 Velasco MC (2013) Mystery shearwater in Canary Islands in December 2012. Dutch Birding 35:186−189 Warham J (1990) The petrels: their ecology and breeding systems. Academic Press, London Weimerskirch H (2007) Are seabirds foraging for unpredictable resources? Deep Sea Res II 54:211−223 Fig. S1. Effect of equinoxes (red dashed lines) on estimation of latitude positions (using threshold method) of Boyd’s shearwater tracked by geolocators in Cape Verde Islands from 2007 – 2012. Horizontal black line refers to mean latitude of two colonies (Ilhéu Raso and Ilhéu de Cima). Positions of all 38 unfiltered tracks are pooled together Supplementary material 42 CHAPTER 1 : Fi g. S 2. E ff ec t o f eq ui no xe s an d so la r an gl e on p os it io n es ti m at io ns ( us in g th re sh ol d m et ho d) o f B oy d’ s sh ea rw at er tr ac ke d by g eo lo ca to rs in C ap e V er de I sl an ds fr om 2 00 7 – 20 12 , f or (A ) l at itu de a nd (B ) l on gi tu de . H or iz on ta l bl ac k li ne r ef er s to m ea n po si ti on o f tw o co lo ni es ( Il hé u R as o a nd Il hé u d e C im a) , v er ti ca l da sh ed li ne s r ef er to sp ri ng (M ar ch ) an d au tu m na l (S ep te m be r) e qu in ox es . F ig ur e vi su al iz e in he re nt s ys te m at ic e rr or o f ge ol oc at io n m et ho d, w he re l at it ud e po si ti on c an b e m or e ac cu ra te ly e st im at ed d ur in g so ls ti ce s an d at h ig h la ti tu de s, b ut m or e pr ob le m at ic d ur in g th e eq ui no x pe ri od a nd a ro un d th e eq ua to r du e to th e li tt le v ar ia ti on in d ay le ng th ( H il l 1 99 4) 43Supplementary material F ur th er m or e, t he e rr or i n la ti tu de i s n ot c on st an t an d is o ve r- o r u nd er es ti m at ed d ep en di ng o n th e pr ox im it y to v er na l/ au tu m na l e qu in ox , h em is ph er e an d m is m at ch b et w ee n li gh t t hr es ho ld v al ue a nd s ol ar a ng le u se d fo r an al ys is ( L is ov sk i e t a t. 20 12 ). S ta rt s on th e da y of d ep lo ym en t a nd e nd s on th e da y of r et ri ev al o f th e lo gg er ( or w he n lo gg er s to pp ed c ol le ct in g da ta ) 44 CHAPTER 1 : Fig. S3. Filtered monthly locations of Boyd’s shearwaters (n = 38 tracks) tracked with geolocators on Cape Verde Islands (Ilhéu Raso and Ilhéu de Cima, marked as yellow squares) from 2007 – 2012, during their breed- ing (September – April) and non-breeding (May – August) period. The lack of locations in March and September is due to filtering process in which positions close to equinoxes were eliminated (see Methods). Bathymetry used as background 45Supplementary material Fig. S4. Migratory corridors of Boyd’s shearwaters tracked with geolocators from 2007 – 2012. All filtered latitudinal positions (see Methods) of (A) postnuptial migration ranged between 2ºS and 24ºN (n = 36), as a migratory corridor we defined a range between mean ± 1 SD of those positions, resulting in corridor between 7° – 15°N. (B) Prenuptial migration positions ranged between 15ºS and 27ºN (n = 20), resulting in corridor be- tween 6º and 15ºN, which overlapped with postnuptial corridor. Solid and dashed lines represent mean ± 1 SD, respectively. Colours refer to different individuals, the same colour does not imply the same individual in (A) and (B). A 10°S 0° 10°N 20°N 30°N 90 100 110 120 130 140 150 160 170 La tit ud e B 10°S 0° 10°N 20°N 30°N 210 220 230 240 250 260 270 280 290 300 Day of year La tit ud e 46 CHAPTER 1 : A B 6.0 7.0 8.0 9.0 10.0 −18.0 −17.0 −16.0 −15.0 −14.0 50°W 40°W 30°W 20°W 50°W 40°W 30°W 20°W Longitude Longitude δ1 5 N (‰ ) δ1 3 C (‰ ) Year 2007 2008 2009 2010 2011 Fig S5. Relation between the (a) δ15N and (b) δ13C values of 8th secondary feather (S8) of Boyd’s shearwaters tracked with geolocators (n=32, 2007-2012) and the longitude of the centroid, reflecting the area exploited during the non-breeding season (May-August). Points represent individual centroids of 50 % kernel utilization distribution during the non-breeding. Dark grey line refers to intercept and slope for fixed factor of linear mixed model (longitude as fixed factor, with year and individual as random, see Results and Fig. 5a-b) for all years pooled together. Dashed lines (coloured by year) refer to intercept and slope of simple linear regression of isotopic values and longitude to visualize that general pattern is maintained over years LITERATURE CITED: Hill RD, Braun MJ (2001) Geolocation by light level—The next step: Latitude. Electron Tagging Track Mar Fish:315–330 Lisovski S, Hewson CM, Klaassen RHG, Korner-Nievergelt F, Kristensen MW, Hahn S (2012) Geolocation by light: accuracy and precision affected by environmental factors. Methods Ecol Evol 3:603–612 47Supplementary material δ15N δ13C Feather Estimate SE p-value Estimate SE p-value P1 - S1 1.146 0.158 < 0.001 -0.640 0.066 < 0.001 P1 - S8 1.183 0.158 < 0.001 -0.762 0.066 < 0.001 P1 - R6 0.247 0.158 0.719 -0.313 0.066 < 0.001 S1 - S8 0.037 0.158 1.000 -0.122 0.066 0.417 S1 - R6 -0.899 0.158 < 0.001 0.327 0.066 < 0.001 S8 - R6 -0.936 0.158 < 0.001 0.448 0.066 < 0.001 δ15N δ13C Random eff. Variance SD % Variance SD % individual 0.322 0.567 39.31 0.029 0.171 20.20 year 0.100 0.317 12.24 0.045 0.213 31.26 residual 0.397 0.623 48.45 0.070 0.265 48.54 δ15N δ13C Model Term Estimate SE t-value Estimate SE t-value Long (Intercept) 12.358 0.632 19.542 -17.941 0.373 -48.100 longitude 0.133 0.018 7.612 -0.055 0.010 -5.398 Long + Lat (Intercept) 12.531 0.739 16.946 -17.933 0.414 -43.314 latitude -0.008 0.020 -0.385 -0.001 0.011 -0.062 longitude 0.136 0.019 7.358 -0.055 0.010 -5.224 Null (Intercept) 7.604 0.153 49.597 -15.990 0.127 -126.377 Lat (Intercept) 7.242 0.318 22.775 -15.888 0.186 -85.536 latitude 0.038 0.031 1.229 -0.011 0.015 -0.715 δ15N δ13C Random effect Estimate SD % Estimate SD % individual 0.103 0.322 69.26 0.000 0.321 0.00 year 0.000 0.000 0.00 0.036 0.000 53.38 residual 0.046 0.214 30.74 0.032 0.214 46.62 Table S1. a) Differences of least squares means and standard errors among sampled feathers (P1, S1, S8 and R6) of Boyd’s shearwater based on the linear mixed effects models (LMM) that included individual and year as random factors (see Results). P-values are adjusted using Bonferroni correction; significant differences are highlighted in bold (b) Variance of random effects explained in models Table S2. (a) Model estimates and standard errors of linear mixed models (LMM) testing for spatial gradient in isotopic values of nitrogen and carbon of Boyd’s shearwaters breeding in Cape Verde Islands. The best-sup- ported models (in bold, see Results and Table 3) include longitude as fixed factor. All models include individual and year as random factors (b) Variance explained by random factors of best-supported longitudinal gradient model Chapter 2: Common Terns on the East-Atlantic Flyway: Temporal-spatial distribution during the non-breeding period Peter H. Becker1, Heiko Schmaljohann1,2, Juliane Riechert1, Götz Wagenknecht1, Zuzana Zajková3, Jacob González-Solís3 1 Institute of Avian Research „Vogelwarte Helgoland“, An der Vogelwarte 21, D-26386 Wilhelmshaven, Germany 2 University of Alaska, Fairbanks, AK, USA 3 Institut de Recerca de la Biodiversitat (IRBio) and Departament de Biologia Animal, Universitat de Barcelona, Av Diagonal 643, 08028 Barcelona, Spain Published in: Becker, P.H., Schmaljohann, H., Riechert, J., Wagenknecht, G., Zajková, Z. & González‐ Solís, J. (2016) Common Terns on the East Atlantic Flyway: temporal–spatial distribution during the non‐breeding period. J. Ornithol. 157: 927– 940 ABSTRACT We studied the temporal–spatial distribution of Common Terns Sterna hirundo along the East Atlantic Flyway. In 2009 and 2010 experienced adults from a colony on the German North Sea coast were tagged with geolocators recording light intensity and saltwater contact. Main objectives were the inter-individual temporal–spatial variation of migration routes and wintering areas, wintering site fidelity, and time spent at sea across the annual cycle. Geolocators had no effects on various traits of breeders, but their reproductive output suffered from egg breakage. This can be avoided by artificially incubating the eggs. Twelve routes of nine individuals were tracked. Transponder read- ings at the breeding site showed that birds left the colony 4 weeks before starting autumn migration. In spring and autumn, Common Terns stopped over around the Canary Islands. Main wintering distribution was the upwelling seas alongside the West African coast and similar between years, but different among individuals. Three females wintered further north and more offshore than six males. Pair mates wintered at different locations. Spring migration was longer (56 ± 8 days) than autumn migration (37 ± 17 days). During both migration and wintering the terns spent more time on salt water than during breeding and post-breeding. In most individuals saltwater contact was higher during the day than at night, reduced at sunrise and sunset likely due to foraging, and peaked about noon possibly related to resting or thermoregulation. Detailed ecological and behavioral studies of common terns during wintering are needed to clarify the results based on geolocators. 50 CHAPTER 2: Common Terns on the East-Atlantic Flyway: INTRODUCTION Seabirds spend most of their non-breeding pe- riod far offshore at the oceans, e.g. Shaffer et al. (2006), Guilford et al. (2009), and Egevang et al. (2010). This makes it difficult studying their behavior during these times. By analyz- ing the stable isotope composition of feath- ers grown outside the breeding area, we gain information about the birds’ diet composition and how this might affect other life-history stages, e.g. Sorenson et al. (2009). Ring recov- eries might give us some indication about the birds’ whereabouts during the non-breeding period, but these recoveries seem to be highly aged-biased in seabirds (Wendeln and Becker 1999; Bairlein et al. 2014). Although both meth- ods can be used to study the ecology and the behavior of seabirds away from their breeding areas to a certain extent, different types of log- gers offer the opportunity to estimate seabirds’ behavior during migration and winter on a more precise scale. After recapture such log- gers provide data about, e.g. GPS coordinates (Weimerskirch et al. 2002), light-level geoloca- tions (Weimerskirch and Wilson 2000), three- dimensional acceleration (Sommerfeld et al. 2013), heart rate (Ropert-Coudert et al. 2006), water depth (Garthe et al. 2000), temperature (Wilson et al. 1992a), saltwater contact (Wil- son et al. 1995), and others (Wilson et al. 2002). So far these studies have been limited to rather large seabirds, because neither the size nor the weight of the specific loggers have allowed de- ploying these devices to small seabirds, i.e., with body mass<100 g. Only little, therefore, was known about the whereabouts and their behavior during the migration and wintering period of such seabirds. The miniaturization of light-level geolocators now allows tracking also these smaller seabirds such as terns (e.g. Egevang et al. 2010; Nisbet et al. 2011a, b; Fijn et al. 2013; van der Winden et al. 2014). Here we add to better knowledge about the ecology of seabirds during the non-breeding period by estimating the temporal–spatial dis- tribution of European Common Terns (Sterna hirundo) along the East Atlantic Flyway. To do so we tagged adult Common Terns with data loggers at a breeding colony site in northwest- ern Germany (e.g. Becker et al. 2008) to record light levels and wet–dry conditions. The main objectives of this study were to estimate the inter-individual temporal–spatial variation of both their migration and wintering period, to explore potential sex-specific and within-pair differences of the wintering area, and to quan- tify the birds’ behavior across the annual cycle in relation to the individual time spent on sea water. METHODS Study site Common Terns considered in this study bred at a monospecific colony of about 400 breeding pairs located at “Banter See” at Wilhelmshav- en on the German North Sea coast (53º36’N, 08º06’E, Becker et al. 2001, 2008; Becker 2010). This colony is the focus of an integrated, long-term population study, and about half of the breeders are aged, sexed, and marked with transponders (e.g. Szostek and Becker 2012). The colony site consists of six rectangular con- crete islands (10.7 x 4.6 m), surrounded by a wall of 60 cm height. The walls are equipped with 44 elevated platforms for terns to land and rest on. Each platform contains an antenna reading transponder codes every 5 s, and half of them contain an electronic balance (accu- racy ±1 g). This allows reliable automatic and remote detection of the birds’ presence at the colony site, arrival, and body mass (Limmer and Becker 2007), with a reencounter probabil- ity of almost 1 (Szostek and Becker 2012). Col- ony site fidelity is very high (adult local return rate ca. 90 %; Ezard et al. 2006; Szostek and Becker 2012). The first and last transponder reading of an individual in a season indicated that the bird had arrived and left the breed- ing colony, respectively (Becker et al. 2008). For simplicity birds are called by individual names. Reproductive performance and output was determined for each clutch including those of geolocator-marked parents using standard 51Temporal-spatial distribution during the non-breeding period protocols (e.g. Becker and Wink 2003; Zhang et al. 2015). For chicks, maximum mass, mass at fledging (±1 g), and age at fledging (±1 day) were recorded (Becker and Wink 2003). Capture and deployment of light-level geo- locators Experienced breeders (9–14 years old, in 2009 and 2010 both pair members; Table S1) were identified by the transponder with a nest anten- na and caught on the nest with an electronically released drop trap (or spring trap in exceptional cases) during incubation, on average 12 days after laying the first egg (Table S1). Before catching the birds, their eggs were replaced by dummy eggs to avoid egg breakage. The cap- tured adults were weighed (±1 g, digital bal- ance), measured (head and bill length ±0.1 mm; wing length 0.5 mm), and tagged with light- level geolocators (Fig. S9). Total handling time was 3–6 min. Most individuals returned to the clutch a few minutes after release and started incubation soon [on average after 13±11(2–38) min, n = 11]. No clutch was deserted owing to catching the breeders. In 2011 when light-level geolocators had to be only recovered, the eggs were removed immediately from the clutch af- ter laying of the identified individuals, put in an incubator and were replaced by dummy eggs. Eggs remained in the incubator until light-level geolocators were retrieved from the adults to avoid any egg breakage. After that original eggs were exchanged again. Captures were performed earlier during incubation than in the previous years. Most individuals were captured in three successive years (Table S1). Light-level geolocators We used miniature light-level geolocators, Mk 10, from the British Antarctic Survey (BAS). They were fixed with layers of self-amalgam- ating tape to a plastic ring with cable tie (Fig. S9; 10 mm height, 5 mm internal diameter,1.0 mm thickness). In 2010, three geolocators were attached to an aluminum ring for a Black-head- ed Gull (Croicocephalus ridibundus, 10 mm height). Mass of the ring and fixing materials was <1.7 g (about 1.3 % of Common Tern body mass). At recapture, the geolocator from the previous year was removed and replaced by a new one (Table S1). During the pre-calibration period light-level geolocators experienced the unhindered natural change in light conditions at the colony site for 7–19 days. After removal a post-calibration was conducted with each light-level geolocator for 5–18 days (in 2011 at the colony, in 2009 and 2010 at the Institute of Avian Research, 53º33’N, 08º06’E). Twelve of the 24 geolocators had failed (see Table S1); reasons for data loss were infiltrated water, non-realistic shift in longitude due to internal clock shifts (Fig. S8), or insufficient lifetime of batteries. Light-level geolocators used in the present study archive maximum light intensity every 10 min. Sunrise and sunset times allow infer- ring length of day and night and the timing of midday and midnight, and finally estimate latitude and longitude twice a day (Wilson et al. 1992b; Hill 1994). As a matter of principle, latitude cannot be estimated on about 10 days around the equinoxes (Wilson et al. 1992b; Hill 1994; Lisovski et al. 2012). The general uncer- tainty of the estimated locations is generally on the order of magnitude of about 150 km (Phil- lips et al. 2004; Fudickar et al. 2012; Lisovski et al. 2012). Light-level geolocation data were analyzed using the statistical software R 3.1.2 (R Core Team 2014) and the freely available SGAT package (https://github.com/SWo therspoon/ SGAT). This packages combines tools of the R package GeoLight (Lisovski and Hahn 2012), which uses the threshold approach (Hill 1994; Ekstrom 2004), and the R package tripEsti- mation (Sumner et al. 2009), which uses the curve-fitting approach (Ekstrom 2004; Nielsen and Sibert 2007) to estimate the animals’ lo- cations. Here a threshold-based approach was used to estimate the birds’ locations via an Estelle model. A probability distribution of these locations is derived from the Markov chain Monte Carlo method with a metropo- lis sampler. In comparison to other methods 52 CHAPTER 2: Common Terns on the East-Atlantic Flyway: of estimating birds’ locations from light-level geolocation data, here a priori knowledge can be used to estimate locations by considering (1) a species-specific movement model, which is described by a bird’s ground speed, (2) a species-specific land mask model, and (3) that the errors in the twilight times, which follow a log normal distribution. Following these as- sumptions, probability distributions of the lo- cations are estimated. The movement model defines the density distribution of travel speed, which is described here by a gamma distribu- tion. As air speed of common terns is about 11 m/s (Bruderer and Boldt 2001; Pennycuick et al. 2013) and as terns in general exploit favor- able wind conditions (Egevang et al. 2010), we arbitrarily set mean ground speed to 15 m/s. To determine the density distribution of ground speeds, all locations of a bird were initially es- timated with the threshold-sensitivity twilight function threshold.path and used to estimate the ground speed for the initial track. This was on average 14.66 ± 1.05 m/s (mean ± SD; n = 11) and similar to the arbitrarily chosen ground speed. In a second step, we excluded extremely high speeds which are associated with errone- ously estimated locations. The mean and SD of these remaining speed values were used to es- timate both the shape parameter (1.51) and rate parameter (0.13) of the corresponding gamma distribution (Becker et al. 1988). This gamma distribution fitted well the density distribution of the ground speed during the tracking period (Fig. S1). The land mask model allows setting different probabilities for the bird being on land or on water. We set the probability of a Com- mon Tern to be near or over water two times higher than being over land because Common Terns are typical seabirds (Harrison 1997; Nis- bet et al. 2011a; Neves et al. 2015) and because the vast majority of ring recoveries from mid- European breeding populations comes from the West African coast and not from inland sites, indicating the wintering grounds to be on or even off the West African coast (Wernham et al. 2002; Bairlein et al. 2014). When sunrise and sunset events are not affected by artificial light, light cannot be detected before sunrise or after sunset by the light sensor. Hence, twilight er- rors are not normally distributed, but described by a lognormal distribution, as twilight error of recorded light cannot be negative (Fig. S2). We considered these assumptions in our anal- yses of estimating birds’ locations (for details and R-code see https://github.com/SWother- spoon/SGAT). The resulting estimates in re- spect of longitude and latitude and their cor- responding 95 % confidence intervals are given for each individual in the electronic supple- mental material (Fig. S3). We defined departures and arrivals from sta- tionary sites, i.e., breeding area, stopover sites, and wintering grounds, as obvious changes in longitude and/or latitude (Fig. S3). In the latter, changes were only considered outside 10 days before and 10 days after the equinoxes. Be- cause of corrupt data and heavy outliers (Fig. S3) the changeLight function of the GeoLight R packages (Lisovski and Hahn 2012) to estimate the migration schedule did not work properly. The values describing the individual migra- tory schedules should be treated cautiously. The estimated start of spring migration, e.g. in Cornelia and Joachim (Table 1; Fig. S3) could also be attributed to the start of movements in the wintering area. Some light-level geoloca- tors broke before detachment, and in some the internal geolocator clock drifted (Figs. S3, S8). The area that was visited during winter time was individually estimated based on light-level geolocation estimates (Fig. S3; Table S2). How- ever, we did not consider location estimates derived before 1 November and after 28 Feb- ruary to minimize the influence of the equi- noxes on the latitudinal estimates (Table S2). The centroid of the wintering ground for each individual was estimated as the mean ± SD of the estimated locations which are all shown in the corresponding figures. Stopover sites could only be determined for three individu- als (Table S2). Kernel densities (45, 75, and 95 %; Epanechnikov kernel) were calculated for wintering grounds of different sets (sex, year) of individuals using the kernelUD function of the R-packages adehabitatHR (Calenge 2006). The ad hoc method was used for the smoothing 53Temporal-spatial distribution during the non-breeding period Table 1 Departure and arrival dates (day–month) of common terns at the breeding and wintering area based on tracks by light-level geolocators and on remote identification by transponders at the colony site Twelve tracks of nine individuals were achieved. Pair mates are italicized ND no data, not analyzed, Diff difference of date of first or last record at colony, date based on geolocator data given in days. Only three individuals had fledged young parameter. The grid was set to 500. The same settings were applied when estimating kernel densities for stopover sites. The distance be- tween the breeding area and the average win- tering ground was calculated as the great circle distance between these locations. Time spent on salt water The Mk 10 BAS geolocators also recorded salt- water immersion every 3 s and stored number of positive records ranging from 0 (continu- ously dry) to 200 (continuously wet) at the end of each 10-min period (“wet– dry” informa- tion). Immersion data were available for eight individual tracks (two females, six males, Ta- ble S3). We estimated the average proportion of time spent on saltwater per hour (0–24 h, Greenwich Mean Time, GMT) and per day (in hours or % of 24 h, and for wintering at the latitude of Dakar, Senegal, we differentiated between daylight (7:30–18:45) and night hours (18:45–7:30). Defining stages of the annual cycle Based on the individual light-level geolocation data combined with data from transponders at the colony site (Table 1; Fig. S3) we defined for each individual six different annual stages: Breeding stage: the bird was at the colony. Post-breeding stage: the bird had left the col- ony, but remained in the vicinity of the German Bight and did not start its autumn migration. Autumn migration: the bird was on the move, but had not reached its wintering area. Bird Departure date at breeding area Wintering area Arrival date at breeding area Name Sex Year Last record colony Geo- locator Diff (days) Arrival date Depar- ture date Geo- locator First record colony Diff (days) Joachim M 2009/10 02-09 02-09 0 22-10 18-02 27-04 28-04 1 2010/11 02-09 21-09 -19 23-10 ND ND 23-04 - Moses M 2009/10 31-07 12-09 -43 08-10 03-03 27-04 03-05 6 2010/11 12-07 06-09 -56 28-10 ND ND ND - Kasimir M 2009/10 12-07 12-09 -62 11-10 ND 28-04 25-04 -3 Cornelia F 2009/10 28-07 28-07 0 08-10 ND ND 26-04 - 2010/11 22-07 22-07 0 29-07 19-02 14-04 14-04 0 Heinera M 2010/11 24-08 06-09 -13 01-11 15-02 11-04 18-04 7 Aylaa F 2010/11 24-08 06-09 -13 25-10 15-02 18-04 14-04 -4 Ernsta M 2010/11 07-08 21-09 -45 28-10 23-02 13-04 14-04 1 Wieland M 2010/11 26-07 11-09 -47 12-10 08-03 23-04 ND - Marianna F 2009/10 15-07 30-08 -46 21-10 ND ND 25-04 - Mean ± SD 04-08 ± 20 02-09 ± 19 -31b ± 23 13-10 ± 25 22-02 ± 8 20-04 ± 7 22-04 ± 7 1b ± 4 a Care of juveniles on migration b Calculated for the individual differences 54 CHAPTER 2: Common Terns on the East-Atlantic Flyway: Wintering: the time after arrival at the win- tering area and before spring migration. Spring migration: the bird started its spring migration and had not reached the colony. Pre-breeding: spring migration was finished, but the colony site not reached (sufficient data only in one individual, Table S3). Defining these stages based on light-level ge- olocation data was a rough estimate, and small differences between these stages with respect to saltwater contact should be interpreted cau- tiously. Statistics Data were analyzed using the statistical soft- ware R 3.1.2 (R Core Team 2014). To assess whether individual birds being tracked for two consecutive winters showed significantly high- er winter area fidelity than the population on average, we performed a randomization test, randomly selecting 10,000 pairs of mean win- tering locations from our data set. We did not allow that a pair of mean locations consisted of the same locations. If the within-individual difference of the two tracked mean wintering locations were shorter than the 250 shortest distances between randomly selected pairs of mean wintering locations, birds were assessed more faithful than expected by chance. We tested for seasonal differences in at-sea activity between stages (without the pre-breed- ing period, owing to insufficient data) using GLMRM (generalized linear model for repeat- ed measurements, SPSS 22). The Mann–Whit- ney U test was applied when comparing non- parametric differences between two groups. The Wilcoxon signed-rank test was used as a non-parametric test for paired samples. If not otherwise stated values are reported as mean ± 1 SD. RESULTS Retrieval of geolocators Twenty-five out of the 29 tagged birds, i.e., 86 %, returned to the breeding colony the year after deployment. All individuals carrying a light-level geolocator bred in their returning year (Table S1). No bird showed any signs of leg injuries when light-level geolocators were removed. One female had lost her light-level geolocator (Table S1). Twelve of the 24 light- level geolocators contained analyzable data by nine adults (three females and six males, in- cluding three pairs). Potential effects of geolocators Carrying light-level geolocators did not signifi- cantly affect both arrival and laying date, mass at arrival, mass at catching, clutch size, body mass growth of chicks, and ability to fledge chicks (see chapter “Additional information about potential effects of geolocators on com- mon terns” in Electronic Supplementary Mate- rial). However, we recorded a strong and sig- nificant deterioration of hatching success from 86 to 43 % reducing reproductive output of pairs marked with geolocators severely (Tables S5, S6). The reduced hatchability was caused by eggshell breakage owing to fine fissures in- creasing with time advancing of incubation by the marked individuals (Figs. S9, S10). In 2011, i.e., the last year of this study, reproductive success of geolocator-birds was successfully increased by exchanging pairs’ original eggs with dummy eggs, and incubating the original eggs in an incubator until geolocators were re- trieved. These measures had increased hatch- ing success to 89 % (for details see Electronic Supplementary Material, Table S6). General temporal–spatial distribution of Common Terns during the non-breeding pe- riod As Common Terns mainly migrated during both equinoxes (Fig. S3), we dispensed with a detailed temporal–spatial analysis of indi- vidual movements between the colony and the wintering areas. Birds left the colony on average on 4 August ±20 days (range 12 July–2 September) and abandoned the German Bight on 2 September 55Temporal-spatial distribution during the non-breeding period Fig. 1 Wintering and stopover locations at Canary Islands of 12 routes of nine Common Terns tracked with light-level geolocators between 2009 and 2011. Breeding site large black dot. Large black triangles (females) and black circles (males) mean wint er locations ±SD. Dotted lines 95 kernel densities; dashed lines 75 kernel densities; solid lines 45 kernel densities. Kernel densities at wintering sites were highlighted in three different shades of grey. Birds migrated to their winter locations by flying mainly over water. Small black dots indicate African ring recoveries during December and January of adult common terns from northwest German breeding sites (Helgoland ringing center, n = 30; age at ringing older than 1 year or period between ringing and recovery date >3 years; cf. Bairlein et al. 2014). Map is Mercator projection 56 CHAPTER 2: Common Terns on the East-Atlantic Flyway: ±19 days (22 July–21 September; Table 1). In general, the data suggested that common terns moved along the East Atlantic Flyway and that they predominantly used offshore migration routes (Fig. S3). The sea around the Canaries was identified as a stopover area (Fig. 1; Table S2): two individuals stopped there during au- tumn migration. One remained in this area ap- proximately for 7 days (Moses in 2010) and the other slightly less than a month (Cornelia in 2009; Table 1, Table S2). Also, during spring migration one individual (Kasimir) stopped there (Table S2). Within 13 days after resuming migration from this stopover area the bird (Ka- simir) reached the colony (Table 1, Table S2). Common Terns arrived at the wintering areas on 13 October ± 25 days (29 July–1 November, Table 1). Mean wintering period lasted 136 ± 34 days (n = 7, calculated by the individual dif- ferences, cf. Table 1). Their preferred winter- ing areas were the upwelling seas alongside the West African coast of Morocco, Western Sa- hara, Mauritania, Senegal, The Gambia, Guin- ea Bissau, Guinea, and Sierra Leone (Fig. 1). Mean great circle distance between the colony and the individual mean wintering locations was 4,782 ± 467 km (range 3,881–5,368 km, n = 12). In autumn, this distance was covered in 41 ± 17 days (n = 12, calculated by the individual differences, cf. Table 1). The mean distances covered per day during southward migration was 158 ± 132 km (n = 12). The four females spent the winter further north (females 20 ± 2.5ºN, range 18–24ºN, males 13 ± 3.8ºN, range 9–19ºN; Mann–Whitney U test: U = 30, p = 0.016; Fig. 1) and seemingly more offshore than the eight males (males 107 ± 57 km, range 30– 217 km; females 293 ± 255 km, range 86–624 km; Mann–Whitney U test: U = 44, p = 0.174). The winter distributions were not obviously different between the 2 years (Fig. S5). There was no indication for significant wintering site fidelity, however, as the within- individual dis- tance of the tracked mean wintering locations were not shorter than expected by chance in comparison to the between-individual distance of the mean wintering locations (Figs. S4–S6). In the three pairs for which light-level geo- location data were available for both partners (Table 1), the general wintering areas and the estimated mean wintering locations did not overlap between the sexes (Fig. 2). There was some spatial overlap of the general winter- ing area of Cornelia and Kasimir (Fig. 2), but they seemed to be temporally separated (Fig. S3). Distance of pair members’ mean winter- ing locations was 897 ± 320 km (530–1,120 km, n = 3) and with longer than the median great circle distance (647 km) of the 10,000 random- ly chosen mean wintering location pairs (Fig. S4). These sex-specific differences in the mean location of the general wintering areas within breeding pairs supported the general picture of females wintering further off-shore and unre- lated to their mates. Spring migration started on average on 22 February ± 8 days (15 February–8 March, Table 1). Common terns arrived at the breeding grounds on 20 April ± 7 days (11–28 April) so that total time of migration was about 56 ± 8 days (mean ± SD, n = 7) in spring. The mean distance covered per day during northward mi- gration was 88 ± 20 km (n = 7). For these seven birds spring migration lasted significantly lon- ger than autumn migration (autumn: 37 ± 17 days; Wilcoxon signed-rank test: V = 0, p = 0.036, n = 7). Common Terns spent about 117 ± 8 days (n = 11) at the breeding colony or in the vicinity of the colony during the reproduc- tive season. Based on transponder data only the tracked common terns stayed 96 ± 23 days (n = 16) at the colony site. The within-individual variation of the migra- tion schedule between 2 years varied in gen- eral by a few days (Table 1). In 2009 Joachim and Cornelia and in 2010 only Cornelia left the colony and the breeding area on the same day, i.e., autumn migration started on the day indi- viduals were last recorded at the colony by their transponder. Cornelia arrived at the wintering area in the beginning of October in 2009, but to the end of July in 2010. This between-year dif- ference in the estimated arrival time at the win- tering area was not explained by the between- year variation in the start of autumn migration (about 1 week). The return of the young of Ayla, 57Temporal-spatial distribution during the non-breeding period Fig. 2 Wintering areas of pair mates tracked during the same winter (Ayla, Heiner 2009/2010; Cornelia, Kasimir 2010/2011) or with male one winter later (Marianna 2009/2010, Wieland 2010/2011). Grey dots female; black dots male locations. Symbols and kernel densities (females highlighted in grey) as described in Fig. 1 Heiner, and Ernst (Table 1) as prospectors to the colony 2 years later showed that post-fledg- ing parental care of these parents was success- ful. The temporal patterns of Ayla’s, Heiner’s, and Ernst’s autumn migration, however, were not distinctively different from the adults fail- ing to produce fledglings (Table 1). Arrival and departure dates at the colony site: a comparison of transponder data and light-level geolocation estimates After leaving the breeding colony (transponder data) it took on average 31 days before Com- mon Terns started their autumn migration (Ta- ble 1; Fig S3). Only two birds had left both the colony site and the breeding area on the same day (Joachim and Cornelia, Table 1; Fig S3). In spring, however, arrival date at the breeding colony detected with the transponder recording system was similar to the estimated arrival date by light-level geolocation data (Table 1). Saltwater contact during the annual cycle The proportion of time spent on salt water var- ied among individuals and stages (Fig. 3, Fig. S7; Table S3). The differences between the 58 CHAPTER 2: Common Terns on the East-Atlantic Flyway: Fig. 3 Seasonal variation in the temporal proportion of saltwater contact across different stages of the an- nual cycle. Means of daily percentage of time eight common terns had contact with salt water recorded by using saltwater immersion data from geolocators (B breeding, PB post-breeding, AM autumn migra- tion, W wintering, SM spring migration) stages of the annual cycle were highly signifi- cant (F = 10.228, p < 0.001, n = 6; 3 stages, F = 11.711, p = 0.002, n = 8; Fig. 3). During breeding and post-breeding, common terns spent only a small proportion of time on saltwater (1.1–3.5 %). During autumn migration, wintering, and spring migration, however, individuals spent significantly more time on salt water (8.6–13.9 %; Fig. 3, Fig. S7; Table S3 with statistics among single periods). Inter-individual differ- ences were consistent between stages: during all periods, e.g. Ayla or Joachim spent more time at sea than, e.g. Heiner and Moses (be- tween subject effects, F = 37.325, p = 0.002, n = 8; Fig S7; Table S3). There was a tendency that individuals wintering more offshore had more water contact than birds wintering closer to the coast (correlation between proportion of time at sea water with distance from the coast, Pear- son, r = 0.624, p = 0.098, n = 8). Furthermore, the daily proportion of time spent at seawater during winter was significantly and positively correlated with the latitude of mean wintering locations of the com- mon terns studied (Pear- son, r = 0.743, p = 0.035, n = 8). The time spent with saltwater contact varied over the course of the day with respect to the stages of the annual cycle (Fig. 4). During both autumn and spring migration and during win- ter, Common Terns spent about 10–15 % of the time on salt water during the night. At times around sunrise and sunset proportion of wa- ter contact was minimal, but highest between these events (Fig. 4), peaking between 11 and 15 GMT. There was no clear daytime pattern for the other stages of the annual cycle (Fig. 4). With respect to day and night differences in winter, five out of seven individuals had more saltwater contact at daylight than during the night, for the other two individuals it was vice versa (Cornelia and Joachim, who also had most saltwater contact in total, cf. Table S3). DISCUSSION Our results show that common terns from the breeding colony in Germany winter in the fish- rich upwelling off the West African coast (Gre- cian et al. 2016; Fig. 1). Females’ wintering ar- eas were situated further to the north by 7º than that of males. The proportion of time birds had direct contact with salt water varied between the different stages of the annual cycle: while at the breeding area saltwater contact was low, it was high during the migration and winter- ing periods (Fig. 3). This difference across the annual cycle might be explained by the daily variation of saltwater contact (Fig. 4). Potential effects of geolocators Despite the phenomenon of egg breakage (Fig. S10 and below) we found no adverse effects of birds being tagged at the tarsus with a light-lev- el geolocator neither on return rate, body con- dition, nor arrival date after spring migration or laying date. Return rate to the colony was 59Temporal-spatial distribution during the non-breeding period Fig. 4 Daily saltwater contact pattern. Mean hourly percentage of time spent on salt water ± standard error of seven common terns recorded using geolocation-immersion loggers during different stages of the annual cycle. Means of values were first calculated for individual birds, then averaged for all birds (without Ayla owing to clock shift, Fig. S8). Vertical lines refer to mean sunrise and sunset hour during wintering. Codes for stages as in Fig. 3 in the range known for this and other colonies of the common tern (Ezard et al. 2006; Szostek and Becker 2012; Nisbet and Cam 2002; Breton et al. 2014; Palestis and Hines 2015). Return rate of tagged birds was also similar to the rates as reported from other light-level geolocation studies of Sterna terns in general (Nisbet et al. 2011a; Fijn et al. 2013). Returned Common Terns equipped with geolocators were in good physical condition like Arctic Terns (Sterna paradisaea, Egevang et al. 2010; Fijn et al. 2013) and showed no reduction of body mass at arrival or when recaptured. This is in contrast to the findings of Nisbet et al. (2011a) in Com- mon Terns and Mostello et al. (2014) in Roseate Terns Sterna dougallii. Neither arrival date of the birds repeatedly measured before, during, or after deployment of the geolocators nor lay- ing date was affected (for further details see Electronic Supplemental Material). Thus, the various parameters recorded in the individu- als tagged with light-level geolocators make us confident that the geolocators did not nega- tively affect the temporal-spatial distribution of the Common Terns during their non-breeding period. After return all experimental birds produced normal clutch sizes (in contrast to Arctic Terns, Egevang et al. 2010), but suffered from in- creased egg breakage (cf. Nisbet et al. 2011a). This was caused by the geolocator and depen- dent on the number of days the eggs were incu- bated by a parent carrying a geolocator. Thus, effects of geolocators on the individual fitness can be serious (cf. Scandolara et al. 2014 for barn swallows Hirundo rustica). This effect, 60 CHAPTER 2: Common Terns on the East-Atlantic Flyway: however, can be minimized by exchanging nat- ural eggs with dummy eggs soon after laying and by artificially incubating the natural eggs until deployment of the geolocator, or even un- til hatching. General temporal–spatial distribution of common terns during the non-breeding pe- riod In agreement with recoveries of adult Com- mon Terns ringed during the breeding period in Germany, this study confirms that individu- als from our study site mainly winter in coastal West Africa (Fig. 1). However, ring recover- ies suggested that the wintering area of adults from eastern, but also from western Germany is further extended to the south of western Af- rica than pictured by the birds from Banter See colony (Fig. 1, cf. Neubauer 1982; Bairlein et al. 2014). Common Terns made use of the up- welling zone supplied by the cold Canary cur- rent off the northwest African coast (Brennink- meijer et al. 2002), where primary productivity is higher than in other areas (McGregor et al. 2007; Ar ı´stegui et al. 2009). Accordingly, the coastline of about 2,200 km along Mauritania, Senegal, Gambia, Guinea Bissau, Guinea, Si- erra Leone to Liberia is a very attractive and important wintering area for many seabird spe- cies (Grecian et al. 2016). To reach and leave this area, Common Terns might make use of stopover sites at the seas around the Canary Islands (Fig. 1), similarly to Black Terns Chli- donias niger (van der Winden et al. 2014). Like other tern species passing West African waters, Common Terns mainly use offshore migration routes (Figs. 3, 4, Figs. S3, S7), cf. Arctic Terns (Fijn et al. 2013) and Black Terns (van der Win- den et al. 2014). Wintering site fidelity is described for some seabird species (Phillips et al. 2005; Guilford et al. 2011; Dias et al. 2013). On average the three birds tracked for two seasons did not revisit the exact same wintering area (see “Results”), sug- gesting a low wintering site fidelity at a narrow spatial scale. However, this may result from a low sample size and indeed site fidelity varied substantially among individuals (Figs. S5, S6). The habitat which common terns seek for win- tering is not fixed to a certain location, because biotic and abiotic environmental conditions are on the move with the actual currents. Hence, we do not predict a similar level of high winter site fidelity as found in terrestrial bird species, e.g. Salewski et al. (2000). The general data indicate that Common Tern females wintered further north than males (Fig. 1), which was supported by within-pair data (Fig. 2). Causes are unknown, but could be re- lated to different nutritional requirements be- tween male and female Common Terns: Nisbet et al. (2002) showed that pair members of Com- mon Terns breeding at Bird Island, MA, USA, had different diets in winter. Females were supposed to feed on a higher trophic level than males. A stable-isotope analysis of feathers from individuals whose gender and wintering site are known could enlighten these interest- ing findings. Based on our light-level geoloca- tion data, we argue that pair mates do not meet during their wintering period and that in con- sequence they likely migrated separately from their mate to the colony. Similar results have been found for other seabird species, e.g. the Cory’s Shearwater Calonectris borealis (Mül- ler et al. 2015). Time schedule of the annual cycle The general timing of the stages within a year was similar between Common Terns on their East and West Atlantic Flyways (Table 1, cf. Nisbet et al. 2011a). In contrast to the more gen- eral pattern that avian spring migration is faster than autumn migration (Nilsson et al. 2013), Common Terns reached their seasonally ap- propriate migratory goal in on average 41 days in autumn, but 55 days in spring. This may be a consequence of prevailing winds, rotating clockwise in the North Atlantic and offering tailwind during autumn migration, but head- wind during spring migration (Liechti 2006). For the few birds tracked along the West Atlan- tic Flyway, however, spring migration was fast- er than autumn migration (Nisbet et al. 2011a) 61Temporal-spatial distribution during the non-breeding period again in agreement with prevailing wind direc- tions. However, these results should be treated cautiously given the location error in light-level geolocation estimates and the low sample sizes. Most adult Common Terns lingered for 4 weeks around the breeding area, as inferred by the time passed between the last detection at the colony site by the transponder system and the first sign of migration from geoloca- tion. A similar pattern was described by Nis- bet et al. (2011a) showing that adult Common Terns stayed about 100–200 km to the east or the west of the breeding colony before starting autumn migration. The reason for this behav- ior remains speculative. Possibly, adults care for their offspring, which they may guard and feed up to several weeks after fledging (Burger 1980; Becker and Ludwigs 2004; Nisbet et al. 2011b: at least until end of September; for oth- er tern species see Ashmole and Tovar 1968). Parents may familiarize their offspring with the extended surroundings of the colony site or to reach more productive feeding grounds (cf. Fijn et al. 2013). Adults may also accumulate energy, in terms of fat and muscle mass, as a preparation for the upcoming migrations. Our light-level geolocator data indicated that the delay until the final departure of adults for mi- gration was independent of sex (Table 1). This is in contrast to the findings of Nisbet et al. (2011a, b) showing that females started earlier than males presumably because the post-fledg- ling guarding is mostly provided by the fathers (Nisbet et al. 2011b). Saltwater contact during the annual cycle Common terns spent small proportions of time resting on saltwater during the breeding period (Figs. 3, 4). This saltwater contact was likely explained by bathing as Common Terns do not swim in the breeding area (PHB personal ob- servations; Nisbet 2002; Nisbet et al. 2011a). During the non-breeding season, however, the birds spent more time on salt water, confirming observations of Common Terns from the West Atlantic Flyway (Nisbet et al. 2011a; Neves et al. 2015). The inter-individual differences in saltwater contact during both migration peri- ods and wintering along the West African coast might be due to individual selection of habi- tats. In contrast to other individuals who spent most time resting at sea water during the day, Cornelia and Joachim showed high saltwater contact during the night, which they obviously had spent offshore (Fig. S8). Perhaps inter-in- dividual variation in wintering habitat selec- tion may be influenced by an extended parental care; hence, wintering on the coast might be beneficial if parents still care for their offspring (e.g. potentially in Heiner, Ayla and Ernst), so that juveniles in poor body condition can eas- ily find sites for resting on beaches or sandbars (e.g. Bugoni et al. 2005; Blokpoel et al. 1982, 1984). Whether Common Terns care for their offspring at wintering sites is still unclear, but juvenile Royal Terns Thalasseus maximus were fed by adults during wintering in Peru in De- cember and January, when they were about 7 months old (Ashmole and Tovar 1968). Changes in the daily routines of Common Terns as suggested by the saltwater contact data could likely be explained to a certain extent by their daily foraging pattern. Radio-tracked Common Terns spending the non-breeding sea- son in southern Brazil usually started foraging from roosting sites on the beach or sandbars in the morning or late afternoon (Bugoni et al. 2005). The low proportion of saltwater contact during sunrise and sunset (Fig. 4) is, therefore, likely to be related to the foraging behavior of Common Terns, considering that during the short plunge dives no saltwater contact was re- corded, cf. in the breeding period (Figs. 3, 4). Another explanation of the high proportion of saltwater contact during the non-breeding pe- riod in common terns on the West (Nisbet et al. 2011a) and East Atlantic Flyways could be ther- moregulatory necessities: during noon at areas close to the equator (Fig. 4) they may cool down their body temperature, which might be heated up considerably by the high solar irradiation. This is corroborated by the significant positive correlation of Common Terns’ saltwater con- tact per day with higher latitude of the winter- ing locations coming along with decreasing sea 62 CHAPTER 2: Common Terns on the East-Atlantic Flyway: water temperatures. Moreover, water contact was highest during spring migration (Figs. 3, 4) when also sunshine duration is highest in Sen- egal and Mauretania, concomitant with lowest sea surface temperatures due to upwelling (19– 20 ºC, February and March; e.g. Hayward and Oguntoyinbo 1987; http://www.iten-online. ch/klima/afrika) that the temperature gradient between birds’ legs and sea water should war- rant body heat release. Another explanation of longer resting times at sea during noon (Fig. 4) may be related to winds, since wind speed is typically higher at midday than at sunrise and sunset, possibly handicapping the terns’ flight. Gannets Sula bassana, too, wintering off West Africa spend more time on the sea water dur- ing daylight than conspecifics wintering at the Bay of Biscaya or the North Sea (Garthe et al. 2012). There is a need of detailed behavioral observations of terns and other seabirds in their wintering areas to clarify these speculations on persisting parental care and thermoregulation by offshore swimming. General migration patterns of Common Tern populations studied by geolocation Our study adds to the three investigations published to date of Common Tern migration based on light-level geolocators (Nisbet et al. 2011a, b; Neves et al. 2015; Moore et al., per- sonal communication). Overall, these stud- ies clearly show a strong east–west separation in their migration routes and wintering areas among breeding populations and connectivity at broad spatial scales (Fig. 5). Some studies on pelagic seabirds have also found a certain degree of migratory connectivity (e.g. Cory’s Shearwater Calonectris diomedea, González- Solís et al. 2007, Bulwer’s petrel Bulweria bul- werii, Ramos et al. 2015), but Common Terns are more coastal seabirds and their longitudinal change in migratory routes parallel those found in terrestrial birds of the Palearctic–Tropical and Nearctic–Neotropical migratory systems (e.g. Trierweiler et al. 2014; Hallworth et al. 2015). Such knowledge is important to under- stand migration strategies and for conservation concerns. Based on information about the mi- gratory connectivity we can recognize and elu- cidate impacts of population-level threats dur- ing the non-breeding period, which may affect demographic rates or traits of migration timing (e.g. in Common Terns: Szostek and Becker 2015; Szostek et al. 2015). The differences in the wintering areas and migratory flyways of Common Terns breeding, in geographical terms, in relative close vicinity to each other are striking for seabirds. Common Terns breed- ing in northwest Germany and on the Azores are separated to a larger scale in winter when visiting the West African coast or the eastern South American coast, respectively, than in summer. A similar pattern exists for the breed- ing populations in North America: Common Terns from the northeast Atlantic coast (Bird Island) spent their winter along the eastern South American coast and mix with birds from the Azores breeding population, whereas Com- mon Terns from the Great Lakes winter along the eastern Pacific coast in South America (Fig. 5). Ring recoveries suggest similar divergence of wintering sites for further common tern pop- ulations (Neubauer 1982; Bairlein et al. 2014; Cohen et al. 2014). The origin and causes of the population-specific migration patterns and wintering areas in Common Terns may be driv- en by geographical structures and barriers such as mountains, coastline courses, wind patterns, currents, water bodies, or oceans. ACKNOWLEDGMENTS We thank Christina Bauch, Alexander Braas- ch, Julia Spieker, Lesley Szostek, Katharina Weißenfels, Silas Wolf, and Christian Wolter for their help with field work. Simeon Lisovski helped with analyzing the light-level geoloca- tion data. Kathrin Hüppop helped preparing the figures. We thank Dave Moore giving access to unpublished data of migration of common terns breeding at the Great Lakes and Olaf Geiter for providing ring recovery data of common terns. The manuscript was improved by helpful com- ments of Franz Bairlein and two anonymous 63Temporal-spatial distribution during the non-breeding period Fig. 5 Breeding grounds (indicated by different symbols), migration routes (diverse lines), and wintering areas (differently shaded areas) of Common Terns tracked with light-level geolocators. Migration routes are rough estimates. Data are from four populations of Common Terns breeding in north Germany (this study), on the Azores (Neves et al. 2015), at MA, USA (Nisbet et al. 2011a, b), and Great Lakes, Canada (Moore et al., per- sonal communication) 64 CHAPTER 2: Common Terns on the East-Atlantic Flyway: reviewers. The studies were performed under license of the Nds. Landesamt für Verbrauch- erschutz und Lebensmittelsicherheit Olden- burg and of the Stadt Wilhelmshaven. H.S. is financed by the Deutsche Forschungsgemein- schaft (SCHM 2647/1-1) which also supported the project (BE 916/8 and 9). REFERENCES Arístegui J, Gasol JM, Duarte CM, Herndl GJ (2009) Microbial oceanography oft he dark ocean’s pelagic realm. Limnol Oceanogr 54:1501-1529 Ashmole NP, Tovar HS (1968) Prolonged pa- rental care in Royal Terns and other birds. Auk 85:90-100 Bairlein F, Dierschke J, Dierschke V, Salewski V, Geiter O, Hüppop K, Köppen U, Field- er W (2014) Atlas des Vogelzugs. AULA. Wiebelsheim Becker PH (2010) Populationsökologie der Flussseeschwalbe: Das Individuum im Blickpunkt. In: Bairlein F, Becker PH (Hrsg.) 100 Jahre Institut für Vogelforsc- hung „Vogelwarte Helgoland“. Aula, Wieb- elsheim, pp 137-155 Becker PH, Dittmann T, Ludwigs J-D, Lim- mer B, Ludwig SC, Bauch C, Braasch A, Wendeln H (2008) Timing of anitial arrival at the breeding site predicts age at first re- production in a long-lived migratory bird. PNAS 105:12349-12352 Becker PH, Ludwigs J-D (2004) Sterna hirun- do Common Tern. In: Parkin D (ed) BWP Update Vol. 6 Nos 1/2 , 93-139. Oxford Uni- versity Press Becker PH, Wendeln H, González-Solís J (2001) Population dynamics, recruitment, individual quality and reproductive strate- gies in Common Terns marked with tran- sponders. Ardea 89 (special issue):239-250 Becker PH, Wink M (2003) Influences of sex, sex composition of brood and hatching order on mass growth in Common Terns (Sterna hirundo). Behav Ecol Sociobiol 54:136-146 Becker RA, Chambers JM, Wilks AR (1988) The New S Language. Wadsworth and Brooks/Cole Blokpoel H, Morris R, Tessier G (1984) Field investigations of the biology of Common Terns wintering in Trinidad. J Field Orni- thol 55:424-434 Blokpoel H, Morris R, Trull P (1982) Winter observations of Common Terns in Trinidad, Guyana, and Suriname. Colonial Water- birds 5:144-147 Brenninkmeijer A, Stienen EWM, Klaas- sen M, Kersten M (2002) Feeding ecology of wintering terns in Guinea-Bissau. Ibis 144:602-613 Breton AR, Nisbet ICT, Mostello CS, Hatch JJ (2014) Age-dependent breeding dispersal and adult survival within a metapopula- tion of Common Terns Sterna hirundo. Ibis 156:534-547 Bruderer B, Boldt A (2001) Flight character- istics of birds: 1. Radar measurements of speeds. Ibis 143:178-204 Bugoni L, Cormons TD, Boyne AW, Hays H (2005) Feeding grounds, daily foraging ac- tivities, and movemnts of Common Terns in Southern Brazil, determined by radio- telemetry. Waterbirds 28:468-477 Burger J (1980) The transition to independence and post fledging parental care in seabirds: In: Burger J, Olla BL, Winn HE (eds.) Ma- rine Birds. Behavior of Marine Animals, Vol.4:367–447. Plenum Press, New York. Calenge C (2006) The package adehabitat for the R software: a tool for the analysis of space and habitat use by animals. Ecol Modelling 197:516-519 Cohen EB, Hostetler JA, Royle JA, Marra PP (2014) Estimating migratory connectivity of birds when re-encounter probabilities are heterogenous. Ecol Evol 4:1659-1670 Core Team R (2014) A language and environ- ment for statistical computing. R Founda- tion for Statistical Computing, Vienna, Austria. URL http://www.R-project.org/. Dias MP, Granadeiro JP, Catry P (2013) Indi- vidual variability in the migratory path and stopovers of a long-distance pelagic mi- grant. Anim Behav 86:359-364 65Temporal-spatial distribution during the non-breeding period Egevang C, Stenhouse IJ, Phillips RA, Peters- en A, Fox JW, Silk JRD (2010) Tracking of Arctic terns Sterna paradisaea reveals lon- gest animal migration. Proc Natl Acad Sci 107:2078-2081 Ekstrom PA (2004) An advance in geolocation by light. Memoirs Natl Inst Polar 58:210– 226 Ezard T, Becker PH, Coulson T (2006) The Contributions of Age and Sex to Variation in Common Tern Population Growth Rate. J Anim Ecol 75:1379-1386 Fijn RC, Hiemstra D, Phillips RA, Winden J van der (2013) Arctic Terns Sterna paradi- saea from the Netherlands migrate record distances across three oceans to Wilkes Land, East Antarctica. Ardea 101:3-12 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 Garthe S, Benvenuti S, Montevecchi WA (2000) Pursuit plunning by northern gannets (Sula bassana) “feeding on capelin (Mallotus vil- losus)”. Proc R Soc B 267:1717-1722 Garthe S, Ludynia K, Hueppop O, Kubetzki U, Meraz JF, Furness RW (2012) Energy budgets reveal equal benefits of varied mi- gration strategies in northern gannets. Mar Biol 159:1907-1915 González-Solís J, Croxall JP, Oro D, Ruiz X (2007) Trans-equatorial migration and mix- ing in the wintering areas of a pelagic sea- bird. Front Ecol Environ 5:297–301 Grecian WJ, Witt MJ, Attrill MJ: Multi-species top predator tracking reveals link between marine biodiversity and ocean productivity. Biol Letters (in revision) Guilford T, Meade J, Willis J, Phillips RA, Boyle D, Roberts S, Collet M, Freeman R, Perrins CM (2009) Migration and stopover in a small pelagic seabird, the Manx shear- water Puffinus puffinus: insights from ma- chine learning. Proc R Soc B 276:1215-1223 Guilford T, Freeman R, Boyle D, Dean B, Kirk H, Phillips R, Perrins C (2011) A dispersive migration in the Atlantic Puffin and its im- plications for migratory navigation. PloS ONE 6:e21336 Hallworth MT, Sillett TS, Van Wilgenburg, SL, Hobson KA, Marra PP (2015) Migratory connectivity of a Neotropical migratory songbird revealed by archival light-level ge- olocators. Ecol Appl 25:336–347 Harrison P (1997) Seabirds of the World. Princ- eton, Princeton Hayward DF, Oguntoyinbo J (1987) The clima- tology of West Africa. Hutchinson, London Hill RD (1994) Theory of geolocation by light levels. Elephant Seals: Population Ecol- ogy, Behavior, and Physiology. In: Burney J, Boeuf BJ, Laws RM (eds)), pp. 227–236. University of California Press, Berkeley Liechti F (2006) Birds: Blowin’ by the wind? J Ornithol 147:202-211 Limmer B, Becker PH (2007) The relative role of age and experience in determining varia- tion in body mass during the early breeding career of the Common Tern (Sterna hirun- do). Behav Ecol Sociobiol 61:1885-1896 Lisovski S, Hahn S (2012) GeoLight–process- ing and analysing light‐based geolocator data in R. Methods Ecol Evol 3:1055-1059 Lisovski S, Hewson CM, Klaassen RHG, Ko- rner-Nievergelt F, Kristensen MW, Hahn S (2012) Geolocation by light: accuracy and precision affected by environmental fac- tors. Methods Ecol Evol 3:603-612 McGregor HV, Dima M, Fischer HW, Mu- litza S (2007) Rapid 20th-century increase in coastal upwelling off Northwest Africa. Science 315:637-639 Mostello CS, Nisbet ICT, Oswald SA, Fox JW (2014) Non-breeding season movements of six North American Roseate Terns Sterna dougallii tracked with geolocators. Seabird 27:1-21 Müller MS, Massa B, Phillips RA, Dell’Omo G (2015) Seabirds mated for life migrate separately to the same places: behavioural coordination or shared proximate causes? Anim Behav 102:267-276 Neubauer W (1982) Der Zug mitteleuropaeisch- er Flusseeschwalben (Sterna hirundo) nach Ringfunden. Ber Vogelwarte Hiddensee 2:59-82 66 CHAPTER 2: Common Terns on the East-Atlantic Flyway: Neves VC, Nava CP, Cormons M, Bremer E, Castresana G, Lima P, Azevedo jun SM, Phillips RA, Magalhães MC, Santos RS (2015) Migration routes and non-breeding areas of Common Terns (Sterna hirundo) from the Azores. Emu 115:158–167 Nielsen A, Sibert JR (2007) State-space model for light-based tracking of marine animals. Can J Fish Aquatic Sci 64:1055-1068 Nilsson C, Klaassen RH, Alerstam T (2013) Differences in speed and duration of bird migration between spring and autumn. Am Nat 181:837-845 Nisbet ICT (2002) Common Tern. The Birds of North America 618: 1-39. Nisbet ICT, Cam E (2002) Test for age-specific- ity in survival of the Common Tern. J Appl Statistic 29:65-83 Nisbet ICT, Montoya JP, Burger J, Hatch JJ (2002) Use of stable isotopes to investigate individual differences in diets and mercury exposures among Common Terns Sterna hirundo in breeding and wintering grounds. Mar Ecol Prog Ser 242:267-274 Nisbet ICT, Mostello CS, Veit RR, Fox JW, Afanasyev V (2011a) Migrations and winter quarters of five Common Terns tracked us- ing geolocators. Waterbirds 34:32-39 Nisbet ICT, Szczys P, Mostello CS, Fox JW (2011b) Female Common Terns Sterna hirundo start autumn migration earlier than males. Seabird 24:103-106 Palestis BG, Hines JE (2015) Adult survival and breeding dispersal of Common Terns (Sterna hirundo) in a declining population. Waterbirds 38:221-228 Pennycuick CJ, Åkesson S, Hedenström A (2013) Air speeds of migrating birds ob- served by ornithodolite and compared with predictions from flight theory. J R Soc In- terface 10:20130419 Phillips RA, Silk JRD, Croxall JP, Afanasyev V, Briggs DR (2004) Accuracy of geoloca- tion estimates for flying seabirds. Mar Ecol Prog Ser 266:265-272 Phillips RA, Silk JRD, Croxall JP, Afanasyev V, Briggs DR (2005) Summer distribution and migration of nonbreeding albatrosses: individual consistencies and implications for conservation. Ecology 86:2386-2396 Ramos R, Sanz V, Militão T, et al. (2015) Leap- frog migration and habitat preferences of a small oceanic seabird, Bulwer’s petrel (Bulweria bulwerii). Journal of Biogeogra- phy 42:1651-1664. Ropert-Coudert Y, Wilson RP, Grémillet D, Kato A, Lewis S, Ryan PG (2006) Electro- cardiogram recordings in free-ranging gan- nets reveal minimum difference in heart rate during flapping versus gliding flight. Mar Ecol Prog Ser 328:275–284 Salewski V, Bairlein F, Leisler B (2000) Recur- rence of some palaearctic migrant passerine species in West Africa. Ringing & Migra- tion 20:29-30 Scandolara C, Rubolini D, Ambrosini R, Capri- oli M, Hahn S, Liechti F, Romano A, Roma- no M, Sicurella B, Saino N (2014) Impact of miniaturized geolocators on Barn Swallow Hirundo rustica fitness traits. J Avian Biol 45:417-423 Shaffer SA, Tremblay Y, Weimerskirch H, Scott D, Thompson DR, Sagar PM (2006) Migratory shearwaters integrate oceanic resources across the Pacific Ocean in an endless summer. Proc Nat Acad Sci 103: 12799-12802 Sommerfeld J, Kato A, Ropert-Coudert Y, Garthe S, Hindell MA (2013) The individu- al counts: within sex differences in foraging strategies are as important as sex-specific differences in masked boobies Sula dacty- latra. J Avian Biol 44:531–540 Sorenson MC, Hipfner JM, Kyser TK, Norris DR (2009) Carry-over effects in a Pacific seabird: stable isotope evidence that non- breeding diet quality influences reproduc- tive success. J Anim Ecol 78:460-467 Sumner MD, Wotherspoon SJ, Hindell MA (2009) Bayesian estimation of animal movement from archival and satellite tags. PLoS One 4:e7324 Szostek KL, Becker PH (2015) Marine primary productivity in the wintering area influenc- es survival and recruitment in a migratory seabird. Oecologia 178:643-657 67Temporal-spatial distribution during the non-breeding period Szostek KL, Bouwhuis S, Becker PH (2015) Are arrival date and body mass after spring migration influenced by large-scale envi- ronmental factors in a migratory seabird? Front Ecol Evol 3:42 Szostek L, Becker PH (2012) Terns in Trouble: Demographic Consequences of Low Breed- ing Success and Recruitment on a Common Tern Population in the German Wadden Sea. J Ornithol 153:313-326 Trierweiler C, Klaassen RGH, Drent RH, Exo K-M, Komdeur J, Bairlein F, Koks BJ (2014) Population specific migration routes and migratory connectivity in a long-distance migratory raptor. Proc Royal Soc B London 281:1471-2954 Winden J van der, Fijn RC, Horssen PW van, Gerritsen-Davidse D, Piersma T (2014) Id- iosyncratic migrations of Black Terns (Ch- lidonias niger): diversity in routes and stop- overs. Waterbirds 37:162-174 Weimerskirch H, Bonadonna F, Bailleul F, Ma- bille G, Dell’Omo G, Lipp H-P (2002) GPS tracking of foraging albatrosses. Science 295:1259 Weimerskirch H, Wilson RP (2000) Oceanic respite for wandering albatrosses. Nature 406:955-956 Wendeln H, Becker PH (1999) Significance of ring removal in Africa for a Common Tern Sterna hirundo colony. Ring Migration 19:210-212 Wernham CV, Toms MP, Marchant JH, Clark JA, Siriwardena GM, Baillie SR (2002) The migration atlas: movements of the birds of Britain and Ireland. T & AD Poyser, Lon- don Wilson RP, Cooper J, Plötz J (1992a) Can we determine when marine endotherms feed? A case study with seabirds. J Exp Biol 167:267-275 Wilson RP, Ducamp JJ, Rees G, Culik BM, Niekamp K (1992b) Estimation of location: global coverage using light intensity. Wild- life Telemetry: Remote Monitoring and Tracking of Animals. In: I.G. Priede IG, Swift SM (eds), pp. 131–134. EllisHoward Ltd, Chichester Wilson RP, Weimerskirch H, Lys P (1995) A Device for Measuring Seabird Activity at Sea. J Avian Biol 26:172-175 Wilson RP, Grémillet D, Syder J, Kierspel MAM, Garthe S, Weimerskirch H, Schäfer- Neth C, Scolaro JA, Bost C-A, Plötz J, Nel D (2002) Remote-sensing systems and seabirds: their use, abuse and potential for measuring marine environmental variables. Mar Ecol Prog Ser 228:241–261 Zhang H, Vedder O, Becker PH, Bouwhuis S (2015) Age-dependent trait variation: the relative contribution of within-individual change, selective appearance and disap- pearance in a long-lived seabird. J Anim Ecol 84:797–807 68 CHAPTER 2 : Supplementary material Ta bl e S1 In fo rm at io n of C om m on T er ns b ei ng ta gg ed w ith li gh t-l ev el g eo lo ca to rs . G iv en w er e th ei r n am e, se x, y ea r o f b irt h, ri ng n um be r ( no .), ge ol oc at or n um be r ( no .), d ep lo ym en t ( de pl oy .) da te , r em ov al d at e, an d na m e o f m at e. / = n o re tu rn o f a du lt; lo st = re tu rn b ut g eo lo ca to r l os t; hi gh - lig ht ed g re y = no n/ br ee di ng m ov em en ts tra ck ed Bi rd 20 08 20 09 20 10 20 11 na m e se x ye ar of bi rth rin g no . ge ol . no . de - pl oy . da te m at e na m e ge ol . no . re - m ov al / de pl oy . da te m at e na m e ge ol . no . re - m ov al / de pl oy . da te m at e na m e re - m ov al da te A lit ze f 19 98 77 31 97 0 71 53 31 -0 5 Ca df ae l / El i f 19 94 77 29 08 8 71 49 31 -0 5 V ic to r / W ie la nd m 19 97 77 32 14 1 71 50 27 -0 5 A nt on ia 74 14 03 -0 6 M ar ia nn a 21 16 6 29 -0 5 M ar ia nn a 18 -0 5 K as im ir m 19 94 77 29 01 4 71 52 27 -0 5 C or ne lia 74 06 20 -0 5 C or ne lia 21 15 7 27 -0 5 C or ne lia / N in a f 19 97 77 31 34 8 71 51 31 -0 5 M os es 74 07 26 -0 5 M os es 21 16 8 31 -0 5 M os es 18 -0 5 Sa lo m e f 20 00 77 83 55 8 74 08 28 -0 5 Er ns t 21 16 4 29 -0 5 Er ns t 13 -0 5 M in na f 19 98 77 00 40 8 74 09 28 -0 5 Jo ac hi m 21 16 7 29 -0 5 Jo ac hi m 18 -0 5 C or ne lia f 19 95 77 31 63 3 74 10 28 -0 5 K as im ir 21 16 2 29 -0 5 K as im ir 01 -0 6 M os es m 19 97 77 32 14 9 74 11 28 -0 5 N in a 21 16 5 29 -0 5 N in a 19 -0 5 Jo ac hi m m 19 99 77 82 15 9 74 13 01 -0 6 M in na 21 15 8 27 -0 5 M in na 19 -0 5 Er ns t m 19 99 77 82 09 9 74 05 20 -0 5 Sa lo m e 21 16 0 27 -0 5 Sa lo m e 12 -0 5 M ar ia nn a f 19 97 77 32 69 6 74 15 05 -0 6 W ie la nd 21 15 9 27 -0 5 W ie la nd 16 -0 5 W er ne r m 19 96 77 29 69 9 74 12 28 -0 5 Cl ai re / Cl ai re f 19 96 77 29 67 8 74 04 20 -0 5 W er ne r lo st H ei ne r m 19 99 77 82 38 0 21 16 3 29 -0 5 Ay la 10 -0 5 Ay la f 20 00 77 83 50 6 21 16 1 27 -0 5 H ei ne r 09 -0 5 1. ADDITIONAL INFORMATION ABOUT LIGHT-LEVEL GEOLOCATION 69Supplementary material Ta bl e S2 E sti m at ed m ea n lo ng itu de a nd m ea n la tit ud e po sit io ns ± s ta nd ar d de vi at io n (S D ) o f i nd iv id ua l w in te rin g ar ea s of 9 c om m on te rn s, 20 09 – 2 01 1. F or th re e bi rd s w in te rin g ar ea s w er e es tim at ed fo r m or e th an o ne y ea r. N um be r o f l oc at io ns c on sid er ed fo r t he e sti m at ed m ea n w in te rin g ar ea (n lo c. ). Fo r t hr ee b ird s a sto po ve r e ith er d ur in g au tu m n or d ur in g sp rin g m ig ra tio n co ul d be e sti m at ed . A s fo r t he w in te rin g ar ea e sti m at ed m ea n lo ng itu de a nd la tit ud e po sit io ns ± st an da rd d ev ia tio ns (S D ) w er e gi ve n. T he p er io d of st op ov er w as in di ca te d. N A = N ot an al ys ed bi rd , y ea r w in te rin g ar ea st op ov er d ur in g au tu m n or s pr in g na m e se x ze ar lo ng itu de m ea n± SD la tit ud e m ea n± SD n lo c. lo ng itu de m ea n± SD la tit ud e m ea n± SD pe rio d Jo ac hi m m 20 09 /1 0 -1 7.7 7± 0. 37 18 .8 ±2 .5 2 23 2 20 10 /1 1 -1 7. 37 ±0 .3 5 18 .4 4± 2. 19 18 3 M os es m 20 09 /1 0 -1 6. 32 ±0 .3 4 9. 39 ±0 ,8 6 22 8 20 10 /1 1 -1 6. 68 ±1 .0 11 .3 7± 2. 21 13 8 -1 3.7 6± 1.1 4 30 .5 8± 0. 52 13 -1 0 – 20 -1 0 K as im ir m 20 09 /1 0 -1 7. 30 ±0 .5 1 15 .14 ±4 .17 24 0 -1 7. 39 ±0 .3 5 29 .16 ±1 .12 N A – 1 2- 04 C or ne lia f 20 09 /1 0 -2 0. 08 ±1 .3 4 19 .13 ±2 .0 8 16 6 -1 7.0 4± 0. 44 25 .8 5± 0. 48 08 -0 8 – 05 -0 9 20 10 /1 1 -1 6. 73 ±0 .7 7 24 .0 6± 2. 04 22 0 H ei ne r m 20 10 /1 1 -1 6. 07 ±0 .4 1 10 .9 9± 2. 23 21 2 Ay la f 20 10 /1 1 -1 7.8 7± 0. 96 20 .3 3± 2. 0 7 7 Er ns t m 20 10 /1 1 -1 5. 51 ±0 .4 4 10 .2 7± 2. 03 22 8 W ie la nd m 20 10 /1 1 -1 5.7 7± 1. 33 10 .4 4± 4. 41 22 2 M ar ia nn a f 20 09 /1 0 -2 2. 42 ±1 .0 8 18 .3 3± 2. 24 8 2 70 CHAPTER 2 : Figure S1 Density distribution of ground speed of the entire track (bars) and the corresponding gamma distribu- tion (shape = 1.51, rate = 0.13, red line) exemplary for one bird (Kasimir tracked in 2009 to 2010) Figure S2 Lognormal distribution of twilight errors (mean of the distribution = 1.792, standard deviation of the distribution = 0.788; both on a log scale) 4    Figure S1. Density distribution of ground speed of the entire track (bars) and the corresponding  gamma distribution (shape = 1.51, rate = 0.13, red line) exemplary for one bird (Kasimir tracked in  2009 to 2010). Figure S2. Lognormal distribution of twilight errors (mean of the distribution = 1.792, standard  deviation of the distribution = 0.788; both on a log scale).  4    Figure S1. Density distribution of ground speed of the entire track (bars) and the corresponding  gamma distribution (shape = 1.51, rate = 0.13, red line) exemplary for one bird (Kasimir tracked in  2009 to 2010). Figure S2. Lognormal distribution of twilight errors (mean of the distribution = 1.792, standard  deviation of the distribution = 0.788; both on a log scale).  71Supplementary material Figures S3 individual light-level geolocation data. Longitude and latitude estimates (red lines) over time for each individual. Blues lines indicate 95% confidence intervals. Black lines indicate obvious changes in bird’s whereabouts. These dates were used to define departures and arrivals of stationary periods, i.e. breeding area, stopover sites, and wintering grounds (see Table 1, Table S2). Grey dashed lines indicate longitude/latitude of the breeding area and first/last reading of transponder at the breeding colony. Outlier estimates are clearly recog- nizable; these were excluded for estimating wintering grounds. Some light-level geolocators produced corrupt data at certain times of the recording which could not be corrected for. This decreased the precision of our data and should be kept in mind when interpreting the data. Grey bars indicate ten days around (± 10 days) each equi- nox. As only latitude estimates were affected by the equinoxes, grey bars are only indicated in the plots showing latitudinal estimates over time. In Table S2 the number of locations considered to estimate mean wintering areas are given for each individual 5    Figure S3. individual light‐level geolocation data. Longitude and latitude estimates (red lines) over  time for each individual. Blues lines indicate 95% confidence intervals. Black lines indicate obvious  changes in bird’s whereabouts. These dates were used to define departures and arrivals of stationary  periods, i.e. breeding area, stopover sites, and wintering grounds (see Table 1, Table S2). Grey  dashed lines indicate longitude/latitude of the breeding area and first/last reading of transponder at  the breeding colony. Outlier estimates are clearly recognizable; these were excluded for estimating  wintering grounds. Some light‐level geolocators pr duce  corrupt data at certain tim  of the  recording which could not be corrected for. This decreased the precision of our data and should be  kept in mind when interpreting the data. Grey bars indicate ten days around (± 10 days) each  equinox. As only latitude estim tes were affect d by the  quinoxes, grey bars are only indicated in  the plots showing latitudinal estimates over time. In Table S2 the number of locations considered to  estimate mean wintering areas are given for each individual.  Figure S3a. LGeolocator number 7406  =  Kasi ir 2009 – 2010      Figure S3a LGeolocator number 7406 = Kasimir 2009 – 2010 72 CHAPTER 2 : Figure S3b LGeolocator number 7410 = Cornelia 2009 – 2010 The internal geolocator clock drifted during the winter so that longitude could not be estimated correctly. Based on the plot longitude over time, we estimated that the internal geolocator clock started drifting on the 28rd of January 2010. We, therefore, did not consider locations after that date for estimating the wintering ground and also not for determining start of spring migration 6      Figure S3b. LGeolocator number 7410  =  Cornelia 2009 – 2010  The internal geolocator clock drifted during the winter so that longitude could not be estimated  correctly. Based on the plot longitude over time, we estimated that the internal geolocator clock  started drifting on the 28rd of January 2010. We, therefore, did not consider locations after that date  for estimating the wintering ground and also not for determining start of spring migration.      73Supplementary material 7    Figure S3c. LGeolocator number 7411  =  Moses 2009 – 2010  Both higher longitude estimates in summer 2009 and lower longitude estimates in spring 2010 in  comparison to longitude of the breeding colony suggest a drift of the internal geolocator clock over  the year. Whether this shift occurred gradually over the season or whether the marked shift in  longitude in the beginning of November 2009 was responsible for the low estimates of longitude in  spring 2010 remained unclear. Therefore, we did not control for this time shift here, but assume that  the bird was at the colony on the 27th of April though longitude estimate was slightly too low. Figure S3c LGeolocator number 7411 = Moses 2009 – 2010 Both higher longitude estimates in summer 2009 and lower longitude estimates in spring 2010 in comparison to longitude of the breeding colony suggest a drift of the internal geolocator clock over the year. Whether this shift occurred gradually over the season or whether the marked shift in longitude in the beginning of November 2009 was responsible for the low estimates of longitude in spring 2010 remained unclear. Therefore, we did not control for this time shift here, but assume that the bird was at the colony on the 27th of April though longitude estimate was slightly too low 74 CHAPTER 2 : 8    Figure S3d. Geolocator number 7413  =  Joachim 2009 – 2010  Figure S3d Geolocator number 7413 = Joachim 2009 – 2010 75Supplementary material Figure S3e Geolocator number 7415 = Marianna 2009 – 2010 Only winter locations between 11th of November and 22nd of December 2009 were considered to estimate mean wintering ground. Light-level geolocator broke at the end of the year 2009 9    Figure S3e. Geolocator number 7415 = Marianna 2009 – 2010  Only winter locations between 11th of November and 22nd of December 2009 were considered to  estimate mean wintering ground. Light‐level geolocator broke at the end of the year 2009.  76 CHAPTER 2 : Figure S3f Geolocator number 21158001 = Joachim 2010 – 2011 Only winter locations between 1st of November 2010 and 4th of February 2011 were considered to estimate mean wintering ground. Light-level geolocator broke in the beginning of February 2011 10    Figure S3f. Geolocator number 21158001 = Joachim 2010 – 2011  Only winter locations between 1st of November 2010 and 4th of February 2011 were considered to  estimate mean wintering ground. Light‐level geolocator broke in the beginning of February 2011. 77Supplementary material Figure S3g Geolocator number 21160001 = Ernst 2010 – 2011 11    Figure S3g. Geolocator number 21160001 = Ernst 2010 – 2011  78 CHAPTER 2 : Figure S3h First figure Geolocator number 21161001 = Ayla 2010 – 2011 The internal geolocator clock drifted, see first figure of this bird. Based on the plot longitude over time, we cor- rected longitude estimates by adding 100°, see second figure of this bird. Data of this bird needs to be treated cautiously. Only winter locations between 10th of December 2010 and 20th of January 2011 were considered to estimate mean wintering ground 12    Figure S3h. Geolocator number 21161001 = Ayla 2010 – 2011  The internal geolocator clock drifted, see first figure of this bird. Based on the plot longitude over  time, we corrected longitude estimates by adding 100°, see second figure of this bird. Data of this  bird needs to be treated cautiously. Only winter locations between 10th of December 2010 and 20th  of January 2011 were considered to estimate mean wintering ground  First figure        79Supplementary material Figure S3h Second figure 13    Second figure    80 CHAPTER 2 : Figure S3i Geolocator number 21162001 = Cornelia 2010 – 2011 14    Figure S3i. Geolocator number 21162001 = Cornelia 2010 – 2011  81Supplementary material Figure S3j Geolocator number 21163001 = Heiner 2010 – 2011 15    Figure S3j. Geolocator number 21163001 = Heiner 2010 – 2011  82 CHAPTER 2 : Figure S3k Geolocator number 21165001 = Moses 2010 – 2011 Light-level geolocator broke in the beginning of January 2011. Last reading of the bird at the breeding colony was 12th of July 2010 16    Figure S3k. Geolocator number 21165001 = Moses 2010 – 2011  Light‐level geolocator broke in the beginning of January 2011. Last reading of the bird at the breeding  colony was 12th of July 2010.       83Supplementary material Figure S3l Geolocator number 21166001 = Wieland 2010 – 2011 17    Figure S3l. Geolocator number 21166001 = Wieland 2010 – 2011        84 CHAPTER 2 : Figure S4 Frequency distribution of great circle distances between 10,000 randomly chosen wintering locations. Within-individual great circle distances between the wintering locations as estimated by light-level geolocation data for the three birds tracked during two winters (Joachim, Moses, Cornelia). Although Joachim’s two winter- ing locations were in close vicinity to each other (58 km), 279 of the 10,000 great circle distances were closer to each other. The dashed orange line indicated the median great circle distance between the randomly chosen wintering locations 18      Figure S4. Frequency distribution of great circle distances between 10,000 randomly chosen  wintering locations. Within‐individual great circle distances between the wintering locations as  estimated by light‐level geolocation data for the three birds tracked during two winters (Joachim,  Moses, Cornelia). Although Joachim’s two wintering locations were in close vicinity to each other     (58 km), 279 of the 10,000 great circle distances were closer to each other. The dashed orange line  indicated the median great circle distance between the randomly chosen wintering locations.      85Supplementary material 19    2. Supplementary results Figure S5. Differences in wintering area between years based on 12 tracks with light‐level geolocators  of 9 Common Terns. Grey dots = 2009/2010 data; small black dots = 2010/2011 data. Dotted lines =  95 kernel densities (highlighted in grey = 2009/2010); dashed lines 75 = kernel densities; solid lines =  45 kernel densities. Large triangles = females (filled white encircled black = 2009/2010, filled black  encircled white = 2010/2011); large squares = males (filled white encircled black = 2009/2010, filled  black encircled white = 2010/2011).      2. SUPPLEMENTARY RESULTS Figure S5 Differenc s in wintering ar a bet een years based on 12 tracks with light-level geoloc tor of 9 Com- mon Terns. Grey dots = 2009/2010 data; small black dots = 2010/2011 data. Dotted lines = 95 kernel densities (highlighted in grey = 2009/2010); dashed lines 75 = kernel densities; solid lines = 45 kernel densities. Large triangles = females (filled white encircled black = 2009/2010, filled black encircled white = 2010/2011); large squares = males (filled white encircled black = 2009/2010, filled black encircled white = 2010/2011) 86 CHAPTER 2 : Figure S6 Wintering areas of individual Common Terns (female Cornelia, males Joachim and Moses) with two years of data. Symbols as in Fig. S5. Great circle distance between mean estimated wintering locations were 647 km (Cornelia), 58 km (Joachim), and 223 km (Moses; see also Fig. S4) 20      Fi ure S6. Wintering areas of individual Common Ter s (fe ale Cornelia, males Joachim and Moses)  with two years of data. Symbols as in Fig. S5. Great circl  distance b tween mean  stimated  wintering locations were 647 km (Cornelia), 58 km (Joachim), and 223 km (Moses; see also Fig. S4).          87Supplementary material 21        Figure S7.  Seasonal variation in the temporal proportion of saltwater contact across different stages  of the annual cycle. Mean daily percentage of time Common Terns had contact with salt water  recorded by using saltwater immersion data from geolocators (B breeding, PB post‐breeding, AM  autumn migration, W wintering, SM spring migration).  Figure S7 Seasonal variation in the temporal proportion of saltwater contact across different stages of the annual cycle. Mean daily percentage of time Common Terns had contact with salt water recorded by using saltwater im- mersion data from geolocators (B breeding, PB post-breeding, AM autumn migration, W wintering, SM spring mig ation) 88 CHAPTER 2 : Figure S8 Diurnal at sea activity during the stages of the annual cycle, based on tracks of 8 Common Terns. The proportion of time at sea per daytime hour is presented. Vertical lines refer to mean sunrise and sunset times during wintering. (B breeding, PB post-breeding, AM autumn migration, W wintering, SM spring migration). The individual Ayla showed an internal clock shift (Fig. S3h) 22    Figure S8. Diurnal at sea activity during the stages of the annual cycle, based on tracks of 8  Common Terns. The proportion of time at sea per daytime hour is presented. Vertical lines  refer to mean sunrise and sunset times during wintering. (B breeding, PB post‐breeding, AM  autumn migration, W wintering, SM spring migration). The individual Ayla showed an  internal clock shift (Fig. S3h).  89Supplementary material Ta bl e S3 S ea w at er c on ta ct (h ou rs p er d ay ± S D , ( n) ) d ur in g th e sta ge s o f t he a nn ua l c yc le , b as ed o n tra ck s o f 8 C om m on T er ns . C f. Fi g. 6 fo r d ai ly p ro po rti on s. S ig ni fi ca nt d if fe re nc es b et w ee n st ag es a re in di ca te d by s ta ge a bb re vi at io ns (c ap it al s p< 0. 05 ); G L M R M , n = 6 in di vi du al s w it h da ta d ur in g ea ch s ta ge , r es pe ct iv el y ID Br ee di ng Po st -b re ed in g Au tu m n m ig ra tio n W in te ri ng W in te ri ng Pr e- br ee di ng Ay la 0. 43 ± 0 .7 7 (1 14 ) 1.1 2 ± 1. 09 (1 2) 4. 92 ± 4 .0 6 (4 9) 4. 05 ± 3 .0 3 (1 12 ) 4. 86 ± 2 .5 1 (5 8) - C or ne lia 0. 35 ± 0 .5 5 (1 01 ) - 0. 65 ± 0 .6 8 (7 ) 4. 75 ± 2 .9 1 (2 05 ) 3. 59 ± 2 .5 9 (5 4) - Er ns t 0. 30 ± 0 .5 2 (1 00 ) 0. 30 ± 0 .2 3 (4 4) 0. 67 ± 0 .76 (3 7) 2. 23 ± 1 .3 3 (1 18 ) 3. 58 ± 1 .5 6 (4 9) 1.1 6 (1 ) H ei ne r 0. 27 ± 0 .4 0 (1 09 ) 0. 24 ± 0 .0 9 (1 2) 0. 49 ± 0 .5 4 (5 6) 1. 55 ± 1 .0 9 (1 06 ) 3. 37 ± 1 .5 1 (5 5) 1. 22 ± 1 .5 9 (7 ) Jo ac hi m 0. 47 ± 0 .6 7 (9 8) 1.7 4 ± 1. 39 (1 8) 5. 37 ± 4 .11 (3 2) 5. 27 ± 3 .2 2 (1 04 ) 2. 00 ± 0 .8 4 ( 2) - K as im ir 0. 14 ± 0 .19 (8 5) 0. 98 ± 0 .8 0 (6 1) 1. 60 ± 1 .71 (2 9) 2. 41 ± 1 .79 (1 72 ) 3. 50 ± 2 .8 8 (2 4) - M os es 0. 10 ± 0 .0 5 (4 4) 1. 07 ± 1 .0 5 (5 5) 1.1 0 ± 1. 60 (5 2) 1. 22 ± 1 .2 2 (8 1) - - W ie la nd 0. 12 ± 0 .3 7 (8 3) 0. 36 ± 0 .4 5 (4 6) 1.7 8 ± 1. 66 (3 1) 3.7 9 ± 2. 96 (1 47 ) 2. 52 ± 1 .3 2 (4 6) - M ea n (n = 6) 0. 29 ± 0 .14 0. 79 ± 0 .5 9 2. 47 ± 2 .13 3. 21 ± 1 .3 9 3. 31 ± 0 .9 9 - p A, W , S A, W , S B, P B, P B, P 90 CHAPTER 2 : Figure S9 Common Tern marked with steel ring (right leg) and geolocator attached to a plastic ring (left leg; Photo: Sabrina Weitekamp) 3. ADDITIONAL INFORMATION ABOUT POTENTIAL EFFECTS OF GEOLOCATORS ON COMMON TERNS Arrival mass In three individuals body mass at arrival could be compared between one or more years before attachment of the geolocator and while the bird was carrying it. Average value before geoloca- tor attachment was 125.3 ± 5.8 g (SD, n=3), and with geolocator 138.5 ± 8.7 g (n=4). Kasimir, before 122 g (2008)/with 127 g (2011); Ayla, be- fore 122 g (2009)/with 148 g (2011); Cornelia, 1999-2002, before, on average 132 g (124-136 g)/with 140 g (138 and 141 g, 2010 and 2011, re- spectively). These values were within the com- mon range of arrival mass of Common Terns at the breeding grounds, i.e.128-136g, depending on age (Limmer & Becker 2007). Mass at catching For 12 breeders no differences were found in mass at catching before deploying the geoloca- tor and after retrieving it (first catch: 128.9 ± 3.3 g, second catch with geolocator: 130.4 ± 2.5 g; t = -0.848, df = 11, p = 0.415, t-test). Arrival date We compared arrival dates as related samples in individuals before, during, and after the geo- locator attachment (Table S4; cf. Table S1). The differences were n.s. and there was no indica- tion that arrival date was impaired by the geo- locator. 24    3. Additional information about potential effects of geolocators on Common Terns Arrival mass In three individuals body mass at arrival could be compared between one or more years before attachment of the geolocator and while the bird was carrying it. Average value before geolocator attachment was 125.3 ± 5.8 g (SD, n=3), and with geolocator 138.5 ± 8.7 g (n=4). Kasimir, before 122 g (2008)/with 127 g (2011); Ayla, before 122 g (2009)/with 148 g (2011); C rnelia, 1999-2002, before, on average 132 g (124-136 )/with 140 g ( 38 and 141 g, 2010 and 2011, respectively). These values were within the common range of arrival mass of Common Terns at the breeding grounds, i.e.128-136g, depending on age (Limmer & Becker 2007). Mass at catching For 12 breeders no differences w r found in mass catching before deploying t e geolocator and after retrieving it (first catch: 128.9 ± 3.3 g, second catch with geolocator: 130.4 ± 2.5 g; t = -0.848, df = 11, p = 0.415, t-test). Figure S9. C mmon Tern mark d with stee ring (right leg) and geolocator attach d to a plastic ring (left leg; Photo: Sabrina Weitekamp) 91Supplementary material Table S4 Arrival dates of individuals in the year be- fore, during or after geolocator attachment, respec- tively. Pairwise Wilcoxon signed-rank tests, two- tailed. Means ± SD Arrival Date (day of year) Before Geoloc. Att. 113.1 ± 6.3 During Geoloc. Att. 111.9 ± 7.9 n 9 U 0.0 p 1.0 During Geoloc. Att. 107.5 ± 9.6 After Geoloc. Att. 108.8 ± 7.1 n 4 U -0.184 p 0.854 Table S5 Annual reproductive success from 2008-2011 in Common Tern pairs equipped with geolocator. In 2008, only one mate per pair was equipped (cf. Table S1). Means ± SE are given Table S6 Reproductive success (means ± SE) of pairs equipped with geolocators. Pairwise comparisons be- tween years before, during, and after geolocator attachment (Wilcoxon signed-rank test, one-tailed; hypothesis: reduced values during geolocator attachment) Year Clutch Size Hatching Success Fledging Success Fledglings Pair-1 N Pairs 2008 3.0 ± 0.0 1.00 ± 0.0 0.47 ± 0.1 1.6 ± 0.2 5 2009 2.8 ± 0.2 0.56 ± 0.2 0.06 ± 0.1 0.2 ± 0.2 6 2010 2.7 ± 0.2 0.33 ± 0.2 0.33 ± 0.2 0.5 ± 0.3 6 2011 2.7 ± 0.3 0.76 ± 0.1 0.52 ± 0.1 1.8 ± 0.3 7 Before Geolocator Att. During Geolo- cator Att.1 After Geolocator Att. 2 P (n) Cluch size 2.9 ± 0.1 2.7 ± 0.2 -- n.s. (7) Hatching success 0.86 ± 0.10 0.43 ± 0.16 0.89 ± 0.07 B/D: <0.05 (7) D/A: <0.05 (6) Fledging success 0.48 ± 0.16 0.22 ± 0.16 0.56 ± 0.06 B/D: n.s. (7) D/A: n.s. (6) Fledglings pair-1 1.0 ± 0.4 0.4 ± 0.3 1.7 ± 0.2 B/D: n.s. (7) D/A: < 0.05 (6) N 7 7 6 Laying date In five pairs with at least one mate carrying a geolocator laying date was compared before, during, and after attachment of the geolocator. Laying date (day of the year) before attachment was 133 ± 4 (SD, range 128 - 139), with geolo- cator 133 ± 7 (123 - 139) and after removing the geolocator 126 ± 2 (125 – 129; Friedman-test, n=5, Chi2 = 4.800, df = 2, p = 0.091). Reproductive success We found no significant influence of carrying light-level geolocators on birds’ reproductive success in terms of clutch size, but on hatch- ing success, fledging success, and fledglings per pair comparing the years birds carried the light-level geolocators with adjacent years (Ta- bles S5 and S6). 1 preferably from year 2010 if data from more than one year were available 2 mainly from year 2011, after deployment of geolocator and clutch management 92 CHAPTER 2 : 27    Figure S10. Clutch of a Common Tern pair tagged each with a geolocator. The number of fine fissures of the egg shell was increasing with incubation time and finally did cause shell breakage, clutch failure, and desertion (Photo: Peter H. Becker). Body mass and fledging age of chicks, subadult return Maximum body mass (126.4 ± 3.8 g), fledging mass (113.0 ± 3.4 g) and fledging age (27.3 ± 0.6 d) of fledglings (n=9) reared by geolocator parents in 2008 – 2010 were in the range typical for Banter See colony, reported e.g. by Becker & Wink (2003). From these juveniles, 5 had returned as prospectors to the Banter See colony two or three years later (55%; at least one prospector from 4 of 6 pairs).   In 2009 reproductive success of the colony was very low. Beyond that in 2009 and especially in 2010, except clutch size, reproductive success of geolocator pairs was much lower than that of other experienced pairs (2009, hatching suc- cess: 0.73 ± 0.04, fledging success: 0.22 ± 0.03, fledglings pair-1: 0.35 ± 0.06, n=77 pairs; 2010, hatching success: 0.84 ± 0.03; fledging success: 0.71 ± 0.03, fledglings pair-1: 1.38 ± 0.09, n=99 pairs; cf. Table S3). In 2011 hatching success was increased successfully by egg exchange with dummy eggs, and incubating the eggs in an incubator for 2-6 d as far as the geolocator was deployed. An inter-annual comparison within the marked pairs showed that hatching success and fledglings pair-1 were significantly different between the treatments (Table S6). Hatching success of the pairs was strongly and negatively affected by the summed no. of days the pair mates were equipped with the geolocator during the incubation period (maximum 22 d per adult; rs=-0.758, N=14, p=0.002; range = 5 – 44 d per pair and year; 2008 – 2010). Main cause of reduced hatching success was egg damage: fine fissures in the egg shell increasing with incubation time of the clutch (Fig. S7). After egg damage four pairs marked with geolocators produced a replace- ment clutch, one in 2009 (2 eggs), three in 2010 (3 eggs, resp.). Also these replacement clutches failed because of egg shell breakage. Body mass and fledging age of chicks, sub- adult return Maximum body mass (126.4 ± 3.8 g), fledging mass (113.0 ± 3.4 g) and fledging age (27.3 ± 0.6 d) of fledglings (n=9) reared by geolocator parents in 2008 – 2010 were in the range typical for Banter See colony, reported e.g. by Becker & Wink (2003). From these juveniles, 5 had re- turned as prospectors to the Banter See colony two or three years later (55%; at least one pros- pector from 4 of 6 pairs). Figure S10 Clutch of a Common Tern pair tagged each with a geolocator. The number of fine fissures of the egg shell was increasing with incubation time and finally did cause shell breakage, clutch failure, and desertion (Photo: Peter H. Becker) Chapter 3: Spatial ecology, phenological variability and moulting patterns of the endangered Atlantic petrel, Pterodroma incerta Marina Pastor-Prieto1, Raül Ramos1, Zuzana Zajková1,2, José Manuel Reyes-González1, Manuel L. Rivas1, Peter G. Ryan3 & Jacob González-Solís1 1 Institut de Recerca de la Biodiversitat (IRBio) and Departament de Biologia Evolutiva, Ecologia i Ciències Ambientals (BEECA), Universitat de Barcelona. 08028 Barcelona, Spain 2 Center for Advanced Studies of Blanes (CEAB-CSIC), 17300 Girona, Spain 3 FitzPatrick Institute of African Ornithology, DST-NRF Centre of Excellence, University of Cape Town, Rondebosch 7701, South Africa Accepted to: Endangered Species Research, Biologging and Conservation Special issue. On 18th of September 2019. ABSTRACT Insights about year-round movement and behaviour of seabirds are essential to better understand their ecology and to evaluate possible threats at sea. The Atlantic petrel (Pterodroma incerta) is an endangered gadfly petrel endemic to the South Atlantic Ocean, with virtually the entire population breeding on Gough Island (Tristan da Cunha archipelago). We describe adult phenology, habitat preferences and at-sea activity patterns for each phenological phase of the annual cycle and refine the current knowledge about its distribution, by using light-level geolocators on 13 adults during one to three consecutive years. We also ascertain its moulting pattern through stable isotope analysis (SIA) of nitrogen and carbon in feathers from 8 carcasses. On average, adults started their post- breeding migration on 25 December, taking 10 days to reach their non-breeding areas on the South American shelf slope. The pre-breeding migration started around 11 April and took 5 days. From phenological data, we found evidence of carry-over effects between successive breeding periods. The year-round distribution generally coincided with the potential distribution obtained from habi- tat modelling, except during the non-breeding and pre-laying exodus periods, when birds only used the western areas of the South Atlantic. Moulting occurred during the non-breeding period, when birds spent more time on the water, and results from SIA helped us to distinguish feathers grown around Gough Island from those grown in the non-breeding area. Overall, our results bring im- portant new insights into the spatial ecology of this threatened seabird, which should help improve conservation strategies in the South Atlantic Ocean. 96 CHAPTER 3 : 1. INTRODUCTION Seabirds are increasingly threatened world- wide, and their populations are subject to a va- riety of threats both on land, where they breed, and at sea, where they rest and forage through- out the year (Croxall et al. 2012, Lewison et al. 2012). Key threats affecting seabird popu- lations include introduction of alien invasive predators to their breeding locations, pollution and habitat degradation, interactions with com- mercial fisheries, climate change and diseases (Lucas & MacGregor 2006, Olmos et al. 2006, Grémillet & Boulinier 2009, Hilton & Cuthbert 2010, Uhart et al. 2018, Philpot et al. 2019). Es- pecially in the case of oceanic seabirds, their sensitive life history traits such as long life, de- layed first breeding, single egg per breeding at- tempt, and strong mate fidelity (Warham 1996, Bried et al. 2003, Rodríguez et al. 2019), make them particularly prone to environmental and human perturbations, which contribute to their current population declines and poor conser- vation status (González-Solís & Shaffer 2009, Croxall et al. 2012). In addition to long-lasting detrimental ef- fects on population dynamics, individual life histories are also shaped by events occurring in geographically disparate places during the breeding, migration and non-breeding periods (Norris & Marra 2007). There is mounting evidence of carry-over effects (i.e. processes that influence individual performance in a sub- sequent season) from the breeding to the non- breeding period, suggesting that migratory, non-breeding and moulting decisions taken by individuals are influenced by their success in previous breeding attempts (Catry et al. 2013). Thus, taking into account the variability of breeding efforts within a population seems ad- visable when trying to define phenology and year-round distributions of long-lived species. Gadfly petrels (Pterodroma spp.) are the larg- est genus of oceanic seabirds, with most species endemic to isolated oceanic archipelagos (Hil- ton & Cuthbert 2010, Croxall et al. 2012). Due to the remote location of their breeding colo- nies, many aspects of gadfly petrels’ ecology remain poorly known (Rodríguez et al. 2019). Few novel studies have generally described their at-sea distribution, showing long-range movements across ocean basins (Rayner et al. 2008, Jodice et al. 2015, Krüger et al. 2016, Ra- mos et al. 2016, Clay et al. 2017, Leal et al. 2017, Ramos et al. 2017). The Atlantic petrel (Pterodroma incerta) is a medium-sized procellariiform seabird (420 – 720 g), with a year-round distribution largely confined to the South Atlantic Ocean (Enticott 1991, Orgeira 2001, Cuthbert 2004). The spe- cies breeds during the austral winter; observa- tions at the breeding islands indicate that they arrive at the colony from mid-March onwards, laying a single egg in June-July, and chicks fledge in December (Richardson 1984, Cuth- bert 2004). Virtually the entire population, esti- mated at approximately 1 million pairs, breeds at Gough Island (40º20’S, 9º53’W) (Cuthbert 2004, Flood & Fisher 2013, Rexer-Huber et al. 2014). In the 1970s, a small remnant population bred on Tristan da Cunha, but the introduction of alien predators, in land habitat modification and hunting by islanders contributed to its pre- sumed extinction as breeder (Richardson 1984, Cuthbert 2004, BirdLife International 2017a). A few pairs also breed on the eastern plateau of Inaccessible Island (Flood & Fisher 2013, P.G. Ryan unpubl. data). The Atlantic petrel is listed as endangered by the IUCN due to its ex- tremely small breeding range, and the high rate of chick predation by introduced house mice (Mus musculus), which has caused the popula- tion decline and may even lead to its extinction, if mice are not eradicated from Gough Island (Cuthbert et al. 2013, Dilley et al. 2015, Bird- Life International 2017a, Caravaggi et al. 2019). The poor conservation status of the Atlantic petrel calls for new insights to better under- stand the species’ ecology and guide conser- vation actions. Most knowledge of its distri- bution at sea comes from ship-based sightings (Enticott 1991, Orgeira 2001). More recently, its general phenology and distribution were summarised together with other gadfly petrels species using tracking data (Ramos et al. 2017). However, Ramos et al. (2017) did not include 97Spatial ecology of the Atlantic petrel detailed descriptions on the phenology and spa- tial ecology and the factors influencing migra- tion schedules within the population or other important aspects of its at-sea ecology, such as habitat preferences, at-sea activity patterns and moulting strategies. This study extends our knowledge about the spatial ecology of adult Atlantic petrels. Our first aim was using geolocation-immersion data to assess in detail phenological phases, at-sea distribution, marine habitat preferences and ac- tivity patterns year-round. Second, we explored whether breeding success might lead to carry- over effects regarding phenology, behaviour or distribution, as previously found in a number of species (Catry et al. 2013, Phillips et al. 2017, Ramos et al. 2018). Since Atlantic petrels suffer high rates of breeding failure (up to 87 % rate of chick predation by introduced house mice) (Wanless et al. 2007, Cuthbert et al. 2013, Dil- ley et al. 2015), we expected to detect, from ge- olocator data, a relatively high number of birds not returning to the colony during the breeding to feed their chick, due to breeding failure. We would then expect these failed at breeding birds leaving the colony earlier than the remaining breeders to adjust their annual phenological calendar. Finally, we investigated the moulting patterns by performing stable isotope analysis (SIA) on feathers from dead specimens. We would expect feathers moulted close to the breeding grounds to show a smaller variability in the isotopic values among individuals than feathers moulted in the wintering areas, since in the latter case a larger spatial segregation of the individual wintering areas would also lead to the integration of disparate baseline isotopic levels in their feathers. 2. MATERIALS & METHODS 2.1. Tag deployment and data filtering We deployed light-level geolocators (mod- els Mk13, Mk14 and Mk19 from ©Biotrack) attached to a PVC ring with cable ties to the tarsus of breeding Atlantic petrels during the incubation period. Between July and August of 2010 2011 and 2012, we deployed 42 geoloca- tors (21, 16 and 5, respectively) on 33 Atlantic petrels attending burrows near the research station at Gough Island. Sex of birds was un- known. Some individuals were tagged in more than one year. Over three years after deploy- ment, 26 of these 33 birds were recaptured, but 5 had lost the device. From the 21 geolocators recovered, 13 provided data. Overall, we gath- ered tracks from 9 individuals for one year, 3 individuals for two years and 1 individual for three years, resulting in 18 year-round tracks from 13 birds. This dataset is already included in Ramos et al. (2017) to provide a general dis- tribution and phenology of the species. Here we analyse these data in more detail, to provide information on habitat preferences, moulting strategies and activity patterns. Geolocators measure light levels every minute and record the maximum value every 5 (model Mk19) or 10 minutes (models Mk13 and Mk14 (Afanasyev 2004)). Based on the photoperiod and sunrise and sunset times, two locations per day can be inferred (one to local midday and other to local midnight) with an average accu- racy of ~186 ± 114 km (standard deviation, SD) (Phillips et al. 2004). Light level curves were supervised using TransEdit from BASTrack software (British Antarctic Survey, BAS). Ge- olocators were calibrated for ~1 week before deployment outside the Gough Island research station. We used calibration data to calculate sun elevation angle for each device (mean ± SD, -3.3 ± 0.44) and applied a threshold value of 20 to estimate sunrise and sunset times. We removed all locations derived from light curves presenting interferences at sunrise or sunset. Those erroneous locations inside a window of 20 days on either side of each equinox (Afa- nasyev 2004) were also removed, as latitude cannot be inferred by light-level geolocation for these periods. We considered locations with flying speeds higher than 55 km h-1 sustained over a 48 h period to be unrealistic and thus they were also removed. Final dataset for fur- ther analysis contained 67 % of all locations and is available in the Seabird Tracking Database of BirdLife International (http://www.seabird- 98 CHAPTER 3 : tracking.org/) at the following address (http:// seabirdtracking.org/mapper/?dataset_id=966; BirdLife International 2017b). 2.2. Phenology and spatial distribution Phenology was determined for each year-round trip by visually inspecting filtered locations in BirdTracker software (BAS) and confirmed using conductivity data, inferred from saltwa- ter immersion data (see below). At this step, unfiltered locations were used to inform lon- gitudinal movements and determine phenol- ogy around the equinoxes, because longitude remains reliable (Hill 1994). Departure and arrival dates from breeding and non-breeding grounds were assessed visually. Departures were identified as the first day that any loca- tion was outside the cluster of locations from the previous 10 days that was followed by a clearly directed movement away from this area. Similarly, arrivals were assessed as the first day any location was inside the cluster of locations, preceded by a directed movement towards that area. Regarding incubation, only entire incu- bation periods were considered (data from 5 birds), excluding those that were not fully re- corded because of the dates of deployment or recovery of devices. We defined an incubation bout as consecutive days without light and with no immersion records preceded and followed by light and immersion records. We inferred chick-rearing when birds made frequent brief visits to the colony at night, without immersion data during several hours, and characteristic of this period (Ojowski et al. 2001). Each visit took place only during night and consecutive visits were typically separated by several days with immersion records (at night and day) at sites where foraging to feed the chick presum- ably occurred. These sites were far enough away from the colony to consider those birds did not visit the colony on consecutive nights. We identified the onset date and duration of the following phenological phases: post-breeding migration, non-breeding, pre-breeding migra- tion, pre-breeding (i.e. from arrival at the col- ony to pre-laying exodus), pre-laying exodus (i.e. period at-sea that extends from mating to egg laying), incubation and chick-rearing. Once those events were identified, we evalu- ated their variability among the year-round trips recorded. A preliminary visual explora- tion of changes in longitude suggested the ex- istence of two phenological groups (see Fig. S1). To typify them objectively, we applied a multivariate hierarchical clustering analy- sis using the function hclust and the method ward.D2 from “stats” R package (R Core Team 2017). We considered seven input variables: the onset of post-breeding migration, non-breed- ing, pre-breeding migration, pre-breeding, pre- laying exodus, incubation (all these dates were included in statistical analyses as the number of days since January 1st), and the duration (in days) of the non-breeding period (Fig. 1A). The start of chick-rearing was not included because 3 birds performed post-breeding migration immediately after incubation, presumably be- cause their breeding attempt failed during in- cubation or around hatching (many chicks are killed by mice within hours of hatching, Dilley et al. 2015). Variables were z-transformed prior to analysis. We performed a silhouette analy- sis, using the function silhouette in the R pack- age “cluster” (Maechler et al. 2017), to evalu- ate within-cluster consistency, i.e. how similar each sample is to the others assigned to the same cluster (Fig. S2) (Rousseeuw 1987). Clustering results showed two well-defined phenological groups, presumably related to breeding success (Figs. 1A and 1B, but see results, Figs. S2 and S3 and discussion for the rationale of this des- ignation), so we termed these groups successful and failed breeders. We tested for differences in phenology between these groups using a U Mann-Whitney-Wilcoxon test, applying Bon- ferroni correction for multiple comparisons. Distribution at population level was deter- mined from filtered positions for each pheno- logical phase through kernel density estima- tion, using the kernelUD function from the “adehabitatHR” R package (Calenge 2011). We used a Lambert Azimuthal Equal Area projec- tion centred in the centroid of all locations and a smoothing parameter equivalent to 186 km 99Spatial ecology of the Atlantic petrel Figure 1. Hierarchical clustering analysis of seven scaled phenological variables of Atlantic petrels. (A) Two groups, assigned as successful (S, blue) and failed (F, light brown) breeders, were identified applying hier- archical clustering analysis on seven phenological variables: starting dates of post-breeding migration (SMig1), non-breeding (SNbre), pre-breeding migration (SMig2), pre-breeding (SPreB), pre-laying exodus (SPreL), in- cubation (SInc) and duration of non-breeding (DNBre); variables were z-transformed prior to analysis. Each row represents individual phenology identified by track ID (birdID_year), the cell colour gradient reflects the value of the z-transformed variable; dark grey shaded cells represent missing values. See Fig. S2 for the results of silhouette analysis of this hierarchical clustering. (B) Phenology of adult Atlantic petrels (successful and failed breeders separately) tracked with geolocators from Gough Island. Thick lines show mean values of each group and thin lines correspond to individual phenologies (see Fig.S3 for detailed individual phenology). Note that starting dates of chick-rearing (Schick) are detailed here, but were not included in the hierarchical clustering analysis because 3 birds performed post-breeding migration immediately after incubation, presumably because they failed either during incubation or at hatching. 100 CHAPTER 3 : (~2º, depending on latitude), in order to account for the average error in geolocation (Phillips et al. 2004). Kernel density contours of 50 and 95 % were considered to represent, respectively, the core areas of activity and the areas of active use for each period (Pinet et al. 2011a). 2.3. At-sea activity analysis Mk13 and Mk14 geolocator models measure the conductivity in saltwater every 3 seconds and summarize the result in 10-minute blocks, with values ranging from 0 (meaning the whole block was continuously dry) to 200 (meaning the whole block was continuously wet) (Afa- nasyev 2004). Mk19 geolocator model pro- vides a different data resolution, storing the time stamp when geolocator recording change from wet to dry and vice versa; data recorded with Mk19 loggers were transformed to match Mk13 and Mk14 data resolution. Saltwater im- mersion data can be used as a proxy to infer activity patterns of seabirds, providing insights into behavioural strategies at different temporal scales (e.g. circadian, daily or seasonal) (Mack- ley et al. 2011, Rayner et al. 2012, Cherel et al. 2016). Activity patterns inform whether spe- cies are mainly diurnal or nocturnal (both situ- ations have been described in petrels, e.g. Bu- goni et al. 2009, Ramos et al. 2015). This may be relevant for species inhabiting oligotrophic oceanic regions, such as gadfly petrels, where diel vertical migration of potential prey can influence seabird behaviour (Dias et al. 2012, Navarro et al. 2013). We explored the activity patterns between day and night throughout the annual cycle based on the time that every log- ger remained in wet mode. Sunrise and sunset times for each day were derived from geoloca- tor transition files (files with extension “trn”). We first evaluated daily time spent on the wa- ter (in %) for successful and failed breeders at each phenological phase, and for day and night separately. For visualisation purposes only, we modelled daily activity at sea during day and night using generalized additive mixed models (GAMMs), separately for successful and failed breeders. We included Julian date as a smooth- ing term and bird identity as a random term. The resulting values show the proportion of day and night spent on the water, to account for the changes of day length throughout the year. We used the “mgcv” R package (Wood & Augustin 2002), based on penalized regression splines and generalized cross-validation, to select the appropriate smoothing parameters. Moonlight can influence activity patterns of petrels, par- ticularly during the non-breeding period (e.g. Yamamoto et al. 2008, Ramos et al. 2016), so we evaluated the effect of moonlight levels on nocturnal activity during the non-breeding pe- riod. We focused on this period to avoid any constraints that breeding might have on activ- ity patterns. We used GAMMs to estimate noc- turnal time on water during the non-breeding period as a response of the number of days since November’s full moon of each year (see Ramos et al. 2016 for more details of the ap- proach) as a smoothing term. This allowed us to determine cyclicity in the time spent on wa- ter during non-breeding in relation to the lunar cycle. Finally, nocturnal time on water during the non-breeding period was regressed against moonlight levels (from 0 during a new moon, to 100 during a full moon) using locally-weight- ed, non-parametric regressions (Jacoby 2000). 2.4. Habitat modelling We used MaxEnt 3.3.3k software to develop habitat suitability models (Phillips et al. 2006, Elith et al. 2011). Taking into account similar studies (Quillfeldt et al. 2013, Ramírez et al. 2013, Ramos et al. 2015), seven environmental variables were selected through jack-knife test for their possible importance for predicting At- lantic petrel distribution: seafloor depth (BAT, m), bathymetric gradient (BATG, %; estimated as proportional change of seafloor depth cal- culated as 100 * (maximum value - minimum value) / (maximum value)), surface chlorophyll a concentration (CHLA, mg m-3 as a proxy of biological production), distance to the colony (DCOL, km), sea surface temperature (SST, ºC), salinity (SAL, ‰) and wind speed (WIND, m s-1). The environmental information layers 101Spatial ecology of the Atlantic petrel were downloaded as monthly averages from the ERDDAP data server in raster format (Simons 2017). In order to select those environmental variables that better explain the distribution of Atlantic petrels we used the function Vari- ableSelection from “MaxentVariableSelection” R package (R Core Team 2019). We first exclud- ed those variables that contributed less than 5 % to the model (contribution threshold = 0.5) and then excluded the correlated environmen- tal variables (Pearson correlation, correlation threshold = 0.7), keeping those with the high- est contribution score. As monthly variables of BAT and BATG were correlated (Table S1), and WIND and SAL explained < 5 % of the distri- bution for all phenological phases, we reduced environmental predictors to four non-redun- dant variables: BATG, CHLA, DCOL and SST. Environmental layers were averaged for each phenological phase and resampled to a spatial resolution of 2º, in order to match the spatial er- ror in geolocation data. With those layers, habi- tat suitability models for each phase were gen- erated using 1,000 possible random locations from inside 50 % kernel density contours. Each final model was the average of 100 models and their fit was evaluated using the area under the curve (AUC) statistic, which measures the abil- ity of model predictions to discriminate species presence from background locations. 2.5. Feather sampling and SIA We analysed the stable isotopes of nitrogen (δ15N) and carbon (δ13C) in the 1st, 3rd, 5th, 7th and 10th primary feathers (P1, P3, P5, P7 and P10), the 13th secondary feather (S13) and the 6th tail feather (rectrix, R6) sampled from 8 dead Atlantic petrels (we cannot distinguish if immature or adults) found on Gough Island in September 2009. As feathers are metabolically inert once formed, they retain the δ15N and δ13C values from the bird’s diet at the time of growth, when they are irrigated by blood. Therefore, stable isotopes from feathers provide informa- tion about trophic levels (δ15N) and foraging areas (δ13C) when feathers were growing (Hob- son et al. 1994, Cherel et al. 2000). All feathers were cleaned in a 0.50M NaOH solution, rinsed twice in distilled water in order to remove any contamination and oven dried at 60ºC to constant mass. Thereafter, feathers were flash frozen with liquid nitrogen and ground using a cryogenic grinder (Spex Certiprep 6850) to obtain a fine powder. Subsamples of 0.30 – 0.32 mg were weighed and placed into tin capsules to be oxidized in a Flash EA1112 and TC/EA coupled to a stable isotope mass spectrometer Delta C through a Conflo III interface (Ther- moFinnigan) in Serveis Científico-Tècnics of the University of Barcelona (Spain). Stable isotope ratios are expressed in δ conventional notation as parts per thousand (‰) according to the following equation: δX = [(Rsample/Rstandard) – 1] x 1000, where X is 15N or 13C and R corre- sponds to ratio 15N/14N or 13C/12C related to the standard values. Rstandard for 15N is atmospheric nitrogen (AIR) and for 13C is Vienna Pee Dee Belemnite (VPDB). The international stand- ards applied (IAEA N 1 , IAEA N2, USGS 34 and IAEA 600 for N; IAEA CH 7 , IAEA CH 6 , USGS 40 and IAEA 600 for C) were inserted every 12 feather samples to calibrate the system and compensate for any drift over time. Values of δ15N and δ13C were compared to those of other petrels (Procellariidae) that overlap their dis- tribution with Atlantic petrel’s non-breeding region. 2.6. Ethics statement All work was conducted in accordance with the appropriate institutional guidelines (Uni- versity of Cape Town Animal Ethics Com- mittee: 2014/V10/PRyan and 2017/V10REV/ PRyan), and with the approval of the Tristan da Cunha government. The weight of tagged birds was > 500 g and the weight of geolocator was ~ 2 g, which was well below the deleterious rec- ommended threshold of 3 - 5 % of body weight for back-mounted devices (Phillips et al. 2003, Igual et al. 2005, Passos et al. 2010). All birds were handled in strict accordance with good animal practice; deployment and recovery of geolocators took < 5 minutes and had no visible deleterious effects on study animals. 102 CHAPTER 3 : 3. RESULTS 3.1. Phenology and spatial distribution All tracked birds remained within the South Atlantic Ocean, with most time spent west of the breeding islands. Successful and failed breeders showed similar spatial distributions in each phenological phase (Fig. 2 shows de- tailed distribution of each phenological group). Adults spent the non-breeding period off north- ern Argentina, Uruguay and southern Brazil; during the pre-laying exodus, they mainly used the waters over the edge of the South American continental shelf, whereas during incubation and chick-rearing they used two main foraging areas, one around Gough Island and another closer to the South American coast (Fig. 2). Multivariate hierarchical clustering based on phenology identified two distinct clusters of birds (Fig. 1). The mean silhouette width, with a value of 0.59, provided reasonable support for the structure (Fig. S2). Three year-round trips presented low widths (< 0.25), which indicated low support for the classification of these sam- Figure 2. Year-round distributions of adult Atlantic petrels (successful and failed breeders separately). Blue for successful breeders and light brown for failed breeders. Filled contours refer to 50 % (darker polygons) and 95 % (lighter polygons) kernel UD (core areas of activity and the areas of active use, respectively) in the South Atlantic Ocean at each phenological phase. Triangle represents breeding colony at Gough Island. 103Spatial ecology of the Atlantic petrel Table 1. Phenology of adult Atlantic petrels detailed separately for successful and failed breeders. Start- ing date (day/month mean ± SD, over all years of study) corresponds to the mean date when each phenological phase or migration starts. Last column resumes results of U Mann-Whitney-Wilcox tests between successful and failed breeders phenological dates, applying Bonferroni correction for multiple comparisons. aDetails about incubation bouts were obtained from 4 successful breeders and 1 failed breeder for which incubation period was not interrupted by deployment or recovery of the logger. This small number of birds prevents us from comparing the duration of incubation. Because only one individual of a couple was tracked, the actual length of the incuba- tion period could be longer. bDuration of chick-rearing was not compared due to the small amount of data for failed breeders (3 birds performed post-breeding migration immediately after incubation). Phenological phase Successful (n = 10) Failed (n = 8) U Mann-Whitney-Wilcox Test Post-breeding migration Start date 25/12 ± 8.1 19/09 ± 32.0 W = 80.0; p-value < 0.001 Duration (d) 9.6 ± 3.1 10.7 ± 3.4 W = 0.6; p-value = 0.591 Non-breeding Start date 04/01 ± 6.1 30/09 ± 31.1 W = 80.0; p-value < 0.001 Duration (d) 97.7 ± 4.2 168.0 ± 29.0 W = 0.0; p-value = 0.001 Pre-breeding migration Start date 11/04 ± 4.7 17/03 ± 10.9 W = 70.0; p-value = 0.001 Duration (d) 5.3 ± 3.0 4.1 ± 2.7 W = 0.4; p-value = 0.373 Pre-breeding Start date 17/04 ± 5.6 21/03 ± 9.1 W = 70.0; p-value = 0.001 Duration (d) 9.4 ± 3.8 24.6 ± 18.2 W = 0.1; p-value = 0.106 Pre-laying exodus Start date 26/04 ± 7.3 15/04 ± 15.1 W = 56.5; p-value = 0.039 Duration (d) 78.3 ± 4.2 77.43 ± 16.8 W = 0.5; p-value = 0.461 Incubation Start date 13/07 ± 5.3 01/07 ± 11.4 W = 61.0; p-value = 0.013 Duration (d) 58.0 ± 7.5a 82.0 ± 0.0a a Chick-rearing Start date 25/08 ± 8.1 03/09 ± 12.7 W = 14.0; p-value = 0.197 Duration (d) 122.1 ± 10.1 33.8 ± 31.4 b ples, but details of their individual phenology support that their classification is more related to failed than to successful breeders (see Figs. S2 and S3 for a detailed explanation). Previous knowledge of breeding phenology based on observations on land (Cuthbert 2004) suggests that late migrants were successful breeders, whereas early migrants were failed breeders (see discussion for an extended explanation). Thus, following the clustering results, we de- scribed the phenology of the Atlantic petrels and detailed the breeding schedules separately for successful (n = 10) and failed breeders (n = 8; Table 1). Successful breeders left Gough Island at the end of December and carried out a post-breeding migration towards South Amer- ica. They arrived at the non-breeding area off northern Argentina, Uruguay and southern Brazil (Fig. 2A) at the beginning of January and stayed in the area for 98 days. Successful breeders started the pre-breeding migration back to the colony in the middle of April, ar- riving 5 days later. In late April, at the begin- ning of the breeding season, they travelled to 104 CHAPTER 3 : off the northern Argentinean coast and the Falkland Islands for the pre-laying exodus (Fig. 2B), returning in the middle of July to lay and incubate the egg (Fig. 2C). Detailed data about incubation (Table 1) were obtained from 4 suc- cessful breeders and 1 failed breeder for which the deployment or recovery of the logger did not interrupt the incubation period. Note how- ever that the total length of the incubation peri- od could be longer than recorded from the geo- locator data because only one bird of each pair was tracked and its partner could have done the first or last bout. Successful breeders incubated the egg in two (3 birds) or three bouts (1 bird), with a median duration ± 95 % confidence interval of 16.0 ± 3.6 days (n = 9 bouts). One failed breeder also incubated in 3 bouts (16.0 ± 3.6 days). Chicks hatched in late August-Sep- tember, when the adults foraged in the same areas used during incubation (one off the Ar- gentinean continental shelf, and one closer to Gough Island; Figs. 2C and D). Failed breeders left for the non-breeding grounds earlier than successful breeders, had a longer non-breeding period, and returned to the colony earlier the following season (Table 1, Fig. 1B). The appar- ent result of failed breeders laying earlier but hatching later than successful breeders (Table 1), would not be taken in consideration because deploying and recovering of geolocators took place during incubation, thus breaking the con- nection between incubation and subsequent chick-rearing (i.e. the consideration as success- ful or failed breeders relate only to the year Figure 3. Year-round at-sea activity patterns of adult Atlantic petrels. Proportion of daily time spent on water (mean ± 95 % confidence interval of the slopes; estimated through generalized additive mixed models; GAMMs) during the day (light blue) and during the night (dark blue) along the annual cycle. Raw data is rep- resented as dots on the background. Data is shown separately for the two cluster groups: (A) successful and (B) failed breeders. Horizontal bars at the top of each subplot show mean phenological dates of each cluster of birds: pre-laying exodus (dark purple), incubation (yellow), chick-rearing (green), non-breeding (blue) and pre-breeding (time from arrival at breeding grounds to pre-laying exodus; light purple). Arrows correspond to post- and pre-breeding migrations. 105Spatial ecology of the Atlantic petrel Figure 4. Effect of moonlight on non-breeding nocturnal activity of Atlantic petrels. (A) The mean of nocturnal time on water estimated through GAMMs is represented by black solid line, and the associated 95 % confidence interval of the slopes corresponds to dark blue region. To compare non-breeding data of different lunar cycles (2010 - 2013), daily hours of nocturnal time spent on water were re-scaled to the first full moon of November of each year. Moon phase is represented with a light grey wavy line (0 representing new moon and 100 full moon). (B) Nocturnal time on water during the non-breeding period as function of moonlight. Dots rep- resent individual observations, thin lines correspond to individual locally-weighted non-parametric regressions, and the thick line corresponds to the mean of the species. right after the logger deployment and cannot be maintained to the next year). 3.2. At-sea activity Both successful and failed breeders spent less time on the water during the breeding period (pre-laying exodus, incubation and chick-rear- ing) than during the non-breeding period (Fig. 3, Table S2). Both successful and failed breed- ers noticeably increased the time on water dur- ing the non-breeding period, although in accor- dance with phenology, failed breeders clearly advanced this pattern in the calendar (Fig. 3). Despite petrels showed similar proportions of time spent on water during day and night with- in each phenological phase, the proportion of time on water was slightly higher during night than during day (Fig. 3), except during the non- breeding period, when nocturnal activity was clearly influenced by moonlight (Fig. 4). Dur- ing this period, tracked birds spent more time on water during nights at new moon and spent more time flying on moonlit nights (Fig. 4). 3.3. Habitat modelling The importance of each environmental variable in the MaxEnt models differed between pheno- logical phases (Table 2, Fig. S4). The most im- portant variables were: SST (20 - 25 ºC) in the non-breeding period; DCOL (2,500 - 4,000 km) during the pre-laying exodus; DCOL (0 - 2,500 km) and SST (0 - 7 ºC) during incubation; and DCOL (0 - 1,900 km) during chick-rearing (Fig. S4; the response curves are detailed in Fig. S5). Fig. 5 compiles the obtained habitat suitability models considering these environmental vari- ables for each phenological phase. During non- breeding, suitable habitats outside the recorded distribution occurred in the southeast Atlantic, especially in the Benguela Upwelling region. 3.4. Stable isotope values Atlantic petrels presented a narrower range of δ15N (13.1 to 15.5 ‰) than δ13C values (-19.3 to -16.1 ‰; Fig. 6, Table 3; see Table S3 for de- tailed values). Both isotopic ranges are wider in 106 CHAPTER 3 : Table 2. Most important environmental variables to the probability of occurrence of adult Atlantic pe- trels. MaxEnt modelling selected gradient of seafloor depth (BATG), chlorophyll a concentration (CHLA) as a proxy of biological production, distance to the colony (DCOL) and sea surface temperature (SST) as the most important environmental variables to predict the occurrence of adult Atlantic petrels within 50 % kernels UD (utilisation distribution) for each phenological phase. Estimates of model fit (as the area under the receiver op- erating characteristic curve; AUC) and relative importance (as percent contribution, in bold values over 15 %) of these environmental variables. Redundant environmental variables (BAT) and those variables explaining < 5 % of the distribution (SAL and WIND) were excluded during the modelling to reduce noise in the outputs. NA when relative importance or percent contribution < 5 %. Figure 5. Habitat suitability of Atlantic petrels for every phenological phase derived from environmental modelling. Habitat suitability ranges from light yellow (less suitable habitat) to dark blue (most suitable habi- tat). Black contour lines indicate 50 % kernel UD of positions of both successful and failed breeders; triangle shows colony location. Phenological phase AUC Relative importance (%) Percent contribution (%) BATG CHLA DCOL SST BATG CHLA DCOL SST Non-breeding 0.915 ± 0.025 NA 11.9 NA 83.6 22.3 11.0 NA 66.6 Pre-laying exodus 0.988 ± 0.002 NA NA 94.0 NA NA 28.7 59.5 11.8 Incubation 0.991 ± 0.002 NA NA 66.5 29.9 NA NA 54.4 28.7 Chick-rearing 0.992 ± 0.002 NA NA 98.1 NA NA NA 90.7 9.3 107Spatial ecology of the Atlantic petrel Table 3. δ15N and δ13C values (mean ± SD) of feathers from several petrel and shearwater species found in the southern Atlantic Ocean. 1st, 3rd, 5th, 7th and 10th primary feathers (P1, P3, P5, P7 and P10), 13th secondary (S13) and 6th rectrix (R6) feathers of Atlantic petrels breeding on Gough Island. Feathers of Great shearwater (Ardenna gravis), Manx shearwater (Puffinus puffinus), Cory’s shearwater (Calonectris borealis) and White-chinned petrel (Procellaria aequinoctialis), are known to be moulted in the Brazil-Falklands Confluence. aValues excluding the outlier. 13.0 13.5 14.0 14.5 15.0 15.5 16.0 P1 P3 P5 P7 P10 S13 R6 Feather δ1 5 N (‰ ) -20.0 -19.5 -19.0 -18.5 -18.0 -17.5 -17.0 -16.5 -16.0 -15.5 P1 P3 P5 P7 P10 S13 R6 Feather δ1 3 C (‰ ) 13.0 13.5 14.0 14.5 15.0 15.5 16.0 P1 P3 P5 P7 P10 S13 R6 Feather δ1 5 N (‰ ) -20.0 -19.5 -19.0 -18.5 -18.0 -17.5 -17.0 -16.5 -16.0 -15.5 P1 P3 P5 P7 P10 S13 R6 Feather δ1 3 C (‰ ) Species Feather n δ15N (‰) δ13C (‰) Source Atlantic petrel P1 8 14.4 ± 0.7 -17.8 ± 1.0 Present study P3 8 14.4 ± 0.7 -17.7 ± 0.8 P5 8 14.4 ± 0.7 -17.6 ± 0.8 P7 7 14.4 ± 0.6 -17.7 ± 0.6 P10 8 14.3 ± 0.3 -17.0 ± 0.4 S13 7a 14.3 ± 0.3 -17.4 ± 0.5 R6 6 14.3 ± 0.3 -17.0 ± 0.4 Great shearwater P1 6 15.6 ± 1.2 -16.7 ± 1.6 T. Militão unpubl. data Manx shearwater R6 13 17.6 ± 1.9 -16.3 ± 0.5 T. Militão unpubl. data Cory's shearwater S13 4 13.9 ± 0.8 -16.4 ± 0.3 T. Militão unpubl. data White-chinned petrel Body feathers 8 - 10 17.6 ± 1.4 -15.5 ± 0.8 (Phillips et al. 2009) Figure 6. Stable isotope signatures (δ15N and δ13C) of primary, secondary and rectrix feathers of Atlantic petrels. (A) δ15N and (B) δ13C of 1st, 3rd, 5th, 7th and 10th primary feathers (P1, P3, P5, P7 and P10), 13th secondary (S13) and 6th rectrix (R6) feathers of Atlantic petrels (n = 8; values in Table 3). Lines connect values corresponding to feathers from the same individual, note that not all sequences are complete. Primary feather replacement is assumed to be simple and descendent in procellariiformes, starting from P1-3 and moulting se- quentially towards P10. Secondary and rectrix feathers (here S13 and R6) are thought to be moulted out of the breeding season, not sequentially, as represented by dashed lines (Bridge 2006, Ramos et al. 2009). 108 CHAPTER 3 : P1-P7, showing higher isotopic variability, than within P10 feathers. Both isotopic signatures and variability of S13 and R6 showed similar values to those of P10. Compared with other petrel species moulting in the Brazil-Falklands Confluence, Atlantic petrels show lower values of δ15N and δ13C (Table 3). 4. DISCUSSION Our study provides new insights into the spatial ecology of the Atlantic petrel. We report for the first time at-sea activity patterns, habitat pref- erences, moulting strategies, and carry-over ef- fects over the entire annual cycle of this endan- gered species. Moreover, we extend previous knowledge about the timing of life-cycle events and migration schedules year-round by quanti- fying phenological variability that arose from presumed breeding success. We present new critical knowledge and refine previous data, providing an ensemble of relevant information for its conservation. However, the small sample size and the lack of immatures in the sample limit the general relevance of our findings. The breeding phenology inferred in this study generally agrees with data reported in previ- ous colony-based studies (Richardson 1984, Cuthbert 2004, Wanless et al. 2012, Dilley et al. 2015) (Table S4). However, our results high- light considerable within-population variabil- ity in phenological events. Multivariate hier- archical clustering based on phenological data allowed us to distinguish between early and late phenological groups. The group with ad- vanced phenology dedicated, on average, about 88 days less to chick-rearing (Table 1), prob- ably as a result of breeding failure. In recent years, a high proportion of chicks is killed by introduced house mice at Gough Island (Dilley et al. 2015). Thus, although breeding outcome was not monitored, the phenological variability found between groups likely is due to breeding success or failure. Both phenological groups differed in the starting date of post-breeding migration and the five subsequent phenological phases (Fig. 1, Table 1). The “early migrants” departed the breeding area between 7 August and 9 November (well before December, when chicks usually fledge (Cuthbert 2004)), indicat- ing that birds showing this early post-breeding migration were likely failed breeders. The “late migrants” started their post-breeding migra- tion in December or later, and therefore pre- sumably, were successful breeders. Interestingly, we found that breeding success influenced subsequent phenological phases of the species. Failed breeders not only departed to the non-breeding area earlier and stayed there longer than successful breeders, they also returned earlier to the colony at the onset of the next breeding period. These results dem- onstrate a carry-over effect on this species not only from the breeding to the non-breeding pe- riod, but also to the subsequent breeding period. It is likely that birds without breeding responsi- bilities that migrate earlier to the non-breeding grounds were able to moult and recover their body condition earlier than successful breeders, potentially improving their chances of breed- ing successfully in the subsequent breeding at- tempt (Kokko 1999). Nevertheless, despite the phenological differences between successful and failed breeders, all birds showed similar flyways and non-breeding areas, probably be- cause of the relatively restricted and consistent non-breeding area for the entire species. This last result was also found in Cory’s shearwater (Calonectris borealis), but their breeding suc- cess did not change their migratory schedule (Ramos et al. 2018). However, our findings con- trast with previous studies also in Cory’s shear- waters and Black-legged kittiwakes (Rissa tri- dactyla), where winter distribution depends on reproductive performance (Bogdanova et al. 2011, Catry et al. 2013). Among Northern gan- nets (Morus bassanus), foraging grounds also differed between failed and successful breed- ers (Votier et al. 2017). However, we are aware of the limitations of our sample size, and the need of an experimental design monitoring the breeding performance of every individual in order to be more conclusive on such carry-over effects (e.g. Harrison et al. 2011). Our geolocation data confirmed that the Southwest Atlantic Ocean is the main distri- 109Spatial ecology of the Atlantic petrel bution range for Atlantic petrels year-round (Ramos et al. 2017). The observed core range is more restricted to the west than tradition- ally considered (i.e. from east coast of South America to west coast of South Africa), but this might be a consequence of our modest sample size, and the fact that only unsexed adults were tracked in this study (Enticott 1991, Orgeira et al. 2013, Carboneras et al. 2017a). Adults were largely confined to oceanic waters of the cen- tral and western South Atlantic. The edge of the South American continental shelf, off northern Argentina, Uruguay and southern Brazil, was exploited during all phenological phases, al- though extension and location of core areas dif- fered between periods (Fig. 2). In this region, the Brazil-Falklands Confluence, where warm waters from the Brazil Current mix with cold waters from the Falklands Current, creates a productive ecosystem that supports a complex community of top-predators, including many seabird species (Croxall & Wood 2002, Olmos 2002, Acha et al. 2004). Although the abun- dance of top-predators results in local competi- tion, the high productivity likely explains why Atlantic petrels exploit this area. Avoidance of competition near Gough Island, where waters are less productive, and richer waters along the South American continental shelf may explain why birds commute around 3,500 km to a more distant and productive area far from the colony. The proportion of time spent on water (both during day and night) was lower while breed- ing than during the non-breeding period (Fig. 3). This pattern is likely explained by moult- ing phenology. Although there is scant infor- mation on the timing of moult in Atlantic pet- rels, most petrels complete an annual primary feathers moult, starting immediately after the breeding season in order to avoid overlapping these metabolically demanding periods (breed- ing and moulting) (Bridge 2006). Moult typi- cally commences with 2 - 4 inner primaries, but only 1 - 2 outer primaries are moulted at a time, because their moult has a greater impact on flight performance (Bugoni et al. 2014). The intense replacement of wing feathers during the non-breeding period (see below) decreases flight capability, forcing birds to spend more time on water (Cherel et al. 2016). The effect of moult on flight time was also observed in failed breeders that advanced both the post-breeding migration and the non-breeding period, and thus likely their moulting period (Fig. 3). An- other possible contributing factor could be the move from central-place foraging while breed- ing (i.e. highly energy investment to meet the breeding demands) to a lower energy demand during the non-breeding period (Mackley et al. 2011, Cherel et al. 2016). To ensure breeding success, seabirds need to increase foraging ef- fort (Lescroël et al. 2010), which likely means to perform both nocturnal and diurnal forag- ing to feed the chicks either more frequently or with a larger variety of prey. However during non-breeding period, compared to the breed- ing period, birds spent more time in flight at night (at least during periods of increased moonlight), when some species of cephalopods become more accessible near the surface due to their diel vertical migrations (DVM) (Imber 1973). As for other Pterodroma petrels, cepha- lopods are the main prey of Atlantic petrels, which may include dead or moribund squid floating at the surface during the day (Rich- ardson 1984, Croxall & Prince 1994, Klages & Cooper 1997, Perez et al. 2019). Nocturnal ac- tivity was clearly influenced by moonlight over the non-breeding period, i.e. petrels spent more time flying with increasing levels of moonlight intensity (Fig. 4). Previous studies have found similar results in other gadfly petrel species and suggested that light intensity during full moon nights could facilitate foraging (e.g. Pinet et al. 2011b, Ramírez et al. 2013, Ramos et al. 2016). However, greater activity levels on well- lit nights may just result from DVM organisms remaining in deeper waters when moonlight is brighter, forcing Atlantic petrels to increase their search effort for prey (Benoit-Bird et al. 2009). We observed a high individual variability in isotopic results on several primary feathers obtained from dead specimens (i.e., P1 - P7 feathers; Fig. 6, Table 3). This likely indicates that these feathers grew in different individual 110 CHAPTER 3 : non-breeding grounds within the general non- breeding area (see Cherel et al. 2000, McMa- hon et al. 2013). By comparison, the low iso- topic variability in P10, S13 and R6 among individuals possibly indicates these feathers were replaced in a common area for all birds, i.e. around the colony site after arrival from the non-breeding area between end of March and mid-April (Fig. 2). Elliot (1957) reported that birds arriving at Tristan da Cunha at the end of March were still in moult, as were birds carried inland in Brazil by hurricane Catarina in March 2004 (Bugoni et al. 2007). Although we cannot distinguish if the 8 dead specimens found at Gough Island were immature or adults, these results indicate similar phenological patterns in their migratory behaviour to those obtained through geolocator data. The isotopic gradient observed along P1 to P7 feathers could reflect a north-south gradient in isotopic baselines, with feathers with lower isotopic values moulted farther north, and those with higher isotopic values, moulted further south, in the Brazil- Falklands Confluence (Figs. 2A and 6). This north-south trend is consistent with prey iso- topic data (see δ15N in Table 4). However, the lack of a detailed zooplankton isoscapes for the non-breeding distribution prevented us from confirming this gradient at lower trophic levels (McMahon et al. 2013). It is clear that the edge of the South American continental shelf is an important foraging area for Atlantic petrels year-round. Shelf slopes are important habitats for many squid species, which are caught by fishing fleets year-round along the outer shelf and upper slope off south- ern Brazil (Haimovici et al. 1998, Arkhipkin et al. 2015). However, we did not find an increase in isotopic values with increasing trophic lev- els when comparing results from flight feath- ers moulted in the Brazil-Falklands Confluence with those from cephalopod species sampled in the same area (e.g. Drago et al. 2015; see Table 4). This mismatch may arise from differential timing of sampling (i.e., different years and/ or seasons within the same year) and from un- specified limitations of using literature isotopic data. Nevertheless, comparisons of δ15N and δ13C values of Atlantic petrel feathers with oth- er shearwater species moulting in the Brazil- Falklands Confluence (e.g., Great shearwater and White-chinned petrel; Table 3) suggest a lower trophic level of the Atlantic petrel, which might reflect the limited use of fisheries dis- cards by this species, and, thus, its lower risk of bycatch compared with other species (Barrett et al. 2007, Bugoni et al. 2008, Phillips et al. 2009, Bugoni et al. 2010). Regarding the Atlantic petrel distribution, oceanic productivity may not be a good predic- Area Prey n δ15N (‰) δ13C (‰) Source Brazil Current Doryteuthis (Loligo) pealeii 5 11.3 ± 0.5 -17.6 ± 0.2 (Drago et al. 2015) Illex argentinus 5 10.0 ± 0.5 -18.1 ± 0.2 (Drago et al. 2015) Loligo sanpaulensis 5 15.2 ± 0.3 -16.3 ± 0.1 (Drago et al. 2015) Ommastrephes bartrami / I. argentinus 8 9.3 ± 0.8 -16.7 ± 0.4 (Bugoni et al. 2010) All species 11.4 ± 0.5 -17.2 ± 0.2 Brazil- Falklands Confluence I. argentinus 5 14.7 ± 0.5 -17.5 ± 0.4 (Drago et al. 2015) I. argentinus 2 13.9 ± 0.7 -18.7 ± 0.2 (Franco-Trecu et al. 2012) L. sanpaulensis 5 18.6 ± 0.2 -16.7 ± 0.2 (Drago et al. 2015) L. sanpaulensis 2 13.7 ± 0.2 -17.9 ± 0.1 (Franco-Trecu et al. 2012) All species 15.2 ± 0.4 -17.7 ± 0.2 Table 4. δ15N and δ13C values (mean ± SD) of several cephalopod species (mantle muscle) in the Brazil Current and Brazil-Falklands Confluence. 111Spatial ecology of the Atlantic petrel tor of its distribution because the species relies on relatively oligotrophic waters for feeding year-round, being a truly oceanic species like most gadfly petrels (Ramos et al. 2016, Ramos et al. 2017). In general, year-round habitat suit- ability models based on several environmental predictors agree well with the observed spe- cies distribution (Fig. 5; Enticott 1991, Orgeira 2001, Carboneras et al. 2017a). However, dur- ing the non-breeding period, only one of the two suitable habitats, the shelf and slopes of the Brazil-Falklands Confluence, fitted well with the core range of Atlantic petrels (Fig. 5A). It is not known why Atlantic petrels are so rare in the Benguela Current region (Enticott 1991). Their distribution contrasts markedly with several other seabird species that use both areas during the non-breeding period, such as Scopoli’s (Calonectris diomedea) and Cory’s shearwaters (González-Solís et al. 2007). Dur- ing the pre-laying exodus, two suitable habitats were identified, one in northern Argentina and Falkland Islands, and another south of Africa (Fig. 5B), which again was not used by any tracked birds, and is an area with few obser- vations at sea (Enticott 1991). During incuba- tion and chick-rearing, an apparently suitable area in the south eastern Atlantic also was not highly used by tracked birds (Figs. 5C and D), but they do occur in reasonable numbers south of Africa (38-42ºS) in November, towards the end of the chick-rearing period (P.G. Ryan pers. obs.). Apart from the small sample size, one possible explanation for these differences could be the competitive exclusion or the “ghost of past” competition with other gadfly petrels in the region (Connell 1980). The Great-winged petrel (Pterodroma macroptera), which shows a similar phenology and diet, is abundant off southern Africa and largely absent from the southwest Atlantic (Ridoux 1994, Brooke 2004, BirdLife International 2017a, Carboneras et al. 2017b). It breeds abundantly at islands in the Southwest Indian Ocean, and used to be com- mon at Tristan and Gough, but has become rare in recent years due to hunting (at Tristan) and introduced predators (at both islands) (Bird- Life International 2017a, Ramos et al. 2017). The smaller Soft-plumaged petrel (Pterodroma mollis) remains abundant at Gough and the un- inhabited Tristan islands, as well as at islands in the Southwest Indian Ocean, and is the most common gadfly petrel in the southeast Atlan- tic, but performs the opposite phenology to the Atlantic petrel (BirdLife International 2017a, Ramos et al. 2017). In addition, the distribution and abundance of squids is poorly known in austral oceans, but commercial squid fisheries are more abundant along the South American shelf and shelf slopes than off South Africa (FAO Marine Resources Service 2005). This fact could indicate a higher abundance of the main prey for gadfly petrels off South America, where Atlantic petrels overlap with other gad- fly petrels, such as the Desertas petrel (Ptero- droma deserta; BirdLife International 2017a, Ramos et al. 2017). This area is important for fishing fleets, and the high fishing intensity may decrease prey abundance for Atlantic petrels and other seabirds (Furness 2003, Bugoni et al. 2008). It also supports large numbers of vessels with their inherent potential threats (as mortal- ity, but also sub-lethal effects) to seabirds and marine life (Finkelstein et al. 2006, Lewison et al. 2012, Krüger et al. 2017, Rodríguez et al. 2017). Since this is the area where all tracked birds spent their non-breeding period, and as Gough Island is virtually the only breeding location for this species, a good conservation strategy for both areas is essential to ensure sustainability of the Atlantic petrel. Indeed, one Ecologically or Biologically Significant Area (EBSA) and several Important Bird Areas (IBAs) overlap with the species’ non-breeding distribution. For the breeding location, one Ma- rine Protected Area (MPA) is designated and several IBAs and MPA are proposed around Tristan da Cunha Island and Gough Island (which is part of an UNESCO World Heritage Site and also Wetlands of International Impor- tance under the Ramsar Convention), which should help to conserve the species (BirdLife International 2017c, Convention on Biological Diversity 2017, Dias et al. 2017, Marine Con- servation Institute 2017, UNESCO 2019). 112 CHAPTER 3 : 5. CONCLUSIONS In this study, we describe important aspects of the spatio-temporal ecology of Atlantic petrels. The non-breeding period of success- ful breeders lasted from the end of December to mid-April. Habitat preferences highlighted the South American continental shelf as an extremely important area for the species. We relate activity patterns with breeding con- straints, foraging behaviour and, together with stable isotope analysis (SIA), provide new in- sights into the timing of wing moult. We also provide evidence of carry-over effects between consecutive breeding attempts. However, fur- ther studies tracking larger numbers of birds of different sexes and ages and monitoring their breeding performance at the colony, would pro- vide more reliable understanding of ecological factors that determine the at-sea distribution and behaviour of this endangered seabird. ACKNOWLEDGEMENTS The Ministerio de Educación y Ciencia and Ministerio de Ciencia e Innovación from the Spanish Government (Projects CGL2006- 01315/BOS, CGL2013-42585-P, CGL2009- 11278/BOS) financially supported this study. Logistical support and financial funding dur- ing field work was provided by the South Afri- can Department of Environmental Affairs and the National Research Foundation, through the South African National Antarctic Programme, with additional support from the Royal Soci- ety for the Protection of Birds. RR, ZZ and JMR-G were supported by Spanish MINECO (Juan de la Cierva postdoctoral programme, JCI-2012-11848), Universitat de Barcelona (UB, APIF/2012) and Spanish MECD (FPU, AP2009-2163), respectively. Permission to conduct research at Gough Island was granted by the Administrator and Island Council of Tristan da Cunha, through Tristan’s Conser- vation Department. We thank R. Ronconi for his help during field work, and the many field assistants who deployed and recovered tags at Gough Island. We thank three anonymous ref- erees for their comments to improve an earlier version of this manuscript. AUTHOR CONTRIBUTIONS The study was conceptualized by RR, ZZ, JMR-G and JG-S. Data were collected by PGR and JG-S. Geolocator and isotope raw data were analysed by MP-P and MLR, respec- tively. Posterior data analysis were carried out by MP-P, RR, ZZ and JMR-G. The manuscript was drafted by MP-P with review and edito- rial contributions by RR, ZZ, JMR-G, PGR and JG-S. JG-S obtained the funding. All authors read and approved the final manuscript. LITERATURE CITED Acha EM, Mianzan HW, Guerrero RA, Favero M, Bava J (2004) Marine fronts at the con- tinental shelves of austral South America: physical and ecological processes. J Mar Syst 44: 83-105 Afanasyev V (2004) A miniature daylight level and activity data recorder for tracking ani- mals over long periods. Memoirs of NIPR 58: 227-233 Arkhipkin AI, Rodhouse PGK, Pierce GJ, Sauer W, Sakai M, Allcock L, Arguelles J, Bower JR, Castillo G, Ceriola L, Chen C-S, Chen X, Diaz-Santana M, Downey N, González AF, Amores JG, Green CP, Guerra A, Hendrickson LC, Ibáñez C, Ito K, Jereb P, Kato Y, Katugin ON, Kawano M, Kidokoro H, Kulik VV, Laptikhovsky VV, Lipinski MR, Liu B, Mariátegui L, Marin W, Medina A, Miki K, Miyahara K, Moltschaniwskyj N, Moustahfid H, Nab- hitabhata J, Nanjo N, Nigmatullin CM, Ohtani T, Pecl G, Perez JAA, Piatkowski U, Saikliang P, Salinas-Zavala CA, Steer M, Tian Y, Ueta Y, Vijai D, Wakabayashi T, Yamaguchi T, Yamashiro C, Yamashita N, Zeidberg LD (2015) World Squid Fisheries. Rev Fish Sci Aquacult 23: 92-252 113Spatial ecology of the Atlantic petrel Barrett RT, Camphuysen KCJ, Anker-Nilssen T, Chardine JW, Furness RW, Garthe S, Hüppop O, Leopold MF, Montevecchi WA, Veit RR (2007) Diet studies of seabirds: a review and recommendations. J Mar Sci, 64: 1675-1691 Benoit-Bird JK, Au WW, Wisdom DW (2009) Nocturnal light and lunar cycle effects on diel migration of micronekton. Limnol Oceanogr 54: 1789-1800 BirdLife International (2017a) IUCN Red List for birds. www.birdlife.org (accessed 09 Sep 2017) BirdLife International (2017b) Seabird Tracking Database. Tracking Ocean Wanderers. http://seabirdtracking.org/ mapper/?dataset_id=966 (accessed 21 Oct 2018). BirdLife International (2017c) Marine IBA e- atlas: delivering site networks for seabird conservation. https://maps.birdlife.org/ marineIBAs/default.html (accessed 4 Jun 2017) Bogdanova MI, Daunt F, Newell M, Phillips RA, Harris MP, Wanless S (2011) Seasonal interactions in the black-legged kittiwake, Rissa tridactyla: links between breeding performance and winter distribution. Proc R Soc B 278: 2412-2418 Bridge ES (2006) Influences of morphology and behavior on wing-molt strategies in seabirds. Mar Ornithol 34: 7-19 Bried J, Pontier D, Jouventin P (2003) Mate fi- delity in monogamous birds: a re-examina- tion of the Procellariiformes. Anim Behav 65: 235-246 Brooke M (2004) Albatrosses and petrels across the world. Oxford University Press, Oxford Bugoni L, Sander M, Costa ES (2007) Effects of the First Southern Atlantic Hurricane on Atlantic Petrels (Pterodroma incerta). Wil- son J Ornithol 119: 725-729 Bugoni L, Mancini PL, Monteiro DS, Nasci- mento L, Neves TS (2008) Seabird bycatch in the Brazilian pelagic longline fishery and a review of capture rates in the southwest- ern Atlantic Ocean. Endang Species Res 5: 137-147 Bugoni L, D’Alba L, Furness R (2009) Marine habitat use of wintering spectacled petrels Procellaria conspicillata, and overlap with longline fishery. Mar Ecol Prog Ser 374: 273-285 Bugoni L, McGill RAR, Furness RW (2010) The importance of pelagic longline fishery discards for a seabird community deter- mined through stable isotope analysis. J Exp Mar Biol Ecol 391: 190-200 Bugoni L, Naves LC, Furness RW (2014) Moult of three Tristan da Cunha seabird species sampled at sea. Antarct Sci 27: 240-251 Calenge C (2011) Home range estimation in R: the adehabitatHR package. Office national de la classe et de la faune sauvage, Saint Be- noist, Auffargis, France Caravaggi A, Cuthbert RJ, Ryan PG, Cooper J, Bond AL (2019) The impacts of introduced House Mice on the breeding success of nesting seabirds on Gough Island. Ibis 161: 648-661. Carboneras C, Jutglar F, Kirwan GM (2017a) Atlantic Petrel (Pterodroma incerta). In: del Hoyo J, Elliott A, Sargatal J, Christie DA, de Juana E (eds) Handbook of the Birds of the World Alive. Lynx Edicions, Barcelona. (https://www.hbw.com/node/52525 ac- cessed 24 Nov 2017) Carboneras C, Jutglar F, Kirwan GM (2017b) Great-winged Petrel (Pterodroma macrop- tera). In: del Hoyo J, Elliott A, Sargatal J, Christie DA, de Juana E (eds) Handbook of the Birds of the World Alive. Lynx Edicions, Barcelona. (www.hbw.com/node/52519 ac- cessed 23 Dec 2017) Catry P, Dias MP, Phillips RA, Granadeiro JP (2013) Carry-over effects from breeding modulate the annual cycle of a long-dis- tance migrant: an experimental demonstra- tion. Ecology 94: 1230-1235 Cherel Y, Hobson KA, Weimerskirch H (2000) Using stable-isotope analysis of feathers to distinguish moulting and breeding origins of seabirds. Oecologia 122: 155-162. Cherel Y, Quillfeldt P, Delord K, Weimerskirch H (2016) Combination of At-Sea Activity, Geolocation and Feather Stable Isotopes 114 CHAPTER 3 : Documents Where and When Seabirds Molt. Front Ecol Evol 4: 3 doi: 10.3389/ fevo.2016.00003 Clay TA, Phillips RA, Manica A, Jackson HA, Brooke MDL (2017) Escaping the oligo- trophic gyre? The year-round movements, foraging behaviour and habitat preferences of Murphy’s petrels. Mar Ecol Prog Ser 579: 139-155 Connell J (1980) Diversity and the coevolution of competitors, or the ghost of competition past. Oikos 35: 131-138 Convention on Biological Diversity (2017) Eco- logically or Biologically Significant Areas (EBSAs), Southern Brazilian Sea. www. cbd.int/ebsa/ (accessed 4 Jun 2017) Croxall, JP, Prince PA (1994) Dead or alive, night or day: how do albatrosses catch squid? Antarct Sci 6: 155-162. Croxall JP, Wood AG (2002) The importance of the Patagonian Shelf for top predator species breeding at South Georgia. Aquatic Conserv: Mar Freshw Ecosyst 12: 101-118 Croxall JP, Butchart SHM, Lascelles B, Stat- tersfield AJ, Sullivan B, Symes A, Taylor P (2012) Seabird conservation status, threats and priority actions: a global assessment. Bird Conserv Int 22: 1-34 Cuthbert RJ (2004) Breeding biology of the Atlantic Petrel, Pterodroma incerta, and a population estimate of this and other bur- rowing petrels on Gough Island, South At- lantic Ocean. Emu 104: 221-228 Cuthbert RJ, Louw H, Lurling J, Parker G, Rex- er-Huber K, Sommer E, Visser P, Ryan PG (2013) Low burrow occupancy and breed- ing success of burrowing petrels at Gough Island: a consequence of mouse predation. Bird Conserv Int 23: 113-124 Dias M, Granadeiro J, Catry P (2012) Working the day or the night shift? Foraging sched- ules of Cory’s shearwaters vary according to marine habitat. Mar Ecol Prog Ser 467: 245-252 Dias MP, Oppel S, Bond AL, Carneiro APB, Cuthbert RJ, González-Solís J, Wanless RM, Glass T, Lascelles B, Small C, Phillips RA, Ryan PG (2017) Using globally threat- ened pelagic birds to identify priority sites for marine conservation in the South Atlan- tic Ocean. Biol Conserv 211: 76-84 Dilley B, Davies D, Bond AL, Ryan PG (2015) Effects of mouse predation on burrowing petrel chicks at Gough Island. Antarct Sci 27: 543-553 Drago M, Franco-Trecu V, Zenteno L, Szteren D, Crespo EA, Riet Sapriza FG, de Olivei- ra L, Machado R, Inchausti P, Cardona L (2015) Sexual foraging segregation in South American sea lions increases during the pre-breeding period in the Río de la Plata plume. Mar Ecol Prog Ser 525: 261-272 Elith J, Phillips SJ, Hastie T, Dudík M, Chee YE, Yates CJ (2011) A statistical explana- tion of MaxEnt for ecologists. Divers Dis- trib 17: 43-57 Elliott HFI (1957) A contribution to the orni- thology of the Tristan da Cunha Group. Ibis 99: 545-586 Enticott JW (1991) Distribution of the Atlantic Petrel Pterodroma incerta at sea. Mar Or- nithol 19: 49-60 FAO Marine Resources Service, Fishery Re- sources Division (2005) Review of the state of world marine fishery resources. In: FAO Fisheries Technical Paper, No. 457. FAO, Rome Flood B, Fisher A (2013) Multimedia Identi- fication Guide to North Atlantic Seabirds Pterodroma Petrels. Scilly Pelagics, Essex Finkelstein M, Keitt BS, Croll DA, Tershy B, Jarman WM, Rodriguez-Pastor S, An- derson DJ, Sievert PR, Smith DR (2006) Albatross species demonstrate regional differences in North Pacific marine con- tamination. Ecol Appl 16: 678-686 Franco-Trecu V, Aurioles-Gamboa D, Arim M, Lima M (2012) Prepartum and postpartum trophic segregation between sympatrically breeding female Arctocephalus australis and Otaria flavescens. J Mammal 93: 514- 521 Furness R (2003) Impacts of fisheries on sea- bird communities. Sci Mar 67: 33-45 González-Solís J, Croxall JP, Oro D, Ruiz X (2007) Trans-equatorial migration and mix- 115Spatial ecology of the Atlantic petrel ing in the wintering areas of a pelagic sea- bird. Front Ecol Environ 6: 297-301 González-Solís J, Shaffer SA (2009) Introduc- tion and synthesis: spatial ecology of sea- birds at sea. Mar Ecol Prog Ser 391: 117-120 Grémillet D, Boulinier T (2009) Spatial ecol- ogy and conservation of seabirds facing global climate change: a review. Mar Ecol Prog Ser 391: 121-137 Haimovici M, Brunetti NE, Rodhouse PG, Csirke J, Leta RH (1998) Illex argentinus. In: Rodhouse PG, Dawe EG, O’Dor RK (eds) Squid recruitment dynamics. The genus Illex as a model. The commercial Il- lex species. Influences on variability. FAO Fisheries Technical Paper. No. 376. Rome, p 27-58 Harrison XA, Blount JD, Inger R, Norris DR, Bearhop S (2011) Carry-over effects as drivers of fitness differences in animals. J Anim Ecol 80: 4-18 Hilton GM, Cuthbert RJ (2010) Review article: The catastrophic impact of invasive mam- malian predators on birds of the UK Over- seas Territories: a review and synthesis. Ibis 152: 443-458 Hill RD (1994) Theory of geolocation by light levels. Elephant Seals: Population Ecol- ogy, Behavior, and Physiology. In: Boeuf L, Burney J, Laws RM (eds) University of California Press, Berkeley Hobson KA, Piatt JF, Pitocchelli J (1994) Using stable isotopes to determine seabird trophic relationships. J Anim Ecol 63: 786-798 Igual JM, Forero MG, Tavecchia G, González- Solis J, Martínez-Abraín A, Hobson KA, Ruiz X, Oro D (2005) Short-term effects of data-loggers on Cory’s shearwater (Calonectris diomedea). Mar Biol 146: 619- 624. Imber MJ (1973) The Food of Grey-Faced Pet- rels (Pterodroma macroptera gouldi (Hut- ton)), with Special Reference to Diurnal Vertical Migration of their Prey. J Anim Ecol 42: 645-662 Jacoby WG (2000) Loess: a nonparametric, graphical tool for depicting relationships between variables. Elect Stud 19: 577-613 Jodice PG, Ronconi RA, Rupp E, Wallace GE, Satgé Y (2015) First satellite tracks of the Endangered black-capped petrel. Endang Species Res 29: 23-33 Klages NTW, Cooper J (1997) Diet of the At- lantic Petrel Pterodroma incerta during breeding at South Atlantic Gough Island. Mar Ornithol 25: 13-16 Kokko H (1999) Competition for early arrival in migratory birds. J Anim Ecol 68: 940- 950 Krüger L, Paiva VH, Colabuono FI, Petry MV, Montone RC, Ramos JA (2016) Year-round spatial movements and trophic ecology of Trindade Petrels (Pterodroma arminjoni- ana). J Field Ornithol 87: 404-416 Krüger L, Ramos JA, Xavier JC, Grémillet D, González-Solís J, Kolbeinsson Y, Mil- itão T, Navarro J, Petry MV, Phillips RA, Ramírez I, Reyes-González JM, Ryan PG, Sigurðsson IA, Van Sebille E, Wanless R, Paiva VH (2017) Identification of candidate pelagic marine protected areas through a seabird seasonal, multispecific and extinc- tion risk-based approach. Anim Conserv 20: 409-424 Leal GR, Furness RW, McGill RA, Santos RA, Bugoni L (2017) Feeding and foraging ecol- ogy of Trindade petrels Pterodroma armin- joniana during the breeding period in the South Atlantic Ocean. Mar Biol 164: 211 doi: 10.1007/s00227-017-3240-8 Lescroël A, Ballard G, Toniolo V, Barton KJ, Wilson PR, Lyver PO, Ainley DG (2010) Working less to gain more: when breeding quality relates to foraging efficiency. Ecol- ogy 91: 2044-2055 Lewison R, Oro D, Godley B, Underhill L, Bearhop S, Wilson R, Ainley D, Arcos JM, Boersma PD, Borboroglu PG, Boulinier T, Frederiksen M, Genovart M, González- Solís J, Green JA, Grémillet D, Hamer KC, Hilton GM, Hyrenbach KD, Martínez- Abraín A, Montevecchi WA, Phillips RA, Ryan PG, Sagar P, Sydeman WJ, Wanless S, Watanuki Y, Weimerskirch H, Yorio P (2012) Research priorities for seabirds: Im- proving seabird conservation and manage- 116 CHAPTER 3 : ment in the 21st century. Endang Species Res 17: 93-121 Lucas Z, MacGregor D (2006) Characteriza- tion and source of oil contamination on the beaches and seabird corpses, Sable Island, Nova Scotia, 1996–2005. Mar Pollut Bull 52: 778-789 Marine Conservation Institute (2017) MPAtlas. www.mpatlas.org (accessed 4 Jun 2017) McMahon KW, Hamady LL, Thorrold SR (2013) A review of ecogeochemistry ap- proaches to estimating movements of ma- rine animals. Limnol Oceanogr 58: 697-714 Mackley EK, Phillips RA, Silk JRD, Wakefield ED, Afanasyev V, Furness RW (2011) At- sea activity patterns of breeding and non- breeding white-chinned petrels Procellaria aequinoctialis from South Georgia. Mar Biol 158: 429-438 Maechler M, Rousseeuw P, Struyf A, Hubert M, Hornik K (2017) Cluster: Cluster Analy- sis Basics and Extensions. R package ver- sion 2.0.6 Navarro J, Votier SC, Aguzzi J, Chiesa JJ, Forero MG, Phillips RA (2013) Ecological Segregation in Space, Time and Trophic Niche of Sympatric Planktivorous Petrels, PLoS One. 8: e62897. doi: 10.1371/journal. pone.0062897. Norris DR, Marra PP (2007) Seasonal inter- actions, habitat quality, and population dynamics in migratory birds. Condor 109: 535-547 Olmos F (2002) Non-breeding seabirds in Bra- zil: a review of band recoveries. Ararajuba - Braz J Ornithol 10: 31-42 Olmos F, Bugoni L, Neves T, Peppes F (2006) Caracterização das aves oceânicas que inte- ragem com a pesca de espinhel no Brasil. In: Neves T, Bugoni L, Rossi-Wongtschowski CLB (eds.) Aves oceânicas e suas intera- ções com a pesca na Região Sudeste-Sul do Brasil. Instituto Oceanográfico - USP, São Paulo, p 37–67 Ojowski U, Editmann C, Furness R, Garthe S (2001) Diet and nest attendance of incu- bating and chick-rearing northern fulmars (Fulmarus glacialis) in Shetland. Mar Biol 139:1193-1200 Orgeira JL (2001) Nuevos registros del Petrel Atlántico (Pterodroma incerta) en el océa- no Atlántico Sur y Antártida. Ornitol Neo- trop 12: 165-171 Orgeira JL, Scioscia G, Torres MA, Dellabian- ca NA (2013) New at-sea records of pelagic seabirds in the South Atlantic Ocean and Antarctica. Polar Res 32: 18972 Passos C, Navarro J, Giudici A, González- Solís J (2010) Effects of extra mass on the pelagic behavior of a seabird. Auk 127: 100- 107 Perez MS, Daudt NW, Tavares M, Ott PH, Santos RA, Fontana CS (2019) Diet of the Atlantic Petrel Pterodroma incerta during the non-breeding season. Mar Ornithol 47: 43-47 Phillips RA, Xavier JC, Croxall JP (2003) Ef- fects of satellite transmitters on albatrosses and petrels. Auk. 120: 1082–1090 Phillips RA, Silk JRD, Croxall JP, Afanasyev V, Briggs DR (2004) Accuracy of geoloca- tion estimates for flying seabirds. Mar Ecol Prog Ser 266: 265-272 Phillips SJ, Anderson RP, Schapire RE (2006) Maximum entropy modelling of species geographic distributions. Ecol Modell 190: 231-259 Phillips RA, Bearhop S, Mcgill RAR, Dawson DA (2009) Stable isotopes reveal individual variation in migration strategies and habitat preferences in a suite of seabirds during the nonbreeding period. Oecol 160: 795-806 Phillips RA, Lewis S, González-Solís J, Daunt F (2017) Causes and consequences of indi- vidual variability and specialization in for- aging and migration strategies of seabirds. Mar Ecol Prog Ser 578: 117-150 Philpot SM, Lavers JL, Nugegoda D, Gilmour ME, Hutton I, Bond AL (2019) Trace ele- ment concentrations in feathers of seven petrels (Pterodroma spp.). Environ Sci Pol- lut Res 26: 9640-9648. Pinet P, Jaquemet S, Pinaud D, Weimerskirch H, Phillips RA, Le Corre M (2011a) Migra- tion, wintering distribution and habitat use of an endangered tropical seabird, Barau’s 117Spatial ecology of the Atlantic petrel petrel Pterodroma baraui. Mar Ecol Prog Ser 423: 291-302 Pinet P, Jaeger A, Cordier E, Potin G, Le Corre M (2011b) Celestial Moderation of Tropical Seabird Behavior. PLoS One 6: e27663. doi: 10.1371/journal.pone.0027663 Quillfeldt P, Masello JF, Navarro J, Phillips RA (2013) Year-round distribution sug- gests spatial segregation of two small petrel species in the South Atlantic. J Biogeogr 40:430-441 R Core Team (2017) R: A language and envi- ronment for statistical computing. Version 3.4.1. R Foundation for Statistical Comput- ing, Vienna, www.R-project.org R Core Team (2019) R: A language and envi- ronment for statistical computing. Version 3.6.0. R Foundation for Statistical Comput- ing, Vienna, www.R-project.org Ramírez I, Paiva VH, Menezes D, Silva I, Phil- lips RA, Ramos JA, Garthe S (2013) Year- round distribution and habitat preferences of the Bugio petrel. Mar Ecol Prog Ser 476: 269-284 Ramos R, González-Solís J, Ruiz X (2009) Linking isotopic and migratory patterns in a pelagic seabird. Oecol 160: 97-105 Ramos R, Sanz V, Militão T, Bried J, Neves VC, Biscoito M, Phillips RA, Zino F, González- Solís J (2015) Leapfrog migration and habi- tat preferences of a small oceanic seabird, Bulwer’s petrel (Bulweria bulwerii). J. Bio- geogr 42: 1651-1664. Ramos R, Ramírez I, Paiva VH, Militão T, Bis- coito M, Menezes D, Phillips RA, Zino F, González-Solís J (2016) Global spatial ecol- ogy of three closely-related gadfly petrels. Sci Rep 6: 23447 Ramos R, Carlile N, Madeiros J, Ramírez I, Paiva VH, Dinis H, Zino F, Biscoito M, Leal GR, Bugoni L, Jodice PGR, Ryan PG, González-Solís J (2017) It is the time for oceanic seabirds: Tracking year-round dis- tribution of gadfly petrels across the Atlan- tic Ocean. Divers Distrib 23: 794-805 Ramos R, Llabrés V, Monclús L, López-Béjar M, González-Solís J (2018) Costs of breed- ing are rapidly buffered and do not affect migratory behavior in a long-lived bird spe- cies. Ecology 99: 2010-2024 Rayner MJ, Hauber M, Clout M, Seldon D, Van Dijken S, Bury S, Phillips R (2008) Forag- ing ecology of the Cook’s petrel Pterodro- ma cookii during the austral breeding sea- son: A comparison of its two populations. Mar Ecol Prog Ser 370: 271-284 Rayner MJ, Taylor GA, Gummer HD, Phil- lips RA, Sagar PM, Shaffer SA, Thompson DR (2012) The breeding cycle, year-round distribution and activity patterns of the en- dangered Chatham Petrel (Pterodroma ax- illaris). Emu 112: 107-116 Rexer-Huber K, Parker GC, Ryan PG, Cuthbert RJ (2014) Burrow occupancy and popula- tion size in the Atlantic Petrel Pterodroma incerta: a comparison of methods. Mar Or- nithol 42: 137-141 Richardson ME (1984) Aspects of the orni- thology of the Tristan da Cunha group and Gough Island 1972-1974. Cormorant 12: 123-201 Ridoux V (1994) The diets and dietary segrega- tion of seabirds at the subantarctic Crozet Islands. Mar Ornithol 22: 1-192 Rodríguez A, Holmes ND, Ryan PG, Wilson K, Faulquier L, Murillo Y, Raine AF, Penni- man, JF, Neves V, Rodríguez B, Negro JJ, Chiaradia A, Dann P, Anderson T, Metzger B, Shirai M, Deppe L, Wheeler J, Hodum P, Gouveia C, Carmo V, Carreira GP, Del- gado-Alburqueque L, Guerra-Correa C, Couzi F, Travers M, Le Corre M (2017), Seabird mortality induced by land-based artificial lights. Conserv Biol 31: 986-1001 Rodríguez A, Arcos JM, Bretagnolle V, Dias MP, Holmes ND, Louzao M, Provencher J, Raine AF, Ramírez F, Rodríguez B, Ron- coni RA, Taylor RS, Bonnaud E, Borrelle SB, Cortés V, Descamps S, Friesen VL, Genovart M, Hedd A, Hodum P, Humphries GRW, Le Corre M, Lebarbenchon C, Mar- tin R, Melvin EF, Montevecchi WA, Pinet P, Pollet IL, Ramos R, Russell JC, Ryan PG, Sanz-Aguilar A, Spatz DR, Travers M, Votier SC, Wanless RM, Woehler E, Chi- aradia A (2019) Future Directions in Con- 118 CHAPTER 3 : servation Research on Petrels and Shear- waters. Front Mar Sci 6: 94 doi: 10.3389/ fmars.2019.00094 Rousseeuw PJ (1987) Silhouettes: A graphical aid to the interpretation and validation of cluster analysis. J Comput Appl Math 20: 53-65 Simons R. (2017) ERDDAP, Monterey, CA: NOAA/NMFS/SWFSC/ERD. https://coast- watch.pfeg.noaa.gov/erddap/index.html (accessed 15 Sep 2017). Uhart MM, Gallo L, Quintana F (2018) Review of diseases (pathogen isolation, direct re- covery and antibodies) in albatrosses and large petrels worldwide. Bird Conserv Int 28: 169-196 UNESCO (2019) World Heritage List. Gough and Inaccessible Islands. https://whc.unes- co.org/en/list/740 (accessed 28 Jul 2019). Votier SC, Fayet AL, Bearhop S, Bodey TW, Clark BL, Grecian J, Guilford T, Hamer KC, Jeglinski JWE, Morgan G, Wakefield E, Patrick SC (2017) Effects of age and re- productive status on individual foraging site fidelity in a long-lived marine predator. Proc R Soc B 284: 20171068. doi:10.1098/ rspb.2017.1068 Wanless RM, Angel A, Cuthbert RJ, Hilton GM, Ryan PG (2007) Can predation by in- vasive mice drive seabird extinctions? Biol Lett 3: 214-244 Wanless RM, Ratcliffe N, Angel A, Bowie BC, Cita K, Hilton GM, Kritzinger P, Ryan PG, Slabber M (2012) Predation of Atlantic Pe- trel chicks by house mice on Gough Island. Anim Conserv 15: 472-479 Warham J (1996) The Behaviour, Population Biology and Physiology of the Petrels. Aca- demic Press, London Wood SN, Augustin NH (2002) GAMs with integrated model selection using penalized regression splines and applications to en- vironmental modelling. Ecol Modell 157: 157-177 Yamamoto T, Takahashi A, Yoda K, Katsumata N, Watanabe S, Sato K, Trathan PN (2008) The lunar cycle affects at-sea behaviour in a pelagic seabird, the streaked shearwater, Calonectris leucomelas, Anim Behav 76: 1647-1652 119Spatial ecology of the Atlantic petrel Figure S1: Changes in longitude (unfiltered locations) of Atlantic petrels throughout the year. Each line represents a different year-round trip, from a breeding episode to the next one (i.e. from July to August next year). Two phenological groups appeared to exist in the data, which led us to carry out a hierarchical clustering to statistically validate the existence of different groups (see Methods for more details). Colours correspond to two clusters obtained after applying the hierarchical clustering (successful breeders in blue; failed breeders in brown; see discussion). Grey shaded regions represent the equinoxes ± 20 days Supplementary material 60°W 40°W 20°W 0° 20°E Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Lo ng itu de Breeding success S F 120 CHAPTER 3 : Figure S2: Silhouette analysis of hierarchical clustering performed on seven phenological variables. Hi- erarchical clustering analysis and phenological variables are detailed in Fig. 1. The mean silhouette width of successful breeders (Group 1, 10 birds, in blue) is 0.76. Failed breeders (Group 2, 8 birds, in light brown) have a mean silhouette width of 0.38. Mean silhouette width is 0.59 (dashed grey line). 3 individuals with low (< 0.25) silhouette widths in the failed cluster (marked with star) may indicate potentially incorrect classification. However, the individual phenology detailed in Fig. S3 shows that although these individuals performed later post-breeding migrations than most other failed breeders (and this could cause low (< 0.25) silhouette widths), their chick-rearing period was far too short for raising a chick to fledging. Each bar corresponds to a year-round trip (n = 18) identified by track ID (birdID_year) and ordered by silhouette width within each group 121Supplementary material Figure S3: Individual phenologies of adult Atlantic petrels. Each bar corresponds to a year-round trip and shows phenological dates of: incubation (yellow), chick-rearing (green), post-breeding migration (red), non- breeding (blue), pre-breeding migration (red), pre-breeding (time from arrival at breeding grounds to pre-laying exodus; light purple) and pre-laying exodus (dark purple). Individual tracks (birdID_year, on the left) are in the same order as in the results of hierarchical clustering in Fig 1A. Stars indicate those individuals with low (< 0.25) silhouette widths. Dashed horizontal line separates birds classified as successful breeders (above) and those classified as failed breeders (below) 122 CHAPTER 3 : Figure S4: Important environmental variables for Atlantic petrel distribution during each phenological phase. Resulting from habitat modelling developed through MaxEnt (by percent contribution; see Table 2 for details), the environmental variables are ordered by importance and contribution to Atlantic petrel distribution (more important variables on left): gradient of seafloor depth (BATG, %), surface chlorophyll a concentration (CHLA, mg m-3, as a proxy of biological production), distance to the colony (DCOL, km) and sea surface tem- perature (SST, ºC). 50 % kernel UD (areas of active use) of each phenological phase are represented over their important environmental variables (A-J). Triangle represents the breeding colony at Gough Island 123Supplementary material Figure S5: Response curves of most important environmental variables for Atlantic petrel distribution at each phenological phase. Probability of occurrence of Atlantic petrel (y-axis) in response to the environmental variables (x-axis), resulting from habitat modelling developed through MaxEnt. The environmental variables are: gradient of seafloor depth (BATG, %), surface chlorophyll a concentration (CHLA, mg m-3, as a proxy of biological production), distance to the colony (DCOL, km) and sea surface temperature (SST, ºC) 124 CHAPTER 3 : A Pre-laying exodus BAT BATG CHLA DCOL SAL SST WIND BAT 1.000 0.813 0.404 0.121 -0.184 -0.270 -0.220 BATG 1.000 0.361 0.119 -0.190 -0.151 -0.239 CHLA 1.000 0.098 -0.120 -0.156 -0.075 DCOL 1.000 -0.207 -0.026 -0.509 SAL 1.000 0.417 -0.147 SST 1.000 -0.487 WIND 1.000 B Incubation BAT BATG CHLA DCOL SAL SST WIND BAT 1.000 0.813 0.396 0.121 -0.322 -0.256 -0.290 BATG 1.000 0.356 0.119 -0.303 -0.114 -0.306 CHLA 1.000 0.104 -0.260 -0.104 -0.148 DCOL 1.000 -0.321 0.053 -0.591 SAL 1.000 0.437 0.200 SST 1.000 -0.359 WIND 1.000 C Chick-rearing BAT BATG CHLA DCOL SAL SST WIND BAT 1.000 0.813 0.415 0.121 -0.350 -0.281 -0.216 BATG 1.000 0.380 0.119 -0.333 -0.176 -0.207 CHLA 1.000 0.110 -0.262 -0.152 -0.046 DCOL 1.000 -0.364 0.162 -0.257 SAL 1.000 0.389 0.094 SST 1.000 -0.456 WIND 1.000 D Non-breeding BAT BATG CHLA SAL SST WIND BAT 1.000 0.813 0.380 -0.216 -0.254 -0.087 BATG 1.000 0.368 -0.206 -0.213 -0.070 CHLA 1.000 -0.152 -0.170 0.018 SAL 1.000 0.473 -0.235 SST 1.000 -0.673 WIND 1.000 Table S1: Pearson correlations for the environmental variables at each phenological phase. The seven en- vironmental variables were selected at the beginning of the modelling. Note that distance to the colony (DCOL) was not selected as important during the non-breeding period. Values in bold indicate significant correlations 125Supplementary material Table S2:Year-round at-sea activity of adult Atlantic petrels. Time spent on water (mean ± SD) during the day and night, for each phenological period, and for successful and failed breeders. Phenological phase n (days) Time spent on water (% of time) Day Night Succ. Fail. Successful Failed Successful Failed Pre-laying exodus 528 285 19.7 ± 17.0 16.4 ± 14.5 30.0 ± 17.8 23.7 ± 16.9 Incubation 62 117 13.9 ± 14.6 10.2 ± 7.9 19.6 ± 12.0 12.5 ± 9.3 Chick-rearing 536 45 11.5 ± 11.5 8.9 ± 7.5 16.7 ± 13.8 17.7 ± 14.9 Non-breeding 385 739 72.7 ± 15.9 62.6 ± 23.6 60.4 ± 22.5 53.4 ± 29.2 126 CHAPTER 3 : ID_Sample Feather δ15N (‰) δ13C (‰) 746 P1 15.41 -17.46 746 P3 15.16 -18.00 746 P5 14.96 -18.12 746 P7 15.40 -17.85 746 P10 14.21 -17.08 746 S13 14.17 -17.52 746 R6 14.03 -16.94 747 P1 13.63 -17.62 747 P3 13.72 -17.44 747 P5 14.15 -17.51 747 P7 14.52 -17.18 747 P10 13.92 -16.76 747 S13 14.05 -16.66 748 P1 13.62 -19.32 748 P3 13.62 -17.22 748 P5 13.10 -18.21 748 P7 13.40 -18.49 748 P10 14.37 -17.02 748 S13 14.21 -16.85 748 R6 14.06 -16.85 749 P1 15.46 -19.23 749 P3 15.54 -18.89 749 P5 15.35 -18.85 749 P7 14.99 -17.93 749 P10 14.64 -17.80 749 S13 14.88 -17.25 749 R6 14.50 -17.86 Table S3: Stable isotope signatures (δ15N and δ13C) of sampled feathers for each Atlantic petrel. When pos- sible, same feathers were sampled from 8 dead Atlantic petrel individuals found at Gough Island in September 2009. These feathers were 1st, 3rd, 5th, 7th and 10th primary feathers (P1, P3, P5, P7 and P10), 13th secondary (S13) and 6th rectrix (R6). a Outlier not included in stable isotope mean values 127Supplementary material ID_Sample Feather δ15N (‰) δ13C (‰) 750 P1 14.76 -17.98 750 P3 14.88 -17.69 750 P5 14.86 -17.64 750 P7 14.51 -17.88 750 P10 14.50 -16.84 750 S13 14.59 -16.74 750 R6 14.48 -16.70 751 P1 14.40 -17.56 751 P3 14.27 -17.70 751 P5 14.07 -17.33 751 P7 14.30 -17.69 751 P10 14.27 -17.00 751 S13a 14.05 -19.47 752 P1 13.82 -16.94 752 P3 13.73 -16.20 752 P5 14.26 -16.59 752 P7 13.97 -16.61 752 P10 14.08 -16.54 752 S13 14.10 -16.06 752 R6 13.99 -16.84 753 P1 14.02 -16.51 753 P3 13.97 -18.35 753 P5 14.19 -16.73 753 P10 14.74 -17.25 753 S13 14.31 -16.64 753 R6 14.53 -17.07 128 CHAPTER 3 : Po st - br ee di ng m ig ra tio n P re - br ee di ng In cu ba tio n C hi ck -r ea rin g St ar t d at e St ar t d at e St ar t d at e D ur at io n St ar t d at e D ur at io n S ou rc e St ud y si te M id D ec em be r M id - M ar ch Fi rs t w ee k of Ju ly 56 B eg in ni ng o f S ep te m be r 10 6 (R ic ha rd so n 19 84 ) Tr is ta n da C un ha - - 29 /0 6 (1 5/ 06 - 2 1/ 07 ) 55 .5 ± 4 .0 20 /0 8 (1 4/ 08 - 1 0/ 09 ) 13 8 (C ut hb er t 20 04 ) G ou gh - - - - 18 /0 8 (0 6/ 08 - 0 1/ 09 ) 15 6 (W an le ss e t a l. 20 12 ) G ou gh 25 /1 2 ± 8. 1 17 /0 4 ± 5. 6 13 /0 7 ± 5. 3 58 .0 ± 7 .5 a 25 /0 8 ± 8. 1 12 2. 1 ± 10 .1 C ur re nt st ud y G ou gh Ta bl e S4 : C om pa ri so n of p he no lo gy o f A tla nt ic p et re ls o bt ai ne d fr om b ib lio gr ap hy a nd c ur re nt st ud y P re vi ou s st ud ie s ar e ba se d on d ir ec t ev id en ce s of t he i nc ub at io n an d ch ic k- re ar in g st ar ti ng d at e (l ay in g an d ha tc hi ng , re sp ec ti ve ly ), w he re as i n th e cu rr en t s tu dy w e es ti m at ed p he no lo gy b as ed o n ge ol oc at io n an d ac ti vi ty d at a; d at a in cl ud ed in th e co m pa ri so n re fe r to s uc ce ss fu l br ee de rs . M ea n ± S D o r r an ge v al ue s ex ce pt in di ca te d. a S in ce th e se x of b ir ds w as u nk no w n an d on ly o ne o f t he p ar en ts w as tr ac ke d, th e en tir e in cu ba tio n sh ar ed b y bo th p ar en ts m ig ht b e lo ng er 129Supplementary material LITERATURE CITED: Cuthbert RJ (2004) Breeding biology of the Atlantic Petrel, Pterodroma incerta, and a population estimate of this and other burrowing petrels on Gough Island, South Atlantic Ocean. Emu 104: 221-228 Richardson ME (1984) Aspects of the ornithology of the Tristan da Cunha group and Gough Island 1972-1974. Cormorant 12: 123-201 Wanless RM, Ratcliffe N, Angel A, Bowie BC, Cita K, Hilton GM, Kritzinger P, Ryan PG, Slabber M (2012) Predation of Atlantic Petrel chicks by house mice on Gough Island. Anim Conserv 15: 472-479 Chapter 4: Air-water behavioural dynamics reveal complex at-sea ecology in global migratory seabirds Zuzana Zajková1,2, José Manuel Reyes-González 1, Teresa Militão1, Jacob González-Solís1, Frederic Bartumeus2,3,4 1 Institut de Recerca de la Biodiversitat (IRBio) and Departament de Biologia Evolutiva, Ecologia i Ciències Ambientals (BEECA), Universitat de Barcelona, Barcelona, Spain 2 Centre d’Estudis Avançats de Blanes (CEAB-CSIC), Blanes, Spain 3 Centre de Recerca Ecològica i Aplicacions Forestals (CREAF), Cerdanyola del Vallès, Spain 4 Institució Català de Recerca i Estudis Avançats (ICREA), Barcelona, Spain ABSTRACT Characterizing the main determinants and variability of animal behaviour in the wild is a daunting task. However, to improve our efforts in the management and conservation of endangered popula- tions it is crucial to assess the underlying sources of behavioural variability, and to identify key behavioural shifts governed by major life-history events. The recent technological revolution allows us to collect behavioural information at an unprecedented level of detail, but novel methodologi- cal protocols are required to bring biologging sensory data to amenable behavioural analyses and descriptions. Our study model, a highly-mobile migratory seabird (the Cory’s shearwater) presents a complex annual cycle that involves central place foraging, ocean-basin long migratory movement and wandering in wintering areas. Our quantitative analysis, based on wet-dry data, reveals the hi- erarchical and modular nature of seabirds’ air-water behavioural interactions, at an unprecedented level of detail. The existence of radically different behavioural contexts linked to phenology, and the need to exploit different marine environments over the year, results in different behavioural preva- lences and transitions both in time and space. We uncover both flexible and structural components of the behavioural organization of Cory’s shearwaters across the annual life cycle. Cory’s shear- waters show complex behavioural sequences and organization during the breeding and wintering stage. On the contrary, migration restricts individual natural variability to a few dominant behav- ioural modes and more predictable behavioural transitions. We also observed a spatial correlation between behavioural diversity and resource hotspots (e.g. upwelling areas). Our framework paves the way for extending behavioural annotation to year-round movements of wildlife, opening new avenues to understand behavioural patterns and the seasonal timing of life-history events of animals spending most of their life out of the human’s sight. 132 CHAPTER 4 : INTRODUCTION Identifying the sources and characteristic scales of behavioural variability is an intricate task, yet it is crucial to address many funda- mental questions in wildlife ecology and con- servation (Krebs & Davies 2009, Berger-Tal et al. 2015). It is well known that gender, age or social status are main drivers of behavioural variability, having a strong impact on popula- tion dynamics (Nilsson et al. 2014, Lecomte et al. 2010). At the same time, major life-history events (e.g. breeding, migration, wintering) can constrain individual-level behavioural repertoire and decision-making. Identifying key behavioural shifts, and understanding the main determinants of a species behavioural or- ganization requires covering the wide range of scales and natural conditions at which behav- iour unfolds. In this context, recent advances in remote tracking and analytical tools has allowed to address fundamental behavioural questions at an unprecedented detail (Block et al. 2011, McIntyre 2014, Kays et al. 20015, Ron- con et al. 2018, Harcourt et al. 2019), fostering our understanding of the adaptive potential of wildlife species to changing environments (Sih et al. 2010, Wong & Candolin 2015). Seabirds are an ideal model to study the influ- ence of different drivers in animal behaviour. These long-lived and highly vagile marine top- predators, nest in-land but fully rely on dynam- ic marine environments to accomplish most biological requirements (Gaston 2004). Indi- vidual movements and behaviour are variable but severely constrained by the seasonal timing of life-history events (Phillips et al. 2017). The changing degree of energetic demands, breed- ing duties and central-place foraging coupled with changes in food availability limit foraging behaviour in time and space to different extent throughout the seasons (Schreiber & Burger, 2001). Therefore, in order to maximize forag- ing success and fitness, individuals must adapt their behaviour over the annual cycle to face the different constraints related to intrinsic (age, sex, breeding status, breeding duties, breed- ing success, migration strategies, moulting strategies, etc.) and extrinsic factors (e.g. food availability, patchy resources, marine habitat, environmental stochasticity, etc.) (Weimer- skirch 2007, Phillips et al. 2017). Seabirds are exposed to a changing number of threads over their annual cycle on land and at sea, making them one of the world’s most rapidly declining vertebrate groups (Croxall et al. 2012). Hence, seabirds represent a particular case of interest where novel technology can revolutionize our understanding of basic behavioural knowledge, urgently needed to improve current conserva- tion and management efforts (Lascelles et al. 2016, Dias et al. 2019). At-sea movement of pelagic seabirds have been addressed over the last 20 years with the wide deployment of light-level geolocation log- gers (global location sensor, GLS) (Burger & Shaffer 2008, Wilson & Vandenabeele 2012). GLS currently remain as the most cost-effec- tive balanced tracking devices to get insights into the movements of pelagic seabird species over the entire annual cycle while ensuring the welfare of tagged individuals (Igual et al. 2010, Vandenabeele et al. 2011, Vandenabeele et al. 2012). Apart from the low-resolution positional data (2 positions per day), some models of GLS, usually referred as geolocation-immersion log- gers, also provide high-resolution wet-dry con- ductivity data, which has been used as a proxy to broadly describe at-sea behaviour of seabirds (e.g. Mackley et al. 2011, Rayner et al. 2012, Gutowsky et al. 2014, Clay et al. 2017). Despite many researchers would acknowledge that the temporal sequence of wet-dry alternating states contains relevant behavioural information (i.e. landings, take-offs, sustained flight, sitting on water) (Weimerskirch et al. 1997, Lecomte et al. 2010, Shaffer 2001), wet-dry dynamics have been rarely used alone to make behav- ioural inferences (Guilford et al. 2009). Yet this information has been used most times as a complement in behavioural characterization approaches relying on speed and turns inferred from movement positional data (Dean et al. 2012, Freeman et al. 2013). In this work, we aim at revealing the drivers and the behavioural complexity of a highly vag- 133Air-water behavioural dynamics reveal complex at-sea ecology in global migratory seabirds ile pelagic seabird species, the Cory’s shear- water (Calonectris borealis). We developed a novel unsupervised protocol to extract the most out of the behavioural information contained in high-resolution wet-dry geolocation-immer- sion data. We showed that one can make key behavioural inferences from air-water seabird interaction patterns. Based on this type of data, the behaviour of Cory’s shearwaters appears to be modular and hierarchically organized, and behavioural modes differing in prevalence and transition probabilities across individuals and phenological stages (namely breeding, migra- tion, and wintering). We analysed carefully the relationship and importance of behaviours across different scales, from individuals to population, from daily to seasonal scales, and revealed the complex and hierarchical nature of seabird behaviour over the entire annual cycle. MATERIAL AND METHODS Model species and fieldwork The Cory’s shearwater Calonectris borealis (Sangster et al. 2012) is a medium-sized free- range pelagic seabird belonging to Order Pro- cellariiformes, which includes albatrosses and petrels. Cory’s shearwaters breed in burrows on islands and islets in the North-East Atlan- tic Ocean. The species perform a complex migratory cycle over the year, as birds take advantage of prevailing winds in their migra- tory flyways, drawing an 8-shape loop over the Atlantic Ocean to move between breeding colonies and wintering areas in the South At- lantic (González-Solís et al. 2007). We stud- ied adult breeders from the colony located at Veneguera (27° 50’ 29” N, 15° 47’ 29” W, Gran Canaria, Canary Islands). Birds from this col- ony are known to commute to the near North- west African shelf during the breeding stage to forage along the enriched cool waters of the Canary Current upwelling system (Navarro & González-Solís, 2009, Reyes-González et al. 2017). Device deployment and data collection We tagged 19 individuals during the breeding stage (June - July 2011) with light-level geo- location-immersion loggers (GLS hereafter). GLS model Mk19 measures light every 60 sec- onds and stores the information in 5-minutes blocks, allowing finally to estimate 2 low-res- olution positions per day (at local midday and midnight) (Phillips et al. 2004, Fox, 2010). GLS additionally record the amount of time a tagged animal is in contact with salt-water recorded ev- ery 3 seconds (see below for more details). We also equipped birds with GPS loggers during the incubation period, deploying the devices before bird departure to a foraging trip and re- covering them after bird arrival to land. Using concurrent GPS tracking we expected to have a proxy of the “ground truth” about displace- ments of birds during the short-term foraging trips. We programmed GPS devices to record a location every 5 minutes to ensure battery life to record complete foraging trips, which usu- ally last several days. We mounted GLS (~ 2.4 g, Mk19 model, Biotrack Ltd ©) on a plastic ring on the leg of each bird. In the case of GPS, we attached the loggers (~ 24 g, 750 mAh bat- tery, Perthold Engineering, Germany) to the back feathers with water-resistant TESA© tape. Body mass of tagged birds ranged from 600 to 900 g so both devices amounted to 2.5 - 3.7 % of the birds’ weight, below the detrimental rec- ommended threshold of 3-5 % (Phillips et al. 2003, Igual et al. 2005, Passos et al. 2010). All fieldwork was conducted under the license ap- proved by the regional committee for scientific capture (Ref.Expt. 2011/0795, Consejería de Medio Ambiente del Cabildo de Gran Canaria; Oficina de Especies Migratorias - Ministry for Ecological Transition, Spain). The final dataset involved 23 complete forag- ing trips of 19 individuals tracked concurrently with GLS and GPS loggers during the incuba- tion period. Average duration of these foraging trips was 12.6 days (range: 6.9 – 16.9 days; Ta- ble S1). At the end of the incubation period, we maintained the GLS on 8 birds to record wet- dry information over the whole annual cycle. 134 CHAPTER 4 : In the following spring/summer (2012), we re- covered GLS. The mean duration of year-round individual tracks recorded was 248 days (range 217 – 270 days; Table S1). For all tracks re- corded (i.e. year-round and short-term foraging trips), we assigned the start and the end of the track as the moment when the animal was in contact with salt-water for the first and the last time after the logger deployment and before the recovery, respectively. Note that therefore nei- ther GPS nor GLS tracks contained incubation stints of birds. Year-round movements and phenology We estimated positional data from twilight events from GLS using the probabilistic al- gorithm implemented in “probGLS” R pack- age (Merkel et al. 2016). We used the function twilight_error_estimation to estimate twi- light events from calibration of light data. The “probGLS” algorithm is based on an iterative forward step selection so for each year-round individual track a median geographic track is calculated from a cloud of weighted possible locations (at each step generated 1,000 par- ticles). We ran 100 iterations for each of the 8 year-round individual tracks, and the particles were weighted by speed in dry (mean ± SD 50 ± 30 m.s-1, max 95 m.s-1) and wet state (0.5 ± 0.25 m.s-1, max 1.7 m.s-1) (see Orben et al. 2018 for details of this methodology). We restricted the selection of estimated locations to sea by applying a land mask. We finally obtained two locations per day, with an overall median er- ror of 185 km during the solstice and 145 km during the equinox periods (Merkel et al. 2016). Next, we applied the ST-DBSCAN algorithm on positional data derived from probGLS to ob- jectively identify statistically coherent spatio- temporal clusters corresponding to different stages of the annual cycle along each year- round individual track. ST-DBSCAN allows clustering spatio-temporal data with arbitrary shape and does not require the predetermina- tion of the number of clusters (see Birant & Kut 2007 for further details). We set the distance parameter to 600 km, the time window to 120 hours and a minimum number of 10 locations to consider our phenological partition, as our exploratory analysis showed this choice of pa- rameters to conform the most meaningful fig- ure in biological terms under Cory’s shearwater expert criteria. Later, for each individual track we visually checked the assignment of each location to the different clusters. As we were interested in 3 main stages (namely breeding, migration and wintering), we identified and maintained only changes indicating the onset/ end of migration and the onset/end of staging in breeding and wintering areas. In the case where stopovers were identified (i.e. a spatio- temporal cluster where an individual spent at least 5 days for refueling along the migratory path, Dias et al. 2010), we included them in the migration phase. ST-DBSCAN failed to iden- tify a reliable spatio-temporal cluster for the breeding area in one individual that did not mi- grate and stayed within the Canary Current all year round. In that case we assigned the end of breeding and onset of wintering as the period from the last nocturnal visit of the nest (iden- tified as prolonged time in dry state over the night) until the start of next breeding stage as identified by ST-DBSCAN. Wet-dry data segmentation The GLS model used in this work checks for the wet/dry state every 3 s, indicating whether the tagged bird was in contact with salt-water. A minimum duration of 6 s within a state is required to record a change between states. To account for natural changes in behaviour, we based our analysis on variable-time segments of the wet-dry data. Therefore, similarly to Meyer et al. (2015), we first coded the wet-dry time series data of each single track (i.e. both short-term foraging trips and year-round trips) as a binary time series, with the wet state as +1 and the dry state as -1, interpolated at 1 s time intervals (Fig. 1A). We then calculated the in- tegrated cumulative sum of wet-dry sequences and later applied a breakpoint algorithm (Knell & Codling 2011). Doing so, we split the vec- tor into homogeneous segments of different 135Air-water behavioural dynamics reveal complex at-sea ecology in global migratory seabirds lengths with positive (prevalence of wet states) and negative (prevalence of dry states) slopes. Abrupt changes in slopes represent strong changes in wet-dry cumulative sums. The breakpoints of the wet-dry times series were obtained based on a local estimation of the slope change under an optimal running win- dow. As smaller this running window, the more sensitive is the breakpoint algorithm in iden- tifying changes in wet or dry states’ temporal correlations, in a similar way to the “tolerance” parameter in the line-simplification algorithm (Douglas-Peucker algorithm, Thiebout & Trembley 2013). Moreover, the larger the run- ning window size, the less jagged (smoother) Figure 1: Segmentation of wet-dry data from geolocation-immersion loggers. Example of one individual short- term foraging trip (duration of 16 days) of Cory’s shearwater tracked simultaneously with geolocation-immer- sion logger and GPS device. A) Original wet-dry data as two states: wet (blue) and dry (yellow). Black line represents the integrated time-series based on cumulative sum of wet (+1) and dry (-1) at 1 second resolution. This vector was used as an input to the breakpoint algorithm. B) Selecting a 60-minutes window for segmenta- tion splits the track into 115 segments of variable lengths. C) Resulting segmentation represented by dark grey vertical lines. Dotted vertical lines represent midnights. D) Histogram of durations of segments (N=115) from the foraging trip, median duration of 88 minutes is marked as vertical dashed line. is the cumulative wet/dry time series, so only large-scale consistent breakpoints are obtained (Fig. S1). Since we had no previous knowledge about the optimal time window size (TWS), we performed a coarse-graining analysis by run- ning the algorithm with varying TWS ranging from 1 – 1440 minutes (24 h) (Fig. S1). We se- lected a TWS of 60 minutes (Fig. 1B and Fig. S2) because it showed the best compromise between the TWS and the number of result- ing breakpoints, that is, the best compromise to obtain fine-scale segments and coarse-scale (i.e. consistent) breakpoints (i.e. elbow point in Fig. S2). Despite the optimal TWS sets an over- all scale of analysis, the breakpoint analysis 136 CHAPTER 4 : procedure accounts much better for the multi (broad)-scale nature of behavioural dynamics reflected in a wide variation of segment dura- tions (Fig. 1D). The fundamental assumption of this analysis is that the nature of air-water behavioural dynamics is intrinsically com- plex and lacks a clear characteristic scale. In other words, wet-dry dynamics, and particu- larly their temporal correlations, show multiple scales (from minutes to hours) which contain crucial behavioural information that is com- monly missed in other approaches. Wet-dry activity metrics For each wet-dry segment identified previous- ly, we calculated various descriptive activity metrics (Table 1). We generated a data matrix entailing 12 856 segments and each character- ized by 11 activity metrics summarizing wet- dry patterns. We did not include the length (to- tal duration) of the segment as input feature in our behavioural mapping and annotation proto- col (see next section) to avoid multicollinearity with some other activity metrics, but we used it later to interpret our behavioural clustering output. Inferring and building up the behavioural space Based on the above multivariate characteriza- tion of trajectory segments (i.e. 11 variables) we built up a behavioural space. We used the unsu- pervised clustering protocol from the “bigMap” R package to map a two-dimensional behav- ioural space (Garriga & Bartumeus 2018). Data matrix refereed in the previous section (size of 12 856 x 11) was used as input in the protocol. We refer to the “bigMap” R package documen- tation and Garriga & Bartumeus (2018) for a detailed explanation of the protocol, therefore Table 1: Metrics calculated from wet-dry activity data at the segment level, which were used as input features to t-SNE algorithm for behavioural annotation. Activity metric Abbreviation Definition Proportion wet Prop.W Duration of time on water divided by total duration of segment (range 0 - 1) Duration wet Dur.W Total time in wet state (s) within segment Duration dry Dur.D Total time in dry state (s) within segment Number of changes Nchanges Total number of changes (transitions) between states (indifferently from wet to dry or dry to wet) Rate of changes Rchanges N of changes (transitions) divided by totalduration of segment (s) Median wet duration Median.W Median duration (s) of wet states withinsegment Median dry duration Median.D Median duration (s) of dry states within segment Standard deviation wet durations SD.W SD of durations (s) of wet states within segment Standard deviation dry durations SD.D SD of durations (s) of dry states withinsegment Maximum wet duration Max.W Maximum duration of wet states (s) withinsegment Maximum dry duration Max.D Maximum duration of dry states (s) within segment 137Air-water behavioural dynamics reveal complex at-sea ecology in global migratory seabirds here we just summarize the steps we followed. We pre-processed the data matrix by comput- ing a principal component analysis (PCA) and a whitening of the rotated data as a standard pro- cedure to homogenize the ranges and weights of the input features in the analysis. We then applied a parallelized and big data adjusted version of the t-stochastic neighbourhood em- bedding algorithm (ptSNE algorithm, func- tion bdm.ptsne at “bigMap”). The t-stochastic neighbourhood embedding algorithm (t-SNE) uses an information-theoretic approach to re- duce the dimensions and to embed multidimen- sional datasets into a 2-dimensional embedding space that represents data point similarities ac- cording to their respective features’ values (see van der Maaten & Hinton 2008 for details). A key parameter of this embedding algorithm is the so-called perplexity (ppx), which defines the neighbouring scale to measure pairwise similarity. It also sets an equilibrium between forcing a highly local analysis of similarity or a much coarse or global view of similarity in- volving all the points in the space. We ran the ptSNE algorithm with a broad range of per- plexities and chose ppx=500 as a compromise value, since it was robust enough to maintain the same embedding space and at the same time discerned both relevant local and global features of the space (ppxs from 250 to 500 did generate statistically similar behavioural spac- es). In such a 2-dimensional embedding space each data point represents a well characterized behavioral segment in terms of wet-dry activ- ity. The segments showing similar character- istics are close together in the space, whereas faraway data points represent non-similar or clearly dissimilar segments (Fig. 2A). To better analyze the overall structure of data point dis- tribution we applied a tailored kernel density (function bdm.pakde at “bigMap”; ppx=250; grid of 200 x 200 cells; Garriga & Bartumeus 2018). The kernel density clearly showed both the largest spatial concentrations of data points and the point-diluted areas in the embedding space. Over the kernel density, we applied a segmentation algorithm (function bdm.wtt at “bigMap”; Garriga & Bartumeus 2018; Fig. 2B) to discretize the embedding space into clus- ters that represent similar behavioral features across segments. Finally, we post-processed the watershed clustering output by merging the initial clusters following a signal-to-noise ratio heuristic that is applied recursively and hi- erarchically (function bdm.s2nr at “bigMap”). This coarse-graining procedure reduces the space complexity and facilitates behavioural annotation by lowering the number of clusters (Fig. 2B-C). Each of the finally obtained clus- ters groups wet-dry activity segments by their similitudes and would correspond to different wet-dry activity-based behavioural modes, to which we will refer hereafter as behavioural clusters (BCs). Activity metric importance for identified BC We used a Random Forest algorithm (RF; see Biau & Scornet 2016 for more details) to rank both the overall importance of input features to assign segments to the different BCs, and the case-wise specific importance of each fea- ture for each different BC identified. We per- formed RF using the function randomForest from the “randomForest” R package (Liaw & Wiener 2002). We split the data set in training and testing (2/3 vs 1/3) and ensured a balanced sampling by stratified selection of equal num- ber of samples (400) at each run. We grew 4 000 trees and selected 6 predictors randomly in each tree. The overall variable importance was measured using the mean decrease in accuracy index (MDA), which reports the MDA over all cross-validated predictions of the model when a given predictor is permuted after the training process and before prediction (Biau & Scornet, 2016). Case-wise variable importance for each BC was calculated with “LocalImp” parameter. For visualization purposes, case-wise variable importance values were rescaled to range be- tween 0-1 within each BC. To measure accu- racy of prediction, we computed the confusion matrix between observed and predicted BC, using functions from “caret” R package (Kuhn 2018). 138 CHAPTER 4 : Sources of variability in behaviour We assessed whether observed behavioural budgets over the annual cycle were shaped by phenology or conversely, they depended more on inter-individual variability. For each BC, within-stages variability represents inter- individual variation, whereas between-stages variability represents phenological variation. If behavioural budgets were constrained by phenology, we would expect the stage means to spread out more than the inter-individual vari- ability within each stage. For each BC, we cal- culated the relative amount of time allocated by individual and stage, and determined the ratio of between-stages to within-stages variances using one-way repeated-measures ANOVA F- test, setting α to 0.05. We used repeated-mea- sures ANOVA to control pseudo-replication since same individuals were represented in the three stages. Behavioural prevalence and transitions across phenology To better understand changes in the behav- ioural space between the different stages of the annual cycle, we used the function bdm.dMap at “bigMap” (Garriga & Bartumeus 2018). This function allowed us to compute and visualize the distribution of behavioural segments from each stage over the 2-dimensional behavioural space. We calculated the probability of belong- ing to one of the three stages (breeding, migra- tion, wintering) for each cell of the behavioural raster or grid. The visual output of the function is composed of 4 plots where the first repre- sents the dominating stage-specific prevalence (breeding, migration, wintering) is shown. More specifically we normalized the probabili- ties for each stage over the behavioural space and calculated 5% contour density lines to de- pict the probability density for each stage. Fi- nally, we calculated and used standardized re- siduals from chi-square test of independence to evaluate the association between behavioural clusters and stage. In order to better understand behavioural strategies, we characterized the structural or- ganization of the observed BCs using network analysis. Network topology was represented by BCs as nodes and the relations between BCs as edges. We constructed an adjacency matrix for each stage, counting the frequency of transi- tions between current BC at time t to next BC at t+1. We converted these matrices into weighted directed networks, with BCs as nodes and tran- sitions as edges. Even though wet-dry activity patterns may be thought to represent a bipartite network structure composed of two indepen- dent sets always alternating each other (mostly wet and mostly dry segments, see Results sec- tion), we cannot treat them as bipartite because our behavioural units are the wet-dry activity segments, but our behavioural description has been statistically aggregated in the form of BCs (nodes of our network). Hence, once we use our protocol to annotate single wet-dry time-series data and later compute network edges, one may find that two consecutive segments could be- long to the same BC or to a BC from the same mostly wet or dry set (though this occurs barely 1.5 % from all transitions measured). To evalu- ate major changes in the structural organiza- tion of behavioural modes across the 3 stages of the annual cycle, we calculated various global (i.e. network level) and local (i.e. node level) quantitative metrics (Table 2). Except for the density metric, we did not consider the weights (accounting for the frequency of transitions be- tween BC). We used “igraph” R package for the analysis and visualization (Csardi & Nepusz 2006). In addition, we examined transitions between BCs at each stage. To account for different du- ration and therefore differences in frequencies of BCs, we normalized the adjacency matrix for each stage by calculating the probability of transition between two BCs (Ti, j ), conditioned by the probability of being in certain stage (S) and in certain BC (C): By this way we obtained for each stage a matrix 139Air-water behavioural dynamics reveal complex at-sea ecology in global migratory seabirds Table 2: Network metrics used in the study. Network metric Level Description Size Global Total number of nodes in the network Density (function edge_density) Global Ratio of the number of edges and the number of possible edges Diameter (function diameter) Global Longest path between two nodes Average path length (function mean_distance) Global The mean of the shortest distance between each pair of nodes in the network Degree centrality (function degree) Local Node’s in- and out-degree (being number of edges that lead into or out of node); indicates the connectivity of node Closeness centrality (function closeness, normalized value) Local Indicates how close is the node to other nodes of the net- work, calculated as the reciprocal of the average length of the shortest paths to/from all other nodes in the network Betweenness centrality (function betweenness) Local Refers to the number of shortest paths (geodesics) between two nodes that go through the node of interest of transition probabilities between BCs, where the sum of all probabilities from BCi equals 1. Later, we used these transition matrices to es- timate entropy rate, using “ccber” R package (Vegetabile et al. 2019). We simulated 10 000 transitions for each stage using function Simu- lateMarkovChain and used the function Cal- cEntropyRate to obtain entropy rate for each stage. Spatial representation of movement and be- haviour We carried out a spatially-explicit approach to visualize behavioural landscapes by iden- tifying the most important BCs exhibited by tracked birds over their entire annual distri- bution range. To do so, we first merged geo- graphical locations from GLS with wet-dry activity segments and linearly interpolated locations at the start of each segment. After that, we regularized the tracks to a location ev- ery 5 minutes, so each location had assigned the BC of the segment to which belonged. To map locations, we used R packages “sf” (Pebesma 2018), “dggridR” (Barnes 2018), and “ggplot2” (Wickham 2016). We used an Ico- sahedral Snyder Equal Area Projection with a cell size of approximately 70 000 km2 and centroids of adjacent cells distanced ~260 km (varying according to latitude). Next, to get a more statistically representative map at popula- tion level in areas intensively used by several individuals, we performed 1 000 iterations of a custom-built randomization procedure to se- lect samples, so at each run and for each grid cell, we chose randomly three-quarters of the locations and quantified the time invested in total and by BC. The final map showed for each grid cell the BC in which the most time was invested over the iterations. We also extended the concept of measuring diversity to evaluate behavioural variability in space, by creating a spatially-explicit behavioural diversity map. We calculated the Shannon diversity index (Krebs 1999) considering the number of seg- ments belonging to each BC within each grid cell. Using a similar bootstrap procedure as before, we built up a final map that shows the average behavioural diversity for each cell. We finally explored behaviour during the breeding period, zooming into the Canary Current and using uniquely wet-dry activity segments re- corded during short-term foraging trips during the incubation period, when animals were con- currently tracked with GPS loggers and thus 140 CHAPTER 4 : spatial locations were accurate. To map main BCs and behavioural diversity, we applied the same bootstrap procedure on a higher resolu- tion grid (~860 km2 each cell and ~30 km be- tween centroids of adjacent cells). Data analysis All data processing, analysis and visualization were conducted in R version 3.4.4 (R Develop- ment Core Team, 2018). RESULTS Movement and phenology During the incubation period in the breeding stage, birds concurrently tagged with GPS and GLS to track their short-term foraging trips rapidly engaged in commuting flights after the trip start, heading towards the upwelling area of the Canary Current in the North-west Af- rican shelf (Fig. 9D). After the breeding, from birds tracked only with GLS year round, one individual did not migrate and remained in the vicinity of the Canary Is. year round. The rest of the birds left the breeding area and started the post-breeding migration between 2nd of No- vember and 22nd of December, arriving to their main wintering area between 23rd of Novem- ber and 4th of January. Birds spent on average 55 days (range 23 – 77 days) in one of main wintering areas located along the South Afri- can waters (Benguela and Anguhlas Currents) and the Central South Atlantic (Fig. 9A). Birds started pre-breeding migration back to the Ca- nary archipelago between 27th of January and 2nd of March and arrived to the breeding colony between 22nd of February and 28th of March. Most birds followed an 8-shaped path to mi- grate over the Atlantic Ocean (Fig. 9A). Behavioural space of Cory’s shearwater at the population-level Overall, we identified 23 clusters from wet-dry segments (Fig. 2C), which we next grouped into 10 behavioural clusters (BCs) based on their similarity (Fig. 2C), each one representing a different behavioural mode. All individuals displayed all BCs over all stages of the annual cycle. Behavioural interpretation of clusters BCs in the upper (BC1, BC3, BC6, BC7, BC10) and lower (BC2, BC5, BC8, BC14, BC19) regions of the behavioural space (Fig. 2) corresponded to mainly dry and mainly wet segments, respectively (Fig. S3). Within the be- havioural space, more similar BCs tended to be positioned closer to each other. Based on a joint view of the input metrics (Table 1, Fig. S4) and the temporal (Fig. S5) and spatial distribution of BCs, we proposed an interpretation to each of the BC (see Table 3 with synthesized semantics and description). We interpreted BC1 as short flights [SF], BC3 as sustained flights [StF], BC6 as transit flights with occasional landings [TFLd], BC7 as com- muting flights with recurrent landings [CF], BC10 as shallow-surface diving [ShD], BC2 as short rests [SRest], BC5 as active sit-wait-dive [ActSWD], BC8 as still sit-wait-dive [StlSWD], BC14 as long sitting [Lsit] and BC19 as resting [Rest]. Dry BCs: We identified 5 flight modes, two related to ballistic displacements (BC1 [SF] and BC3 [StF]) and three presumably includ- ing foraging activities (BC6 [TFLd], BC7 [CF] and BC10 [ShD]) (Table 1, Fig. S4). Although the average median duration (~ 70 minutes) and IQR of segments assigned to BC3 [StF], BC6 [TFLd] and BC10 [ShD] were similar, we observed substantial differences in other important activity metrics defining each clus- ter, such as the rate of wet-dry transitions and the proportion of time in wet, leading to dif- ferent behavioural interpretation of those clus- ters (Table 3). We interpreted BC6 [TFLd] and BC10 [ShD] as mainly related to foraging due to the high rate of air/water transitions (i.e. landings and take-offs). BC6 [TFLd] was char- acterized by longer flights interrupted by very short periods on water, likely related to ex- tensive search within foraging grounds. BC10 141Air-water behavioural dynamics reveal complex at-sea ecology in global migratory seabirds Figure 2: Behavioural space of Cory’s shearwater constructed from segments of wet-dry data. (A) A multidi- mensional dataset (11 activity metrics and > 12 000 segments) was embedded into a 2-dimensional space result- ing from the parallelized t-SNE algorithm at “bigMap” R package. Each point represents one wet-dry activity segment. (B) Probability density estimation over the 2-dimensional space (colours from light yellow to black). By a discretization algorithm the space was divided into 23 clusters; thin grey lines delimit cluster borders. (C) Original 23 clusters were merged into 10 behavioural clusters; thin grey lines delimit cluster borders. The area size of cluster region indicates the variability within the cluster. (D) Resulting behavioural space, each colour represent one of 10 main behavioural clusters. Numbers correspond to cluster identity. Black triangles represent peaks in the density. 142 CHAPTER 4 : BC Behavioural semantics Wet-dry metrics Behavioural mode description 1 Short flights [SF] Duration = 34 min (31 - 39) Rchanges = 0 (i.e. without any air/water transition) Prop.W = 0 CWVI: Prop.W, Dur.D Occurred all over the day. Animals inverted less than 1% of time in this BC. Accounting for 4.9% in terms of frequency. 3 Sustained flights [StF] Duration = 70 min (55 - 99) Rchanges = 0 Prop.W = 0 CWVI: Prop.W, Nchanges Occurred all over the day. Occasion- ally included a limited number of very short wet states of few seconds within some segments. During the breeding stage (particularly pre-breeding stage), this BC included nocturnal visits of the colony of prolonged duration (4.7 hours on average, up to 10 hours). Individuals invested 2.8% of time in this BC. Ac- counting for 5.7% in terms of frequency. 6 Transit flights with occasional landings [TFLd] Duration = 69 min (45 - 111) Rchanges = 3 h-1 (1.9 - 4.3) Median.D = 19 min (11 - 30) Median.W = 1 min (0 - 3) Prop.W = 3% (1 - 9) CWVI: Prop.W, Nchanges, Rchanges Occurred all over the day. Includes oc- casional landings. Individuals invested around 5% of time in this BC on aver- age. Accounting for 13.8% in terms of frequency. 7 Commuting flights with recurrent landings [CF] Duration = 318 min (206 - 562). Rchanges = 6.7 h-1 (3.9 - 12) Max.D = 96 min (66 - 150) Max.W = 10 min (6 - 16) Prop.W = 9.5% (6 -15) CWVI: Dur.D, Nchanges Occurred all over the day. Individuals engage most intensively in this BC on sunrise and sunset. Animals invested around 25.6% of time in this BC on average, but more than 40% during the migration. Accounting for 15% in terms of frequency. Table 3: Description of behavioural clusters (BCs) corresponding to behavioural modes of Cory’s shearwaters. Using wet-dry data obtained from geolocation-immersion loggers, we developed a protocol to build up a behav- ioural space composed of 10 BCs. See Material and Methods and Fig. S4 in Supplementary Material for more details. In this summary we highlight quantitative metrics that best characterize each BC, adding a description derived from interpreting BCs in different contexts (see Supplementary Material). Note that duration of seg- ments was not used as input variable due to multi-collinearity. Values presented in the table denote median and interquartile range. CWVI indicates the most important metrics based on case-wise variable importance obtained from Random Forest (see Material and Methods and Fig. S8 in Supplementary Material for more de- tails). Note that here we express Rate of changes (Rchanges) as wet-dry transitions h-1 to ease the interpretability. 143Air-water behavioural dynamics reveal complex at-sea ecology in global migratory seabirds 10 Shallow- surface diving [ShD] Duration = 73 min (48 - 106) Rchanges = 17.7 h-1 (11.3 - 33.2) Median.D = 50 s (24 - 138) Median.W = 21 s (12 - 39) Max.D = 26 min (16 - 39) Max.W = 4 min (2 - 8) Prop.W = 17% (9 - 24) CWVI: Prop.W, Dur.D, Rchanges Transit with high landing rate (the high- est among all BCs). Occurred all over the day, most intensively during daylight hours. Individuals invested around 4.3% of time in this BC. Accounting for 11.3% in terms of frequency. 2 Short rests [SRest] Duration = 57 min (40 - 90) Rchanges = 0 CWVI: Prop.W Occurred all over the day, with preva- lence during daylight hours. This BC oc- casionally included very short dry states. Individuals invested around 1.8% of time in this BC. Accounting for 5.5% in terms of frequency. 5 Active sit-wait- dive [ActSWD] Duration = 85 min (51 - 147) Rchanges = 5.7 h-1 (2.2 - 13.6) Median.D = 22 s (12 - 63) Median.W = 11 min (2 - 31) Max.D = 2.5 min (0 - 8) Max.W = 39 min (24 - 62) Prop.W = 94% (85 - 99) CWVI: Median.W, Prop.W, SD.W Occurred all over the day; those start- ing in the afternoon hours tend to last until late night hours. Time on water highly variable, and combined with short flights. Animals invested around 9.9% of time in this BC. Accounting for 19.9% in terms of frequency. 8 Still sit- wait-dive [StlSWD] Duration = 615 min (473 - 786) Rchanges = 4.5 h-1 (3.1 - 7) Median.D = 18 s (12 - 28) Median.W = 105 s (48 - 280) Max.W = 270 min (212 - 367) Prop.W = 95% (92 - 97) CWVI: Max.W The longest median duration among all BCs. Generally including long wet state and various dry and wet short states of variable durations. Occurred all over the day, with prevalence to start in the afternoon/dusk, long over the night and finishing before the dawn. Individuals invested around 11.8% of time in this BC. Accounting for 4.5% in terms of fre- quency. 14 Long sitting [LSit] Duration = 567 min (393 - 846) Rchanges = 5.2 h-1 (3.3 - 10.5) Median.D = 18 s (12 - 33) MedianW = 4 min (1 - 12) Max.D = 15 min (9 - 21) Prop.W = 92% (88 - 95) CWVI: Dur.W, Nchanges Occurred all over the day, with preva- lence of the start after the sunrise and be- fore sunset. Individuals invested around 27.8% of time in this BC. Accounting for 9.4% in terms of frequency. Includes segments with up to 127 min of continu- ously wet. 19 Resting [Rest] Duration = 232 min (149 - 336) Rchanges = 3.4 h-1 (2.1 - 5.5) Median.D = 21 s (12 - 51) Median.W = 6 min (2 - 15) Max.D = 6 min (1 - 14) Max.W = 115 min (87 - 165) Prop.W = 95.5% (90 - 99) CWVI: Max.W, SD.W, Prop.W Occurred all over the day, except time around sunrise. With prevalence to start in the afternoon/dusk, long over the night (nocturnal resting) and finishing before the dawn. Individuals invested around 10.5% of time in this BC. Accounting for 10% in terms of frequency. 144 CHAPTER 4 : [ShD] encompassed segments with the high- est rate of transitions per hour, indicating high foraging activity, probably related to active area-restricted search within foraging patches, including short shallow dives to catch prey near the surface. We interpreted BC7 [CF] as com- muting flights. The long continuous duration of dry state (~ 100 minutes) and rate of changes clearly indicated relocation movements. In- deed, concurrent GPS tracking during incu- bation pinpointed BC7 to be mainly restricted to commuting corridors between the breeding colony and the main foraging grounds in the North-west African shelf (Fig. 6). Year-round GLS tracking also supported BC7 as commut- ing, as it was prevalent in most part of the mi- gratory routes between breeding and wintering sites (Fig. 7, Fig. S7). However, a small sub- cluster of segments within the BC7 [CF] corre- sponding to the pre-breeding and chick-rearing presumably included prolonged nocturnal vis- its to the colony (Fig. S5). Similarly, within the BC3 [StF] during the pre-breeding stage, we identified segments of prolonged duration of several hours during the night (Fig. S5), which we assumed to correspond also with visits to the colony. Wet BCs: Segments in these other 5 BCs are characterized by high proportion of time spent on water (> 90% of segment duration on aver- age). Based on short duration and any take- offs/landings, we interpreted BC2 as short rests [Srest]. We propose BC5 to correspond to active sit-wait-dive behaviour [ActSWD], characterized by a high rate of wet-dry transi- tions, high variability in duration of wet states and short flights. This BC might be related to intense local feeding in patches with abundant prey, where birds sit on water and dive to cap- ture prey within birds’ reach (i.e. fish schools near the surface). Generally, segments in BC8 [StlSWD], BC14 [LSit] and BC19 [Rest] were longer (> 3.5 hours on average) than the other BCs, indicating that once in wet state, birds tended to persist in it. These BCs likely includ- ed periods of rafting/drifting when birds sit- ing on the sea surface are passively carried by ocean currents and seldom interrupted by short periods of flights. BC8 [StlSWD] was char- acterized by long durations and by contain- ing at least one long wet state lasting several hours, presumably related to nocturnal resting, though surface-foraging events during those long bouts should not be discarded. Variable importance in behavioural classifica- tion We obtained an overall accuracy of 97% of pre- diction of BC based on the 11 activity metrics using RF. Metrics that mostly contributed to the accuracy of the classification of BCs were those reflecting the wet-dry activity variability at the segment level (e.g. number and rate of changes), together with the total durations of wet and dry states and proportion of wet (Fig. S6A). At the cluster level, case-wise variable importance varied between BCs, being pro- portion of wet the most important for five BC (Table 3, Fig. S6B). Changes in the behavioural space over the annual cycle Computation of density maps by stage con- firmed a different prevalence of BCs per stage (Fig. 3A-C). The time budget allocated to each BC (Fig. 5A) and a significant statisti- cal association between certain BC and stages (Supplementary Material, Fig. S7) also support this result. Essentially, different regions of the behavioural space, involving different BCs, were dominant at the different stages (Fig. 3A- C, Fig. S8). During the breeding period BC1 [SF], BC2 [SRest] and BC3 [StF] dominated the behavioural space, alongside with BC5 [ActSWD], BC6 [TFLd] and BC14 [LSit]. Dur- ing the migration, BC5 [ActSWD] and BC7 [CF] emerged as most dominant, although BC1 [SF], BC2 [SRest] and BC3 [StF] maintained their dominance. During the wintering, BC6 [TFLd], BC10 [ShD] and BC8 [StlSWD] were dominant. Moreover, we observed that even within the same BC region, different parts of the area were dominant in different stages (e.g. BC5 [ActSWD], BC10 [ShD]) (Fig. 3). 145Air-water behavioural dynamics reveal complex at-sea ecology in global migratory seabirds Figure 3: Changes in the behavioural space of Cory’s shearwater over the annual cycle. Probability density estimation over the behavioural space for separated stages: A) breeding, B) migration and C) wintering. Dashed lines represent 5% contours. Peaks correspond to dominant behavioural clusters (BCs) after grouping to 10 clusters. D-F) Transition probability rates between behavioural clusters for separated stages: D) breeding, E) migration and F) wintering. Layout of nodes reflect the 10 peaks of BCs in the behavioural space. Grey lines represent connections between BCs, the width of line is relative to the transition probability between BCs. BCs semantics: BC1 = SF, BC3 = StF, BC6 = TFLd, BC7 = CF, BC10 = ShD, BC2 = SRest, BC5 = ActSWD, BC8 = StlSWD, BC14 = Lsit, BC19 = Rest. Behavioural networks At global level, we did not find differences in network metrics (Table S2), indicating similar global network structure in the three stages of the annual cycle (Fig. 4). At local (node) level, mostly dry BC6 [TFLd], BC7 [CF] and BC10 [ShD] were the most important BCs in all three stages, as they were directly connected to the majority of BCs (see in- and out-degree central- ity values, Table S2). In contrast, mostly wet BC2 [SRest], BC5 [ActSWD], BC8 [StlSWD], BC14 [Lsit] and BC19 [Rest] had less connec- tions (i.e. lowest in- and out-degree centrality values). All BCs were closely connected regard- less of the stage. BC10 [ShD], BC7 [CF], BC6 [TFLd] and BC3 [StF] were the most impor- tant in terms of closeness centrality as shortest paths connected them to other BCs (Table S2). Conversely, BCs with lower closeness centrali- ty values (mostly wet BCs) needed longer paths to connect to other BCs (Table S2). Two BCs acted as major “hubs” in the networks: BC10 [ShD] was the most central node in terms of be- tweenness centrality during the breeding (high betweenness centrality), but during wintering both BC10 [ShD] and BC7 [CF] had equal im- portance. During the migration, however, BC7 [CF] was the most central node (Table S2, Fig. 4D-F). In terms of transition probabilities be- tween BCs (Fig. 4, Fig. S9), especially during the migration there was a high probability of 146 CHAPTER 4 : Figure 4: Behavioural networks of Cory’s shearwaters at the three different phenological stages over the annual cycle. Each column, from left to right, correspond to breeding, migration and wintering. The layout reflects the 10 peaks in the behavioural space. Grey lines represent connections between BCs. The width of these lines is relative to the transition probability between BCs. BCs semantics: BC1 = SF, BC3 = StF, BC6 = TFLd, BC7 = CF, BC10 = ShD, BC2 = SRest, BC5 = ActSWD, BC8 = StlSWD, BC14 = Lsit, BC19 = Rest. A-C) Average time invested in each behavioural cluster (BC) is represented by the size of the point, for separated stages: D-F) Betweenness centrality metric on node level represented by the size of the point. Two BCs, BC10 and BC7, acted as main “hubs” over all stages. mostly dry BCs to transit to BC5 [ActSWD] (range of transition probabilities 0.36-0.54) and mostly wet clusters to transit to BC7 [CF] (range 0.40-0.49). During the breeding and the wintering, these transitions were more evenly distributed (see transition matrices in Supple- mentary Material, Fig. S9). During wintering BC8 [StlSWD] gained importance compared to breeding and migration (transition probabilities ranged between 0.11-0.17 in wintering whereas was < 0.1 in breeding and migration). Finally, estimated entropy rate was the lowest for the migration stage, but values in breeding and wintering were only slightly above (breeding: 2.21, migration: 2.09, wintering: 2.23; see Table S2). Sources of variability in behavior We found between-stage variance significant- ly greater than within-stage variance in BC3 [StF], BC7 [CF], BC8 [StlSWD] and BC14 [Lsit] (BC3: F2,13 = 7.4, p = 0.007; BC7: F2,13 = 18.49, p < 0.001; BC8: F2,13 = 9.46, p = 0.003; BC14: F2,13 = 11.33, p= 0.001), indicating that between- stage behavioural variability was greater than inter-individual variability in those BCs (Fig. 5, Table S3). Birds were constrained in the amount of time invested in each BC particu- larly during the migration, when the amount of time in BC7 [CF] increased and time in BC14 [Lsit] decreased in all individuals, compar- ing to breeding and wintering. We observed a 147Air-water behavioural dynamics reveal complex at-sea ecology in global migratory seabirds Figure 5: Changes in the amount of time invested in each behavioural cluster (BC) over the annual cycle of Cory’s shearwater derived from wet-dry data. A) Mean proportion of time invested in each BC of 8 individuals in three stages. B) Individual variability in the amount of time invested in each BC. Points refer to individual proportions. Solid lines connect individual values over stages (B=breeding, M=migration, W=wintering). Black squares and solid vertical lines refer to mean ± SD values; dashed lines connect mean values over stages. Co- lours refer to BCs. BCs semantics: BC1 = SF, BC3 = StF, BC6 = TFLd, BC7 = CF, BC10 = ShD, BC2 = SRest, BC5 = ActSWD, BC8 = StlSWD, BC14 = Lsit, BC19 = Rest. Note the variable scales on y-axis. Asterisks in the upper right corner identify BCs where between-stage variability was significantly greater than inter-individual variability. gradual decrease in time invested in BC3 [StF] from breeding through migration to wintering, and contrarily, an increase in time invested in BC8 [StlSWD] from breeding to wintering for all individuals. Conversely, in the rest of BCs inter-individual variability was high across stages (Fig. 5, Table S3). The amount of time invested in these BCs (BC1 [SF], BC2 [SRest], BC5 [ActSWD], BC6 [TFLd], BC10 [ShD] and BC19 [Rest]) indicated higher flexibility of birds in time allocation for these behaviours and thus likely not so constrained by stages of the annual life cycle. 148 CHAPTER 4 : Behavioural space of Cory’s shearwater at the individual-level We showed that changes in time allocation to BCs varied between stages and individuals on a coarse-scale. We also visually evaluated the interpretation of BCs and time allocation at in- dividual level, at various spatial and temporal scales. For visualization purposes, all subse- quent figures refer to a single individual (ID: 6175726; Fig. 6, Fig. 7 and Fig. 8)). During the foraging trip of 13 days, the tagged bird tracked with GPS (Fig. 6), was involved in BC7 [CF] on the outward and inward commuting flights to foraging grounds over the African shelf. Both during the day and night bird was involved in foraging behaviours BC5 [ActSWD], BC6 [TFLd], BC10 [ShD] and sustained flights BC3 Figure 6: Movement and behaviour of one individual of Cory’s shearwater over one short-term foraging trip to the North-west African shelf. The plot illustrates a GPS track of 13 days of duration, composed of 117 behav- ioural segments of variable durations. Each point on the map, representing a GPS position, is coloured by the corresponding behavioural mode (i.e. behavioural cluster, BC) identified from unsupervised clustering of wet- dry data. In the zoom to a foraging area on the left, the positions are split into day (white border of the point), twilight (grey border) and night (black border). Note, for example, that animal was engaged in foraging BCs (BC10, BC6, BC5) mostly during the day, and in resting BCs during the night (BC14, BC8), although not ex- clusively. Black star indicates the breeding colony. Grey arrows indicate the direction of the commuting flights. BCs semantics: BC1 = SF, BC3 = StF, BC6 = TFLd, BC7 = CF, BC10 = ShD, BC2 = SRest, BC5 = ActSWD, BC8 = StlSWD, BC14 = Lsit, BC19 = Rest. [StF]. Resting behaviours BC8 [StlSWD] and BC14 [LSit] dominated during the night, but BC 19 [Rest] during the day. Year-round track over the whole breeding cycle from GLS (Fig. 7) reveals that, for example, on southward migra- tion to wintering grounds the tagged bird was involved mostly in BC7 [CF] during the day and rested during the night, more engaged in BC14 [Lsit] and BC19 [Rest]. However, on the northward migration to the breeding grounds in Canary Islands the tagged bird was involved in BC7 [CF] also during the night, since ear- lier arrival to breeding sites is advantageous for males to defend the nest. We present an acto- gram plot of a selected example (Fig. 8, see also Fig. S10 for other individual actograms). We can observe some clear circadian rhythms in shearwater’s behaviour, also adjusted over the 149Air-water behavioural dynamics reveal complex at-sea ecology in global migratory seabirds Figure 7: Movement and behaviour of Cory’s shearwater derived from GLS and wet-dry data over the year- round annual cycle. Plots show an example of one individual year-round track. For illustration purposes, posi- tions were linearly interpolated from 2 daily GLS positions to the start of each behavioural segment and later at 5-minutes intervals until the start of next behavioural segment, therefore should be treated with caution. Positions are split into day (left) and night (right). Each colour represents the behavioural mode (i.e. behav- ioural cluster, BC) identified from unsupervised clustering of wet-dry data from GLS. Grey arrows indicate the direction of the migratory flyway. Note, for example, active flight (BC7) during the night while engaged the pre-breeding migration to the colony (right panel), when the animal is likely pressured to arrive to the colony to defend the nest and start mating. Code on the top indicates individual and track identity. BCs semantics: BC1 = SF, BC3 = StF, BC6 = TFLd, BC7 = CF, BC10 = ShD, BC2 = SRest, BC5 = ActSWD, BC8 = StlSWD, BC14 = Lsit, BC19 = Rest. year-round cycle. For example, several behav- iours tended to be more diurnal (BC6 [TFLd], BC10 [ShD], BC5 [ActSWD]), although not exclusively. At the end of the breeding season (mid-September), this tagged bird was involved in mostly wet BCs, particularly BC14 [Lsit]. BC7 [CF] and BC3 [StF] presented bimodal pattern: while during the migration and winter- ing stage these BCs were mostly restricted to daylight hours, during breeding they occurred also during the night, likely representing noc- turnal visits to the nest. We can observe a clear shift in the circadian rhythm during the win- tering spent in the South-African coast (Fig. 8), when the bird adjusted behaviour to earlier sunrise and sunset. Especially at the beginning of the breeding season (March), before enter- ing the breeding colony after the sunset, bird was engaged in wet BCs, particularly BC5 [ActSWD] and BC19 [Rest]. Spatial and temporal distribution of behav- iours At global scale, mostly “wet” BCs [Srest, ActSWD, StlSWD, Lsit, DRest] predominat- ed within the wintering areas (Fig. 9B). BC14 [LSit] dominated in the southernmost part of the Canary Current, west Gulf of Guinea, Na- mibia off-shore and Agulhas Current. How- ever, in the wintering area of the Mozambique Channel, BC8 [StlSWD] was dominant. In the wintering area located in the pelagic zone of the South Atlantic BC8 and BC19 were similar- ly dominant. Mostly “dry” BCs [SF, StF, TFLd, CF, ShD] predominated along the migratory 150 CHAPTER 4 : Figure 8: Individual actogram of year-round behaviour of a Cory’s. Each coloured segment, of variable length, represents the behavioural mode (i.e. behavioural cluster, BC) identified from unsupervised clustering of wet-dry data from GLS. Each column represents one single day (0-24h). On the x-axis data starts on the day of deployment of the logger and ends on the day of recovery the next year. Black horizontal solid and dashed lines refer to time of local sunrise/sunset and nautical twilight at bird location, respectively. Vertical black-white lines delimit stages of annual life cycle: onset of post-breeding migration, arrival to the main wintering area, onset of pre-breeding migration and arrival to the breeding area, respectively. Note, for example, a clear shift in the circadian rhythm during the wintering spent in the South-African coast, when the bird adjusted behaviour to earlier sunrise and sunset. Code on the top indicates individual and track identity. BCs semantics: BC1 = SF, BC3 = StF, BC6 = TFLd, BC7 = CF, BC10 = ShD, BC2 = SRest, BC5 = ActSWD, BC8 = StlSWD, BC14 = Lsit, BC19 = Rest. Figure 9 (next page): Spatially-explicit behavioural landscapes of Cory’s shearwaters tracked with GLS (A- C) and concurrently with GPS and GLS (D-F). A) Year-round movements and distribution of 8 individuals tracked with GLS across the Atlantic Ocean, all individuals pooled together. Blue areas represent main wintering areas in the central South Atlantic and South African coast (wintering area of a resident individual around Canary Is. is ex-cluded for clarity). Purple area represents the distribution during the breeding stage according to GLS positional data. Yellow lines represent migratory flyway; grey arrows indicate the direction of the trip. D) Short-term foraging trips (n=23) of 19 individuals tracked concurrently with GPS and GLS during the breeding period. Black dashed line delimits the continental shelf. B and E) Main behavioural modes inferred from wet-dry data. Each grid cell shows the BC in which birds invested most of the time. C and F) Map of behavioural diversity based on Shannon Index. Black point refers to the breeding colony. See Material and Methods for details. BCs semantics: BC1 = SF, BC3 = StF, BC6 = TFLd, BC7 = CF, BC10 = ShD, BC2 = SRest, BC5 = ActSWD, BC8 = StlSWD, BC14 = Lsit, BC19 = Rest. flyway, particularly the BC7 [CF]. Though wet BCs were dominant in the northernmost part of the flyway in the North Atlantic. Regard- ing behavioural diversity, maximum values of Shannon index rose up in the breeding and the wintering areas were animals spent most of the time over the annual cycle (Fig. 9B). Intermedi- ate diversity was maintained in the pelagic win- tering area of the South Atlantic and also along the southern section of the migratory 8-shape loop flyway. However, lowest values were in the northwest section of the 8-shape loop. Fo- cusing in the Canary Current during the incu- bation period, mostly wet BCs were dominant, especially BC14 [Lsit] over the neritic domain of the African continental shelf (Fig. 9D-F). In 151Air-water behavioural dynamics reveal complex at-sea ecology in global migratory seabirds 152 CHAPTER 4 : contrast, BC7 [CF] was largely predominant in the pelagic domain out of the continental shelf. Behavioural diversity reached the highest val- ues over the continental shelf, especially in the southernmost region visited, whereas interme- diate values were reached in the commuting flyways connecting the foraging grounds with the breeding colony in Gran Canaria Is. The lowest diversity values were located out of the commuting flyways and the continental shelf. DISCUSSION Based on wet-dry data, we uncovered both flexible and structural components of the be- havioural organization of Cory’s shearwaters across the annual life cycle. Our study model, a highly-mobile migratory bird, presents a com- plex annual cycle that involves central place foraging, ocean-basin long migratory move- ment, and wandering in wintering areas. The existence of radically different behavioural contexts linked to phenology, and the need to exploit different marine environments over the year, results in different behavioural preva- lence and transitions both in time and space. A multi-scale wet-dry behavioural dynamics Wet-dry data (i.e. GLS wet-dry dynamics) have high behavioural content of at-sea movement behaviour, spanning scales from elementary motion patterns (e.g. minutes, hours) to com- plex ecological interactions (e.g. seasonality, annual life cycle). However, the majority of studies using wet-dry data rely on fixed-time segments (e. g. 10s, 10 minutes, 1 hour, etc.), depending on the sampling and aggregation time-scale of the logger. The time-scale is optimized by manufacturers or researchers, mainly as a trade-off between memory storage capacity and battery life (Johnson et al. 2017). Moreover, these values are traditionally aggre- gated and activity budgets are reported as the proportion of total time spent on water/in flight or splitting into day and night (Phalan et. 2007, Mackley et al. 2010, Dias et al. 2012a). The fixed-time approach ignores the natural dynamics and transitions between potentially different behavioural modes, limiting the be- havioural representation of wet-dry patterns (Bom et al. 2014). Despite some GLS loggers can store continuous data by registering each change and duration of a wet-dry state, analy- ses of these time-series have been limited to calculation of flight durations and number of landings/take-offs over different time periods (Catry et al. 2004, Shaffer et al. 2001). Indeed, only few studies have accounted for continu- ous transitions in order to split wet-dry data and identify foraging bouts (Dias et al. 2012b, Gutowsky et al. 2014, Ponchon et al. 2019). In this study, we revealed the presence of strong disruptions on wet-dry cumulative dynamics, and we used this as the basis of our behavioural segmentation and our quantitative description. We suggest this is a better way to capture the inherent multi-scale character of air-water be- havioural dynamics in seabirds, ranging from minutes to many hours. More generally, the be- havioural mapping protocol used in this work, namely (i) breakpoint analysis to segment wet- dry time-series, (ii) multidimensional charac- terization of the segments, (iii) unsupervised classification and embedding based on wet-dry metrics similarity, and (iv) interpretation of clusters as behavioural modes, goes much be- yond the standard analysis of wet-dry activity data and opens the door to obtain further be- havioural information from biologging. A rich wet-dry behavioural space Our quantitative analysis reveals the hierarchi- cal and modular nature of seabirds’ air-water behavioural dynamics at an unprecedented level of detail. The resulting behavioural space covers a wide range of behavioural scales and can be analysed at different levels of coarse- graining. In this work we identified 10 statisti- cally significant behavioural clusters (BCs) for Cory’s shearwaters corresponding to 10 behav- ioural modes, a number fairly greater than re- ported in previous studies using wet-dry data for any seabird species (Guilford et al. 2009, Dean et al. 2013, Gutowsky et al. 2014, Con- 153Air-water behavioural dynamics reveal complex at-sea ecology in global migratory seabirds ners et al. 2015), yet low enough to ease inter- pretability. Two large regions emerged in the behavioural space containing mostly wet and mostly dry behavioural modes, respectively (Fig. 2). Based on the activity metrics measured we were able to distinguish 5 modes within each region. Dry BCs: BC1 [SF] and BC3 [StF] involved essentially short flight displacements, whereas BC6 [TFLd], BC10 [ShD], and BC7 [CF] in- volved different types of foraging strategies. Whether covering more extensive areas (BC6 [TFLd]) or showing more intense and localized activity (shallow-surface diving in high-re- source patches (BC10 [ShD]), foraging behav- iour can be identified by high air-water transi- tions rates (i.e. landings and take-offs). BC10 [ShD] is likely related to active area-restricted search within foraging patches, including pur- suit diving and short shallow-surface dives to catch prey near the surface (Cianchetti-Bene- detti et al., 2017). Large distance commuting flights from the breeding colony to foraging grounds or during migration also incorpo- rate short wet periods to rest or forage (BC7 [CF]). This observation is in line with the pre- viously reported ‘‘fly-and-forage’’ strategy of Cory’s shearwaters during migration (Dias et al. 2012b), also found in related species such as Grey-headed albatross or Manx shearwater (Catry et al. 2004, Dean et al. 2013). The me- dian of the maximum duration of continuous flight state in BC7[CF] was about 100 minutes, which is close to flight bout durations described previously for Calonectris species (Dias et al. 2012b, Yoda et al. 2017). Despite in our analysis we omitted incubation behaviour, “dry” states include not only the pe- riods of flight, but also visits to the colony for mating and nest attendance while breeding. We identified prolonged nocturnal visits to the col- ony in BC3 [StF] and BC7 [CF]. The duration and aim of these visits may be related to differ- ent breeding duties according to the date along the breeding stage, from nest defence and mat- ting to brood-guarding and food-provisioning of the chick (Navarro et al. 2007). Wet BCs: In wet BCs birds spent most of the time on water (> 90% of segment duration on average). While on water, seabirds can show a wide spectrum of behaviours, from feeding to rafting/drifting, resting, bathing and plum- age maintenance (Catry et al. 2004, Weimer- skirch et al. 2010, Carter et al. 2016, Johnson et al. 2017, Granadeiro et al. 2018). Prey capture, handling, ingestion and digestion also occur in water (Harper 1987) and differences in wet BCs may reflect different feeding strategies ac- cording to the ecology and behaviour of target- ed prey (Elliot et al. 2008, Weimerskirch 1997, Davoren 2000). Except for BC2 [SRest] that we interpret as short resting periods on water, the rest of wet BCs can be related to (i) surface- foraging strategies (sit-wait-and-dive, surface seizing), e.g. BC5 [ActSWD], BC8 [StlSWD], where birds target prey while sitting on the sea surface (Weimerskirch et al. 1997), or (ii) resting behaviour, e.g. BC14 [LSit] and BC19 [Rest], when birds sit on the sea surface for long times but seldom interrupted by short flights. Sit-wait-dive activity can be characterized by high wet-dry transitions rates and variable wet state durations (classified as “active”, BC5 [ActSWD]) or else, by low wet-dry transition rates and more regular wet state durations (clas- sified as “still”, BC8 [StlSWD]). BC5[ActSWD] suggests active fishing by sitting on water in patches with abundant prey. Moreover, as ob- served in individual actograms, BC5 [ActSWD] occurred frequently at dusk, particularly dur- ing the breeding season (March-April), indicat- ing also rafting behaviour of shearwaters in the vicinity of the breeding colony, before entering the nest. Rafting has been previously described for other seabird species (Wilson et al. 2008, Weimerskirch et al. 2010, Rubolini et al. 2015) and thus pointed out the importance of waters nearby the colony for seabirds (Carter et al. 2016, Granadeiro et al. 2017). BC8 [StlSWD] suggests less active or sporad- ic fishing and may include also night foraging behaviour combined with resting. This type of behaviour has been recorded using biologging techniques in several albatross species that feed during the night on small-sized prey (Catry et al. 2004, Louzao et al. 2014). 154 CHAPTER 4 : Long periods sitting on water (BC14 [LSit]) both during daylight and darkness (Fig. 8, Fig. S10, Fig. S5) suggest important constraints forcing birds to reduce energy expenditure. Moulting is a high-energetically demanding process for birds since it alters flight capabil- ity and requires energy allocation to feathers replacement (Bridge et al. 2006, Ramos et al. 2009, Cherel et al. 2016). In Cory’s shearwater, moulting of primary feathers starts at the end of the breeding stage (mid-September, Alonso et al. 2008, Ramos et al. 2018) and ends in the wintering area (Camphuysen & Van Der Meer 2001). We found many birds intensively en- gaged in BC14 [LSit] in the southern-most part of the Canary Current from mid-September to mid-November (Fig. 8, Fig. S10), then mi- grating to the wintering area (engaged mostly in BC7 [CF]) and then again spending most of the time in BC14 [LSit] upon arrival. This find- ing suggests that BC14 may be related to ac- tive moulting of flight feathers, when birds are forced to sit on the water for prolonged periods of time, and that birds may interrupt moulting to account migration and restart it after arrival to the wintering area. Behavioural space use: changes in beha- vioural strategies, organization and com- plexity Animals need to cope with different biologi- cal constrains over the different stages of the annual cycle, which can lead to a different arrangement of activity budgets (Maclkey et al. 2011, Rayner et al. 2012, Gutowsky et al. 2014, Clay et al. 2017). Moreover, changes of behavioural organization may also reflect envi- ronmental and habitat conditions that animals faced (Perón et al. 2010, Freeman et al. 2013). Despite all BCs appeared in all stages of the annual cycle, we found the relative prevalence was markedly different, and some BCs were dominant at breeding (BC1 [SF], BC2 [SRest], BC3 [StF], BC5 [ActSWD], BC6 [TFLd], BC14 [LSit]), at migration (BC7 [CF] and BC5 [ActSWD]), or at wintering (BC8 [StlSWD], BC6 [TFLd], BC10 [ShD]). Our results are con- sistent with the idea that during breeding most activity has to do with different feeding strat- egies and complex high air-water transition rates, whereas during migration birds spent more than 40% of their time in flight (BC7 [CF]) but refill their energy on stopover sites (Dias et al. 2012b, Freeman et al. 2013). Win- tering period is important to adult seabirds, especially to restore energy after the breed- ing and migration and to prepare for the next breeding season. Hence, a generalized decrease in flight and an increase in drifting on water surface are probably related to the stronger bio- logical constrains while wintering (Perón et al. 2010, Mackley et al. 2011, Rayner et al. 2012, Gutowsky et al. 2014, Clay et al. 2017), includ- ing the energetically-costly moulting process (Ramos et al. 2009, 2018). These results are clearly observed in our network analysis, which suggests that the behavioural strategy observed during the breeding stage, where animals com- bine flight and water activity in relatively simi- lar proportions (e.g. time invested in BC7 [CF] and BC14 [LSit], respectively, Fig. 4A-C), is clearly a ‘mixture’ of the behavioural strategies observed during the migration and wintering periods. Cory’s shearwaters showed more complex behavioural sequences and organization during the breeding and wintering stage, as the larger the entropy rate the larger the behavioural com- plexity. Migration forces shearwaters to restrict their behavioural organization to sequences be- tween several dominant behavioural modes, mainly dry BC7 [CF] alternated by some wet BC (foraging/resting in stopovers). Transitions between behaviours became more frequent and predictable than during breeding and winter- ing, as seen by stronger fluxes in our network analysis, and larger values of betweenness cen- trality of mostly wet BCs (i.e. BC2, BC5, BC8, BC14 and BC19). The latter BCs become rel- evant during migration as transitions between other BCs pass systematically through them, see Fig. 4). When evaluating the potential sources of be- havioural variability in each BC, we showed that the phenology shapes the variability of the most dominant BCs (BC7, BC14, BC8, and also 155Air-water behavioural dynamics reveal complex at-sea ecology in global migratory seabirds BC3) over the annual cycle, yet for non-dom- inant BCs inter-individual variability is larger compared to the variability introduced by the phenological stages. Our results suggest a mul- ti-level (i.e. individual vs. population) and com- plex behavioural response of seabirds to both intrinsic and extrinsic (environmental) signals, which expand over a wide range of scales, from daily to seasonal scales. Global spatially-explicit behavioural land- scapes The construction of spatially-explicit behav- ioural maps allowed us to reveal differences in the prevalence and diversity of BCs over dif- ferent areas. Both at global (year-round trips) and local (short-term foraging trips) scales we found the highest behavioural diversity to over- lap with upwelling regions of the Atlantic, both during breeding (Canary Current) and winter- ing (Benguela and Agulhas Currents), This result suggests a positive correlation between hotspots of behavioural diversity and important foraging grounds, so that wet-dry data alone can be used to identify major feeding areas for the species. Contrarily, areas related to migra- tory routes and transits to foraging grounds showed low behavioural diversity. More gener- ally, these findings indicate a positive correla- tion between behavioural complexity and habi- tat complexity, but further investigations are needed to confirm this result with more data and refined spatial statistical methodologies. Actograms: revealing individual daily and seasonal behavioural patterns As far as we know, this is the first study show- ing detailed year-round behaviour of seabird at such detail inferred uniquely from wet-dry data. Actograms (Bäckman et al. 2017) allows for detailed examination of time allocated to the different behavioural modes simultane- ously on both daily and seasonal scales. From inspecting temporal changes in behavioural modes we can infer the timing of major an- nual life cycle events, such as migration (e.g. increase in BC7 [CF]), wintering (prevalence of mostly wet BCs) and return to breeding grounds (nocturnal visits to the colony, e.g. BC7 [CF] and BC3 [StF]). Our protocol com- bined with the use of actograms may also assist to evaluate the existence of carry-over effects (i.e. processes that influence individual perfor- mance in a subsequent season) since deviation from common timing of phenological events and associated changes in behavioural budg- ets may be expected when individuals fail to success in events such as breeding (Harrison et al. 2011, Catry et al. 2013, Fayet et al. 2016). For example, actograms can easily allow us to infer breeding failure according to time budget allocated to behavioural modes related to nest attendance (BC7 [CF] and BC3 [StF]). Indeed, they can even help to detect an advancement of the moulting period, presumably due to breed- ing failure (Ramos et al. 2018), when preva- lence of the behavioural mode likely related to moult (BC14 [LSit]) advances in the calendar. Therefore, our protocol to decipher behavioural modes combined with appropriated data visu- alization, such as actograms, provide a power- ful tool to depict the timing and time allocation of behaviours over the entire annual cycle. Framework transferability: applicability on biologging data Despite we develop our framework with wet- dry data from geolocator-immersion loggers, it could be suited for different tracking data. As a multi-step process, the whole or part of the protocol can be applied on multiple sources of biologging data. Steps of segmentation, dimen- sionality reduction and clustering carried out to build up a behavioural space and discretize behavioural complexity into interpretable units allows for analysing multidimensional data, such as data recorded by multi-sensor devices. Moreover, functions provided by the “bigMap” R package that we applied here (Garriga & Bar- tumeus 2018) are especially designed to work with big data, as such generated with multi- sensor devices equipped with accelerometer and able to work for prolonged periods of time. 156 CHAPTER 4 : CONCLUDING REMARKS In this work we provided a novel framework for behavioural annotation based on wet-dry data from geolocator-immersion loggers, which al- lows for exploring behavioural organization and diversity in behavioural repertoires at sev- eral scales of complexity, from daily to annual scale and from individual to population level. We illustrated the protocol analysing behav- ioural complexity over the annual cycle of a long-range migratory seabird species tracked with geolocator-immersion sensors, yet the multi-step protocol may be suited to other dif- ferent sources of biologging data. This frame- work paves the way for extending behavioural annotation to year-round movements of wild- life, opening new avenues to understand be- havioural patterns and the seasonal timing of life-history events of animals spending most of their life out of the human’s sight. AUTHORS’ CONTRIBUTION ZZ, JMRG and FB developed the conceptual framework and design of the study. ZZ analysed data and developed visualizations. JMRG analysed data and developed behavioural landscapes visualization. TM analysed GLS raw data. JGS provided funding for fieldwork and FB for computational resources (Computational Biology Lab, CEAB-CSIC). TM and JMRG carried out the fieldwork. ZZ, JMRG and FB wrote the first version of the manuscript, with later contributions of JGS. All authors revised the last version of this manuscript. AKNOWLEDGEMENTS We are thankful to Joan Garriga, the developer of the “bigMap” R package, for the guidance and recommendations on the use of specific functions and parameter selection along the analytical framework. BIBLIOGRAPHY Alonso, H., Matias, R., Granadeiro, J.P. & Catry, P. (2009) Moult strategies of Cory’s Shearwaters Calonectris diomedea borea- lis: the influence of colony location, sex and individual breeding status. Journal of Orni- thology, 150(2): 329-337. Anderson, D.J. & Perona, P. (2014) Toward a science of computational ethology. Neuron, 84(1): 18-31. Bäckman, J., Andersson, A., Pedersen, L., Sjö- berg, S., Tøttrup, A.P. & Alerstam, T. (2017) Actogram analysis of free-flying migratory birds: new perspectives based on accelera- tion logging. Journal of Comparative Phys- iology A, 203(6-7): 543-564. Barnes, R. (2018) dggridR: Discrete Global Grids. R package version 2.0.3. Beever, E.A., Hall, L.E., Varner, J., Loosen, A.E., Dunham, J.B., Gahl, M. K., Smith, F.A. & Lawler, J.J. (2017) Behavioral flex- ibility as a mechanism for coping with cli- mate change. Frontiers in Ecology and the Environment, 15(6): 299-308. Berger‐Tal, O., Blumstein, D.T., Carroll, S., Fisher, R.N., Mesnick, S. L., Owen, M.A., Saltz, D., Cassady, C. & Swaisgood, R.R. (2016) A systematic survey of the integra- tion of animal behavior into conservation. Conservation Biology, 30(4): 744-753. Berman, G.J., Bialek, W. & Shaevitz, J.W. (2016) Predictability and hierarchy in Dro- sophila behavior. Proceedings of the Na- tional Academy of Sciences, 113(42): 11943- 11948. Berman, G.J., Choi, D.M., Bialek, W. & Shae- vitz, J.W. (2014) Mapping the stereotyped behaviour of freely moving fruit flies. Jour- nal of The Royal Society Interface, 11(99): 20140672. Biau, G. & Scornet, E. (2016) A random forest guided tour. Test, 25(2): 197-227. Birant, D. & Kut, A. (2007) ST-DBSCAN: An algorithm for clustering spatial–tempo- ral data. Data & Knowledge Engineering, 60(1): 208-221. Block, B.A., Jonsen, I.D., Jorgensen, S.J., Win- ship, A.J., Shaffer, S.A., Bograd, S.J. et al. (2011) Tracking apex marine predator movements in a dynamic ocean. Nature, 475(7354): 86. 157Air-water behavioural dynamics reveal complex at-sea ecology in global migratory seabirds Bom, R.A., Bouten, W., Piersma, T., Ooster- beek, K. & van Gils, J.A. (2014) Optimiz- ing acceleration-based ethograms: the use of variable-time versus fixed-time segmen- tation. Movement ecology, 2(1): 6. Bridge, E.S. (2006) Influences of morphology and behavior on wing-molt strategies in seabirds. Marine Ornithology, 34: 7-19. Burger, A.E. & Shaffer, S.A. (2008) Perspec- tives in ornithology application of tracking and data-logging technology in research and conservation of seabirds. The Auk, 125(2): 253-264. Camphuysen, C.J. & Van Deer Meer, J.A. (2001) Pelagic distribution, moult and (sub-) specific status of Cory’s Shearwaters Calo- nectris [d.] diomedea/borealis wintering off southern Africa. Marine Ornithology, 29: 89-96. Carter, M.I., Cox, S.L., Scales, K.L., Bicknell, A.W., Nicholson, M.D., Atkins, K.M., Mor- gan, G. Morgan, L. Grecian, J.W., Patrick, S. & Votier, S.C. (2016) GPS tracking re- veals rafting behaviour of Northern Gan- nets (Morus bassanus): implications for foraging ecology and conservation, Bird Study, 63(1): 83-95. Catry, P., Dias, M.P., Phillips, R.A. & Grana- deiro, J.P. (2013) Carry‐over effects from breeding modulate the annual cycle of a long‐distance migrant: an experimental demonstration. Ecology, 94(6): 1230-1235. Catry, P., Granadeiro, J.P., Ramos, J., Phillips, R.A. & Oliveira, P. (2011) Either taking it easy or feeling too tired: old Cory’s Shear- waters display reduced activity levels while at sea. Journal of Ornithology, 152(3): 549- 555. Catry, P., Phillips, R. A., Phalan, B., Silk, J.R. & Croxall, J.P. (2004) Foraging strategies of grey-headed albatrosses Thalassarche chrysostoma: integration of movements, ac- tivity and feeding events. Marine Ecology Progress Series, 280: 261-273. Cherel, Y., Quillfeldt, P., Delord, K. & Weimer- skirch, H. (2016) Combination of at-sea activity, geolocation and feather stable iso- topes documents where and when seabirds molt. Frontiers in Ecology and Evolution, 4: 3. Chimienti, M., Cornulier, T., Owen, E., Bolton, M., Davies, I.M., Travis, J.M.J. & Scott, B.E. (2016) The use of an unsupervised learning approach for characterizing latent behaviors in accelerometer data. Ecology and Evolution, 6(3): 727–741. Cianchetti-Benedetti, M., Catoni, C., Kato, A., Massa, B. & Quillfeldt, P. (2017) A new algorithm for the identification of dives reveals the foraging ecology of a shallow- diving seabird using accelerometer data. Marine Biology, 164(4): 77. Clay, T., Phillips, R.A., Manica, A., Jackson, H. & Brooke, M. (2017) Escaping the oligo- trophic gyre? The year-round movements, foraging behaviour and habitat preferences of Murphy’s petrels. Marine Ecology Prog- ress Series, 579, 139-155. Conners, M.G., Hazen, E.L., Costa, D.P. & Shaffer, S.A. (2015) Shadowed by scale: subtle behavioral niche partitioning in two sympatric, tropical breeding albatross spe- cies. Movement Ecology, 3(1): 28. Croxall, J.P., Butchart, S.H., Lascelles, B.E.N., Stattersfield, A.J., Sullivan, B.E.N., Symes, A. & Taylor, P.H.I.L. (2012) Seabird conser- vation status, threats and priority actions: a global assessment. Bird Conservation In- ternational, 22(1): 1-34. Csardi, G. & Nepusz, T. (2006) The igraph software package for complex network re- search, InterJournal, Complex Systems, 1695(5): 1-9. Davoren, G.K. (2000) Variability in foraging in response to changing prey distributions in rhinoceros auklets. Marine Ecology Prog- ress Series, 198: 283-291. Dean, B., Freeman, R., Kirk, H., Leonard, K, Phillips, R.A., Perrins, C.M. & Guilford, T. (2013) Behavioural mapping of a pelagic seabird: combining multiple sensors and a hidden Markov model reveals the distri- bution of at-sea behaviour. Journal of the Royal Society Interface, 10(78): 20120570. Dias, M.P., Granadeiro, J.P. & Catry, P. (2012a) Working the day or the night shift? Forag- 158 CHAPTER 4 : ing schedules of Cory’s shearwaters vary according to marine habitat. Marine Ecol- ogy Progress Series, 467: 245-252. Dias, M.P., Granadeiro, J.P. & Catry, P. (2012b) Do seabirds differ from other migrants in their travel arrangements? On route strategies of Cory’s shearwater during its trans-equatorial journey. PLoS One, 7(11): e49376. Dias, M.P., Granadeiro, J.P., Phillips, R.A., Alonso, H. & Catry, P. (2010) Breaking the routine: individual Cory’s shearwaters shift winter destinations between hemispheres and across ocean basins. Proceedings of the Royal Society B: Biological Sciences, 278(1713): 1786-1793. Dias, M.P., Martin, R., Pearmain, E.J., Burf- ield, I.J., Small, C. Phillips, R.A., Yates, O., Lascelles, B., Garcia Borboroglu, P. & Croxall, J.P. (2019) Threats to seabirds: A global assessment. Biological Conserva- tion, 237: 525-537. Dominoni, D.M., Åkesson, S., Klaassen, R., Spoelstra, K. & Bulla (2017) Methods in field chronobiology. Philosophical Transac- tions of the Royal Society B: Biological Sci- ences, 372(1734): 20160247. Egnor, S.R., & Branson, K. (2016) Computa- tional analysis of behavior. Annual Review of Neuroscience, 39: 217-236. Elliott, K.H., Woo, K., Gaston, A.J., Benve- nuti, S., Dall’Antonia, L. & Davoren, G. K. (2008) Seabird foraging behaviour indi- cates prey type. Marine Ecology Progress Series, 354: 289-303. Fayet, A.L., Freeman, R., Shoji, A., Kirk, H.L., Padget, O., Perrins, C.M. & Guilford, T. (2016) Carry‐over effects on the annual cycle of a migratory seabird: an experimen- tal study. Journal of Animal Ecology, 85(6): 1516-1527. Fox, J. W. (2010) Geolocator manual v8. British Antarctic Survey, Cambridge. Freeman, R., Dean, B., Kirk, H., Leonard, K., Phillips, R.A., Perrins, C.M. & Guilford, T. (2013) Predictive ethoinformatics reveals the complex migratory behaviour of a pe- lagic seabird, the Manx Shearwater. Jour- nal of the Royal Society Interface, 10(84): 20130279. Garriga, J. & Bartumeus, F. (2018) bigMap: Big Data Mapping with parallelized t-SNE. arXiv preprint, arXiv: 1812.09869. Gaston, A.J. (2004) Seabirds: a natural history. Yale University Press. González-Solís, J., Croxall, J.P., Oro, D. & Ruiz, X. (2007) Trans‐equatorial migration and mixing in the wintering areas of a pe- lagic seabird. Frontiers in Ecology and the Environment, 5(6): 297-301. Granadeiro, J. P., Campioni, L. & Catry, P. (2018) Albatrosses bathe before departing on a foraging trip: implications for risk as- sessments and marine spatial planning. Bird Conservation International, 28(2): 208-215. Guilford, T., Meade, J., Willis, J., Phillips, R.A., Boyle, D., Roberts, S., Collett, M., Freeman, R. & Perrins, C.M (2009) Migra- tion and stopover in a small pelagic seabird, the Manx shearwater Puffinus puffinus: in- sights from machine learning. Proceedings of the Royal Society B: Biological Sciences, 276(1660): 1215-1223. Gutowsky, S.E., Gutowsky, L.F., Jonsen, I.D., Leonard, M.L., Naughton, M.B., Romano, M.D. & Shaffer, S.A. (2014) Daily activity budgets reveal a quasi-flightless stage dur- ing non-breeding in Hawaiian albatrosses. Movement Ecology, 2(1): 23. Han, S., Taralova, E., Dupre, C. & Yuste, R. (2018) Comprehensive machine learning analysis of Hydra behavior reveals a stable basal behavioral repertoire. eLife, 7: e32605. Harcourt, R., Sequeira, A.M.M., Zhang, X., Roquet, F., Komatsu, K., Heupel, M. et al. (2019) Animal-Borne Telemetry: An Inte- gral Component of the Ocean Observing Toolkit. Frontiers in Marine Science, 6:326. Harper, P. C. (1987) Feeding behaviour and oth- er notes on 20 species of procellariiformes at sea. Notornis, 34: 169-192. Harrison, X.A., Blount, J.D., Inger, R., Nor- ris, D.R. & Bearhop, S. (2011) Carry‐over effects as drivers of fitness differences in animals. Journal of Animal Ecology, 80(1): 4-18. 159Air-water behavioural dynamics reveal complex at-sea ecology in global migratory seabirds Helm, B., Visser, M. E., Schwartz, W., Kro- nfeld-Schor, N., Gerkema, M., Piersma, T. & Bloch, G. (2017) Two sides of a coin: eco- logical and chronobiological perspectives of timing in the wild. Philosophical Trans- actions of the Royal Society B: Biological Sciences, 372(1734): 20160246. Hill, R.D. & Braun, M.J. (2001) Geolocation by light level, electronic tagging and tracking in marine fisheries. Proceedings of the sym- posium on tagging and tracking marine fish with electronic devices. East-West Center. University of Hawaii. Berlin: Springer. pp. 315–330. Igual, J.M., Forero, M.G., Tavecchia, G., González-Solís, J., Martínez-Abraín, A., Hobson, K.A., Ruiz, X. & Oro, D. (2005) Short-term effects of data-loggers on Cory’s shearwater (Calonectris diomedea). Marine Biology, 146(3): 619-624. Johnson, L.R., Boersch-Supan, P.H., Phil- lips, R.A. & Ryan, S.J. (2017) Changing measurements or changing movements? Sampling scale and movement model iden- tifiability across generations of biologging technology. Ecology and Evolution. 7(22): 9257-9266. Kays, R., Crofoot, M.C., Jetz, W. & Wikelski, M. (2015) Terrestrial animal tracking as an eye on life and planet. Science, 348(6240): aaa2478. Knell, A.S. & Codling, E.A. (2011) Classifying area-restricted search (ARS) using a partial sum approach. Theoretical Ecology, 5: 325- 339. Krebs, C.J. (1999) Ecological Methodology. Addison-Wesley. Krebs, J. R. & Davies, N. B. (2009) Behav- ioural ecology: an evolutionary approach. John Wiley & Sons. Kuhn M. et al. (2018) caret: Classification and Regression Training. R package version 6.0-79. Lascelles, B.G., Taylor, P.R., Miller, M.G.R., Dias, M.P., Oppel, S. et al. (2016) Apply- ing global criteria to tracking data to define important areas for marine conservation. Diversity and Distributions, 22(4): 422-431. Lecomte, V.J., Sorci, G., Cornet, S., Jaeger, A., Faivre, B., Arnoux, E., Gaillard, M., Trou- vé, C., Besson, D., Chastel, O. & Weimer- skirch, H. (2010) Patterns of aging in the long-lived wandering albatross. Proceed- ings of the National Academy of Sciences of the United States of America, 107(14): 6370–6375. Liaw, A. & Wiener, M. (2002) Classification and regression by randomForest. R news, 2(3): 18-22. Louzao, M., Wiegand, T., Bartumeus, F. & Weimerskirch, H. (2014) Coupling instan- taneous energy-budget models and behav- ioural mode analysis to estimate optimal foraging strategy: an example with wander- ing albatrosses. Movement Ecology, 2(1): 8. Mackley, E.K., Phillips, R.A., Silk, J.R., Wake- field, E.D., Afanasyev, V., Fox, J.W. & Fur- ness, R.W. (2010) Free as a bird? Activity patterns of albatrosses during the nonbreed- ing period. Marine Ecology Progress Se- ries, 406, 291-303. Mackley, E.K., Phillips, R.A., Silk, J.R., Wake- field, E.D., Afanasyev, V. & Furness, R.W. (2011) At-sea activity patterns of breeding and nonbreeding white-chinned petrels Procellaria aequinoctialis from South Georgia. Marine Biology, 158(2): 429-438. Marques, J.C., Lackner, S., Félix, R.L. & Org- er, M.B. (2018) Structure of the Zebrafish Locomotor Repertoire Revealed with Un- supervised Behavioral Clustering. Current Biology, 28: 181-195.e5. McIntyre, T. (2014) Trends in tagging of ma- rine mammals: a review of marine mammal biologging studies. African Journal of Ma- rine Science, 36(4): 409-422. Merkel, B., Phillips, R.A., Descamps, S., Yoc- coz, N. G., Moe, B. & Strøm, H. (2016) A probabilistic algorithm to process geoloca- tion data. Movement Ecology, 4(1): 26. Meyer, X., MacIntosh, A.J., Kato, A., Chiara- dia, A. & Ropert-Coudert, Y. (2015) Hy- drodynamic handicaps and organizational complexity in the foraging behavior of two free-ranging penguin species. Animal Bio- telemetry, 3(1): 25. 160 CHAPTER 4 : Nathan, R., Getz, W.M., Revilla, E., Holyoak, M., Kadmon, R., Saltz, D. & Smouse, P.E. (2008) A movement ecology paradigm for unifying organismal movement research. Proceedings of the National Academy of Sciences, 105(49): 19052-19059. Navarro, J. & González-Solís, J. (2009) Envi- ronmental determinants of foraging strate- gies in Cory’s shearwaters Calonectris dio- medea. Marine Ecology Progress Series, 378: 259-267. Navarro, J., González-Solís, J. & Viscor, G. (2007) Nutritional and feeding ecology in Cory’s shearwater Calonectris diomedea during breeding. Marine Ecology Progress Series, 351: 261-271. Nilsson, J.Å., Bronmark, C., Hansson, L.A. & Chapman, B.B. (2014) Individuality in movement: The role of animal personality. In: Hansson et al. (Eds.) Animal Movement Across Scales, Oxford university press. Norris, K. (2004) Managing threatened spe- cies: the ecological toolbox, evolutionary theory and declining‐population paradigm. Journal of Applied Ecology, 41: 413-426. Orben, R. A., Kokubun, N., Fleishman, A. B., Will, A. P., Yamamoto, T., Shaffer, S. A., Paredes, R., Takahashi, A. & Kitaysky, A. S. (2018) Persistent annual migration pat- terns of a specialist seabird. Marine Ecol- ogy Progress Series, 593: 231-245. Passos, C., Navarro, J., Giudici, A. & González- Solís, J. (2010) Effects of extra mass on the pelagic behavior of a seabird. The Auk, 127(1): 100-107. Pebesma, E. (2018) Simple Features for R: Standardized Support for Spatial Vector Data. The R Journal, 10 (1): 439-446. Péron, C., Delord, K., Phillips, R.A., Char- bonnier, Y., Marteau, C., Louzao, M. & Weimerskirch, H. (2010) Seasonal variation in oceanographic habitat and behaviour of white-chinned petrels Procellaria aequi- noctialis from Kerguelen Island. Marine Ecology Progress Series, 416: 267-284. Phalan, B., Phillips, R.A., Silk, J. R., Afa- nasyev, V., Fukuda, A., Fox, J., Catry, P., Higuchi, H. & Croxall, J. P. (2007) Foraging behaviour of four albatross species by night and day. Marine Ecology Progress Series, 340: 271-286. Phillips, R.A., Lewis, S., González-Solís, J. & Daunt, F. (2017) Causes and consequences of individual variability and specializa- tion in foraging and migration strategies of seabirds. Marine Ecology Progress Series, 578: 117-150. Phillips, R.A., Silk, J.R.D., Croxall, J.P., Afa- nasyev, V. & Briggs, D.R. (2004) Accuracy of geolocation estimates for flying seabirds. Marine Ecology Progress Series, 266: 265- 272. Phillips, R.A., Xavier, J. C. & Croxall, J.P. (2003) Effects of satellite transmitters on albatrosses and petrels. The Auk, 120(4): 1082-1090. Ponchon, A., Cornulier, T., Hedd, A., Grana- deiro, J.P. & Catry, P. (2019) Effect of breed- ing performance on the distribution and ac- tivity budgets of a predominantly resident population of black‐browed albatrosses. Ecology and Evolution, 9(15): 8702-8713. R Core Team (2019) R: A language and envi- ronment for statistical computing. R Foun- dation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/. Ramos, R., Llabrés, V., Monclús, L., López‐Bé- jar, M. & González‐Solís, J. (2018) Costs of breeding are rapidly buffered and do not af- fect migratory behavior in a long‐lived bird species. Ecology, 99(9): 2010-2024. Ramos, R., Militão, T. González-Solís, J. & Ruiz, X. (2009) Moulting strategies of a long-distance migratory seabird, the Medi- terranean Cory’s Shearwater Calonectris diomedea diomedea. Ibis, 151:151–159. Rayner, M.J., Taylor, G.A., Gummer, H.D., Phillips, R.A., Sagar, P.M., Shaffer, S.A. & Thompson, D.R. (2012) The breeding cycle, year-round distribution and activity patterns of the endangered Chatham Petrel (Pterodroma axillaris). Emu - Austral Orni- thology, 112(2): 107-116. Reyes-González, J. M., Zajková, Z., More- ra-Pujol, V., De Felipe, F., Militão, T., Dell’Ariccia, G., Ramos, R., Igual, J. M., 161Air-water behavioural dynamics reveal complex at-sea ecology in global migratory seabirds Arcos, J. M. & González-Solís, J. (2017) Migración y ecología espacial de las po- blaciones españolas de pardela cenicienta. Monografía n.º 3 del programa Migra. SEO/BirdLife. Madrid. Roncon, G., Bestley, S., McMahon, C.R., Wienecke, B. & Hindell, M.A. (2018) View from below: inferring behavior and physi- ology of Southern Ocean marine predators from dive telemetry. Frontiers in Marine Science, 5: 1-23. Rubolini, D., Maggini, I., Ambrosini, R., Impe- rio, S., Paiva, V.H., Gaibani, G., Saino, N. & Cecere, J.G. (2015) The effect of moon- light on Scopoli’s Shearwater Calonectris diomedea colony attendance patterns and nocturnal foraging: a test of the foraging ef- ficiency hypothesis. Ethology, 121(3): 284- 299. Sánchez-Román, A., Gómez-Navarro, L., Fab- let, R., Oro, D., Mason, E., Arcos, J.M., Ruiz, S. & Pascual, A. (2019) Rafting be- haviour of seabirds as a proxy to describe surface ocean currents in the Balearic Sea. Scientific Reports, 9(1): 17775. Sangster, G., Collinson, J.M. Crochet, P.A., Knox, A.G., Parkin, D.T. & Votier, S.C. (2012) Taxonomic recommendations for British birds: eighth report. Ibis, 154: 874- 883. Schreiber, E.A. & Burger, J. (2001) Biology of marine birds. CRC press. Shaffer, S.A., Costa, D.P. & Weimerskirch, H. (2001) Behavioural factors affecting forag- ing effort of breeding wandering albatross- es. Journal of Animal Ecology, 70: 864-874. Shamoun-Baranes, J., Bouten, W., Camphuy- sen, C.J. & Baaij, E. (2011) Riding the tide: intriguing observations of gulls resting at sea during breeding. Ibis, 153: 411-415. Sih, A., Stamps, J., Yang, L.H., McElreath, R. & Ramenofsky, M. (2010) Behavior as a key component of integrative biology in a human-altered world. Integrative and com- parative biology, 50(6): 934-944. Thiebault, A. & Tremblay, Y. (2013) Splitting animal trajectories into fine-scale behavior- ally consistent movement units: breaking points relate to external stimuli in a forag- ing seabird. Behavioral Ecology and Socio- biology, 67(6): 1013-1026. Todd, J.G., Kain, J.S. & Bivort, B.L. (2017) Systematic exploration of unsupervised methods for mapping behavior. Physical Biology, 14(1): 015002. van der Maaten, L. & Hinton, G. (2008). Visu- alizing data using t-sne. Journal of Machine Learning Research, 9: 2579-2605. Vandenabeele, S.P., Shepard, E.L., Grogan, A. & Wilson, R.P. (2012) When three per cent may not be three per cent; device-equipped seabirds experience variable flight con- straints. Marine Biology, 159(1): 1-14. Vandenabeele, S.P., Wilson, R.P. & Grogan, A. (2011) Tags on seabirds: how seriously are instrument-induced behaviours consid- ered? Animal Welfare-The UFAW Journal, 20(4): 559. Vegetabile, B. G., Stout-Oswald, S.A., Davis, E.P., Baram, T.Z. & Stern, H.S. (2019) Es- timating the entropy rate of Finite Markov Chains with application to behavior studies. Journal of Educational and Behavioral Sta- tistics, 44(3): 282-308. Weimerskirch, H. (2007) Are seabirds foraging for unpredictable resources? Deep Sea Re- search Part II: Topical Studies in Oceanog- raphy, 54(3-4): 211-223. Weimerskirch, H., Bertrand, S., Silva, J., Marques, J.C. & Goya, E. (2010) Use of so- cial information in seabirds: compass rafts indicate the heading of food patches. PloS one, 5(3): e9928. Weimerskirch, H., Wilson, R.P. & Lys, P. (1997) Activity pattern of foraging in the wander- ing albatross: a marine predator with two modes of prey searching. Marine Ecology Progress Series, 151: 245-254. Wickham, H. (2016) ggplot2: Elegant Graph- ics for Data Analysis. Springer-Verlag New York, 2016. Wilson, L.J., McSorley, C.A., Gray, C.M., Dean, B.J., Dunn, T. E., Webb, A. & Reid, J. B. (2008) Rafting behaviour of Manx Shearwaters Puffinus puffinus. Seabird, 21: 85-92. 162 CHAPTER 4 : Wilson, R.P. & Vandenabeele, S.P. (2012) Tech- nological innovation in archival tags used in seabird research. Marine Ecology Prog- ress Series, 451: 245-262. Wong, B. & Candolin, U. (2015) Behavioral re- sponses to changing environments. Behav- ioral Ecology, 26(3): 665-673. Yoda, K., Shiomi, K. & Sato, K. (2014) Forag- ing spots of streaked shearwaters in relation to ocean surface currents as identified using their drift movements. Progress in Ocean- ography, 122: 54-64. Yoda, K., Yamamoto, T., Suzuki, H., Matsumo- to, S., Müller, M. & Yamamoto, M. (2017) Compass orientation drives naïve pelagic seabirds to cross mountain ranges. Current Biology, 27(21): R1152-R1153. Supplementary material SUPPLEMENTARY FIGURES Fig S1: Segmentation of wet-dry data applying different time windows Fig S2: Selection of time window Fig S3: Values of activity variables mapped on the behavioural landscape Fig S4: Summary metrics plot per wet-dry activity metrics Fig S5: Temporal distribution of behaviours. Fig S6: Random forest variable importance Fig S7: Association between BCs and phonological stage (Pearson’s residuals of chi-square test of independence) Fig S8: Changes in the behavioural landscape of Cory’s shearwater over the breeding cycle: fuzzy clustering of pixels per stages Fig S9: Transition prob. Matrices Fig. 10: Individual actograms of year-round behavior of 8 Cory’s shearwater SUPPLEMENTARY TABLES Table S1: Tracking details and deployments Table S2: Summary table of network metrics results for each stage. Table S3: Results from the one-way repeated-measures ANOVA to test the effect of stage on amount of time invested in each behavioral cluster (BC) 163Supplementary material Fi g S1 : E ffe ct o f t im e w in do w se le ct io n on se gm en ta tio n of w et -d ry d at a fro m C or y’ s s he ar w at er , s ho w ed o n ex am pl e of o ne fo ra gi ng tr ip o f 1 7 da ys o f d ur at io n (s ee F ig . 1 a in th e m ai n te xt ). Ea ch fa ce t r ep re se nt s se gm en ta tio n us in g th e tim e w in do w s ho w ed a bo ve (i n se co nd s) , f ro m 1 m in to 2 4 h. S m al le r t im e w in do w re su lts in h ig he r n um be r o f s eg m en ts (N se g. o n th e le ft sid e of th e pl ot ). Th e ru g lin es o n th e bo tto m re fe r t o th e lo ca tio n of b re ak po in ts be tw ee n se gm en ts on th e tim e ax is. T he li gh tg re y ve rti ca l l in es re pr es en t d iv isi on b et w ee n da ys . F or a na ly sis p re se nt ed in th is w or k w e ch os e th e tim e w in do w o f 6 0 m in (3 60 0 s) . T o fa ci lit at e th e di sti nc tio n of se gm en ts, th ey a re c ol ou r-c od ed ra nd om ly . 164 CHAPTER 4 : Figure S2: Selection of time window used for segmentation of wet-dry cumulative time series. Each line repre- sents one track, GLS and GPS tracks pooled together. We used various time windows (from 1 min up to 24 h, see Fig. S1) to study the relation between the time window size applied and resulting number of breakpoints. Inset represent whole range of values. For simplicity of the presented approach, after visual examination we used 60 min time window to segment all tracks (darkgrey vertical dashed line in the plot). 165Supplementary material Figure S3: Values of activity variables mapped on the behavioural space, as default output from the “bigMap” R package. Each point represents a segment obtained from wet-dry behavioural dynamics data. Values of each variable are scaled to range from low (blue) to high values (red). Activity metrics: Prop.W = proportion wet, Dur.W = duration wet, Dur.D = duration dry, Nchanges = number of changes, Rchanges = rate of changes, Median.W = median wet duration, Median.D = median dry duration, SD.W = standard deviation wet durations, SD.D = standard deviation dry durations, Max.W = maximum wet duration, Max.D = maximum dry duration. 166 CHAPTER 4 : 167Supplementary material Figure S4 (previous page): Median values (black point and numerical) and interquartile range of 11 activity metrics and total duration of the segment (not included as input variable due to multi-collinearity) averaged over all segments for 10 behavioural clusters. Black point and numeric value refer to medians, point range to interquartile range. Coloured points in the background represent segments obtained from breakpoint algorithm applied to wet-dry data. Points are jittered for clarity. Note the variable range and logarithmic scale on y-axis. Note that rate of changes is expressed here as the number of changes per hour (N of changes h-1) to ease inter- pretability. A – J illustrates the different variable metrics. BCs semantics on the y axis: BC1 = SF, BC3 = StF, BC6 = TFLd, BC7 = CF, BC10 = ShD, BC2 = SRest, BC5 = ActSWD, BC8 = StlSWD, BC14 = Lsit, BC19 = Rest. Activity metrics: Prop.W = proportion wet, Dur.W = duration wet, Dur.D = duration dry, Nchanges = number of changes, Rchanges = rate of changes, Median.W = median wet duration, Median.D = median dry du- ration, SD.W = standard deviation wet durations, SD.D = standard deviation dry durations, Max.W = maximum wet duration, Max.D = maximum dry duration. Activity metrics: Prop.W = proportion wet, Dur.W = duration wet, Dur.D = duration dry, Nchanges = number of changes, Rchanges = rate of changes, Median.W = median wet duration, Median.D = median dry duration, SD.W = standard deviation wet durations, SD.D = standard deviation dry durations, Max.W = maximum wet duration, Max.D = maximum dry duration. 168 CHAPTER 4 : 169Supplementary material Figure S5 (previous page): Temporal distribution of behaviours of Cory’s shearwater derived from GLS and wet-dry immersion data. Each facet corresponds to one behavioural mode (i.e. behavioural cluster, BC). Each black line represents one behavioural segment, the length corresponds to the duration of the segment. The or- ange vertical dashed lines correspond to the average nautical twilight over the year. BCs in the left and right column correspond to mostly dry and wet segments, respectively. BCs semantics: BC1 = SF, BC3 = StF, BC6 = TFLd, BC7 = CF, BC10 = ShD, BC2 = SRest, BC5 = ActSWD, BC8 = StlSWD, BC14 = Lsit, BC19 = Rest. Figure S6: Variable importance plots of activity metrics used for identification of behavioural clusters (BCs) of Cory’s shearwater. Values reflect the mean decrease in accuracy (MDA), resulting from Random Forest models (see Methods for more details). Higher values indicate variables that contribute more to the identification of BCs. (A) Overall variable importance of 11 activity metrics used as input variables in the protocol to classify wet-dry activity segments into BCs (see Methods). (B) Case-wise variable importance for each BC, values were rescaled to range between 0-1 for each BC. BCs semantics: BC1 = SF, BC3 = StF, BC6 = TFLd, BC7 = CF, BC10 = ShD, BC2 = SRest, BC5 = ActSWD, BC8 = StlSWD, BC14 = Lsit, BC19 = Rest. Activity met- rics: Prop.W = proportion wet, Dur.W = duration wet, Dur.D = duration dry, Nchanges = number of changes, Rchanges = rate of changes, Median.W = median wet duration, Median.D = median dry duration, SD.W = standard deviation wet durations, SD.D = standard deviation dry durations, Max.W = maximum wet duration, Max.D = maximum dry duration. 170 CHAPTER 4 : Figure S7: Standardized Pearson’s residuals of chi-squared test of independence (chi.sq = 297.524, df = 18, p < 0.001) testing for association between behavioural clusters and phenological stage. Values below -2 and above 2 indicate significant association. BC1, BC2 and BC3 are highly associated with breeding stage (observed more than expected), BC5 and BC7 with migration and BC8, BC10 and BC14 with wintering. On the other side, BC10 and BC14 were observed less than expected during breeding, similarly BC1, BC8, BC10 and BC14 dur- ing migration and BC2, BC3, BC5 and BC7 during wintering. BCs semantics: BC1 = SF, BC3 = StF, BC6 = TFLd, BC7 = CF, BC10 = ShD, BC2 = SRest, BC5 = ActSWD, BC8 = StlSWD, BC14 = Lsit, BC19 = Rest. 171Supplementary material Figure S8: Changes in the behavioural space of Cory’s shearwater over the annual cycle (breeding, migration, wintering). (A) Colour of each cell of the grid represents the dominant stage of the annual cycle, estimated as the highest probability to belong to one of three stages: breeding (yellow), migration (purple), wintering (darkblue). Black triangles and numbers correspond to the peaks of 10 dominant behavioural clusters. (B - D) Density esti- mation (colour scale from low yellow to high purple) over the behavioural space of three stages: (B) breeding, (C) migration (pooled together postnuptial and prenuptial migration) and (D) wintering. Darkgrey lines delimi- tate the borders of 10 behavioural clusters. BCs semantics: BC1 = SF, BC3 = StF, BC6 = TFLd, BC7 = CF, BC10 = ShD, BC2 = SRest, BC5 = ActSWD, BC8 = StlSWD, BC14 = Lsit, BC19 = Rest. 172 CHAPTER 4 : 173Supplementary material Figure S9 (previous page): Transitions between behavioural clusters of Cory’s shearwater. Left panels (first column) reflects counts of transitions from and to behavioral clusters (BCs) for each stage of breeding cycle represented in rows: A) breeding, B) migration and C) wintering. Right panels (second column) reflects transi- tion probabilities from BC at time t to BC at t+1. BCs semantics: BC1 = SF, BC3 = StF, BC6 = TFLd, BC7 = CF, BC10 = ShD, BC2 = SRest, BC5 = ActSWD, BC8 = StlSWD, BC14 = Lsit, BC19 = Rest. 174 CHAPTER 4 : Fig. S10 (also in next three pages): Individual actograms of the year-round behavior of 8 individuals of Cory’s shearwater. Each coloured segment, of variable length, represents a behavioural mode (i.e. behavioural cluster, BC) identified from unsupervised clustering of wet-dry data from geolocator-immersion loggers. Each column represents one single day (0-24h). On the x-axis data starts on the day of deployment of the logger and ends on the day of recovery the next year. Black horizontal solid and dashed lines refer to time of local sunrise/sunset and nautical twilight at bird location, respectively. Vertical black-white lines delimit stages of annual life cycle: onset of post-breeding migration, arrival to the main wintering area, onset of pre-breeding migration and arrival to the breeding area. Individual and track identity, and corresponding breeding success are indicated in the top border. BCs semantics: BC1 = SF, BC3 = StF, BC6 = TFLd, BC7 = CF, BC10 = ShD, BC2 = SRest, BC5 = ActSWD, BC8 = StlSWD, BC14 = Lsit, BC19 = Rest. 175Supplementary material Fig. S10 176 CHAPTER 4 : Fig. S10 177Supplementary material Fig. S10 178 CHAPTER 4 : Table S1: Deployments of devices to 19 unique birds (Bird ID) of Cory's shearwater Calonectris borealis during the study at the breeding colony at Veneguera, Gran Canaria (Canary Is.). Abbreviations: GI – only geolocation- immersion logger deployed year-round, GI+GPS – geolocation-immersion logger and GPS device deployed during short-term foraging trip in incubation period. Nseg – number of segments after applying breakpoint algorithm (see Methods for details) on each track. Table is ordered by Bird ID within the Device used. Track ID Bird ID Device Start End Dura-tion N seg. VE6106933_10062011_25062011_c 6106933 GI + GPS 10/06/2011 25/06/2011 14,9 94 VE6134701_15072011_26072011_c 6134701 GI + GPS 15/07/2011 26/07/2011 11,0 124 VE6140497_14072011_24072011_c 6140497 GI + GPS 14/07/2011 24/07/2011 10,0 97 VE6140719_03072011_14072011_c 6140719 GI + GPS 03/07/2011 14/07/2011 11,2 70 VE6140719_23072011_30072011_c 6140719 GI + GPS 23/07/2011 30/07/2011 6,9 24 VE6140756_06062011_20062011_c 6140756 GI + GPS 06/06/2011 20/06/2011 14,7 114 VE6140855_03072011_19072011_c 6140855 GI + GPS 03/07/2011 19/07/2011 16,0 138 VE6143070_28062011_14072011_c 6143070 GI + GPS 28/06/2011 14/07/2011 16,0 115 VE6143090_05072011_17072011_c 6143090 GI + GPS 05/07/2011 17/07/2011 12,7 81 VE6143090_11062011_23062011_c 6143090 GI + GPS 11/06/2011 23/06/2011 12,0 84 VE6143093_08062011_22062011_c 6143093 GI + GPS 08/06/2011 22/06/2011 14,7 125 VE6175726_10062011_22062011_c 6175726 GI + GPS 10/06/2011 22/06/2011 12,7 117 VE6175730_04062011_18062011_c 6175730 GI + GPS 04/06/2011 18/06/2011 14,7 113 VE6175776_15072011_27072011_c 6175776 GI + GPS 15/07/2011 27/07/2011 11,1 83 VE6175776_19062011_05072011_c 6175776 GI + GPS 19/06/2011 05/07/2011 16,9 97 VE6175784_13072011_24072011_c 6175784 GI + GPS 13/07/2011 23/07/2011 10,7 88 VE6175784_20062011_01072011_c 6175784 GI + GPS 20/06/2011 01/07/2011 11,9 65 VE6188609_14072011_25072011_c 6188609 GI + GPS 14/07/2011 25/07/2011 11,1 108 VE6188705_07072011_18072011_c 6188705 GI + GPS 07/07/2011 18/07/2011 11,0 99 VE6195159_11072011_22072011_c 6195159 GI + GPS 11/07/2011 22/07/2011 11,7 109 VE6198153_03062011_16062011_c 6198153 GI + GPS 03/06/2011 16/06/2011 13,7 95 VE6198156_15062011_30062011_c 6198156 GI + GPS 15/06/2011 30/06/2011 15,8 115 VE6198172_08072011_17072011_c 6198172 GI + GPS 08/07/2011 17/07/2011 9,7 77 YR_6134701_19095002 6134701 GI 27/07/2011 17/04/2012 266 1504 YR_6140497_23353001 6140497 GI 26/07/2011 18/03/2012 237 1252 YR_6143070_23348001 6143070 GI 23/07/2011 25/02/2012 217 1044 YR_6143090_23319001 6143090 GI 19/07/2011 22/02/2012 218 1172 YR_6175726_19179002 6175726 GI 24/07/2011 14/04/2012 265 1543 YR_6175730_23326001 6175730 GI 29/06/2011 25/03/2012 270 1344 YR_6175776_19177002 6175776 GI 27/07/2011 11/04/2012 259 1456 YR_6175784_19089002 6175784 GI 25/07/2011 03/04/2012 253 1314 179Supplementary material Table S2: Network metrics results for the different stages of the annual cycle (breeding, migration, wintering). A) Global metrics of networks. Lower value indicates more stable networks. B-D) Local metrics at node level (i.e. by BC). Values per stage and BC of in- and out- degree centrality, closeness centrality and betweenness centrality. For each stage, the number indicates the BC identity and the number within parenthesis refers to the value of the metric. Degree centrality measures the number of edges (i.e. connections to other BCs) of each BC. Closeness centrality quantifies how close a BC is to all other BCs. Betweenness centrality measures the number of shortest paths that pass through each BC. In the case of equal number of edges for various BCs, they are listed by descending BC identity. S2.A. GLOBAL METRICS Breeding Migration Wintering Size 10 10 10 Diameter 2 2 2 Edge density 0.72 0.67 0.71 Av. path length 1.31 1.36 1.31 Entropy rate 2.21 2.09 2.23 S2.B. LOCAL (NODE) METRICS Degree centrality Breeding Migration Wintering in out all in out all in out all 10 (9) 10 (10) 10 (19) 7 (9) 7 (8) 7 (17) 7 (10) 7 (9) 7 (19) 7 (8) 7 (9) 7 (17) 10 (8) 10 (8) 10 (16) 10 (10) 10 (9) 10 (19) 3 (7) 5 (7) 5 (14) 6 (7) 6 (7) 6 (14) 3 (7) 1 (7) 3 (14) 5 (7) 6 (7) 6 (14) 19 (6) 1 (6) 1 (11) 1 (6) 3 (7) 1 (13) 6 (7) 3 (6) 3 (13) 1 (5) 5 (6) 5 (11) 6 (6) 6 (7) 6 (13) 1 (6) 19 (6) 19 (12) 2 (5) 2 (5) 19 (11) 2 (5) 2 (5) 2 (10) 19 (6) 1 (5) 1 (11) 3 (5) 3 (5) 2 (10) 5 (5) 5 (5) 5 (10) 2 (5) 2 (5) 2 (10) 5 (5) 8 (5) 3 (10) 8 (5) 8 (5) 8 (10) 8 (5) 8 (5) 8 (10) 8 (5) 14 (5) 8 (10) 14 (5) 14 (5) 14 (10) 14 (5) 14 (5) 14 (10) 14 (5) 19 (5) 14 (10) 19 (5) 19 (5) 19 (10) 180 CHAPTER 4 : S2.C. LOCAL (NODE) METRICS Closeness centrality Breeding Migration Wintering in out all in out all in out all 10 (0.90) 10 (1.00) 10 (1.00) 7 (0.90) 6 (0.82) 7 (0.90) 7 (1.00) 7 (0.90) 7 (1.00) 3 (0.82) 7 (0.90) 7 (0.9) 6 (0.82) 7 (0.82) 6 (0.82) 10 (1.00) 10 (0.90) 10 (1.00) 6 (0.82) 6 (0.82) 3 (0.82) 10 (0.82) 10 (0.82) 10 (0.82) 3 (0.82) 1 (0.82) 1 (0.82) 7 (0.82) 3 (0.75) 6 (0.82) 19 (0.75) 1 (0.75) 1 (0.75) 1 (0.75) 3 (0.82) 3 (0.82) 1 (0.75) 5 (0.75) 1 (0.75) 1 (0.69) 5 (0.75) 5 (0.75) 6 (0.75) 6 (0.82) 6 (0.82) 5 (0.75) 19 (0.75) 5 (0.75) 2 (0.69) 2 (0.69) 19 (0.75) 2 (0.69) 2 (0.69) 2 (0.69) 19 (0.75) 1 (0.69) 19 (0.75) 3 (0.69) 3 (0.69) 2 (0.69) 5 (0.69) 5 (0.69) 5 (0.69) 2 (0.69) 2 (0.69) 2 (0.69) 5 (0.69) 8 (0.69) 3 (0.69) 8 (0.69) 8 (0.69) 8 (0.69) 8 (0.69) 8 (0.69) 8 (0.69) 8 (0.69) 14 (0.69) 8 (0.69) 14 (0.69) 14 (0.69) 14 (0.69) 14 (0.69) 14 (0.69) 14 (0.69) 14 (0.69) 19 (0.69) 14 (0.69) 19 (0.69) 19 (0.69) 0.69) S2.D. LOCAL (NODE) METRICS Betweenness centrality Breeding Migration Wintering 10 (4.58) 7 (4.13) 10 (4.79) 7 (3.74) 1 (3.80) 7 (4.79) 1 (3.60) 3 (3.80) 1 (4.00) 3 (3.60) 6 (3.80) 3 (4.00) 6 (3.60) 10 (3.80) 6 (4.00) 2 (1.78) 2 (2.53) 2 (1.29) 5 (1.78) 5 (2.53) 5 (1.29) 8 (1.78) 8 (2.53) 8 (1.29) 14 (1.78) 14 (2.53) 14 (1.29) 19 (1.78) 19 (2.53) 19 (1.29) 181Supplementary material BC BC description Df stage Df resid Within variance Between variance F p-value 1 SF 2 12 0.046 0.084 1.830 0.202 3 StF 2 13 1.143 8.461 7.400 0.007 6 TFLd 2 13 1.878 3.920 2.087 0.164 7 RFLd 2 13 82.752 1530.475 18.495 < 0.001 10 ShD 2 13 4.879 11.720 2.402 0.130 2 SRest 2 13 0.827 0.331 0.400 0.678 5 ActSWD 2 13 8.893 0.282 0.032 0.969 8 StlSWD 2 13 28.179 266.663 9.463 0.003 14 Lsit 2 13 75.543 855.911 11.330 0.001 19 Rest 2 13 10.138 0.623 0.061 0.941 Table S3: Results from the one-way repeated-measures ANOVA to test the effect of stage on the amount of time invested in each behavioral cluster (BC) by each of 8 individuals of Cory's shearwater tracked by wet-dry im- mersion loggers. BCs where significant effect was found are marked in bold. GENERAL DISCUSSION This thesis comprises 4 chapters addressing different topics in the context of seabird ecology. Every chapter provides new insights into the factors shaping the at-sea behaviour of pelagic seabirds from the Atlantic Ocean. Through this thesis I have shown and highlighted the usefulness of wet-dry data, a source of information that can greatly enrich our knowledge of seabird ecology in a diversity of dimensions. In Chapters 1, 2 and 3, I and my co-authors provide new insights about at-sea ecology of little-known seabird species so far, reporting year-round movements and migratory schedules. In Chapters 2 and 3 we further revealed differences in activity budgets between different groups (males vs females, successful vs failed breeders, respectively), discussing the causes and conse- quences of these differences. In the Chapter 4, we revealed the complexity of seabird behaviour, and at the same time we presented a set of new analytical techniques and data visualization tools that allowed us to get the most from wet-dry data, which may open new avenues to understand the complexity of seabird behavioural patterns from manifold perspectives. Using geolocators in seabird research Understanding movements and spatial ecology over the annual cycle is nowadays the most studied topic in seabird research using geolocators (see Box 1 in the Introduction of this thesis). But before entering in the era of biologging, movements and distribution of many seabirds at sea were mostly studied from shipboard or coastal observations (Louzao et al. 2006, Ballance 2007). However, this approach does not inform about intrinsic factors of the individuals observed, such as origin or breeding status, precluding to properly interpret movement, behaviour and seasonal timing at indi- vidual level. Paraphrasing Nathan et al. (2008), two of the main fundamental questions regarding causes and consequences of animal movement are to know where and when the animals go. Moreo- ver, these questions are also fundamental to address conservation actions (Lascelles et al. 2016). Therefore, the use of tracking devices is essential to understand seabirds’ movements and timing of life-history events. The increasing miniaturization of geolocators has allowed to progressively address the track- ing of medium-to-small sized species, overcoming the initial bias towards research carried out on large-sized seabirds such as albatrosses (e.g. Weimerskirch & Wilson, 2000). In the last decade a numerous research undertaken uncovered spectacular migrations and non-breeding areas of several species (e.g. Shaffer et al. 2006, González-Solís et al. 2007, Egevang et al. 2010). Yet basic informa- tion regarding migratory movements and seasonal timing is still lacking for a few seabird species (Grémillet & Boulinier, 2009), including medium-to-small sized species from polar to temperate to tropical regions, such as the species studied in this thesis. Seabirds’ year-round movements and seasonal timing of life-history events can be inferred from positional and wet-dry data Along this thesis I have shown that wet-dry data from geolocator-immersion loggers provide a 184 powerful source of information to describe not only the movements but also the seasonal timing of life-history events. Our articles contribute to the published literature supporting the usefulness of wet-dry data to this end (e.g. Hedd et al. 2014, Rayner et al. 2016, Militão et al. 2017). This utility is even more relevant for elusive species and for those where location of breeding sites impedes periodic on-site nest monitoring. In this thesis, I and my co-authors have provided new insights about the year-round movements and the seasonal timing of life-history events in three little-studied seabird species of medium-to-small body size: Boyd’s shearwater, Atlantic petrel and Common tern (Chapters 1 to 3). For many medium-to-small sized seabirds, identification of closely related species at sea is chal- lenging or even unreliable (Ballance 2007), precluding to know their actual distribution, even at broad scale, just from at-sea observations from vessels. One example in this regard is the Little− Audubon’s shearwater complex, which species are hard to distinguish at sea (Flood & van der Vliet 2019). Some of them, such as the Boyd’s shearwater, are distributed in the tropical latitudes in the North Atlantic Ocean. In Chapter 1 I revealed for the first time the year-round movements of this little-known tropical seabird using geolocators. This study allowed us to place on the map the wintering areas and migratory routes of the Boyd’s shearwater. Unexpectedly, the findings were quite contrary to the movements previously known from closely related species. Barolo shearwaters from the Azores and Salvagens Islands (Neves et al. 2012, Paiva et al. 2016) disperse mostly in the vicinity of their breeding colonies and forage also over the rich cool waters of the African shelf. Similarly, Audubon’s shearwaters breeding in Caribbean archipelagos forage year round over the continental shelf (Precheur 2015). In contrast, our results revealed that Boyd’s shearwaters spent their non-breeding season in oligotrophic waters in the centre of the Atlantic Ocean. In addition to the positional data, I could also define the timing of major life-history events thanks to wet-dry data. The timing of important life-history events, such as migration or arrival to the breeding colonies, among others, is influenced by many different factors, both intrinsic (e.g. sex, breeding status) and extrinsic (e.g. habitat seasonality, inter-annual environmental variability) (Cubaynes et al. 2010, Vo- tier et al. 2009, Keogan et al. 2018). In this regard, we observed high inter-individual variability in the timing of various aspects of breeding biology of Boyd’s shearwaters. This variation might arise from the marked seasonal and inter-annual variability in the abundance of food resources in tropical oligotrophic waters (Catry et al. 2013a, Hennicke & Weimerskirch 2014). This may lead to differ- ences in the breeding success among individuals and years, which can influence the onset of post- breeding migration (Catry et al. 2013b, Ramos et al. 2018). Analysing wet-dry data in depth would provide important insights in this regard, as we did in the study of the Atlantic petrel (see below). Another example of closely related species difficult to distinguish at sea are the gadfly petrels (Pterodroma sp.). They are medium-sized seabirds with a broad distribution across the world oceans, and their movements at sea have remained unknown until recently (Ramos et al. 2017). In Chapter 3 we analysed in detail geolocator data from the Atlantic petrel, a gadfly petrel endemic as breeder to Gough Island and Tristan da Cunha archipelago. We provided a detailed description of year-round movements and spatial distribution of tracked individuals thanks to the combined use of positional and wet-dry data. As in our study with Boyd’s shearwaters, I also found high 185GENERAL DISCUSSION inter-individual variability in the timing of events related to breeding biology, which again made me suspect that such variability may arise from breeding success. In fact, Atlantic petrels suffer an extremely high rate of breeding failure due to predation of chicks by the introduction of house mice (Mus musculus) in Gough Island (Caravaggi et al. 2019). As we did not have information regarding the breeding success from nest monitoring, I used exploratory data visualization to evaluate breed- ing success based on phenology, and classified the individuals as presumed successful and failed breeders using multivariate clustering. It could be expected that movement, behaviour and timing of major life-history events differ according to breeding output (Catry et al. 2013b). Indeed, for each group (successful and failed) we found the annual timing of these events to correlate in time, i.e. failed breeders advanced their post-breeding migration, stayed longer in the wintering areas, and returned earlier to the breeding colony in the next breeding stage. However, we did not find differ- ences in spatial distribution, as all individuals wintered in the same area on the South American shelf slope. These results may suggest carry-over effects at some extent in relation to breeding suc- cess, as reported for other seabird species (Catry et al. 2013b, Schultner et al. 2014, Shoji et al. 2015, Fayet et al. 2016, Ramos et al. 2018). The impact of invasive species on seabirds is well known (Dias et al. 2019). However, our results also suggest that predation and subsequent breeding failure could also affect the timing of life-history events over the annual cycle mediated by carry-over effects, an unexpected impact of invasive species on seabirds yet not reported in literature and that should receive more attention in the future. Unlike the species discussed previously, some other species have more coastal habits, which make them easier to observe and study, both using direct observation and ringing. This is the case of Common terns, for which ring recoveries pointed out the importance of West African coast during winter for individuals breeding in Europe (Wernham et al. 2002, Bairlein et al. 2014). Nevertheless, ring recoveries and coastal observations alone cannot inform about year-round movements and behaviour in detail, and therefore these aspects have remained unknown so far. In Chapter 2 we unveiled for the first time the timing of migration along the East Atlantic flyway and the importance of the West African Coast for Common terns breeding in continental Europe, throughout the use of geolocators. Moreover, using wet-dry data we showed that wintering habitat in Common terns differs between sexes, as females tended to winter in more offshore areas, contrary to males winter- ing nearby the coastline. Since males usually care the offspring during migration and might be at wintering sites (Nisbet et al. 2011), differences in wintering habitat would be reflecting constraints related to sex and parental care. We also found that pairs did not overlap in their wintering areas. Sexual segregation in wintering areas has been reported for several large-to-medium sized seabirds using tracking devices (e.g. González-Solís et al. 2007), but our study is one of the few reporting such segregation in small sized seabird species, if not the unique so far, to the best of my knowledge. Circadian and circa-annual at-sea activity rhythms of seabirds In this thesis, we have explored how seabirds adjust their activity budgets and change their behav- iour in response to biological and environmental constraints. The different constraints over the an- nual life cycle should be reflected in behavioural budgets and activity rhythms (Phillips et al. 2017). 186 Wet-dry data have been used by many researchers to evaluate this expectation (see Box 1 in Intro- duction of this thesis). In this thesis we have confirmed that individuals adapt their activity budgets over the different stages of the annual cycle. We verified this expectation in Chapter 2, Chapter 3 and Chapter 4. In the case of Common terns (Chapter 2), we found that inter-individual variabil- ity in post-breeding (autumn) migration but especially during pre-breeding (spring) migration was much lower than during winter (see Fig. 3 in Chapter 2). Similarly, in the case of Cory’s shearwater (Chapter 4), we found inter-individual variability in those behaviours where most time was invested to be overall much lower during migration than during winter (see Fig. 5 in Chapter 4). In the case of Atlantic petrels (Chapter 3), we did not account directly to inter-individual variability, but time spent on water was overall much higher during wintering than during migration (see Fig. 3 in Chap- ter 3). Thus, the findings in the three species support the importance of phenology in constraining individual behaviour, shaping circa-annual at-sea activity rhythms. During breeding, central place foraging and particularities of each species also shaped behavioural budgets, complicating possible comparisons. On a daily basis, Common terns showed circadian activity rhythms that varied across the stages of the annual cycle (see Fig. 4 in Chapter 2). The same occurs for Boyd’s shearwaters. De- spite in Chapter 1 and the related published article we did not include an explicit analysis of wet-dry data, I carried out such analysis independently for illustrating purposes, and found a similar pattern (see Fig. 1 below). Therefore, we found clear evidences that activity budgets are shaped by circadian and circa-annual rhythms. Wet-dry data for behavioural annotation reveal the complexity of behavioural organization in seabirds It is obvious that a greater capacity to interpret behavioural patterns will allow us to link behav- ioural strategies with the rest of individuals’ traits, enhancing our understanding about the causes and consequences of behavioural decisions within the life-history of animals (Sih et al. 2010). Along this thesis I have highlighted the usefulness of wet-dry data to decipher behavioural patterns. How- ever, I have remarked in the Introduction that most studies have used raw data to quantify duration of a state (wet or dry) (e.g. Phalan et. 2007, Mackley et al. 2010, Dias et al. 2012, Rayner et al. 2012), without accounting for the inherent temporal correlations contained in the structure of wet-dry data, which is in fact a valuable information to infer behaviours (e.g. Phalan et. 2007, Mackley et al. 2010, Dias et al. 2012, Rayner et al. 2012). At most, some studies using this source of data were limited to identify the basic behaviours ‘foraging’, ‘flying’ and ‘sitting on water’ (Guilford et al 2009, Dias et al. 2012, Gutowsky et al. 2014, Ponchon et al. 2019). Guilford et al. (2009) used unsupervised clus- tering using solely wet-dry data but aggregating data on predefined daily blocks. Even in this thesis, in Chapter 2 and Chapter 3 we also used an approach aggregating information in predefined blocks (stage, day, etc.) to calculate the amount of time birds spent on water or in flight. Thus, despite its po- tential utility to distinguish behaviours, wet-dry data have never been used so far to annotate more complex behaviours. In Chapter 4 we filled this gap, extending the use of wet-dry data for behav- ioural annotation. We took advantage of machine learning (Valletta et al. 2017) to analyze within a multidimensional unsupervised framework an array of metrics derived solely from wet-dry data. Multidimensionality reduction techniques allowed us to map samples on a behavioural space and 187GENERAL DISCUSSION Fig. 1. This figure illustrates activity budgets of Boyd’s shearwaters on a daily basis and for the different stages of the annual cycle, supporting that birds change the behaviour according to circadian and annual rhythms. Analysis performed with wet-dry data from 37 individuals from Raso and Ilheu de Cima (Cape Verde), tracked with geolocator-immersion loggers. identify 10 different behaviours based on wet-dry data, thus surpassing the basics ‘foraging’, ‘fly- ing’ and ‘sitting’ behaviours. However, the greater the number of behaviours, the more difficult is their interpretation. We initially found 23 behaviours, but we grouped them later in 10 by proximity, facilitating their interpretability. We used a combination of data visualization and statistical tools to interpret and give semantics to each behaviour. Hence, it is important to underline that interpreta- tion may not be trivial and that a good knowledge on biology of the model species may be required for a successful depicting of behaviours. Quantifying complexity of seabirds’ behavioural strategies Studying how individuals allocate their time budgets to different behaviours is an important key- stone in ecology, as it could enhance the interpretation of behavioural strategies of animals under variable conditions within a life-history context (Sih et al. 2010, Wong & Candolin 2012, 2015). In the Introduction of this thesis I exposed how wet-dry data have been used to quantify activity bud- gets in seabirds. I also highlighted that most research intended to analyze activity patterns typically settled for aggregated wet-dry data at different scales (daily, monthly, by day/night). In that way, 188 behavioural budgets are too broad to allow a proper and detailed investigation of constraints, causes and consequences of behavioural strategies. In contrast with previous research, in Chapter 4 we did not quantify broadly the foraging effort but went beyond, first identifying a variety of behaviours related to foraging, flying and resting and later quantifying their relative importance within the behavioural budgets over time. This quantifi- cation provided a variety of insights about the variability of behavioural strategies over the annual cycle and its constraints. For example, we found that most dominant behaviours in terms of time invested are also those more constrained by phenology, limiting inter-individual variability and thus shaping behavioural patterns at population scale. These insights open the door to consider differences in detailed behavioural budgets when comparing populations within meta-population approaches (Frederiksen et al. 2012, Ramos et al. 2013, Dean et al. 2015) or even to compare closely related species (Ramos et al. 2017), enriching the repertoire of possible ecological dimensions ana- lyzed to evaluate aspects such as competition or diversification. Moreover, we inspected the results from a multidimensional view, tackling individual variabil- ity trough time, i.e. seasonal variability, circadian and circa-annual rhythms, by using actograms. Actograms potentially allow linking behavioural budgets and strategies trough time with a whole range of intrinsic (age, sex, breeding status, breeding timing, breeding success, migration strate- gies, moulting strategies, etc.) and extrinsic factors (photoperiod, moonlight phase, environmental seasonality, etc.). Similarly, carrying out detailed behavioural annotation of positional data (e.g. GPS tracking, see Fig. 6 in Chapter 4) enhance our ability to interpret the role of extrinsic factors in behavioural budgets and strategies. In this regard, our approach for behavioural annotation at fine scale or at large spatio-temporal scale could be combined with currently available tools for track annotation, i.e. merging trajectories with environmental data provided from a variety of satellite- derived data sources (Kemp et al. 2012, Dodge et al. 2013, White et al. 2019) in order to understand the actual landscape faced by the birds. While this has been addressed at some extent using other devices and sources of data with some species (Vansteelant et al. 2017, Dodge et al. 2014), to date no research has been addressed in such detail using year-round data of pelagic and diving seabird spe- cies, since detrimental effects of long-lasting device deployment methods impede their use in such species. It is well known that environmental drivers such as wind (González-Solís et al. 2009) or human activities such as fisheries (Bartumeus et al. 2010) shape behaviour of seabirds. Therefore, the use of our approach for behavioural annotation combined with environmental data will assist to decipher the role of extrinsic factors such as marine habitat, wind direction and speed, food availability or fishing activity, among others, in shaping year-round behaviour of seabirds in an unprecedented detail (Obringer et al. 2017). As commented above, to interpret every behaviour and give semantics we inspected results from manifold perspectives. Inspecting data through actograms allowed us to notice one behaviour likely corresponding to moult. The state of plumage and feathers depend on a physiological process decisive for seabirds’ fitness (Cherel et al. 2016, Weimerskirch et al. 2019. As evidences of moulting have been rarely addressed explicitly in seabird tracking studies (Cherel et al. 2016), our 189GENERAL DISCUSSION results open the door to inspect in depth the impact of moulting strategies on behaviour, including the identification of the moulting period and moulting areas in different species. As highly gregarious species, social interaction plays an important role for seabirds (Gaston 2004). For example, bearing of individuals departing from waters surrounding the colony seems to be used a cue by conspecifics to head towards foraging grounds (Weimerskirch et al. 2010). Behaviours displayed by individuals in the proximity of breeding colonies, such as rafting (Wilson et al. 2008, Carter et al. 2016) or bathing (Granadeiro et al. 2018) can be important for social interaction. Thus, our protocol for behavioural annotation and visualization may shed light on the way individuals’ behaviour relates with social interaction in the vicinity of the breeding colonies. Among seabird species, we can find a complete spectrum from mostly diurnal to those mostly nocturnal species. This variability is related to different strategies of foraging. Several species of al- batross and petrels, such as the Boyd’s shearwaters or the Atlantic petrel, are more active at twilight or at night (Shealer 2002). This has been related to availability of their potential prey, which become more available at night during dial vertical migrations (Elliott & Gaston 2015). On the other side of the day-night spectrum, we may find seabird species mostly restricted to daylight foraging activi- ties, such as Cory’s shearwaters or Common terns, which greatly rely on vision to localize prey (Fauchald 2009). Within seabird species, nocturnal/diurnal behaviour can also change across time and space, such as the Atlantic petrel, according to prey availability (Regular et al. 2011, Dias et al. 2012). These changes can be easily detected by plotting the time spent on water during the daylight and darkness over the different stages of the annual cycle (see for example Fig 1 in this Discussion or Fig. 3 in Chapter 3). Nevertheless, behavioural annotation and actograms may allow to evaluate the nocturnal/diurnal behaviours from a richer perspective, showing for instance how a same behaviour could be displayed during daylight or darkness depending on seasons or how specific behaviours are more displayed in oceanic environments during darkness, indicating specific foraging tactics to take advantage of resources more available at night (Regular et al. 2011, Krüger et al. 2017), while others are more exhibited during daylight in neritic waters, indicating foraging tactics relying on vision (Collet et al. 2015). Data visualization and network analysis in animal movement and behavioural ecology Data visualization is an essential part of the scientific process, although many times only remains restricted to report results in hypothesis-driven studies. Nevertheless, data visualization should become an important part of the scientific process, also in movement and behavioural studies. Ap- propriate data visualization may lead to the discovery of new patterns in data, thus promoting the generation of new hypothesis previously not “visible” (Williams et al. 2019). Recently, increasing interest in visual movement analysis have led researchers to gather at specific workshops (Shamoun- Baranes et al. 2011), to promote an interdisciplinary research network (Demšar et al. 2015) or to cover the topic in the special section of a journal (Demšar et al. 2019). Moreover, several innovative tools for exploration and visualization of animal movement from tracking devices have been devel- oped (Kavathekar et al. 2013, Slingsby & Van Loon, 2016, Dodge et al. 2018, Konzack et al. 2019, 190 Schwalb-Willmann 2019). Similarly, in visualization of behaviour, new techniques have been devel- oped to ease the visualization and enhance the understanding of behaviours inferred from complex data such as that from accelerometers or magnetometers (Grundy et al. 2009, Williams et al. 2017). Within studies presented in this thesis, data visualization has represented an essential tool, applied at various aspects along the research process, from positional data to phenology to wet-dry data analysis. As it has been mentioned previously, the estimation of positions from GLS comes inher- ently with a certain error, variable especially around the equinoxes and in tropical regions. Even though the newly emerged analytical tools may significantly increase the accuracy of positional estimates from geolocation and therefore minimize the errors (Lisovski et al. 2019), it is always es- sential to pair these analyses with additional data visualizations. In Chapter 1, we initially inferred from positional data that the distribution of Boyd’s shearwaters during the incubation and chick- rearing period shifted from north to south, respectively. However, the visualization of longitudinal and latitudinal positions as time series indicated clear effect of equinoxes and therefore prevented us to come up with misleading conclusions (see Fig. S1-S3 in Supplementary Material of Chapter 1). In Chapter 3 data visualization of phenology at individual level together with longitudinal move- ments of Atlantic petrels drove our hypothesis of the possible existence of two different groups, which we later classified and related with successful and failed breeders. Furthermore, data visual- ization helped us to consider the likely existence of carry-over effects related to breeding success. In Chapter 4 I presented effective visualizations of wet-dry data and inferred behaviours. Acto- gram plots have been used for many years in ethology and chronobiology to present behavioural or activity rhythms of animals from behavioural studies in laboratory (Aschoff 1979, Numata & Helm 2014). However, new detailed information obtained from biologging devices currently allow us to record, visualize and analyse information in much more detail even over long periods of time (Zúñiga et al. 2016, Bäckman et al. 2017). I used actogram plots to represent time series of behav- iours inferred from wet-dry data on daily and seasonal scales at the same time. Actograms allow for visualization of behavioural budgets and behavioural strategies (i.e. how the different behaviours are arranged over time) simultaneously. I acknowledge that visualizing raw wet-dry data already provides us with several hints indicating changes in behaviour (see Box 2). However, by applying the protocol proposed in Chapter 4 and visualizing inferred behaviours using actograms, we can explore in much more detail circadian and circannual rhythms in behaviour of seabirds at individual level, which could greatly help to infer the biological meaning of the different behaviours and deci- pher the different constraints that shape them. Similar visualizations illustrating detailed daily and seasonal activity or movement based on data from biologging devices can be find in several studies, i.e. flight and activity of swifts (Liechti et al. 2013), activity of lynxes (Heurich et al. 2014), flight behaviour of shrikes (Bäckman et al. 2017) or as spatial chronogram of fishes (Aspillaga et al. 2016). Freeman et al. (2013) visualized time invested in 3 behaviours (foraging, resting, flight) of seabirds year-round, yet showing only daily aggregates of those basic behaviours. To my knowledge, the work I present in Chapter 4 is the first study where a diverse array of behaviours has been presented and visualized in such detail. 191GENERAL DISCUSSION Visualizing behaviour in a spatially explicit framework is particularly useful to study animals that move over vast areas and present different space-use (Papastamtiou et al. 2018, Dodge et al. 2018). Moreover, their behaviour may change over time even when remain in the same area. Therefore, not only it is important to know where the animals go and how they use the space (which is commonly addressed using different approaches of kernel density estimations), but also in which behaviours they mostly engage across the areas used. According to this, in Chapter 4 we constructed spatially explicit behavioural landscapes as a data visualization tool, based on behavioural annotation of trajectories from behaviours inferred from wet-dry data (see Fig. 9 in Chapter 4). These “activity seascapes” or “behavioural seascapes” mapped the spatial distribution and prevalence of behaviours over the annual life cycle of tracked seabirds. Some other approaches have been carried out to vi- sualize behaviour in a spatial framework. For example, some authors used multi-sensor data from Manx shearwater (Puffinus puffinus) and kernel density estimation to point out regions were birds mostly engaged in one of three basic behaviours (foraging, resting and flying, Guilford et al. 2009, Freeman et al. 2013). However, when a great spatial overlap between those basic behavioural modes occurs, such approach does not achieve an effective visualization to highlight important areas for each behaviour. Our approach copes with a great number of behaviours and can highlight at global scale prevalent behaviours at each region and stage of the annual cycle, assisting for a quick iden- tification of areas important for different foraging modes, moulting, commuting within migratory flyways or refuelling at stop-over sites (see Fig. 7 and 9 in Chapter 4). Other authors have focused to visualize the rate of nocturnal versus diurnal activity, through calculating the so-called “night flight index” using wet-dry data (Dias et al. 2012, Ramos et al. 2015). We could go beyond with our method and visualizations, since through disaggregating by day and night we could for example differentiate areas of nocturnal foraging, which kind of nocturnal foraging behaviour birds use, or evaluate whether those areas overlap with resting areas, at both coarse and fine scale (see Fig. 6 and 9 in Chapter 4). Lastly, since our approach can cope with high diversity of behaviours, it allows for mapping even behavioural diversity at global scale (see Fig. 9 in Chapter 4), which might provide a tool to evaluate the capacity of populations to cope with changing environments (Wong & Candolin 2015). Other tool widely used in behavioural studies is network visualization and analysis. It is well established in biology and ecology, in contexts encompassing studies of social interactions (Hasen- jager & Dugatkin 2015), molecular biology (Barabási & Oltvai, 2004), trophic dynamics and inter- actions (Bascompte et al. 2003, Oshima & Leaf 2018) and space-use (Jacoby et al. 2012, Stehfest et al. 2013). Its potential in the field of movement ecology has been acknowledged by some authors (Jacoby & Freeman, 2016). Moreover, in behavioral studies, the transition rates between described behaviours are also visualized as network graphs (Dankert et al. 2009, Berman et al. 2016) or as transition matrices (Dragon et al. 2012, Chimienti et al. 2016). However, the analysis of the proper- ties of behavioural networks is not so settled, even despite it may reveal in more detail the structure of behavioural strategies and changes in their organization and complexity (Bradbury & Vehren- camp 2014, Todd et al. 2017). Thus, in Chapter 4, taking advantage of the variety of behaviours raised from our protocol, I explored the analysis of network properties and network visualization of relations (i.e. transitions) between inferred behaviours. This approach allowed me to reveal that 192 changes in the prevalence of certain transitions and behaviours are shaped by breeding stages, but the method could be easily applied to compare behavioural strategies between individuals from dif- ferent groups such as sexes, ages, populations or species (Stauss et al. 2012, de Grissac et al. 2017, Mendez et al. 2017, Ramos 2017). BOX 2: From heatmaps to actograms: a visualization journey through complexity in wet-dry data The structure of wet-dry data recorded by geolocation-immersion sensors has change over time according to the models developed. Models manufactured by the British Antarctic Sur- vey were widely used, most of them recording wet-dry data in 0-200 schedule. Later models, provided for example by Biotrack Ltd. and also widely deployed on seabirds, record changes in wet-dry state in a continuous way. Figures included in this Box illustrate the extent of in- ference about individual behaviour that we could achieve from wet-dry data. Fig. B1 and Fig. B2 are heatmaps displaying wet-dry states over a year-round cycle from raw data recorded in 0-200 format. Note that Fig. B1 corresponds to an individual of Boyd’s shearwater and Fig. B2 to an individual of Cory’s shearwater. Fig. B3 is a replica of the actogram previously shown (Fig. 8 in Chapter 4), where behavioural annotation on a year-round trip of Cory’s shearwater is represented. Fig. B4 shows wet-dry data recorded in continuous format corresponding to a short-term foraging trip of Cory’s shearwater. Finally, for comparison purposes, Fig. B5 refers to the same individual foraging trip than Fig. B4 but illustrates an actogram after applying our method exposed in Chapter 4. Data visualization highlights the insight enrichment accom- plished with our behavioural annotation method. 193GENERAL DISCUSSION Fig. B1. Heatmap illustrating year-round activity patterns from an individual of Boyd’s shearwater. Data come from geolocation-immersion loggers recording wet-dry data in 0-200 format. Each pixel corre- sponds to a 10-minutes block. The colour code from yellow to blue correspond to the scale from totally dry to totally wet. The data start on the date of logger deployment at the end of breeding stage and end on the date of logger recovery in the next breeding season. We can clearly identify prolonged periods in dry state (yellow) as colony attendance events: diurnal and nocturnal visits and incubation stints, which identification is essential to define timing of annual cycle life-history events. 194 Fig. B2. Heatmap illustrating year-round activity patterns from an individual of Cory’s shearwater. Data come from geolocation-immersion loggers recording wet-dry data in 0-200 format. Each pixel corre- sponds to a 10-minutes block, so the colour scale codes from totally dry (yellow) to totally wet (blue). The data start on the date of logger deployment at the end of breeding stage and end on the date of logger recovery in the next breeding season. We can clearly observe a circadian pattern (0-24h on y-axis), as birds adjust their activity according to local sunrise and sunset. 195GENERAL DISCUSSION Fig. B3. Actogram illustrating year-round behavioural patterns from an individual of Cory’s shearwa- ter. Raw data come from geolocation-immersion loggers recording wet-dry data in continuous format. We applied on such data our method for behavioural annotation explained in Chapter 4, so each colour represents a different inferred behaviour. Each column represents one single day (0-24h). On the x-axis data start on the day of deployment of the logger and end on the day of recovery the next year. Black horizontal solid and dashed lines refer to time of local sunrise/sunset and nautical twilight at bird location, respectively. Vertical black-white lines delimit stages of annual life cycle: onset of post-breeding migra- tion, arrival to the main wintering area, onset of pre-breeding migration and arrival to the breeding area, respectively. Note, for example, a clear shift in the circadian rhythm in December, indicating that the bird arrived at the wintering area -in this case in South-African waters-, and adjusted the behaviour to local sunrise and sunset times. 196 Fig. B4. Wet-dry activity patterns during a short-term foraging trip of a Cory’s shearwater. Raw data come from geolocation-immersion loggers recording wet-dry data in continuous format. Each column repre- sents one single day (0-24h). On the x-axis data start with the bird leaving the colony for a foraging trip and end when the bird returns to the colony 15 days later. In contrast with 0-200 wet-dry heatmaps, this actogram illustrates the actual changes between states, but getting insights in terms of behaviour based on frequency, duration and time sequence of states is an arduous task. 197GENERAL DISCUSSION Fig. B5. Actogram illustrating behavioural patterns during a short-term foraging trip of a Cory’s shear- water, the same trip than the one illustrated in Fig. 4. Raw data come from geolocation-immersion log- gers recording wet-dry data in continuous format. We applied on such data our method for behavioural annotation explained in Chapter 4, so each colour represents a different inferred behaviour. Each column represents one single day (0-24h). On the x-axis data start with the bird leaving the colony for a foraging trip and end when the bird returns to the colony 15 days later. As we showed in Chapter 4, interpretation of behaviours arisen from our protocol allowed us to give semantics to each of them. In this actogram we can visualize in which behaviour the bird engaged along the trip and how it allocated the time between behaviours. 198 Concluding remarks and future directions Along the four chapters of this thesis I have provided new analytical and visualization approaches to show how wet-dry data can bring new insights in several dimensions of seabird ecology. A very ba- sic source of data, the wet-dry data recorded by geolocator-immersion loggers, can assist to under- stand movements and behaviour at multiple scales. Despite its utility, wet-dry data are underused by the seabird research community. In this thesis, I have presented a new protocol using state-of-the-art analytical techniques to show the incredible extent to which wet-dry data can help us to understand behavioural patterns of elusive species. Moreover, as it is based on multidimensional techniques, the protocol is also suitable for other sources of data. That is, it could also handle multi-sensor data together with external sources of data such as, for example, environmental annotation. For example, we could address which behaviours the different individuals display, how they differently arrange behavioural budgets and how their behavioural strategies change throughout life stages and in dif- ferent environmental contexts. At population level, we could evaluate whether a more diverse array of behaviours exhibited in population relates to a greater resilience to face changes in the environ- ment (Wong & Candolin 2015, Beever et al. 2017). At species level, we could study in detail whether distinct species arrange their behaviours to avoid competition without spatial or temporal segrega- tion. Our framework paves the way to use behavioural annotations for addressing a repertoire of old and new questions of interest in ecology from new perspectives, and from individuals to popula- tions to species level. Considering geolocator-immersion sensors continue to be the most extended loggers to track year-round movements of seabirds, and based on results compiled in this thesis, I encourage researchers to incorporate the use of wet-dry data within hypothesis-driven frameworks, which would surely contribute to increase our knowledge of seabird ecology at sea. CONCLUSIONS 1. Wet-dry data from geolocator-immersion loggers constitute a powerful and irreplaceable source of information to study movement, at-sea behaviour and timing of life-history events of seabirds, but such data have been largely underused by the seabird research community. 2. Analysis of wet-dry data clearly highlights the circadian and circa-annual rhythms of behav- iour. Wet-dry data allowed us to verify that migratory species adjust their internal biological clock to local conditions. We found evidences across four different seabird species with con- trasting migratory patterns and spread over the Atlantic Ocean. 3. Wet-dry data enhance our ability to describe timing of major life-history events. We revealed the previously unknown phenology of two pelagic seabird species that perform longitudinal migratory movements, Boyd’s shearwater and Atlantic petrel. 4. Behavioural patterns are shaped by a diverse array of intrinsic factors (age, sex, breeding status, breeding timing, breeding success, migration strategies, moulting strategies, etc.). We found sex to condition behaviour in Common tern, and breeding success to influence year-round be- haviour in Atlantic petrel, including timing of migration. We also found moonlight intensity to shape behaviour during winter in the Atlantic petrel. 5. Wet-dry dynamics have great behavioural content of at-sea movement and behaviour, spanning scales from elementary motion patterns (e.g. at scale of minutes, hours) to complex ecological interactions (e.g. seasonality, annual life cycle). We used cutting-edge techniques to build up a new analytical protocol that evidences how a broad repertoire of behaviours can be deciphered uniquely from wet-dry data. 6. Through this protocol, we uncovered both flexible and structural components of the behav- ioural organization of a highly-mobile migratory seabird, the Cory’s shearwater, a pelagic spe- cies with a complex annual cycle that involves central place foraging, ocean-basin long migra- tory movement and wandering in wintering areas. 7. Knowing the actual landscape faced by individuals will lead to understand the role of extrinsic factors shaping behavioural strategies and decision-making in elusive species. In this regard, our approach allows for behavioural annotation over long periods and large scales, which combined with currently available tools for environmental annotation from satellite-derived data sources (e.g. winds, fisheries) will provide new insights at an unprecedented detail. 8. Data visualization is a powerful tool to reveal insights from complex multidimensional infor- mation, such as data from animal movement and behaviour collected through biologging. We devised effective data visualizations such as actograms, behavioural landscapes and networks to assist the research proccess and provide new insights, otherwise hard to find, about behav- ioural patterns of species spending most of their live out of the human’s sight. BIBLIOGRAPHY Altmann, J. (1974) Observational study of behavior: sampling methods. Behaviour, 49(3): 227-267. Aschoff, J. (1979) Circadian rhythms: influences of internal and external factors on the period mea- sured in constant conditions. Zeitschrift für Tierpsychologie, 49(3): 225-249. Aspillaga, E., Bartumeus, F., Linares, C., Starr, R. M., López-Sanz, À., Díaz, D., Zabala, M. & Hereu, B. (2016) Ordinary and extraordinary movement behaviour of small resident fish within a Mediterranean marine protected area. PloS one, 11(7): e0159813. Bäckman, J., Andersson, A., Pedersen, L., Sjöberg, S., Tøttrup, A.P. & Alerstam, T. (2017) Acto- gram analysis of free-flying migratory birds: new perspectives based on acceleration logging. Journal of Comparative Physiology A, 203(6-7): 543-564. Bairlein, F., Dierschke, J., Dierschke, V., Salewski, V., Geiter, O., Hüppop, K., Köppen, U. & Fielder, W. (2014) Atlas des Vogelzugs. AULA,Wiebelsheim. Ballance, L.T. (2007) Understanding seabirds at sea: why and how? Marine Ornithology, 35: 127- 135. Barabasi, A. L. & Oltvai, Z. N. (2004) Network biology: understanding the cell’s functional organi- zation. Nature Reviews Genetics, 5(2): 101. Bartumeus, F., Giuggioli, L., Louzao, M., Bretagnolle, V., Oro, D. & Levin, S.A. (2010) Fishery dis- cards impact on seabird movement patterns at regional scales. Current Biology, 20(3): 215-222. Bascompte, J., Jordano, P., Melián, C. J. & Olesen, J.M. (2003) The nested assembly of plant–animal mutualistic networks. Proceedings of the National Academy of Sciences, 100(16): 9383-9387. Beever, E.A., Hall, L.E., Varner, J., Loosen, A.E., Dunham, J.B., Gahl, M. K., Smith, F.A. & Lawler, J.J. (2017) Behavioral flexibility as a mechanism for coping with climate change. Frontiers in Ecology and the Environment, 15(6): 299-308. Berman, G.J., Bialek, W. & Shaevitz, J.W. (2016) Predictability and hierarchy in Drosophila behav- ior. Proceedings of the National Academy of Sciences, 113(42): 11943-11948. BirdLife International (2015) Puffinus lherminieri. The IUCN Red List of Threatened Species 2015: eT45959182A84678618. BirdLife International (2018a) Calonectris borealis. The IUCN Red List of Threatened Species 2018: e.T22732244A132661794. BirdLife International (2018b) Pterodroma incerta. The IUCN Red List of Threatened Species 2018: e.T22698084A132623707. Block, B.A., Jonsen, I.D., Jorgensen, S.J., Winship, A.J., Shaffer, S.A., Bograd, S.J. et al. (2011) Tracking apex marine predator movements in a dynamic ocean. Nature, 475(7354): 86. 204 Bouten, W., Baaij, E.W., Shamoun-Baranes, J. & Camphuysen, K.C. (2013) A flexible GPS tracking system for studying bird behaviour at multiple scales. Journal of Ornithology, 154(2): 571-580. Boyd, I.L., Kato, A. & Ropert-Coudert, Y. (2004) Bio-logging science: sensing beyond the bounda- rie. Memoirs of National Institute of Polar Research (Special Issue), 58: 1-14. Bradbury, J.W. & Vehrencamp, S.L. (2014) Complexity and behavioral ecology. Behavioral Ecol- ogy, 25(3): 435-442. Brlík, V., Koleček, J., Burgess, M. et al. (2019) Weak effects of geolocators on small birds: A meta- analysis controlled for phylogeny and publication bias. Journal of Animal Ecology, 2019; 00: 1-14. Brooke, M. (1990) The Manx shearwater. Academic Press, London. Brooke, M. (2004) Albatrosses and petrels across the world. Oxford University Press, Oxford Bugoni, L. & Vooren, C.M. (2004) Feeding ecology of the Common Tern Sterna hirundo in a win- tering area in southern Brazil. Ibis, 146: 438-453. Burger, A. E. & Shaffer, S. A. (2008) Perspectives in ornithology application of tracking and data- logging technology in research and conservation of seabirds. The Auk, 125(2): 253-264. Candolin, U. & Wong, B.B. (Eds.) (2012) Behavioural responses to a changing world: mechanisms and consequences. Oxford University Press. Caravaggi, A., Cuthbert, R.J., Ryan, P.G., Cooper, J. & Bond, A.L. (2019) The impacts of introduced House Mice on the breeding success of nesting seabirds on Gough Island. Ibis, 161(3): 648-661. Carboneras, C. (1992) Family Procellariidae (petrels and shearwaters). In: del Hoyo, J. et al. (Eds.) Handbook of the birds of the world, Vol 1. Lynx Edicions, Barcelona, p 216−271. Carboneras, C., Jutglar, F. & Kirwan, G.M. (2016) Audubon’sshearwater (Puffinus lherminieri). In: del Hoyo J. et al. (Eds.) Handbook of the birds of the world alive. Lynx Edicions, Barcelona. Carter, M. I., Cox, S. L., Scales, K. L., Bicknell, A. W., Nicholson, M. D., Atkins, K. M., Morgan, G., Morgan, L., Grecian, W.J., Patrick, S.M. & Votier, S.C. (2016) GPS tracking reveals rafting behaviour of Northern Gannets (Morus bassanus): implications for foraging ecology and con- servation. Bird Study, 63(1): 83-95. Catry, P., Dias, M.P., Phillips, R.A. & Granadeiro, J.P. (2011) Different means to the same end: long- distance migrant seabirds from two colonies differ in behaviour, despite common wintering grounds. PLoS One, 6(10): e26079. Catry, T., Ramos, J.A., Catry, I., Monticelli, D. & Granadeiro, J. P. (2013a) Inter-annual variability in the breeding performance of six tropical seabird species: influence of life-history traits and relationship with oceanographic parameters. Marine Biology, 160(5): 1189-1201. Catry, P., Dias, M.P., Phillips, R.A. & Granadeiro, J.P. (2013b) Carry-over effects from breeding 205BIBLIOGRAPHY modulate the annual cycle of a long-distance migrant: an experimental demonstration. Ecology, 94(6): 1230-1235. Catry, P., Granadeiro, J.P., Ramos, J., Phillips, R.A. & Oliveira, P. (2011) Either taking it easy or feeling too tired: old Cory’s Shearwaters display reduced activity levels while at sea. Journal of Ornithology, 152(3): 549-555. Catry, P., Phillips, R. A., Phalan, B., Silk, J.R. & Croxall, J.P. (2004) Foraging strategies of grey- headed albatrosses Thalassarche chrysostoma: integration of movements, activity and feeding events. Marine Ecology Progress Series, 280: 261-273. Cherel, Y., Quillfeldt, P., Delord, K. & Weimerskirch, H. (2016) Combination of at-sea activity, geolocation and feather stable isotopes documents where and when seabirds molt. Frontiers in Ecology and Evolution, 4: 3. Chimienti, M., Cornulier, T., Owen, E., Bolton, M., Davies, I.M., Travis, J.M.J. & Scott, B.E. (2016) The use of an unsupervised learning approach for characterizing latent behaviors in accelerom- eter data. Ecology and Evolution, 6(3): 727–741. Chmura, H.E., Glass, T.W. & Williams, C.T. (2018) Biologging physiological and ecological re- sponses to climatic variation: New tools for the climate change era. Frontiers in Ecology and Evolution, 6: 92. Cianchetti-Benedetti, M., Catoni, C., Kato, A., Massa, B. & Quillfeldt, P. (2017) A new algorithm for the identification of dives reveals the foraging ecology of a shallow-diving seabird using ac- celerometer data. Marine Biology, 164(4): 77. Clay, T.A., Pearmain, E. J., McGill, R.A., Manica, A. & Phillips, R.A. (2018) Age-related variation in non-breeding foraging behaviour and carry-over effects on fitness in an extremely long-lived bird. Functional Ecology, 32(7): 1832-1846. Collet, J., Patrick, S.C. & Weimerskirch, H. (2015) Albatrosses redirect flight towards vessels at the limit of their visual range. Marine Ecology Progress Series, 526: 199-205. Cooke, S.J, Hinch, S.G., Wikelski, M., Andrews, R.D., Kuchel, L.J., Wolcott, T.G. & Butler, P.J. (2004) Biotelemetry: a mechanistic approach to ecology. Trends in Ecology and Evolution, 19: 334-343. Cramp, S. & Simmons, K.E.L. (1977) Handbook of the birds of Europe, the Middle East and North Africa: the birds of the Western Palearctic, Vol I. Oxford University Press, London. Croxall, J.P., Butchart, S.H., Lascelles, B.E.N., Stattersfield, A.J., Sullivan, B.E.N., Symes, A. & Taylor, P.H.I.L. (2012) Seabird conservation status, threats and priority actions: a global assess- ment. Bird Conservation International, 22(1): 1-34. Cubaynes, S., Doherty, P.F., Schreiber, E. A. & Gimenez, O. (2010) To breed or not to breed: a sea- bird’s response to extreme climatic events. Biology Letters, 7(2): 303-306. 206 Cuthbert, R.J. (2004) Breeding biology of the Atlantic Petrel, Pterodroma incerta, and a population estimate of this and other burrowing petrels on Gough Island, South Atlantic Ocean. Emu, 104: 221-228. Damschen, E.I., Brudvig, L.A., Haddad, N.M., Levey, D.J., Orrock, J.L. & Tewksbury, J.J. (2008) The movement ecology and dynamics of plant communities in fragmented landscapes. Proceed- ings of the National Academy of Sciences, 105(49): 19078-19083. Dankert, H., Wang, L., Hoopfer, E.D., Anderson, D.J. & Perona, P. (2009) Automated monitoring and analysis of social behavior in Drosophila. Nature Methods, 6: 297-303. Dean, B., Freeman, R., Kirk, H., Leonard, K, Phillips, R.A., Perrins, C.M. & Guilford, T. (2013) Behavioural mapping of a pelagic seabird: combining multiple sensors and a hidden Markov model reveals the distribution of at-sea behaviour. Journal of the Royal Society Interface, 10(78): 20120570. Dean, B., Kirk, H., Fayet, A., Shoji, A., Freeman, R., Leonard, K., Perrins, C.M. & Guilford, T. (2015) Simultaneous multi-colony tracking of a pelagic seabird reveals cross-colony utilization of a shared foraging area. Marine Ecology Progress Series, 538: 239-248. de Grissac, S., Bartumeus, F., Cox, S.L., & Weimerskirch, H. (2017) Early-life foraging: Behavioral responses of newly fledged albatrosses to environmental conditions. Ecology and Evolution, 7(17): 6766-6778. del Hoyo, J., Collar, N. & Kirwan, G.M. (2019) Cory’s Shearwater (Calonectris borealis) In: del Hoyo et al. (Eds.) Handbook of the Birds of the World Alive. Lynx Edicions, Barcelona. Demšar, U., Buchin, K., Cagnacci, F., Safi, K., Speckmann, B., Van de Weghe, N., Weiskop, D. & Weibel, R. (2015) Analysis and visualisation of movement: an interdisciplinary review. Move- ment ecology, 3(1): 5. Demšar, U., Slingsby, A. & Weibel, R. (2019) Introduction to the special section on Visual Move- ment Analytics. Information Visualization, 18(1): 133-137. Dias, M.P., Granadeiro, J.P. & Catry, P. (2012) Working the day or the night shift? Foraging sched- ules of Cory’s shearwaters vary according to marine habitat. Marine Ecology Progress Series, 467: 245-252. Dias, M.P., Martin, R., Pearmain, E.J., Burfield, I.J., Small, C. Phillips, R.A., Yates, O., Lascelles, B., Garcia Borboroglu, P. & Croxall, J.P. (2019) Threats to seabirds: A global assessment. Bio- logical Conservation, 237: 525-537. Dias, M.P., Romero, J., Granadeiro, J.P., Catry, T., Pollet, I.L. & Catry, P. (2016) Distribution and at- sea activity of a nocturnal seabird, the Bulwer’s petrel Bulweria bulwerii, during the incubation period. Deep Sea Research Part I: Oceanographic Research Papers, 113: 49-56. Dodge, S., Bohrer, G., Bildstein, K., Davidson, S.C., Weinzierl, R., Bechard, M.J., Barber, D., Kays, R., Brandes, D., Han, J. & Wikelski, M. (2014) Environmental drivers of variability in the move- 207BIBLIOGRAPHY ment ecology of turkey vultures (Cathartes aura) in North and South America. Philosophical Transactions of the Royal Society B: Biological Sciences, 369(1643): 20130195. Dodge, S., Bohrer, G., Weinzierl, R., Davidson, S. C., Kays, R., Douglas, D., Cruz, S. Han, J. Brandes, D. & Wikelski, M. (2013) The environmental-data automated track annotation (Env- DATA) system: linking animal tracks with environmental data. Movement Ecology, 1(1): 3. Dodge, S., Xavier, G. & Wong, W.Y. (2018) DynamoVis - Dynamic Visualization of Animal Move- ment Data. Retrieved from the Data Repository for the University of Minnesota, https://doi. org/10.13020/D6PH49. Dragon, A.C., Bar-Hen, A., Monestiez, P. & Guinet, C. (2012) Horizontal and vertical movements as predictors of foraging success in a marine predator. Marine Ecology Progress Series, 447: 243-257. Durant, J.M., Hjermann, D.Ø., Frederiksen, M., Charrassin, J.B., Le Maho, Y., Sabarros, P.S., Craw- ford, R.J.M. & Stenseth, N.C. (2009) Pros and cons of using seabirds as ecological indicators. Climate Research, 39(2): 115-129. Egevang, C., Stenhouse, I. J., Phillips, R.A., Petersen, A., Fox, J.W. & Silk, J.R. (2010) Tracking of Arctic terns Sterna paradisaea reveals longest animal migration. Proceedings of the National Academy of Sciences, 107(5): 2078-2081. Elliott, K.H. & Gaston, A.J. (2015) Diel vertical migration of prey and light availability constrain foraging in an Arctic seabird. Marine Biology, 162(9): 1739-1748. Enticott, J.W. (1991) Distribution of the Atlantic Petrel Pterodroma incerta at sea. Marine Ornithol- ogy, 19: 49-60. Fauchald, P. (2009) Spatial interaction between seabirds and prey: review and synthesis. Marine Ecology Progress Series, 391: 139-151. Fayet, A.L., Freeman, R., Shoji, A., Kirk, H.L., Padget, O., Perrins, C.M. & Guilford, T. (2016) Carry-over effects on the annual cycle of a migratory seabird: an experimental study. Journal of Animal Ecology, 85(6): 1516-1527. Flood, R.L. & van der Vliet, R. (2019) Variation and identification of Barolo Shearwater and Boyd’s Shearwater. Dutch Birding, 41: 215-237. Fox, J. W. (2010) Geolocator manual v8. British Antarctic Survey, Cambridge. Frederiksen, M., Moe, B., Daunt, F., Phillips, R.A., Barrett, R. T., Bogdanova, M.I. et al. (2012) Multicolony tracking reveals the winter distribution of a pelagic seabird on an ocean basin scale. Diversity and Distributions, 18(6): 530-542. Freeman, R., Dean, B., Kirk, H., Leonard, K., Phillips, R.A., Perrins, C.M. & Guilford, T. (2013) Predictive ethoinformatics reveals the complex migratory behaviour of a pelagic seabird, the Manx Shearwater. Journal of the Royal Society Interface, 10(84): 20130279. 208 Furness, R.W. & Camphuysen, K. (1997) Seabirds as monitors of the marine environment. ICES Journal of Marine Science, 54(4): 726-737. Gaston, A.J. (2004) Seabirds: a natural history. Yale University Press. Gochfeld, M., Burger, J., Christie, D.A. & Garcia, E.F.J. (2019) Common Tern (Sterna hirundo) In: del Hoyo et al. (Eds.) Handbook of the Birds of the World Alive. Lynx Edicions, Barcelona. González-Solís, J., Croxall, J.P. & Afanasyev, V. (2007) Offshore spatial segregation in giant petrels Macronectes spp.: differences between species, sexes and seasons. Aquatic Conservation: Ma- rine and Freshwater Ecosystems, 17(S1): S22-S36. González-Solís, J., Croxall, J.P., Oro, D. & Ruiz, X. (2007) Trans-equatorial migration and mixing in the wintering areas of a pelagic seabird. Frontiers in Ecology and the Environment, 5(6): 297-301. González-Solís, J., Felicísimo, A., Fox, J.W., Afanasyev, V., Kolbeinsson, Y. & Muñoz, J. (2009) In- fluence of sea surface winds on shearwater migration detours. Marine Ecology Progress Series, 391: 221-230. Granadeiro, J. P., Campioni, L. & Catry, P. (2018) Albatrosses bathe before departing on a foraging trip: implications for risk assessments and marine spatial planning. Bird Conservation Interna- tional, 28(2): 208-215. Granadeiro, J.P., Monteiro, L.R., Silva, M.C. & Furness, R.W. (2002) Diet of common terns in the Azores, Northeast Atlantic. Waterbirds, 25(2): 149-156. Grémillet, D. & Boulinier, T. (2009) Spatial ecology and conservation of seabirds facing global climate change: a review. Marine Ecology Progress Series, 391: 121-137. Grundy, E., Jones, M.W., Laramee, R.S., Wilson, R.P. & Shepard, E.L. (2009) Visualisation of sen- sor data from animal movement. Computer Graphics Forum, 28(3): 815-822. Guilford, T., Meade, J., Willis, J., Phillips, R.A., Boyle, D., Roberts, S., Collett, M., Freeman, R. & Perrins, C.M (2009) Migration and stopover in a small pelagic seabird, the Manx shearwater Puffinus puffinus: insights from machine learning. Proceedings of the Royal Society B: Biologi- cal Sciences, 276(1660): 1215-1223. Gutowsky, S.E., Gutowsky, L.F., Jonsen, I.D., Leonard, M.L., Naughton, M.B., Romano, M.D. & Shaffer, S.A. (2014) Daily activity budgets reveal a quasi-flightless stage during non-breeding in Hawaiian albatrosses. Movement Ecology, 2(1): 23. Halpern, B.S., Selkoe, K.A., Micheli, F. & Kappel, C.V. (2007) Evaluating and ranking the vul- nerability of global marine ecosystems to anthropogenic threats. Conservation Biology, 21(5): 1301-1315. Harcourt, R., Sequeira, A.M.M., Zhang, X., Roquet, F., Komatsu, K., Heupel, M. et al. (2019) An- imal-Borne Telemetry: An Integral Component of the Ocean Observing Toolkit. Frontiers in 209BIBLIOGRAPHY Marine Science, 6, 326. Hasenjager, M.J. & Dugatkin, L.A. (2015) Social network analysis in behavioral ecology. In: Na- guib, M. et al. (Eds.) Advances in the Study of Behavior - Vol. 47. pp. 39-114. Academic Press. Hays, G.C., Ferreira, L.C., Sequeira, A.M., Meekan, M.G., Duarte, C.M., Bailey, H. et al. (2016) Key questions in marine megafauna movement ecology. Trends in Ecology & Evolution, 31(6): 463-475. Hazevoet, C.J. (1995) The birds of the Cape Verde Islands. An annotated check list. British Ornitolo- gists Union, London. Hedd, A., Montevecchi, W.A., Phillips, R.A. & Fifield, D.A. (2014) Seasonal sexual segregation by monomorphic sooty shearwaters Puffinus griseus reflects different reproductive roles during the pre-laying period. PLoS One, 9(1): e85572. Hennicke, J. C. & Weimerskirch, H. (2014) Coping with variable and oligotrophic tropical wa- ters: foraging behaviour and flexibility of the Abbott’s booby Papasula abbotti. Marine Ecology Progress Series, 499: 259-273. Hertel, F. & Ballance, L.T. (1999) Wing ecomorphology of seabirds from Johnston Atoll. The Con- dor, 101(3): 549-556. Heurich, M., Hilger, A., Küchenhoff, H., Andrén, H., Bufka, L., Krofel, M., Mattisson J., Odden, J., Persson, J., Rauset, G.R., Schmidt, K. & Linnell, D. C., J. (2014) Activity patterns of Eurasian lynx are modulated by light regime and individual traits over a wide latitudinal range. PLoS One, 9(12): e114143. Hill, R.D. & Braun, M.J. (2001) Geolocation by light level, electronic tagging and tracking in ma- rine fisheries. Proceedings of the symposium on tagging and tracking marine fish with electronic devices. East-West Center. University of Hawaii. Berlin: Springer. pp. 315–330. Igual, J.M., Forero, M.G., Tavecchia, G., González-Solís, J., Martínez-Abraín, A., Hobson, K.A., Ruiz, X. & Oro, D. (2005) Short-term effects of data-loggers on Cory’s shearwater (Calonectris diomedea). Marine Biology, 146(3): 619-624. Jacoby, D.M., Brooks, E.J., Croft, D.P. & Sims, D.W. (2012) Developing a deeper understanding of animal movements and spatial dynamics through novel application of network analyses. Meth- ods in Ecology and Evolution, 3(3): 574-583. Jacoby, D. M. & Freeman, R. (2016) Emerging network-based tools in movement ecology. Trends in Ecology & Evolution, 31(4): 301-314. Jeltsch, F., Bonte, D., Pe’er, G., Reineking, B., Leimgruber, P., Balkenhol, N. et al. (2013) Integrating movement ecology with biodiversity research-exploring new avenues to address spatiotemporal biodiversity dynamics. Movement Ecology, 1(1): 6. Kays, R., Crofoot, M.C., Jetz, W. & Wikelski, M. (2015) Terrestrial animal tracking as an eye on life 210 and planet. Science, 348(6240): aaa2478. Kavathekar, D., Mueller, T. & Fagan, W.F. (2013) Introducing AMV (Animal Movement Visual- izer), a visualization tool for animal movement data from satellite collars and radiotelemetry. Ecological informatics, 15: 91-95. Kemp, M.U., Emiel van Loon, E., Shamoun-Baranes, J. & Bouten, W. (2012) RNCEP: global weath- er and climate data at your fingertips. Methods in Ecology and Evolution, 3(1): 65-70. Keogan, K., Daunt, F., Wanless, S., Phillips, R.A., Walling, C.A. et al. (2018) Global phenological insensitivity to shifting ocean temperatures among seabirds. Nature Climate Change, 8(4): 313. Klages, N.T.W. & Cooper, J. (1997) Diet of the Atlantic petrel Pterodroma incerta during breeding at South Atlantic Gough Island. Marine Ornithology, 25: 13-16. Konzack, M., Gijsbers, P., Timmers, F., van Loon, E., Westenberg, M. A. & Buchin, K. (2019) Vi- sual exploration of migration patterns in gull data. Information Visualization, 18(1): 138-152. Krüger, L., Paiva, V.H., Petry, M.V. & Ramos, J. A. (2017) Strange lights in the night: using abnor- mal peaks of light in geolocator data to infer interaction of seabirds with nocturnal fishing ves- sels. Polar Biology, 40(1): 221-226. Kürten, N., Vedder, O., González-Solís, J., Schmaljohann, H. & Bouwhuis, S. (2019) No detectable effect of light-level geolocators on the behaviour and fitness of a long-distance migratory sea- bird. Journal of Ornithology. Lascelles, B.G., Taylor, P.R., Miller, M.G.R., Dias, M.P., Oppel, S. et al. (2016) Applying global criteria to tracking data to define important areas for marine conservation. Diversity and Dis- tributions, 22(4): 422-431. Lewison, R., Oro, D., Godley, B.J., Underhill, L., Bearhop, S., Wilson, R.P. et al. (2012) Research priorities for seabirds: improving conservation and management in the 21st century. Endangered Species Research, 17(2): 93-121. Liechti, F., Witvliet, W., Weber, R. & Bächler, E. (2013) First evidence of a 200-day non-stop flight in a bird. Nature Communications, 4: 2554. Lisovski, S., Bauer, S., Briedis, M., Davidson, S.C., Dhanjal-Adams, K.L., Hallworth, M.T. et al. (2019) Light-Level Geolocator Analyses: A user’s guide. Journal of Animal Ecology. López-López, P. (2016) Individual-based tracking systems in ornithology: welcome to the era of big data. Ardeola, 63(1): 103-137. Louzao, M., Hyrenbach, K.D., Arcos, J.M., Abelló, P., Sola, L.G. & Oro, D. (2006) Oceanographic habitat of an endangered Mediterranean procellariform: implications for Marine Protected Ar- eas. Ecological Applications, 16: 1683-1695. Mackley, E.K., Phillips, R.A., Silk, J.R., Wakefield, E.D., Afanasyev, V., Fox, J.W. & Furness, R.W. (2010) Free as a bird? Activity patterns of albatrosses during the nonbreeding period. Marine 211BIBLIOGRAPHY Ecology Progress Series, 406: 291-303. Mallory, M.L. & Gilbert, C.D. (2008) Leg-loop harness design for attaching external transmitters to seabirds. Marine Ornithology, 36: 183-188. Mattern, T., Masello, J.F., Ellenberg, U. & Quillfeldt, P. (2015) Actave.net – a web-based tool for the analysis of seabird activity patterns from saltwater immersion geolocators. Methods in Ecology and Evolution, 6(7): 859-864. McIntyre, T. (2014) Trends in tagging of marine mammals: a review of marine mammal biologging studies. African Journal of Marine Science, 36(4): 409-422. McKnight, A., Irons, D.B., Allyn, A.J., Sullivan, K.M. & Suryan, R.M. (2011) Winter dispersal and activity patterns of post-breeding black-legged kittiwakes Rissa tridactyla from Prince William Sound, Alaska. Marine Ecology Progress Series, 442: 241-253. Mendez, L., Prudor, A. & Weimerskirch, H. (2017) Ontogeny of foraging behaviour in juvenile red- footed boobies (Sula sula). Scientific Reports, 7(1): 13886. Militão, T., Dinis, H.A., Zango, L., Calabuig, P., Stefan, L.M. & González-Solís, J. (2017) Popula- tion size, breeding biology and on-land threats of Cape Verde petrel (Pterodroma feae) in Fogo Island, Cape Verde. PloS one, 12(4): e0174803. Missagia, R.V., Ramos, J.A., Louzao, M., Delord, K., Weimerskirch, H. & Paiva, V.H. (2015) Year- round distribution suggests spatial segregation of Cory’s shearwaters, based on individual expe- rience. Marine Biology, 162(11): 2279-2289. Nathan, R., Getz, W.M., Revilla, E., Holyoak, M., Kadmon, R., Saltz, D. & Smouse, P.E. (2008) A movement ecology paradigm for unifying organismal movement research. Proceedings of the National Academy of Sciences, 105(49): 19052-19059. Nathan, R. & Giuggioli, L. (2013) A milestone for movement ecology research. Movement Ecology, 1:1. Neves, V.C., Bried, J., González-Solís, J., Roscales, J.L. & Clarke, M.R. (2012) Feeding ecology and movements of the Barolo shearwater Puffinus baroli baroli in the Azores, NE Atlantic. Marine Ecology Progress Series, 452: 269-285. Nisbet, I.C.T., Szczys, P., Mostello, C.S. & Fox, J.W. (2011) Female Common Terns Sterna hirundo start autumn migration earlier than males. Seabird, 24: 103-106. Numata, H. & Helm, B. (2014) Annual, lunar, and tidal clocks. Springer Verlag. Obringer, R., Bohrer, G., Weinzierl, R., Dodge, S., Deppe, J., Ward, M., Brandes, D., Kays, R. & Wikelski, M. (2017) Track Annotation: Determining the Environmental Context of Movement Through the Air. In: Chilson P. et al. (Eds.) Aeroecology. Springer, Cham O’Connor, C. M., Norris, D. R., Crossin, G. T. & Cooke, S. J. (2014) Biological carryover effects: linking common concepts and mechanisms in ecology and evolution. Ecosphere, 5(3): 1-11. 212 Orgeira, J.L. (2001) Nuevos registros del Petrel Atlántico (Pterodroma incerta) en el océano Atlán- tico Sur y Antártida. Ornitología Neotropical, 12: 165-171. Oshima, M.C. & Leaf, R.T. (2018) Understanding the structure and resilience of trophic dynamics in the northern Gulf of Mexico using network analysis. Bulletin of Marine Science, 94(1): 21-46. Paiva, V.H., Fagundes, A.I., Romao, V., Gouveia, C. & Ramos, J.A. (2016) Population-scale foraging segregation in an apex predator of the North Atlantic. PloS one, 11(3): e0151340. Papastamatiou, Y.P., Watanabe, Y.Y., Demšar, U., Leos-Barajas, V., Langrock, R., Bradley, D., Weng, K., Lowe, C., Friedlander, A. & Caselle, J. (2018) Activity seascapes highlight central place refuging strategies in marine predators that never stop swimming. Movement Ecology, 6(1): 9. Phalan, B., Phillips, R.A., Silk, J. R., Afanasyev, V., Fukuda, A., Fox, J., Catry, P., Higuchi, H. & Croxall, J. P. (2007) Foraging behaviour of four albatross species by night and day. Marine Ecol- ogy Progress Series, 340: 271-286. Phillips, R.A., Lewis, S., González-Solís, J. & Daunt, F. (2017) Causes and consequences of indi- vidual variability and specialization in foraging and migration strategies of seabirds. Marine Ecology Progress Series, 578: 117-150. Phillips, R.A., Silk, J.R.D., Croxall, J.P., Afanasyev, V. & Briggs, D.R. (2004) Accuracy of geoloca- tion estimates for flying seabirds. Marine Ecology Progress Series, 266: 265-272. Pinet, P., Jaquemet, S., Phillips, R.A. & Le Corre, M. (2012) Sex-specific foraging strategies throughout the breeding season in a tropical, sexually monomorphic small petrel. Animal Be- haviour, 83(4): 979-989. Pollet, I.L., Hedd, A., Taylor, P.D., Montevecchi, W.A. & Shutler, D. (2014) Migratory movements and wintering areas of Leach’s Storm-Petrels tracked using geolocators. Journal of Field Orni- thology, 85(3): 321-328. Ponchon, A., Cornulier, T., Hedd, A., Granadeiro, J.P. & Catry, P. (2019) Effect of breeding perfor- mance on the distribution and activity budgets of a predominantly resident population of black- browed albatrosses. Ecology and Evolution, 9(15): 8702-8713. Precheur, C. (2015) Ecologie et conservation du puffin d’Audubon (Puffinus lherminieri lher- minieri) de la réserve naturelle des îlets de Sainte-Anne (Martinique). These de Doctorat de l’Universite des Antilles. Quillfeldt, P., Masello, J.F., Navarro, J. & Phillips, R.A. (2013) Year-round distribution suggests spatial segregation of two small petrel species in the South Atlantic. Journal of Biogeography, 40(3): 430-441. Ramos, R., Carlile, N., Madeiros, J., Ramírez, I., Paiva, V. H., Dinis, H.A., Zino, F., Biscoito, M., Leal, G.R., Bugoni, L., Jodice, P.G.R., Ryan, P.G. & González-Solís, J. (2017) It is the time for oceanic seabirds: Tracking year-round distribution of gadfly petrels across the Atlantic Ocean. 213BIBLIOGRAPHY Diversity and Distributions, 23(7): 794-805. Ramos, R., Granadeiro, J.P., Rodríguez, B., Navarro, J., Paiva, V.H., Bécares, J., Reyes-González, J.M., Fagundes, I., Ruiz, A., Arcos, J.M., González-Solís, J. & Catry, P. (2013) Meta-population feeding grounds of Cory’s shearwater in the subtropical Atlantic Ocean: implications for the definition of Marine Protected Areas based on tracking studies. Diversity and Distributions, 19(10): 1284-1298. Ramos, R., Llabrés, V., Monclús, L., López-Béjar, M. & González-Solís, J. (2018) Costs of breeding are rapidly buffered and do not affect migratory behavior in a long-lived bird species. Ecology, 99(9): 2010-2024. Ramos, R., Ramírez, I., Paiva, V.H., Militão, T., Biscoito, M., Menezes, D., Phillips, R.A., Zino, F. & González-Solís, J. (2016) Global spatial ecology of three closely-related gadfly petrels. Scientific Reports, 6: 23447. Ramos, R., Sanz, V., Militão, T., Bried, J., Neves, V. C., Biscoito, M., Phillips, R.A., Zino, F. & González-Solís, J. (2015) Leapfrog migration and habitat preferences of a small oceanic seabird, Bulwer’s petrel (Bulweria bulwerii). Journal of Biogeography, 42: 1651-1664. Rayner, M.J., Carlile, N., Priddel, D., Bretagnolle, V., Miller, M.G.R., Phillips, R.A., Ranjard, L., Bury, S.J. & Torres, L.G. (2016) Niche partitioning by three Pterodroma petrel species during non-breeding in the equatorial Pacific Ocean. Marine Ecology Progress Series, 549:217-229. Rayner, M.J., Taylor, G.A., Gummer, H.D., Phillips, R.A., Sagar, P.M., Shaffer, S.A. & Thompson, D.R. (2012) The breeding cycle, year-round distribution and activity patterns of the endangered Chatham Petrel (Pterodroma axillaris). Emu - Austral Ornithology, 112(2): 107-116. Regular, P.M., Hedd, A. & Montevecchi, W.A. (2011) Fishing in the dark: a pursuit-diving seabird modifies foraging behaviour in response to nocturnal light levels. PLoS One, 6(10): e26763. Reyes-González, J.M. & González-Solís, J. (2016) Pardela cenicienta atlántica – Calonectris borea- lis. In: Salvador, A. & Morales, M.B. (Eds.) Enciclopedia Virtual de los Vertebrados Españoles. Museo Nacional de Ciencias Naturales, Madrid. Richardson, M.E. (1984) Aspects of the ornithology of the Tristan da Cunha group and Gough Is- land 1972-1974. Cormorant, 12: 123-201. Rodríguez, A., Arcos, J. M., Bretagnolle, V., Dias, M. P., Holmes, N. D., Louzao, M. et al. (2019) Future directions in conservation research on petrels and shearwaters. Frontiers in Marine Sci- ence, 6: 94. Roncon, G., Bestley, S., McMahon, C.R., Wienecke, B. & Hindell, M.A. (2018) View from below: inferring behavior and physiology of Southern Ocean marine predators from dive telemetry. Frontiers in Marine Science, 5: 1-23. Ropert-Coudert, Y., Kato, A., Grémillet, D. & Crenner, F. (2012) Bio-logging: recording the eco- physiology and behavior of animals moving freely in their environment. In: Le Galliard, J.F. et 214 al. (Eds.) Sensors for ecology: towards integrated knowledge of ecosystems. CNRS INEE, Paris. Ropert-Coudert, Y. & Wilson, R.P. (2005) Trends and perspectives in animal-attached remote sens- ing. Frontiers in Ecology and the Environment, 3: 437-444. Rutz, C. & Hays, G.C. (2009) New frontiers in biologging science. Biology Letters, 5(3): 289-292. Schultner, J., Moe, B., Chastel, O., Tartu, S., Bech, C. & Kitaysky, A.S. (2014) Corticosterone medi- ates carry-over effects between breeding and migration in the kittiwake Rissa tridactyla. Marine Ecology Progress Series, 496: 125-133. Schwalb-Willmann, J. (2019) moveVis: Movement Data Visualization. R package version 0.10.2. https://CRAN.R-project.org/package=moveVis Shaffer, S.A., Costa, D.P. & Weimerskirch, H. (2001) Behavioural factors affecting foraging effort of breeding wandering albatrosses. Journal of Animal Ecology, 70: 864-874. Shaffer, S.A., Tremblay, Y., Weimerskirch, H., Scott, D., Thompson, D.R., Sagar, P.M., Moller, H., Taylor, G.A., Foley, D.G., Block, B.A. & Costa, D. P. (2006) Migratory shearwaters integrate oceanic resources across the Pacific Ocean in an endless summer. Proceedings of the National Academy of Sciences, 103(34): 12799-12802. Shamoun-Baranes, J., van Loon, E.E., Purves, R.S., Speckmann, B., Weiskopf, D. & Camphuysen, C.J. (2011) Analysis and visualization of animal movement. Biology Letters, 8: 6-9. Shealer, D. A. (2001) Foraging behavior and food of seabirds. In: Schreiber, E.A. & Burger, J. (Eds.) Biology of marine birds. CRC press. Shoji, A., Aris-Brosou, S., Culina, A., Fayet, A., Kirk, H., Padget, O., Juarez-Martinez, I., Boyle, D., Nakata, T., Perrins, C.M. & Guilford, T. (2015) Breeding phenology and winter activity predict subsequent breeding success in a trans-global migratory seabird. Biology Letters, 11(10): 20150671. Sih, A., Stamps, J., Yang, L.H., McElreath, R. & Ramenofsky, M. (2010) Behavior as a key com- ponent of integrative biology in a human-altered world. Integrative and Comparative Biology, 50(6): 934-944. Slingsby, A. & Van Loon, E. (2016) Exploratory visual analysis for animal movement ecology. Com- puter Graphics Forum, 35: 471–480. Stauss, C., Bearhop, S., Bodey, T.W., Garthe, S., Gunn, C., Grecian, W.J., Inger, M.E., Knight, J. Newton, S.C. Patrick, Phillips, R.A., Waggitt, J.J. & Votier, S.C. (2012) Sex-specific foraging behaviour in northern gannets Morus bassanus: incidence and implications. Marine Ecology Progress Series, 457: 151-162. Stehfest, K.M. Patterson, T.A., Barnett, A. & Semmens, J.M. (2015) Markov models and network analysis reveal sex-specific differences in the space-use of a coastal apex predator. Oikos, 124: 307-318. 215BIBLIOGRAPHY Sutherland, W.J., Freckleton, R.P., Godfray, H.C.J., Beissinger, S.R., Benton, T. et al. (2013) Identi- fication of 100 fundamental ecological questions. Journal of Ecology, 101(1): 58-67. Tavares, D.C., Moura, J.F., Acevedo-Trejos, E. & Merico, A. (2019) Traits shared by marine mega- fauna and their relationships with ecosystem functions and services. Frontiers in Marine Sci- ence, 6: 262. Todd, J.G., Kain, J.S. & Bivort, B.L. (2017) Systematic exploration of unsupervised methods for mapping behavior. Physical Biology, 14(1): 015002. Tukey, J.W. (1977) Exploratory Data Analysis. New York: Addison-Wesley. Valletta, J.J., Torney, C., Kings, M., Thornton, A. & Madden, J. (2017) Applications of machine learning in animal behaviour studies. Animal Behaviour, 124: 203-220. Van Buskirk, J. (2012) Behavioural plasticity and environmental change. In: Candolin, U. & Wong, B.B.M. (Eds.) Behavioural Responses to a Changing World. Mechanisms and Consequences. Oxford University Press. Vandenabeele, S.P., Shepard, E.L., Grogan, A. & Wilson, R.P. (2012) When three per cent may not be three per cent; device-equipped seabirds experience variable flight constraints. Marine Biol- ogy, 159: 1-14. Vandenabeele, S.P., Wilson, R.P. & Grogan, A. (2011) Tags on seabirds: how seriously are instru- ment-induced behaviours considered? Animal Welfare-The UFAW Journal: 20(4): 559. Vansteelant, W.M., Shamoun-Baranes, J., van Manen, W., van Diermen, J. & Bouten, W. (2017) Seasonal detours by soaring migrants shaped by wind regimes along the East Atlantic Flyway. Journal of Animal Ecology, 86(2): 179-191. Votier, S.C., Hatchwell, B.J., Mears, M. & Birkhead, T.R. (2009) Changes in the timing of egg- laying of a colonial seabird in relation to population size and environmental conditions. Marine Ecology Progress Series, 393: 225-233. Warham, J. (1990) The petrels: their ecology and breeding systems. Academic Press, London. Weimerskirch, H. (2007) Are seabirds foraging for unpredictable resources? Deep Sea Research Part II: Topical Studies in Oceanography, 54(3-4): 211-223. Weimerskirch, H., Bertrand, S., Silva, J., Marques, J. C. & Goya, E. (2010) Use of social information in seabirds: compass rafts indicate the heading of food patches. PloS one, 5(3): e9928. Weimerskirch, H., Pinet, P., Dubos, J., Andres, S., Tourmetz, J., Caumes, C., Caceres S, Riethmuller, M. & Corre, M.L. (2019) Wettability of juvenile plumage as a major cause of mortality threatens endangered Barau’s petrel. Journal of Avian Biology, 50(1). Weimerskirch, H. & Wilson, R.P. (2000) Oceanic respite for wandering albatrosses. Nature, 406(6799): 955. 216 Wendeln, H. & Becker, P.H. (1999) Significance of ring removal in Africa for a Common tern Sterna hirundo colony. Ringing & Migration, 19: 210-212. Wernham, C.V., Toms, M.P., Marchant, J.H., Clark, J.A. Siriwardena, G.M. & Baillie, S.R. (2002) The migration atlas: movements of the birds of Britain and Ireland. T & AD Poyser, London. White, T.D., Ferretti, F., Kroodsma, D.A., Hazen, E.L., Carlisle, A.B., Scales, K.L., Bograd, S.J. & Block, B.A. (2019). Predicted hotspots of overlap between highly migratory fishes and industrial fishing fleets in the northeast Pacific. Science Advances, 5(3): eaau3761. Williams, H.J., Holton, M.D., Shepard, E.L., Largey, N., Norman, B., Ryan, P.G. et al. (2017) Identi- fication of animal movement patterns using tri-axial magnetometry. Movement Ecology, 5(1): 6. Williams, H.J., Taylor, L.A., Benhamou, S., Bijleveld, A.I., Clay, T.A. et al. (2019) Optimising the use of bio-loggers for movement ecology research. Journal of Animal Ecology. Wilmers, C.C., Nickel, B., Bryce, C.M., Smith, J.A., Wheat, R.E. & Yovovich, V. (2015) The golden age of bio-logging: how animal-borne sensors are advancing the frontiers of ecology. Ecology, 96: 1741-1753. Wilson, L.J., McSorley, C.A., Gray, C.M., Dean, B.J., Dunn, T.E., Webb, A. & Reid, J.B. (2008) Rafting behaviour of Manx Shearwaters Puffinus puffinus. Seabird, 21: 85-92 Wilson, R.P. (1992) Estimation of location: global coverage using light intensity. In: Priede, S. (Ed.) Wildlife telemetry: remote monitoring and tracking of animals. Ellis Horwood, New York. Wilson, R.P. & Vandenabeele, S.P. (2012) Technological innovation in archival tags used in seabird research. Marine Ecology Progress Series, 451: 245-262. Wilson, R.P., Weimerskirch, H. & Lys, P. (1995) A device for measuring seabird activity at sea. Journal of Avian Biology, 26: 172-175. Wong, B. & Candolin, U. (2015) Behavioral responses to changing environments. Behavioral Ecol- ogy, 26(3): 665-673. Yamamoto, T., Takahashi, A., Yoda, K., Katsumata, N., Watanabe, S., Sato, K. & Trathan, P.N. (2008) The lunar cycle affects at-sea behaviour in a pelagic seabird, the streaked shearwater, Calonectris leucomelas. Animal Behaviour, 76(5): 1647-1652. Yoda, K. (2019) Advances in bio-logging techniques and their application to study navigation in wild seabirds. Advanced Robotics, 33(3-4): 108-117. Zúñiga, D., Falconer, J., Fudickar, A.M., Jensen, W., Schmidt, A., Wikelski, M. & Partecke, J. (2016) Abrupt switch to migratory night flight in a wild migratory songbird. Scientific Reports, 6: 34207. ANNEX 1: Published articles of Chapter 1. Zajková Z, Militão T, González-Solís J (2017) Year- round movements of a small seabird and oceanic isotopic gradient in the tropical Atlantic. Mar Ecol Prog Ser 579:169-183 of Chapter 2. Becker, P.H., Schmaljohann, H., Riechert, J., Wagenknecht, G., Zajková, Z. & González‐Solís, J. (2016) Common Terns on the East Atlantic Flyway: temporal–spatial distribution during the non‐breeding period. J. Ornithol. 157: 927– 940 The article Zajková et al. (2017) Year-round movements of a small seabird and oceanic isotopic gradient in the tropical Atlantic, Mar Ecol Prog Ser 579:169–183, https://doi.org/10.3354/meps12269, is reproduced in this thesis under special license from the copyright holder, Inter-Research, with the following restriction: the complete article may not be further copied and distributed from this source separately from the copying and distribution of the thesis. This restriction ends on September 14, 2022. 219ANNEX 1: Published articles: CHAPTER 1 220 221ANNEX 1: Published articles: CHAPTER 1 222 223ANNEX 1: Published articles: CHAPTER 1 224 225ANNEX 1: Published articles: CHAPTER 1 226 227ANNEX 1: Published articles: CHAPTER 1 228 229ANNEX 1: Published articles: CHAPTER 1 230 231ANNEX 1: Published articles: CHAPTER 1 232 233ANNEX 1: Published articles: CHAPTER 1 Reprinted by permission from Springer Nature: Springer Nature - Journal of Ornithology. Becker, P. H., Schmaljohann, H., Riechert, J., Wagenknecht, G., Zajková, Z., & González-Solís, J. (2016) Common Terns on the East Atlantic Flyway: temporal–spatial distribution during the non-breeding period. Journal of Ornithology, 157(4): 927-940. License Number 4673200175200 (2016). https://link.springer.com/article/10.1007/s10336-016-1346-2 235ANNEX 1: Published articles: CHAPTER 2 236 237ANNEX 1: Published articles: CHAPTER 2 238 239ANNEX 1: Published articles: CHAPTER 2 240 241ANNEX 1: Published articles: CHAPTER 2 242 243ANNEX 1: Published articles: CHAPTER 2 244 245ANNEX 1: Published articles: CHAPTER 2 246 247ANNEX 1: Published articles: CHAPTER 2 248 ANNEX 2: Stable isotope analysis 251ANNEX 2: Stable isotope analysis Stable isotope analysis Some chapters of this thesis include the use of stable isotopes. In the following lines I introduce and extend the basics about stable isotope analysis to ease reading for those who are not familiar with this technique. Intrinsic markers, such as stable isotope analysis (SIA), have become an essential tool in ecologi- cal studies, allowing researchers to reveal aspects of animals’ spatial and trophic ecology across different spatial and temporal scales (Ramos & González-Solís 2012). SIA allows to infer trophic relations between predators and prey across the food webs, since stable isotopes are directly trans- ferred from prey to predator tissues throughout diet, which allows their traceability. As stable iso- topes are already present in the sampled tissue and hold the information needed at the moment of sampling, there is no need to recapture the individual. In marine environments, carbon δ13C values (the ratio 13C/12C) have been established as relevant indicator of inshore vs. offshore feeding dis- tribution (Hobson et al. 1994). On the other side, nitrogen δ15N values (the ratio 15N/14N) are used to infer trophic level position of consumers. This is especially relevant to study seabirds during the non-breeding season, when conventional diet analysis cannot be carried out. Therefore, SIA ap- proach can provide information to infer potential changes in seabird diet and trophic position over the annual cycle. In seabird studies, SIA of feathers have been commonly used, as they become metabolically inert and integrate information about when and where they were grown (Inger & Bearhop 2008). Moult represent a critical period for birds, as feather growth is energetically demanding and also reduce flight efficiency (Rayner and Swaddle 2000). Moulting pattern in many seabird species remains unknown because in most cases the replacement of feathers occurs during the non-breeding season, after birds leave their breeding grounds and thereafter become hard to observe in detail. However, differences and similarities in isotope values among feathers can be used to reveal the moulting sequence in relation to the annual cycle (Cherel et al. 2000). Moreover, the knowledge of the moult pattern of studied species is important to properly link the timing and area where feathers were grown and hence reveal movements within and between breeding and non-breeding grounds (Hob- son 2005). Isotopic values vary geographically according to baseline levels specific of each zone. The gradi- ent of this values across space can be represented in ‘isoscape’ maps. While for terrestrial territories various isoscape maps are currently available (West et al. 2008, Hobson et al. 2012a, Hobson et al. 2012b), there exists a lack for such maps for seas and oceans (Somes et al. 2010, McMahon et al. 2013a). Such isoscape maps represent an important tool for linking movements of animals with the marine environment, and to examine migratory connectivity of different populations and species (Rubenstein & Hobson 2004, Hobson et al. 2010). Overall, SIA can complement other techniques, such as animal tracking, opening new areas of research with potential for identifying foraging areas and potential prey of seabirds (Meier et al. 252 2017), spatial and trophic segregation of multiple species (Roscales et al. 2011, Navarro et al. 2013) or revealing moulting areas of seabirds (Cherel et al. 2016). REFERENCES Cherel, Y., Hobson, K. A. & Weimerskirch, H. (2000) Using stable-isotope analysis of feathers to distinguish moulting and breeding origins of seabirds. Oecologia, 122(2): 155-162. Cherel, Y., Quillfeldt, P., Delord, K. & Weimerskirch, H. (2016) Combination of at-sea activity, geolocation and feather stable isotopes documents where and when seabirds molt. Frontiers in Ecology and Evolution, 4: 3. Hobson, K. A., Piatt, J. & Pitocchelli, J. (1994). Using Stable Isotopes to Determine Seabird Trophic Relationships. Journal of Animal Ecology, 63(4): 786-798. Hobson, K. A. (2005) Using stable isotopes to trace long-distance dispersal in birds and other taxa. Diversity and Distributions, 11(2): 157-164. Hobson, K. A., Barnett-Johnson, R. & Cerling, T. (2010) Using isoscapes to track animal migra- tion. In Isoscapes (pp. 273-298). Springer, Dordrecht. Hobson, K. A., Van Wilgenburg, S. L., Wassenaar, L. I., & Larson, K. (2012a) Linking hydrogen (δ2H) isotopes in feathers and precipitation: sources of variance and consequences for assign- ment to isoscapes. PloS one, 7(4): e35137. Hobson, K. A., Van Wilgenburg, S. L., Wassenaar, L. I., Powell, R. L., Still, C. J. & Craine, J. M. (2012b). A multi-isotope (δ13C, δ15N, δ2H) feather isoscape to assign Afrotropical migrant birds to origins. Ecosphere, 3(5): 1-20. Inger, R. & Bearhop, S. (2008) Applications of stable isotope analyses to avian ecology. Ibis, 150(3): 447-461. McMahon KW, Hamady LL & Thorrold SR (2013a) A review of ecogeochemistry approaches to es- timating movements of marine animals. Limnol Oceanogr 58: 697−714 Meier, R. E., Votier, S. C., Wynn, R. B., Guilford, T., McMinn Grivé, M., Rodríguez, A., ... & Trueman, C. N. (2017) Tracking, feather moult and stable isotopes reveal foraging behaviour of a critically endangered seabird during the non-breeding season. Diversity and Distributions, 23(2): 130-145. Navarro, J., Votier, S. C., Aguzzi, J., Chiesa, J. J., Forero, M. G. & Phillips, R. A. (2013) Ecological 253ANNEX 2: Stable isotope analysis segregation in space, time and trophic niche of sympatric planktivorous petrels. PloS one, 8(4): e62897. Rayner, J. M. V. & Swaddle, J. P. (2000) Aerodynamics and behaviour of moult and take-off in birds. Biomechanics in animal behaviour, 125-157. Ramos, R. & González-Solís, J. (2012) Trace me if you can: the use of intrinsic biogeochemical markers in marine top predators. Frontiers in Ecology and the Environment, 10(5): 258-266. Roscales, J. L., Gómez-Díaz, E., Neves, V. & González-Solís, J. (2011) Trophic versus geographic structure in stable isotope signatures of pelagic seabirds breeding in the northeast Atlantic. Ma- rine Ecology Progress Series, 434: 1-13. Rubenstein, D. R. & Hobson, K. A. (2004) From birds to butterflies: animal movement patterns and stable isotopes. Trends in ecology & evolution, 19(5): 256-263. Somes, C. J., Schmittner, A., Galbraith, E. D., Lehmann, M. F., Altabet, M. A., Montoya, J. P., ... & Eby, M. (2010) Simulating the global distribution of nitrogen isotopes in the ocean. Global Biogeochemical Cycles, 24(4). West, J. B., Sobek, A., & Ehleringer, J. R. (2008) A simplified GIS approach to modeling global leaf water isoscapes. PLoS one, 3(6): e2447. ANNEX 3: Outreach contributions 257ANNEX 3: Outreach contributions POSTER PRESENTED AT 22ND SPANISH CONFERENCE OF ORNITHOLOGY, MADRID, 2014 New insights into behavioural strategies and time cycles in a small oceanic seabird: Boyd’s shearwater (Puffinus boydi) Zuzana Zajková*, Santiago Guallar & Jacob González-Solís Many aspects of seabird behaviour at sea remain unknown. Recently miniaturized devices that combine different kind of sensors allow us to reveal the movements and at-sea behaviour of highly pelagic species in more detail than ever before. Salt-water immersion sensors provide continuous data during long periods that can be used as a proxy to infer activity patterns. In combination with light geolocation positioning we have revealed the phenology, migratory movements and behaviour of Boyd’s Shearwater for the annual cycle. Institut de Recerca de la Biodiversitat (IRBio) and Dept Biologia Animal, Universitat de Barcelona, Barcelona, Spain *E-mail: zuzulaz@gmail.com Material & Methods • Boyd’s Shearwater (Puffinus boydi) • Cape Verde Islands (Ilhéu Raso, Ilhéu de Cima) • Six years of tracking (2007-2013) • Data from 37 geolocators (BAS & Biotrack) with salt-water immersion sensors of 30 individuals. Raw immersion data in 10 minutes blocks range from 0 (dry) to 200 (wet). As a measure of activity for each individual we calculated percentage of time spent on the water: • per day (consecutive light and dark period) • per day that occured during daylight and darkness • per hour within 24 hours Time calculations excluded periods spent in burrows (whole daylight or whole darkness duration). Introduction Results Conclusions (1) Activity during year The proportion of time spent on the water during the day varied between different stages of life cycle (p < 0.01). Birds spend more time on water during non-breeding than during other stages. During the breeding period they are more active (less time on water, more time flying), probably due to foraging effort related to breeding duties. (2) Activity by daylight & darkness When we split up the time spent on water by darkness and daylight for every day, we find out Boyd’s Shearwater to be a slightly nocturnal species (p < 0.01). Comparing between daylight and darkness, birds spend more time on the water during daylight. (3) Activity by 24-hours Activity daily patterns show clear differences among phenological phases. Birds consistently are more active during sunrise and sunset for the whole year since they spend more time in flight (i.e. less time on water). (4) Activity by moon phase Birds do not show an increase in activity (i.e. time flying) in relation to the moon phase during the non-breeding stage. This finding, that differs to previous publications about relation between other seabird species and the moon, may be related to a highly use of the sit-and-wait foraging strategy, which implies less time flying. Our results show Boyd’ Shearwater: - spends non-breeding period in oligotrophic waters of Central North Atlantic Ocean, - is more active (i.e. invest more time in flying) during breeding period, - is slightly more active during darkness along the year, - is more active in twilight periods of day, - presents inter-individual differences in activity which may reflect individual specialization in different foraging strategies. Differences are present even at intra-individual level among years (individuals marked with a grey rectangle). Since the whole population winter in oligotrophic waters where marine productivity (and its variability) is low, these individual differences suggest a possible specialization in foraging strategies at individual level. Analysis of activity at individual level, presented by line graphs corresponding to individual and year (5a, colours correspond to phenological phases), remarks differences in phenology and behaviour during the annual cycle. 0 20 40 60 80 100 0 4 8 12 16 20 24 Time of day (GMT) P ro po rti on o f t im e on w at er p er h ou r ( % ) Stage of life cycle Breeding Postnup. mig. Prenup. mig. Non−breeding 0 20 40 60 80 100 0 30 60 90 120 150 180 210 240 270 300 330 360 Day of year P ro po rti on o f t im e on w at er p er d ay (% ) Stage of life cycle Breeding Postnup. mig. Non−breeding Prenup. mig. 0 20 40 60 80 100 0 30 60 90 120 150 180 210 240 270 300 330 360 Day of year P ro po rti on o f t im e on w at er (% ) Time of day Darkness Daylight ● ● ● ● ● ●● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ●● ●●●●●●●● ● ● ● ● ● ● ● ● ● ● ● ●● ● ●●● ● ●●●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ●● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ●●● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ●● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ●●● ●● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ●● ●●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ●● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ●● ● ●● ● ● ● ● ● ● ● ●● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ●● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ●● ● ● ● ● ●● ● ● ● ● ● ●● ● ● ● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ●● ● ● ● ● ● ● ● ● ● ● ● ● ●●●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● R2 = 0.0314 p < 0.010 20 40 60 80 100 0 25 50 75 100 Moon phase (% iluminated) P ro po rti on o f w et p er d ar kn es s ( % ) (5) Behavioural strategies at individual level 5500007_2223001_2009 5500007_2246001_2008 5500019_2243001_2008 5500025_2251001_2008 5500028_2247001_2008 5500036_2214001_2009 5500040_2215001_2008 5500042_2167001_2009 5500047_2234001_2008 5500056_2166001_2008 5500124_17844001_2013_1year 5500124_17844001_2013_2year 5500183_17014002_2013 5500195_17032002_2012 5500438_17024001_2011 5500444_17167002_2013 5500444_17840001_2012 5500446_17020001_2011 5500446_17050001_2012 5500446_17842002_2013 5500452_17032001_2011 5500454_17019001_2011 5500455_17017001_2011 5500455_17170002_2013 5500455_17261001_2012 5500456_17852001_2012 5500457_17011001_2011 5500458_17029001_2012_1year 5500458_17029001_2012_2year 5500473_17841001_2012 5500475_17843001_2012 5500510_8609001_2011_1year 5500512_8594001_2011_1year 5500512_8594001_2011_2year 5500517_8595001_2011_1year 5500517_8595001_2011_2year 5500518_8601001_2010 5500519_17016001_2011 5500519_8599001_2010 5500520_17014001_2011 5500522_8598001_2010 0 20 40 60 80 100 Proportion of total activity type per day (%) In di vi du al Activity type Wet daylight Dry daylight Wet darknes Dry darknes 5500007_2223001_2009 5500007_2246001_2008 5500019_2243001_2008 5500025_2251001_2008 5500028_2247001_2008 5500036_2214001_2009 5500040_2215001_2008 5500042_2167001_2009 5500047_2234001_2008 5500056_2166001_2008 5500124_17844001_2013_1year 5500124_17844001_2013_2year 5500183_17014002_2013 5500195_17032002_2012 5500438_17024001_2011 5500444_17167002_2013 5500444_17840001_2012 5500446_17020001_2011 5500446_17050001_2012 5500446_17842002_2013 5500452_17032001_2011 5500454_17019001_2011 5500455_17017001_2011 5500455_17170002_2013 5500455_17261001_2012 5500456_17852001_2012 5500457_17011001_2011 5500458_17029001_2012_1year 5500458_17029001_2012_2year 5500473_17841001_2012 5500475_17843001_2012 5500510_8609001_2011_1year 5500512_8594001_2011_1year 5500512_8594001_2011_2year 5500517_8595001_2011_1year 5500517_8595001_2011_2year 5500518_8601001_2010 5500519_17016001_2011 5500519_8599001_2010 5500520_17014001_2011 5500522_8598001_2010 0 20 40 60 80 100 Propor ion of total activity type per day (%) In di vi du al Activity type Wet daylight Dry daylight Wet darknes Dry darknes Analysis of spatial data revealed Boyd’s shearwater is a species with oceanic distribution all year round. After the breeding, in the beginning of May, shearwaters migrated on average 1 482 km to oligotrophic waters in the Central North Atlantic Ocean (5 – 15º N/ 30 - 40º W), where they spent aprox. 115 days. Birds started the prenuptial migration on 2nd of September, but because of equinox effect, exact route remains unknown. Distribution & Phenology (5a) (5b) (5c) (5d) Activity during non-breeding stage compared among individuals shows clear differences (example of two individuals in 5b, 5c; all individuals in 5d), especially regarding the nocturnal activity. Non-breeding Non-breeding