Linking different worlds: migration ecology in two species of hirundines

1 Linking different worlds: migration ecology in two species of hirundines PhD Thesis Cosme López Calderón 2019 2 Cover design by Cosme López Calderón 3 Conectando mundos distintos: ecología de migración en dos especies de hirundínidos Tesis doctoral Cosme López Calderón 2019 Directores: Dr. Javier Balbontín Arenas (Universidad de Sevilla) Dr. Alfonso Marzal Reynolds (Universidad de Extremadura) 4 5 Agradecimientos No podría empezar de otra forma que agradeciendo a mi madre, Isabel Calderón, pues ella fomentó que desarrollara desde niño mi pasión por la zoología y la investigación. Agradezco a mi familia y amigos por soportar reiteradamente conversaciones sobre pájaros y por aguantarme en periodos de estrés. Especialmente doy gracias a Álvaro Merino y Aleksander Kubicki, que sin duda han contribuido en mi formación como científico. Tengo muy buen recuerdo de mi etapa en la Universidad de Extremadura, donde además de formarme en estudios de medio ambiente, conocí a buenos amigos y aprendí a valorar nuestra tierra. En especial, quiero agradecer a Juan M. Sánchez su labor de mentor más allá de un simple trabajo fin de grado. Agradezco también la influencia de Juan M. Pleguezuelos, con quien di mis primeros pasos en la investigación. Gracias a Blas Torres y José Manuel Guerra, supe del proyecto de investigación “Migración, conectividad y malaria en aves”. Así conocí al director de mi tesis, Javier Balbontín, que sin duda es la persona que más tengo que agradecer. Gracias a su proyecto, financiado por la Junta de Andalucía, he contado con el apoyo económico necesario para dedicarme por completo a esta tesis doctoral. Además de los innumerables conocimientos científicos que 6 Javier me ha trasmitido, le agradezco los buenos momentos que hemos pasado juntos. Continúo dando gracias a mi co-director Alfonso Marzal, a quien debo mucho también. Gracias a Alfonso, hoy en día trabajo en algunas temáticas que no hubiera podido imaginar años atrás. También agradezco a Florentino de Lope, que compagina una extensa labor científica con un buen sentido del humor. Sigo agradeciendo al resto de componentes de nuestro grupo de investigación: Sergio, Luz, Maribel, Yolanda, Manu y Jaime. A ellos les debo el buen ambiente de trabajo y que siempre respetaron mis hábitos de alimentación. En especial doy gracias a Sergio, que fue la primera persona que me enseñó a manejar un pájaro, momento que nunca se olvida. Muchas gracias también a nuestros colaboradores externos Keith Kobson y Anders Møller, que he podido conocer durante mis estancias en Canadá y Francia. Siempre os consideraré influencias fundamentales en mi desarrollo científico. Además os agradezco la buena comida servida. Tengo mucho que agradecer también a Xiomara, que hizo todo lo posible para facilitar mi estancia en Canadá. Gracias también a Steven Van Wilgenburg y Kevin Kardynal, que me permitieron avanzar en programación. Aunque en la Universidad de Sevilla no tenemos aún grupo de investigación consolidado, he de agradecer la compañía ofrecida por los miembros del Laboratorio de Biología Marina. Gracias a Guerrita, Free, Carlos, Maca, Pablo, Sedano, Altai, Íñigo, Gema, Ale,Marta 7 buenasfotografías. y Juanmi entre otros, que siempre tuvieron conversación para comer y tomar uncafé. A Francisco Sedano© además tengo que agradecer las Por último, agradezco a las personas que han hecho posible nuestro seguimiento de la golondrina común. Mil gracias a las familias Sangrán, Lora y Marzal, así como a Plácido, que nos han permitido trabajar en sus propiedades. Juan Sangrán merece un agradecimiento especial ya que nos ayudó a buscar nuevas zonas de estudio. También doy gracias a todos los trabajadores de las fincas, que tuvieron una paciencia ilimitada para recibirme a cualquier hora del día: Fernando, Antonio, Martín, Manuel, Yosi, Javier, Gabriel, Antonio Jesús y Yolanda. Pido disculpas ante cualquier molestia que os hemos podido causar y agradezco de corazón vuestra buena voluntad. También tengo que dar gracias a Marina Bollo y Sergio Reinoso por acompañarme en el trabajo de campo, su ayuda se hizo esencial en la etapa final de mi doctorado. Finalmente, agradezco también a todos los voluntarios que participaron en el trabajo de campo: Valentín, Carlos, Julia, Luis, Alejandro y Victoria. Resumen El comportamiento migratorio constituye una de las adaptaciones más espectaculares de los seres vivos frente a los cambios ambientales del planeta. Este comportamiento ha evolucionado en diversos grupos taxonómicos, desde mariposas a mamíferos, pero sin duda es en las aves donde ha encontrado su mayor desarrollo. Conocer dónde migran las aves siempre ha sido un misterio y no fue hasta finales del siglo XIX cuando comenzó el anillamiento científico. Desde los años ochenta, el desarrollo de nuevas técnicas de marcaje ha permitido obtener datos de gran precisión sobre el comportamiento migratorio. Esta información es necesaria para poder conservar adecuadamente poblaciones de aves migratorias en declive. En esta tesis doctoral, hemos utilizado análisis de isótopos estables y geolocalizadores para estudiar la ecología de migración en la golondrina común (Hirundo rustica) y el avión común (Delichon urbicum). En primer lugar, hemos descrito en detalle el comportamiento migratorio de la golondrina común reproductora en el sur de España. Además hemos identificado que la migración prenupcial es el principal nexo de unión entre lo que ocurre en África durante el invierno y lo que ocurre en Europa durante la temporada de reproducción. A continuación, hemos detectado que las condiciones ambientales experimentadas durante el invierno pueden tener consecuencias sobre la reproducción, lo que resulta sorprendente teniendo en cuenta que ambos eventos se encuentran separados varios meses en el tiempo y miles de kilómetros 13 1 General introduction Migratory behaviour constitutes one of the most amazing adaptations of life to face the changing conditions on Earth (Newton 2008; Hobson & Wassenaar 2008). This behavior has evolved in a wide variety of animals, from zooplankton to mammals (e.g. Hays 2003; Bolger et al. 2008). Every time we are impressed by new studies describing extraordinary movements, such as butterflies travelling about 4,000 km in several generations (Wassenaar & Hobson 1998), juvenile marine turtles crossing entire oceans (Luschi et al. 2003) or songbirds moving from Alaska to eastern Africa (Bairlein et al. 2012). Undoubtedly, migratory behaviour has developed to the greatest extent in birds, thanks to the domination of flying (Newton 2008). Bird migration has been always a mystery and people of the past thought very curious ways to explain why some species disappear after spring. For instance, it was supposed that breeding species transformed themselves into wintering species, or that swallows hibernated within the mud of lakes. Interestingly, these unbelievable explanations came first than thinking of birds travelling to other continents (Hobson & Wassenaar 2008). The first evidence of bird migration was discovered in 1822 by chance, when a white stork was found in Germany carrying an African spear (Kinzelbach 2005). During subsequent years, more than twenty storks were found in the same way, confirming the discovery. General introduction 2 Later in 1899, the first bird was ringed with an aluminium ring (Preuss 2001), and since then this has been the most extended method to unravel the whereabouts of migratory birds. Over a century of bird banding we have gained many information so far (du Feu et al. 2009), but recapture rates of ringed birds is very low (e.g. Hill et al. 1997) and so the information given by ringing recoveries is very limited. Since the 1980s, advances in satellite telemetry have allowed scientists to obtain high precision data on migratory behaviour (Perras & Nebel 2012; Kays et al. 2015). Once a bird is capture and attached with a satellite transmitter, researchers can monitor movements for long periods without having to recapture it. However, these devices are very expensive, which in turn restrict the sample size in most of studies. And even more importantly, due to their high size and weight, satellite transmitters cannot be deployed on small migratory birds (Kays et al. 2015; but see Hallworth & Marra 2015). An alternative solution to study migratory movements emerged after 1988 from stable isotope analyses (Hobson & Wassenaar 2008). Under this approach, the measurement of stable isotope ratios of a given tissue (e.g. feather) is matched to predicted values in maps, because the tissue incorporates the isotopic signature from the place where it was grown. The main advantages of the isotope method are that only one capture is needed, their lower cost compared to other methods, and that there is not restriction on body size (Hobson 1999). Nonetheless, stable isotope analyses can only 3 provide broad geographical information for certain periods of the migration cycle (e.g. moult areas), so other geographic information such as migratory routes or stopover sites are missing. This strong limitation encouraged the development of solar geolocation as the only alternative to obtain precise information for migratory songbirds (McKinnon et al. 2013). The first migratory songbird was tracked with light-level geolocators on 2007 (Stutchbury et al. 2009), and since then, thousands of birds have been tagged with these devices all over the world (e.g. Finch et al. 2017). Geolocators record light levels, which are then transferred into position estimates on the basis that photoperiod is associated with latitude and the time of solar noon or midnight with longitude (Ekstrom 2004; 2007). Within the short life of solar geolocation, this method has provided more information than banding recoveries over a century (McKinnon et al. 2013). Nevertheless, this technology is still far more expensive than stable isotope analyses, and furthermore ethical considerations must be taken into account (Costantini & Møller 2013; Brlík et al. in press). Therefore, both stable isotope analyses and light-level geolocators are entirely complementary (Hobson & Kardynal 2016) and they represent the best options to study migration in passerines. For instance, if the researcher is just interested in generalizing geographical origins (e.g. winter areas), then stable isotope analyses should provide the most cost-effective choice. By contrast, if the General introduction 4 purposes of the study require to know full migratory routes and schedules, then solar geolocation is the best option. Moreover, the combination of both tracking techniques has already provided refined results and so they are prone to feedback from each other (Contina et al. 2013; Hobson & Kardynal 2016; Imlay et al. 2018). To date, few studies have tested if moult origins inferred by stable isotopes agree with position estimates given by geolocators (González-Solís et al. 2011; Hallworth et al. 2013; Seifert et al. 2018). The degree of consistency between these methods depends on the structure of underlying isoscapes, accuracy of geolocators, and also to biological traits of the study species such as habitat or foraging behavior (Oppel et al. 2011; Hobson & Kardynal 2016; Seifert et al. 2018). Consequently, it is hard to generalize at this point, and so previous studies have found mixed results (González-Solís et al. 2011; Hallworth et al. 2013; Seifert et al. 2018). The recent development of solar geolocation has made possible to ground-truth geographical assignations based on stable isotopes, which was necessary to identify the flaws of that approach. In this line, we expect that future studies will introduce new statistical methods to increase the match between both tracking techniques. Independently of the geographical assignment of origin, stable isotopes analyses have been proved to be very powerful in the study of migration ecology (Hobson 1999). Precisely, the forensic isotope approach has contributed profoundly to establish links between 5 different worlds (Webster et al. 2002; Norris & Marra 2007; Harrison et al. 2011). Since the late 1990s, it has become widely accepted that events occurring in one part of the migration cycle may severely affect subsequent performance in another, i.e. “carry-over effect” or “seasonal interaction” (Hobson & Wassenaar 2008). The first clear example of a carry-over effect in a migratory songbird was described by Marra et al. (1998) using stable isotopes analyses. These authors demonstrated for the American redstart (Setophaga ruticilla), that winter habitat quality in Jamaica was related to spring arrival to the breeding grounds in New Hampshire. Since then, many other groundbreaking studies have described carry-over effects for passerines using stable isotope analyses (e.g. Norris et al. 2004a, 2004b; Bearhop et al. 2005; Gunnarson et al. 2005). More recently, the development of solar geolocation has inaugurated a new golden age for the study of carryover effects (e.g. Gils et al. 2016; Saino et al. 2017; Kramer et al. 2017; Patchett et al. 2018). Traditional ornithology has been mainly focused on the breeding period, but the development of new tracking technologies has stimulated a shift of paradigm (Faaborg et al. 2010). Before the 1980s, it was assumed that migratory birds are temperate breeders that overwinter in warmer latitudes feeding on "excess" available resources. Currently, our model of migration includes birds with an evolutionary tropical origin that started to breed northwards as a seasonal reproductive strategy. In fact, migratory birds usually spend General introduction 6 more time travelling and wintering than actually breeding (Faaborg et al. 2010; Cornelius et al. 2013), therefore knowing the whereabouts of migrants and the extent of carry-over effects is essential for conservation and management (Norris & Marra 2007; Harrison et al. 2011; McKinnon et al. 2013). First, it is necessary to know where do migratory birds stopover and winter; second, it is important to determine the most suitable non-breeding habitats that may drive subsequent fitness advantages; and finally, we should prioritize the protection of these key non-breeding habitats. This is because the protection of certain suitable habitats along the migratory route should have a profound impact on widespread population trends (Sillett & Holmes 2002; Norris et al. 2004; Gunnarson et al. 2005; Delmore et al 2012; Laughlin et al. 2013). In addition, any disturbance at these specific non-breeding habitats will have different consequences depending on the pattern of connections with the breeding areas, i.e. “migratory connectivity” (Webster et al. 2002; Webster & Marra 2005; Finch et al. 2017). Thus, if individuals of the same breeding population congregate together at localized non-breeding habitats (strong connectivity), that breeding population should be more vulnerable to environmental changes that another with widespread non-breeding distributions (weak connectivity). This is because gene flow among wintering populations must be higher in a situation of weak migratory connectivity, what leads to lower rates of local adaptation and thus provides more 7 resilience against rapid environmental changes (Webster & Marra 2005). It is well known that migratory behaviour differs across sex and age classes (Cristol et al. 1999; Newton 2008). Females usually select partners that are established in suitable nesting habitats, and thus competition for earlier arrival to breeding areas is higher among males (Møller 1992; Lozano et al. 1996; Hasselquist 1998). As a consequence, males have developed different strategies for this purpose, such as wintering in high quality habitats to improve physical condition and depart earlier from winter areas (Marra et al. 1998). In addition, young birds breeding for the first time usually arrive later to Recently, it has been shown an outstanding example that integrates different aspects of migration ecology to provide guidelines for conservation (Kramer et al. 2018). By using light level geolocators, these authors identified the wintering areas of goldenwinged warblers (Vermivora chrysoptera) and blue-winged warblers (Vermivora cyanoptera). They found a strong migratory connectivity for the golden-winged warbler, which was associated with historical trends in breeding populations. Interestingly, the steepest population decline at the breeding grounds was found for birds overwintering in habitats that had undergone severe deforestation. As predicted by Webster et al. (2002), the lower migratory connectivity was found for the blue-winged warbler, who has already replaced golden-winged warblers in many regions (Gill 1980). General introduction 8 the breeding areas and have lower reproductive success than more experienced birds. This lower performance of younger individuals could be either because they are less competitive than socially dominant older individuals, because of their lower experience, or due to the selective disappearance of poor quality individuals (Forslund & Pärt 1995; Newton 2008). Consequently, carry-over effects along different stages of the life cycle also change according to age and sex (Saino et al. 2004; Norris et al. 2004; Balbontín et al. 2009; Drake et al. 2013; Sergio et al. 2014). Indeed this is another source of variance to take into account in migration ecology, and also in management decisions regarding conservation efforts. In addition to intrinsic factors such as age and sex, environmental factors play an important role in shaping migration strategies. For instance, weather conditions strongly affects migrants, determining when they can travel, their energy requirements, potential risks encountered during the journey, and the visibility of any celestial or geographical cue used for navigation (Newton 2008). Less obvious are the effects of other extrinsic factors such as parasitism pressure. Because diversity and prevalence of parasites is usually higher in the tropical wintering grounds than in the temperate breeding areas (Bensch et al. 2009; Sehgal 2015), migratory behaviour may have evolved as a mean to escape infections (Møller & Szép 2010). Alternatively, the exposure to more parasites in the tropics could be seen as a cost to migration (Figuerola & Green 2000; Waldenström et 15 Hasselquist, D. Polygyny in great reed warblers: a long-term study of factors contributing to male fitness. Ecology 79, 2376–2390 (1998). Hays, G. C. A review of the adaptive significance and ecosystem consequences of zooplankton diel vertical migrations. Hydrobiologia 503, 163–170 (2003). Hill, L. A. Trans-Sahara recoveries of house martins Delichon urbica, with discussion on ringing, roosting and sightings in Africa. Safring News 26, 7–12 (1997). Hobson, K. A. & Kardynal, K. J. An isotope (δ34 S) filter and geolocator results constrain a dual feather isoscape (δ2H, δ13 C) to identify the wintering grounds of North American Barn Swallows. The Auk 133, 86– 98 (2016). Hobson, K. A. & Wassenaar, L. I. Tracking animal migration with stable isotopes (Academic Press, 2008). Hobson, K. A. Tracing origins and migration of wildlife using stable isotopes: a review. Oecologia 120, 314–326 (1999). Imlay, T. L., Hobson, K. A., Roberto-Charron, A. & Leonard, M. L. Wintering Areas, Migratory Connectivity and Habitat Fidelity of Three Declining Nearctic- Neotropical Migrant Swallows. Animal Migration 5, 1–16 (2018). Kays, R., Crofoot, M. C., Jetz, W. & Wikelski, M. Terrestrial animal tracking as an eye on life and planet. Science 348, aaa2478-1–aaa2478-7 (2015). General introduction 16 Kinzelbach, R. K. Das Buch vom Pfeilstorch (Basilisken: Marburg, Germany, 2005) Kramer, G. R. et al. Population trends in Vermivora warblers are linked to strong migratory connectivity. P. Natl. Acad. Sci. 115, E3192–E3200 (2018). Laughlin, A. J. et al. Integrating information from geolocators, weather radar, and citizen science to uncover a key stopover area of an aerial insectivore. The Auk 130, 230–239 (2013). Lozano, G. A., Perreault, S. & Lemon, R. E. Age, arrival date and reproductive success of male American redstarts Setophaga ruticilla.J. Avian Biol. 27, 164–170 (1996). Luschi, P., Hays, G. C. & Papi, F. A review of long-distance movements by marine turtles, and the possible role of ocean currents. Oikos 103, 293– 302 (2003). Marra, P. P, Hobson, K. A. and Holmes, R. T. Linking winter and summer events in a migratory bird by using stable-carbon isotopes. Science 282, 1884–1886 (1998). McKinnon, E. A., Fraser, K. C. & Stutchbury, B. J. M. New discoveries in landbird migration using geolocators, and a flight plan for the future. The Auk 130, 211–222 (2013). Møller, A. P. Sexual selection in the monogamous swallow Hirundo rustica. II. Mechanisms of intersexual selection. J. Evol. Biol. 5, 603–624 (1992). 17 Møller, A. P. & Szép, T. The role of parasites in ecology and evolution of migration and migratory connectivity. J. Ornithol. 152, 141–150 (2010). Newton, I. The migration ecology of birds (Academic Press, 2008). Norris, D. R. & Marra, P. P. Seasonal interactions, habitat quality, and population dynamics in migratory birds. The Condor 109, 535 (2007). Norris, D. R., Marra, P. P, Montgomerie, R, Kyser, T. K., Ratcliffe, L. M. Reproductive Effort, Molting Latitude, and Feather Color in a Migratory Songbird. Science 306, 2249–2250 (2004a). Norris, D. R., Marra, P. P., Kyser, T. K., Sherry, T. W. & Ratcliffe, L. M. Tropical winter habitat limits reproductive success on the temperate breeding grounds in a migratory bird. Proc. R. Soc. Lond. B 271, 59–64 (2004b). O’Connor, C. M., Norris, D. R., Crossin, G. T. & Cooke, S. J. Biological carryover effects: linking common concepts and mechanisms in ecology and evolution. Ecosphere 5, 28 (2014). Oppel, S. et al. High variation reduces the value of feather stable isotope ratios in identifying new wintering areas for aquatic warblers Acrocephalus paludicola in West Africa. J. Avian Biol. 42, 342–354 (2011). Parmesan, C. Ecological and Evolutionary Responses to Recent Climate Change. Annu. Rev. Ecol. Evol. Syst 37, 637–669 (2006). General introduction 18 Patchett, R., Finch, T. & Cresswell, W. Population consequences of migratory variability differ between flyways. Curr. Biol. 28, R340–R341 (2018). Perras, M. & Nebel, S. Satellite Telemetry and its Impact on the Study of Animal Migration. Nature Education Knowledge 3, 4 (2012). Preuss, N. O. Hans Christian Cornelius Mortensen: aspects of his life and of the history of bird ringing. Ardea 89, 1-6 (2001). Saino, N. et al. Ecological conditions during winter predict arrival date at the breeding quarters in a trans-Saharan migratory bird. Ecol. Letters 7, 21–25 (2004). Saino, N. et al. Sex-dependent carry-over effects on timing of reproduction and fecundity of a migratory bird. J. Anim. Ecol. 86, 239–249 (2017). Sehgal, R. N. M. Manifold habitat effects on the prevalence and diversity of avian blood parasites. IJP-PAW 4, 421–430 (2015). Seifert, N. et al. Matching geographical assignment by stable isotopes with African non-breeding sites of barn swallows Hirundo rustica tracked by geolocation. PLoS ONE 13, e0202025 (2018). Sergio, F. et al. Individual improvements and selective mortality shape lifelong migratory performance. Nature 515, 410–413 (2014). Sillett, T. S. & Holmes, R. T. Variation in survivorship of a migratory songbird throughout its annual cycle. J. Anim. Ecol. 71, 296–308 (2002). 19 Stutchbury, B. J. M. et al. Tracking Long-Distance Songbird Migration by Using Geolocators. Science 323, 896–896 (2009). Waldenström, J., Bensch, S., Kiboi, S., Hasselquist, D. & Ottosson, U. (2002). Cross-species infection of blood parasites between resident and migratory songbirds in Africa. Mol. Ecol. 11, 1545–1554. Walther, G. R. et al. Ecological responses to recent climate change. Nature 416, 389 (2002). Wassenaar, L. I. & Hobson, K. A. Natal origins of migratory monarch butterflies at wintering colonies in Mexico: New isotopic evidence. P. Natl. Acad. Sci. 95, 15436–15439 (1998). Webster, M. S. & Marra, P. P. The importance of understanding migratory connectivity and seasonal interactions in Birds of two worlds: the ecology and evolution of migration (eds Greenberg, R. & Marra, P. P.) 199-209 (JHU Press, 2005). Webster, M. S., Marra, P. P., Haig, S. M., Bensch, S. & Holmes, R. T. Links between worlds: unraveling migratory connectivity. TREE 17, 76–83 (2002). Chapter 1 20 Description of migratory behaviour 21 Chapter 1 Chapter 1 22 The migration system of barn swallows (Hirundo rustica) breeding in southwestern Spain Cosme López-Calderón1, Alfonso Marzal2and Javier Balbontín1 1Department of Zoology, Faculty of Biology, Seville, Spain 2Department of Anatomy, Cellular Biology and Zoology, University of Extremadura, Badajoz, Spain Description of migratory behaviour 23 Abstract Animal migration has always fascinated biologists. Beyond the inherent scientific knowledge, it is necessary to know the whereabouts of migrants as a prior requisite for their conservation. The recent development of solar geolocation has allowed ornithologists to track extensively small sized birds, which make up the majority of bird abundance. Here we used light-level geolocators to describe in detail the migration system of barn swallows breeding in southwestern Spain. We used the most up to date statistical framework to reconstruct migratory routes from light data in order to determine migratory schedules, stopover and wintering areas, migratory connectivity, factors determining variability in migration and carryover effects. We compare our results with previous literature and provide ecological interpretations. Solar geolocation has already filled traditional gaps of knowledge within an unprecedented time. Chapter 1 24 Introduction Migratory behavior has evolved in a variety of taxonomic groups, such as those constituting zooplankton, insects, fish, amphibians, reptiles and mammals (e.g. Wassenaar & Hobson 1998; Hays 2003; Luschi et al. 2003; Quinn 2005; Bolger et al. 2008; Kovar et al. 2008). However, with no doubts, migratory behavior has developed to the greatest extent in birds, yet they can fly and move with relatively low energy requirements (Newton 2008). Thus, more than 50 billion birds perform regular annual movements between breeding and wintering areas, often involving thousands of kilometers (Berthold 1993). Specifically, an estimate of two billion birds migrate each autumn from Europe to sub-Saharan Africa (Hahn et al. 2009). This bulk of individuals constitutes a strong ecological link between different continents across the world, demanding resources along migratory routes and potentially transporting associated species such as pollen, seeds, parasites and invasive species (Kays et al. 2015). Beyond the inherent scientific knowledge that has always fascinated ornithologists, it is necessary to know where migratory birds stay during the non-breeding period as a prior requisite for their conservation (Faaborg et al. 2010). Indeed, many migratory bird species are declining in population numbers, especially long-distance migrants (Sanderson et al. 2006), aerial insectivores (Nebel et al. 2010) and species with low variability in migratory performance Description of migratory behaviour 31 corrected for clock drift with Decompressor (British Antarctic Survey). Twilight transitions were individually inspected with library BAStag (Wotherspoon et al. 2016) of R(R Core Team 2016). Transitions with light peaks or non-linear transitions that substantially disrupted twilight time were manually excluded from further analysis. This process resulted in the rejection of 4-29% of twilight events in each tag record. We never edited non-natural twilights, we delete them, but we did not fill these gaps with an expected twilight time. This criterion should lead to higher uncertainty in the estimated positions during periods of low quality light data, but we were confident that this is the most conservative approach. A minimum dark period of five hours was considered, and the arbitrary threshold value defining twilight was set at 16. Migratory routes were reconstructed from light data with library FLightR (Rakhimberdiev et al. 2017), which provides refined accuracy and precision of calculated positions by using a hidden Markov chain model and a template fit observation model (Rakhimberdiev et al. 2015). We used two different calibration periods of approximately one week each, and we accounted for the loss of transparency in the surface of the light sensor. The use of cavities for nesting had a strong impact in the light profiles (Liechti et al. 2014; Hobson et al. 2015), and thus the first calibration period was taken just after reproduction but when the bird was still at a breeding colony before departure. The second calibration period was taken after Chapter 1 32 recapture in the following year, exposing the tags to the sun light on a rooftop. We used a spatial-mask, constraining our model to consider states of residence only over land, with a buffer distance of 30 km into the sea. Following Rakhimberdiev et al. (2015), we incorporated the priors of 0.1 for probability of migratory behavior and 300 ± 150 km for distance covered between consecutive twilights. Key dates of the migration cycle (i.e. departure from breeding area, arrival to winter area, departure from winter area and arrival to breeding area) were calculated using the function find.times.distribution from the library FLightR (Rakhimberdiev et al. 2017). This function derives the time at which an animal crossed a specified latitude, together with the measure of its uncertainty (we always used Q.50 value). We considered departure/arrival from/to breeding areas when a given swallow crossed the boundary latitude of 36º (i.e. Gibraltar Strait). Regarding arrival/departure to/from wintering areas, we considered specific boundary latitudes for each individual. We defined boundary latitude of arrival to winter area as the latitude when the tracked bird stopped its southward journey, and boundary latitude of departure from winter area as the latitude when the given bird started its northward journey. We defined the wintering areas for each tracked swallow as polygons generated by the function plot_util_distr from the library FLightR (Rakhimberdiev et al. 2017). This function calculates Description of migratory behaviour 33 the accumulated probabilities of occurrence during a specified period (i.e. between arrival and departure date from wintering areas). For example, the polygons obtained with this function setting percentiles=0.5 indicated the grid points which accumulated more than 50% of probability of occurrence during winter. These polygons were circles of 85 km diameter because FLightR discretizes the space to reduce computation time (Rakhimberdiev et al. 2015). In order to identify the stopover locations for each tracked swallow, we generated our own function in R(R Core Team 2016). We considered a stopover as any location during autumn (between departure date from breeding area and arrival date to winter area) or spring migration (between departure date from winter area and arrival date to breeding area) in which the given bird stayed for at least two days. If the coordinates (being latitude and longitude rounded to one decimal) of twilight "i" matched the coordinates of twilight "i + n" but there were missing twilights among them due to poor quality light data, we considered this location as a stopover only if the bird remained here for at least two consecutive days, and we obtained the stopover duration from the first and last day at this stopover (even if during this period some twilights were missing). To depict the wintering and stopover areas used by our whole population under study we rasterized respectively, polygons with accumulated probabilities of occurrence during winter and error Chapter 1 34 squares of stopover locations (as defined by upper and lower 95% Credible Intervals for latitude and longitude). Pixels within these polygons were coded as one and all others as zero, resulting in one binary map per individual. Finally, for each stage of the migration, we summed over each individual raster layer. All GIS operations were conducted in R version 3.3.1 (R Core Team 2016) with packages raster (Hijmans 2016) and sp (Pebesma & Bivand 2005). Statistical analysis First, as complementary descriptive information, we quantified migratory connectivity following Finch et al. (2017). This was accomplished by calculating two new variables: population spread and inter-population mixing. The former was simply the mean pairwise distance between individual wintering areas (i.e. centroids) for each breeding colony, and the latter was obtained by Mantel correlation coefficient between pairwise distance matrices of individual wintering and breeding locations (Ambrosini et al. 2009). This coefficient corresponds to a Pearson product moment correlation coefficient between two matrices, whose significance is assessed with randomized permutation. The mantel coefficient ranges from -1 to +1, and it quantifies whether distances between individual breeding colonies are maintained during the winter. Strong positive Mantel coefficients indicate that individuals which breed close together also Description of migratory behaviour 35 spend the winter season relatively close together (i.e. low interpopulation mixing), whereas weak or negative coefficients indicate that individuals which breed close together spend the winter season apart from each other (i.e. high inter-population mixing). Second, we tested if barn swallows performed migration differently according to year, breeding colony and sex. For this purpose, we fit five Generalized Linear Models for each stage of the migratory cycle. During autumn migration, our dependent variables of interest were: departure date from breeding areas, number and mean duration of stopovers (days), arrival date to wintering areas and migration period (days). During winter, our dependent variables of interest were: latitude and longitude of the wintering area centroid, size of the wintering area (km2), great circle migratory distance (kilometres between winter area centroids and breeding colonies) and wintering period (days). All these dependent variables were obtained for wintering areas defined as grid points which accumulated >50% probability of occurrence during winter. During spring migration, our dependent variables of interest were: departure date from wintering areas, number and mean duration of stopovers (days), arrival date to breeding areas and migration period (days). For every dependent variable, we fit our models including year of tag detachment, breeding colony and sex as fixed effects. We did not account for any interaction term because of our low sample size on migratory tracks. We always used Gaussian distribution of errors and the identity link function. Chapter 1 36 Residuals of each GLM were visually inspected for deviation from normality using normal “qqplots”. Third, we tested if differences in departure date from breeding areas were consistently accumulated through the migration cycle to finally affect arrival date to breeding areas. Because key dates during migration flow sequentially in a time series fashion, we used a Structural Equation Model (hereafter: SEM). This is a probabilistic model that hypothesizes a causal network with multiple variables that can appear as both predictor and response variables (Lefcheck 2016). We used “confirmatory path analysis” or piecewise SEM, based on applications from graph theory (Shipley 2016). In piecewise SEM, the causal network is translated to a set of linear equations (e.g. linear mixed models), which are then evaluated individually thus allowing a wide range of distributions and sampling designs. Our causal network was made up of three linear mixed models: the first model tested the effect of departure date from breeding areas on arrival date to wintering areas, the second model tested the effect of arrival date to wintering areas on departure date from wintering areas and the third model tested the effect of departure date from wintering areas on arrival date to breeding areas. In every linear mixed model, we included detachment year and breeding colony as cross-random intercepts, thus controlling for inter-annual effects and among breeding site variation. We always used Gaussian distribution of errors and the identity link function. To qualitatively evaluate each Description of migratory behaviour 37 individual model fit, we visually inspected residuals for deviation from normality using normal “qqplots”. Finally, we quantified the goodness of fit of our entire causal network with directed separation test (“d-separation test”; Shipley 2016), which tests the assumption that all variables are conditionally independent (i.e. that there are no missing relationships among unconnected variables). Results Description of migratory behavior In 2017, we recaptured 21 out of 64 tagged swallows. However, in 2018 we recaptured only 7 out of 59 tagged swallows. Four of the tags recovered in 2017 did not provide any useful light data. Thus we count for this study with a total sample size of 24 migratory tracks. In 2017, the apparent survival rate of tagged birds (32.81%) was significantly lower than that of control birds (56.25%) (χ2= 6.152; df = 1; p= 0.013). In 2018, the apparent survival rate of tagged birds (11.86%) was also lower than that of control birds (20%), but this difference was not statistically significant (χ2= 1.227; df = 1; p= 0.268). Four swallows were equipped with geolocators in both attachment years of our study, but only one of them survived the second migration (ring: Y65702). Thus, we only could obtain longitudinal data on migration for a single individual. To avoid Chapter 1 38 pseudo-replication, we excluded the first recapture of this individual from subsequent analyses (we retained the second recapture to increase sample size in our second migration year). Individual migratory routes are presented in Figures 1-24 of Supplementary Materials. Generally, barn swallows used the Strait of Gibraltar to cross the sea. Considering the limitations of our light data (i.e. precision and missing twilights), few individuals crossed open sea at the Gulf of Cádiz or the Alboran Sea. Depending on each track, individuals used either similar or different routes for autumn and spring migration. Roughly, half of our tracked individuals conducted a loop migration. In many cases, migratory routes followed in autumn were closer to the Atlantic coast than routes followed in spring. Nevertheless, it seems that our study individuals did not show consistent soaring behavior and rather they usually crossed large regions of Sahara desert. Our studied barn swallows left their breeding areas generally in August (Fig. 1 and 2). One third of them stayed to refuel in northwestern Morocco (Fig. 3), and they took between 4 - 60 days to conduct the autumn migration (mean ± SD = 20.79 ± 18.01 days; n= 24). Around 1st of September (Fig. 1 and 2), barn swallows arrived to their wintering areas (Fig. 4), where they spent approximately five months (mean ± SD = 152.88 ± 23.53 days; n= 24). Wintering areas were largely spread across West Africa, roughly from 5º to 25º in latitude and from -15º to 5º in longitude. The main core of habitats Description of migratory behaviour 39 used for wintering (i.e. grid points which accumulated more than 50% of probability of occurrence) was concentrated along Ivory Coast, Guinea, Republic of Sierra Leone, Mali, Burkina Faso and Ghana. These areas were mainly dominated by savannah and cropland mixed with natural vegetation (Allen 2016). Considering also the areas used occasionally during winter (i.e. grid points which accumulated more than 95% of probability of occurrence), our study populations spread out through other different African biomes, such as desert, grassland and broadleaf forest. From January to March, barn swallows left their wintering areas (Fig. 1 and 2). Two thirds of them stayed to refuel from western Algeria to northern Morocco (Fig. 5), and they took between 20 - 103 days to conduct the spring migration (mean ± SD = 48.92 ± 25.73 days; n= 24). Finally, our studied individuals arrived back to their breeding colonies from February to May (Fig. 1 and 2). Interestingly, a single individual did not migrate to Africa, but instead it travelled northwards during autumn migration (ring: 2L12590; Fig. 23 Supplementary Materials). This swallow took two months to reach Manchester (UK) and then only one week to come back to Spain. Then, this individual spent about three months wintering in the northwestern Iberian Peninsula (from Central Portugal to Galicia). After this stage, it travelled northwards again, crossing 1,000 km of open sea and reaching Ireland only in two days. Then, after a few days, this bird started the southward journey back to Spain and it arrived to its breeding colony in about 20 days. This time, Chapter 1 40 it made a detour through Cornwall and Brittany, but still crossing open sea at the Bay of Biscay. Because this was a very especial case in our study population, we excluded this single individual from our subsequent statistical analyses. Migratory connectivity Population spread during winter was 672.69 km, 653.26 km, 521.19 km and 733.41 km for “Las Coladas”, “La Calera”, “La Alegría” and “Asesera”, respectively. Inter-population mixing on the wintering grounds was high (Mantel r= 0.022; p= 0.343). Therefore, the distance between two individuals during the breeding season did not correspond with the distance between the same individuals during the winter (i.e. weak migratory connectivity; Fig. 6). Description of migratory behaviour 47 Factors influencing migratory behavior Regarding the factors that could influence our migratory variables of interest, we found widespread inter-annual effects and some differences across breeding colonies during spring migration. Sex was never a significant predictor of migratory performance. During autumn migration, individuals left their breeding areas earlier, stopped to refuel more times and thus took more days to complete the journey in the second migration year (Table 1). Mean duration of stopovers in autumn was higher in the second migration year, although these differences were not statistically significant. On the other hand, arrival date to winter areas was very similar across years. Individual wintering areas showed consistent differences across years (Table 2). Specifically, individuals spread out through a larger surface, wintered northerly in the African winter quarter and consequently the migratory distance was lower in the second migration year. In addition, individuals wintered further to the east and during a longer period in the second migration year, although these differences were not statistically significant. During spring migration, we found widespread differences according to breeding colony (Table 3). Individuals breeding at “Las Coladas” stopped to refuel more times (mean ± SD = 5.2 ± 2.3 days) Chapter 1 48 than individuals breeding at other sites (mean ± SD = 1.75 ± 1.87 days), stopped over for longer periods (mean ± SD = 5.83 ± 2.81 days) than swallows from other populations (mean ± SD = 2.75 ± 0.71 days), arrived later to the breeding colony (mean ± SD = 23th April ± 18.26 days) than the other swallows (mean ± SD = 8th March ± 19.5 days), and therefore took more days to accomplish the journey (mean ± SD = 73 ± 19.5) than the other individuals (mean ± SD = 33 ± 11.28). Besides, barn swallows arrived later to their breeding colonies during the second migration year. Additionally, individuals left their wintering areas later in the second migration year, but these differences were not statistically significant (Table 3). Description of migratory behaviour 49 Table 1. F statistic and associated pvalue of GLM developed for each migration variable in autumn. Dependent variables are underlined. Independent variables with significant effects are highlighted in bold. Averages and standard deviations are given for dependent variables by year. Sample size was 22 migratory tracks. Model F (p) Mean (SD): 2016 Mean (SD): 2017 Departure date: Year 7.05 (0.017) 17/08 (4.38) 01/08 (20.03) Breeding colony 0.12 (0.947) Sex 0.04 (0.849) Nº stopovers: Year 4.76 (0.044) 0.88 (1.09) 2.67 (2.5) Breeding colony 0.29 (0.835) Sex 1.03 (0.325) Duration stopovers: Year 1.57 (0.258) 3.85 (1.77) 5.69 (3.64) Breeding colony 0.83 (0.523) Sex 0.66 (0.447) Arrival date: Year 0.41 (0.530) 31/08 (6.81) 02/09 (7.63) Breeding colony 0.55 (0.658) Sex 1.44 (0.248) Migration period: Year 7.44 (0.015) 14.00 (5.91) 31.83 (21.89) Breeding colony 0.02 (0.996) Sex 0.66 (0.430) Chapter 1 50 Table 2. F statistic and associated pvalue of GLM developed for each migration variable in winter. Dependent variables are underlined. Independent variables with significant effects are highlighted in bold. Averages and standard deviations are given for dependent variables by detachment year. Sample size was 22 migratory tracks. Model F (p) Mean (SD): 2017 Mean (SD): 2018 Lat. of winter area: Year 16.14 (0.001) 10.27 (1.40) 13.36 (1.85) Breeding colony 0.32 (0.812) Sex 0.56 (0.465) Long. of winter area: Year 3.28 (0.089) -7.72 (3.68) -3.65 (3.38) Breeding colony 0.03 (0.994) Sex 0.18 (0.676) Winter area size: Year 7.58 (0.014) 102,921 (61,128) 212,410 (77,319) Breeding colony 0.05 (0.984) Sex 0.020 (0.887) Migratory distance: Year 18.74 (0.001) 3,086 (134) 2,725 (181) Breeding colony 0.03 (0.992) Sex 0.60 (0.448) Wintering period: Year 2.75 (0.117) 150.06 (11.10) 170.83 (28.44) Breeding colony 0.12 (0.947) Sex 0.26 (0.619) Description of migratory behaviour 51 Table 3. F statistic and associated pvalue of GLM developed for each migration variable in spring. Dependent variables are underlined. Independent variables with significant effects are highlighted in bold. Averages and standard deviations are given for dependent variables by year. Sample size was 22 migratory tracks. Model F (p) Mean (SD): 2017 Mean (SD): 2018 Departure date: Year 3.68 (0.073) 28/01 (9.85) 20/02 (30.93) Breeding colony 0.15 (0.929) Sex < 0.01 (0.955) Nº stopovers: Year 1.10 (0.311) 3.13 (2.96) 3.83 (1.83) Breeding colony 5.34 (0.009) Sex 0.20 (0.658) Duration stopovers: Year 0.003 (0.960) 4.13 (2.93) 4.67 (1.42) Breeding colony 4.12 (0.027) Sex 2.16 (0.164) Arrival date: Year 11.86 (0.003) 16/03 (26.42) 25/04 (15.64) Breeding colony 12.98 (0.001) Sex 0.87 (0.366) Migration period: Year 0.56 (0.464) 46.50 (22.47) 64.17 (30.82) Breeding colony 9.62 (0.001) Sex 0.78 (0.389) Chapter 1 52 Carry-over effects across migratory schedules We also found that differences in departure date from breeding areas did not accumulate throughout the migration cycle (Fig. 7). In other words, carry-over effects of migratory schedules did not propagate during the whole migration cycle. Nevertheless, we found that departure date from wintering areas affected arrival date to breeding areas (Fig. 7). This latter result highlights that the wintering period is linked to the subsequent breeding season through mediation of spring migration. Finally, our path analysis provided robust fit to data, as revealed by d-separation test (Fisher’s C = 9.05; df = 6; p= 0.171). Thus, we concluded that the hypothesized causal relationships we examined were consistent with the data. By further inspection of “qqplots”, we determined that every single linear mixed model was adequately fitted. Description of migratory behaviour 53 Figure 7. Path diagram of the causal model explaining migratory schedules. One-headed arrows represent a causal effect of one variable on another (i.e. a path). The width of arrows reflects the magnitude of standardized path coefficients, which are presented above each arrow (with standard errors in brackets). Black arrows indicate significant effects and grey arrows non-significant effects. Departure date breeding areas Arrival date wintering areas Arrival date breeding areas 0.30 (0.10) 0.23 (0.19) 0.07 (0.22) Departure date wintering areas Chapter 1 54 Discussion In this study, we have unveiled the migration system of barn swallows breeding in southwestern Spain. We utilized the most updated statistical framework to estimate positions from light data. We have determined that our studied individuals started autumn migration generally in August, and arrived to their African wintering areas around the beginning of September. From January to March, barn swallows departed from their wintering sites and arrived back to their breeding colonies from February to May. The main stopover area used during autumn migration was located in northwestern Morocco. The highest overlap of individual wintering areas was found in Ivory Coast and surroundings countries, places dominated by savannahs and crops. Regarding spring migration, most of individuals stopped over to refuel from western Algeria to northern Morocco. Surprisingly, one swallow migrated in autumn to the British Islands, then it stayed for winter in the northwestern Iberian Peninsula and afterwards it migrated again to the north. We also found a weak migratory connectivity for our breeding populations. In addition, our results highlight strong interannual effects on migratory performance. We only detected differences across breeding colonies for the spring migration. By contrast, we did not find differences between sexes in migratory performance. Finally, we showed that migration schedules did not accumulate during the full migration cycle, but only departure date Description of migratory behaviour 55 from wintering areas significantly predicted arrival date to breeding sites. In the first migration year of our study, we found a significant negative effect of tag attachment on apparent survival rates, which agrees with previous studies on aerial insectivores (Scandolara et al. 2014; Szép et al. 2017; Morganti et al. 2018). In order to reduce this negative effect, in the second migration year we attached geologgers without light stalk. This had been recommended in previous studies highlighting that light stalk should considerably increase air drag, with especial detrimental effects for aerial insectivores (Bowlin et al. 2010; Costantini & Møller 2013; Morganti et al. 2018). In agreement, during our second migration year, the negative effect of tag deployment on return rates was not significant. We successfully reduced the impact of geologgers on apparent survival rates, but this came at the cost of increased noisy light data (Bowlin et al. 2010). Consequently, it can be appreciated that uncertainty in estimated position is usually wider for migratory tracks during the second migration year (see Suppl. Mat.). As it is the basis in solar geolocation (Ekstrom 2007; Bridge et al. 2013; Rakhimberdiev et al. 2015; 2017), we found larger uncertainties for latitude than for longitude, especially when getting closer to equinoxes (e.g. two weeks after and before). In addition, we acknowledge that our position errors should be higher than others found for species moving at northern latitudes, because day length at the equator is Chapter 1 56 constant through the year. We noticed that barn swallows spent nights within rural buildings some time before they actually arrived to their breeding colonies. Therefore, strong shading events started to appear usually in February or March, corresponding to estimated positions across northern Africa. Following previous studies (Liechti et al. 2014; Hobson et al. 2015), we excluded these periods from our raw light data. This latter source of error together with the spring equinox provided in some cases extremely large uncertainties for our estimated positions (e.g. Figures 9, 10, 12 from Supp. Mat.). Most of our studied barn swallows started autumn migration in August, which is about a month earlier than previous findings on hirundines (Liechti et al. 2014; Hobson et al. 2015; Szép et al. 2017). This difference should be explained by latitude of breeding colony, which was lower in our study compared to others. In agreement, Hobson et al. (2015) sampled barn swallow breeding populations across a wide geographical range, and they found the earliest departure date for the southernmost breeding area. Actually, July and August in southwestern Spain can be extremely warm and dry, so it is expected that barn swallows leave their breeding areas by the time when resources decline. During autumn migration, roughly half of our study individuals stopped over to refuel for at least two days, up to six times and during a maximum period of 12 days. In a previous study, five out of six common swifts (Apus apus) made stopovers in autumn migration, up to five times and staying between 10 and 56 days Description of migratory behaviour 63 These encountered results must be driven by different geographical scales (Mckinnon et al. 2013). Because our study breeding colonies were located close to each other, it is easy for these birds to spread and mix together at the wintering grounds. However, when considering breeding populations sampled across all Europe, barn swallows ringed in southwestern countries were usually recovered nearby the Gulf of Guinea, but individuals ringed up to the north were mainly recaptured in South Africa (Ambrosini et al. 2009). The same pattern was found for the American barn swallows, yet migratory connectivity was strong at a broad geographical scale (Hobson et al. 2015; Finch et al. 2017), but it was weak at a local scale (Imlay et al. 2018). Some previous studies have advocated quantifying migratory connectivity only when using multiple sampling locations throughout the distribution range (e.g. Hallworth et al. 2015). Nevertheless, we think that quantifying migratory connectivity at a local scale would be useful depending on the purposes of the study or given economic and logistic limitations. We have found in this study generalized inter-annual effects in migratory performance, which is already known for barn swallows breeding in Italy (Liechti et al. 2014). As we did in our study, these authors detected significant variability among the years for departure date from breeding colonies, duration of autumn migration, locations of wintering areas and arrival date to breeding colonies. Such strong inter-annual effects were hypothesized to be weather related, and for Chapter 1 64 example, departure date from the Italian breeding colonies coincided with heavy rains (whose timing differed among years). Food availability for aerial foragers should decrease with heavy rain and therefore barn swallows may use rainfall as an environmental index to start migration, in addition to other triggers of circannual rhythms such as photoperiod (Coppack et al. 2008). Moreover, our results showed differences in spring migration across breeding colonies. Interestingly, barn swallows breeding in the largest colony under study stopped over to refuel more times and for longer periods, took more days to finish the journey and arrived later to the breeding colony than barn swallows from smaller colonies. Future studies may address the relationship between colony size and spring migration, because a higher intra-specific competition could select certain migratory behaviors (e.g. Studds & Marra 2005). In agreement with Hobson et al. (2015), we did not find any difference in migratory performance between sexes, what should be driven by small sample size on tracked individuals (Liecthi et al. 2014). This is because protandry, i.e. the earlier arrival to breeding colonies of males compared to females, has been proved for the barn swallow using much larger datasets (e.g. Møller 2004). Previous studies have found either that duration of both migrations is similar or that autumn migration last longer than spring migration (Stutchbury et al. 2009; Åkesson et al. 2012; McKinon et al. 2013; Liechti et al. 2014; Szép et al. 2017; Cooper et al. 2017). Description of migratory behaviour 65 Shorter duration of spring migration could be expected taking into account the competition among individuals to arrive earlier at the breeding grounds, given the benefits associated with earlier arrival (Marra et al. 1998; Winkler & Allen 1996; Smith & Moore 2004; Newton 2008; Marzal et al. 2013). In contrast, we found a longer duration for the spring migration, but also a higher variability in migratory schedules or the range of stopover locations. In addition, migratory timetables did not accumulate during the whole migration cycle (Senner et al. 2014), but only departure date from wintering areas significantly predicted arrival date to breeding grounds. This finding may also reflect competition for earlier arrival to the breeding grounds; but more importantly, it highlights the tight link between the winter and the breeding stage and provides a mechanism for carryover effects to occur in spring (McKinon et al. 2013; Lemke et al. 2013; Szép et al. 2017; Cooper et al. 2017). Summarizing, we have extensively deployed light-level geolocators to describe important details of the migration system of barn swallows breeding in southwestern Europe. Such information relates to distribution of non-breeding habitats, phenology of migration, migratory connectivity, factors determining migration and carry-over effects. These have been traditional gaps of knowledge in ornithology, and we have started to fill them within an unprecedented time thanks to solar geolocation. Chapter 1 66 Acknowledgements This study was funded by research projects of the Junta de Andalucía (P12-RNM-2144), Ministerio de Economía y Competitividad (CGL2015-646509) and Junta de Extremadura (IB16121). The first author was supported by an operating grant from Junta de Andalucía (P12-RNM-2144). We thank Juan Sangran and his family, Borja Lora and his family, Plácido and Marzal family for allowing us to work on their properties. Thanks also to Martin, Manuel, Antonio, Fernando, Javier, Gabriel and other staff that work in the farms. We also thank Marina Bollo and Sergio Reinoso for helping in fieldwork and data base management. Finally, we thank Álvaro Merino Alejandre for helping to elaborate R scripts. References Åkesson, S., Klaassen, R., Holmgren, J., Fox, J. W. & Hedenström, A. Migration Routes and Strategies in a Highly Aerial Migrant, the Common Swift Apus apus, Revealed by Light-Level Geolocators. PLoS ONE 7, e41195 (2012). Allen, J. Land cover classification (1 year). NASA Earth Observations http://neo.sci.gsfc.nasa.gov/view.php?datasetId=MCD12C1_T1 (2016). Ambrosini, R., Møller, A. P. & Saino, N. A quantitative measure of migratory connectivity. J. Theor. Biol. 257, 203–211 (2009). Description of migratory behaviour 67 Arizaga, J., Maggini, I., Hama, F., Crespo, A. & Gargallo, G. Site- and species-specific fuel load of European–Afrotropical passerines on arrival at three oases of southeast Morocco during spring migration. Bird Study 60, 11–21 (2013). Bächler, E. et al. Year-Round Tracking of Small Trans-Saharan Migrants Using Light-Level Geolocators. PLoS ONE 5, e9566 (2010). Berthold, P. Bird migration. A general survey (Oxford University Press, Oxford, 1993). BirdLife International. Hirundo rustica. The IUCN Red List of Threatened Species http://www.iucnredlist.org/details/22712252/0 (2016). Bolger, D. T., Newmark, W. D., Morrison, T. A. & Doak, D. F. The need for integrative approaches to understand and conserve migratory ungulates. Ecol. Lett. 11, 63–77 (2008). Bowlin, M. S. et al. The effects of geolocator drag and weight on the flight ranges of small migrants. Methods Ecol. Evol. 1, 398–402 (2010). Bradley, D. W. et al. Trans-Gulf of Mexico loop migration of tree swallows revealed by solar geolocation. Current Zoology 60, 653–659 (2014). Bridge, E. S. et al. Advances in tracking small migratory birds: a technical review of light-level geolocation: Light-Level Geolocation Dataloggers. J. Field Ornithol. 84, 121–137 (2013). Brlík, V. et al. Weak effects of geolocators on small birds: a meta-analysis controlled for phylogeny and publication bias. J. Anim. Ecol. (in press). Chapter 1 68 Cooper, N. W., Hallworth, M. T. & Marra, P. P. Light-level geolocation reveals wintering distribution, migration routes, and primary stopover locations of an endangered long-distance migratory songbird. J. Avian Biol. 48, 209–219 (2017). Coppack, T. et al. Can long-distance migratory birds adjust to the advancement of spring by shortening migration distance? The response of the pied flycatcher to latitudinal photoperiodic variation. Glob. Change Biol. 14, 2516–2522 (2008). Costantini, D. & Møller, A. P. A meta-analysis of the effects of geolocator application on birds. Curr. Zool. 59, 697–706 (2013). Drake, A., Rock, C., Quinlan, S. P., Martin, M. & Green, D. J. Wind speed during migration influences the survival, timing of breeding, and productivity of a Neotropical migrant, Setophaga petechia. PLoS ONE 9, e97152 (2014). Ekstrom, P. A. An advance in geolocation by light. Mem. Natl Inst. Polar Res., Spec. Issue 58, 210–226, (2004). Ekstrom, P. Error measures for template-fit geolocation based on light. Deep. Sea Res. Part 2: Top. Stud. Oceanogr. 54, 392–403 (2007). Eraud, C. et al. Migration Routes and Staging Areas of Trans-Saharan Turtle Doves Appraised from Light-Level Geolocators. PLoS ONE 8, e59396 (2013). Faaborg, J. et al. Conserving migratory land birds in the New World: Do we know enough? Ecol. Appl. 20, 398–418 (2010). Description of migratory behaviour 69 Finch, T., Butler, S. J., Franco, A. M. A. & Cresswell, W. Low migratory connectivity is common in long-distance migrant birds. J. Anim. Ecol. 86, 662–673 (2017). Gilroy, J. J., Gill, J. A., Butchart, S. H. M., Jones, V. R. & Franco, A. M. A. Migratory diversity predicts population declines in birds. Ecol. Lett. 19, 308–317 (2016). Hahn, S., Bauer, S. & Liechti, F. The natural link between Europe and Africa - 2.1 billion birds on migration. Oikos 118, 624–626 (2009). Hallworth, M. T. & Marra, P. P. Miniaturized GPS Tags Identify Nonbreeding Territories of a Small Breeding Migratory Songbird. Sci. Rep. 5, (2015). Hallworth, M. T., Sillett, T. S., Van Wilgenburg, S. L., Hobson, K. A. & Marra, P. P. Migratory connectivity of a Neotropical migratory songbird revealed by archival light‐level geolocators. Ecol. Appl.25, 336-347 (2015). Hama, F., Gargallo, G., Benhoussa, A., Zerdouk, S. & Idrissi, H. R. Autumn body condition of Palaearctic trans-Saharan migrant passerines at an oasis in southeast Morocco. Ringing & Migration 28, 77–84 (2013). Harrison, X. A., Blount, J. D., Inger, R., Norris, D. R. & Bearhop, S. Carryover effects as drivers of fitness differences in animals. J. Anim. Ecol. 80, 4–18 (2011). Chapter 1 70 Hays, G. C. A review of the adaptive significance and ecosystem consequences of zooplankton diel vertical migrations. Hydrobiologia 503, 163–170 (2003). Hijmans, R. J. raster: Geographic Data Analysis and Modeling. R package version 2.5-8 https://CRAN.R-project.org/package=raster (2016). Hobson, K. A. et al. A Continent-Wide Migratory Divide in North American Breeding Barn Swallows (Hirundo rustica). PLoS ONE 10, e0129340 (2015). Imlay, T. L., Hobson, K. A., Roberto-Charron, A. & Leonard, M. L. Wintering Areas, Migratory Connectivity and Habitat Fidelity of Three Declining Nearctic- Neotropical Migrant Swallows. Animal Migration 5, 1–16 (2018). Kays, R., Crofoot, M. C., Jetz, W. & Wikelski, M. Terrestrial animal tracking as an eye on life and planet. Science 348, aaa2478-1–aaa2478-7 (2015). Kovar, R., Brabec, M., Bocek, R. & Vita, R. Spring migration distances of some Central European amphibian species. Amphibia-Reptilia 30, 367– 378 (2009). Lefcheck, J. S. piecewiseSEM : Piecewise structural equation modelling in R for ecology, evolution, and systematics. MEE 7, 573–579 (2016). Lemke, H. W. et al. Annual Cycle and Migration Strategies of a Trans- Saharan Migratory Songbird: A Geolocator Study in the Great Reed Warbler. PLoS ONE 8, e79209 (2013). Description of migratory behaviour 71 Liechti, F. et al. Timing of migration and residence areas during the nonbreeding period of barn swallows Hirundo rustica in relation to sex and population. J. Avian Biol. 46, 254–265 (2015). Lope, F. de. La invernada de las Golondrinas Comunes ibéricas en África (II Reunión Iberoamericana para la Conservación y Zoología de Vertebrados, Cáceres, 1980). Luschi, P., Hays, G. C. & Papi, F. A review of long-distance movements by marine turtles, and the possible role of ocean currents. Oikos 103, 293– 302 (2003). Maggini, I. & Bairlein, F. Body condition and stopover of trans-Saharan spring migrant passerines caught at a site in southern Morocco. Ringing & Migration 26, 31–37 (2011). Marra, P. P, Hobson, K. A. and Holmes, R. T. Linking winter and summer events in a migratory bird by using stable-carbon isotopes. Science 282, 1884–1886 (1998). Marzal, A. et al. Malaria infection and feather growth rate predict reproductive success in house martins. Oecologia 171, 853–861 (2013). McKinnon, E. A., Fraser, K. C. & Stutchbury, B. J. M. New discoveries in landbird migration using geolocators, and a flight plan for the future. The Auk 130, 211–222 (2013). Møller, A. P. Sexual selection and the barn swallow (Oxford University Press, 1994). Chapter 1 72 Morganti, M. et al. Effect of light-level geolocators on apparent survival of two highly aerial swift species. J. Avian Biol. e01521 (2018). Nebel, S., Mills, A., McCracken, J. D. & Taylor, P. D. Declines of aerial insectivores in North America follow a geographic gradient. Avian Conserv. Ecol. 5, (2010). Newton, I. The migration ecology of birds (Academic Press, 2008). Norris, D. R. & Marra, P. P. Seasonal interactions, habitat quality, and population dynamics in migratory birds. The Condor 109, 535 (2007). Nus, T. van & Neto, J. M. Urban Roost of Wintering Barn Swallows Hirundo rustica in Aveiro, Portugal. Ardea 105, 1–6 (2017). O’Connor, C. M., Norris, D. R., Crossin, G. T. & Cooke, S. J. Biological carryover effects: linking common concepts and mechanisms in ecology and evolution. Ecosphere 5, 28 (2014). Ouwehand, J. et al. Light-level geolocators reveal migratory connectivity in European populations of pied flycatchers Ficedula hypoleuca.J. Avian Biol. 47, 69–83 (2016). Patchett, R., Finch, T. & Cresswell, W. Population consequences of migratory variability differ between flyways. Curr. Biol. 28, R340–R341 (2018). Pebesma, E. J. & Bivand, R. S. Classes and methods for spatial data in R. R News 5 (2) https://cran.r-project.org/doc/Rnews/ (2005). Wintering areas and breeding phenology 79 Abstract Understanding connections between breeding, stopover and wintering grounds for long-distance migratory birds can provide important insight into factors influencing demography and the strength of carryover effects among various periods of the annual cycle. Using previously described, multi-isotope (δ13C, δ15N, δ2H) feather isoscapes for Africa, we identified the most probable wintering areas for house martins Delichon urbica breeding at Badajoz in southwestern Spain. We identified two most-probable wintering areas differing in isotopic signature in West Africa. We found that the probability to winter in the isotopic cluster two was related to age and sex of individuals. Specifically, experienced males (i.e. two years or older) winter in the isotopic cluster two with a greater probability than experienced females, whereas first-year females winter in the isotopic cluster two with a greater probability than first-year males. In addition, wintering area was correlated with breeding phenology, with individuals wintering in the isotopic cluster two initiating their clutches earlier than those wintering in the isotopic cluster one. For birds wintering in the isotopic cluster two, there was no relationship between age and clutch initiation date. In contrast, young birds wintering in the isotopic cluster one initiated their clutches earlier than experienced birds wintering in this area. There was no significant correlation between wintering area and clutch size or the number of fledglings produced. We hypothesize that the relationship among social status, population Chapter 2 80 density and winter habitat quality should be the most important driver of the carry-over effect we found for this population. Introduction Migratory connectivity, or the degree of connection among individuals between breeding and wintering sites and along migratory routes, has important consequences for the ecology, evolution and conservation of migratory species and populations (Webster et al. 2002). Populations of migratory birds with strong connectivity may be more vulnerable to habitat disturbance occurring at their wintering areas compared to populations consisting of individuals wintering across several wintering locations (Webster et al. 2002; Fraser et al. 2012). Examination of such connected ecosystems can provide important insights into how environmental changes may affect migrants, since different habitats along the migratory route may experience different perturbations or environmental changes (Both et al. 2006). Many migratory birds, especially aerial insectivores and long-distance migrants, have shown general declining population trends (Sanderson et al. 2006; Nebel et al. 2010; but see Michel et al. 2015). Specifically, the Palearctic population of the house martin Delichon urbica has declined during 1990–2000 and the global population is also declining (Sanderson et al. 2006; Birdlife International 2012). However, causes for these declines and the degree Wintering areas and breeding phenology 81 of migratory connectivity in this species are poorly known (Hobson et al. 2012a). In spite of the huge amount of information provided by ring recoveries, this traditional approach to studying migration has been insufficient to identify the wintering sites of the Palearctic house martin. The overall trans-Sahara recovery ratio for house martins ringed in the European breeding quarters is 1:51,000. Unfortunately, for the 23,948 individuals ringed in Spain, none have been recaptured in their African winter areas (Hill 1997). Probably, this was because this area is greatly impoverished and it is involved in multiple political conflicts which prevent ornithological research in the area. Fortunately, the analysis of endogenous markers like stable isotope ratios in feathers, allow estimation of the wintering areas of birds that are otherwise not easily studied by other methods (Chamberlain et al. 1997, 2000; Hobson & Wassenaar 1997, 2001, 2008; Rubenstein et al. 2002; Rubenstein & Hobson 2004; Robinson et al. 2009).For instance, Hobson et al. (2012a) assigned a house martin breeding population from the Netherlands to wintering sites in West Africa based on stable isotope ratios in winter-grown feathers and constrained their assignment longitudinally based on ring recoveries. These authors proposed a method of combining isotope data with ring recoveries to estimate wintering ranges among breeding and wintering geographical locations in Europe and Africa, respectively. Other European breeding populations of house martins were also assigned with highest Chapter 2 82 probability to winter in West Africa, by examining concordance between the normalized difference vegetation index (NDVI) in Africa and local breeding population trends (Ambrosini et al. 2011), and by linking winter rainfall values with survival rate (Robinson et al. 2008). Establishing migratory connectivity using isotopic and other methods has also provided information on how individuals differing in their age or sex also differ in their migration strategies or their use of winter habitat. For instance, male American redstarts Setophaga ruticilla wintering in high-quality winter habitats (inferred from stable-carbon isotope analyses) arrived earlier to their breeding areas and in better physical condition than males wintering in low quality winter habitats (Marra et al. 1998). Furthermore, Norris et al. (2004) predicted for redstarts that individuals wintering in higher-quality habitats also reared more fledglings at the breeding grounds; moreover the strength of this carry-over effect was higher for females than for males. Another example is found in yellow warblers Setophaga petechial, where only yearling females wintering in high-quality habitats (inferred from carbon and nitrogen stable-isotope analyses), initiated reproduction earlier and thus raised more fledglings than yearling females wintering in drier habitats (Drake et al. 2013). Comparable results were obtained utilizing NDVI, a large-scale measure of plant productivity on the wintering grounds. For example, in the Palearctic barn swallow Hirundo rustica, Saino et al. (2004) detected earlier spring arrival dates in Italy for older individuals after Wintering areas and breeding phenology 83 winters with more favorable conditions in Africa. Similarly, Balbontin et al. (2009) found an advance in arrival dates of barn swallows with improved ecological conditions at stopover areas in North Africa. The extent to which winter habitats cause carry-over effects on subsequent breeding performance may depend on age and sex (Marra et al. 1998; Norris et al. 2004; Saino et al. 2004; Balbontin et al. 2009; Drake et al. 2013), but also on the specific migratory species. For instance, it has been shown that delayed arrivals or departures at a given stage of the migratory route, do not accumulate and therefore do not affect subsequent breeding success in Hudsonian godwit Limosa haemastica or wood thrush Hylocichla mustelina (Senner et al. 2014; McKinnon et al. 2015). In addition, for 19 migratory bird species breeding in the United Kingdom, climatic conditions on the breeding grounds were more important drivers of breeding phenology than carry-over effects from the winter areas (Ockendon et al. 2013). Thus, it is necessary to be cautious when generalizing carry-over effects across different age and sex categories, or across different taxa. Here we used stable isotope methods to examine migratory connectivity with the aim to 1) identify general wintering areas in Africa of two house martin colonies breeding in southwest Spain, 2) to determine if age or sex were correlated with winter habitat use, and 3) to determine whether African winter habitat use had any carry-over effect during the subsequent breeding season in Europe by investigating breeding phenology (i.e. laying date) and breeding Chapter 2 84 success (i.e. clutch size and number of fledglings). Based on the previous studies, we expected our house martins to winter in West Africa. We predicted that experienced birds (i.e. two years old or older), which have already accomplished the migratory journey at least on one occasion, would winter in habitats of higher quality than young birds that had never accomplished the migratory journey before. Also, we predicted that males would be more competitive than females with respect to winter habitat selection since they are prone to arrive earlier than females to the breeding areas to increase mating and reproductive success. We also predicted that those individuals wintering in high quality habitats could begin to reproduce earlier and could have a greater reproductive success than birds wintering in low quality habitats. Materials and methods Field procedures Our study took place in two breeding colonies of house martins at Badajoz (38°52’N, 7°05’W), southern Spain, during an 8-yr period (2005–2013, excluding 2006). Adult house martins were captured at dawn in nests or using mist nets placed at the colony. Birds captured for the first time were marked with numbered metal rings, and recaptured birds were identified by their ring. From February to July, Wintering areas and breeding phenology 85 we followed reproductive events inspecting nests every second day to determine the start of laying and hatching. When nestlings were nine days old, we captured adult house martins at dawn in their nests and recorded tarsus length with a digital caliper to the nearest 0.01 mm, and body mass with a Pesola spring balance to the nearest 0.5 g. From each adult we took the outermost rectrix for isotopic analysis (original sample size of 300 birds). Each year we performed more than one hundred capture sessions from March to late June. When recapture rate of individuals captured in a given session was >90%, capture sessions were stopped. We aged house martins confidently because this species shows high breeding site philopatry, and recruitment occurs at the age of one year (de Lope and da Silva 1988). Briefly, breeding dispersal is negligible in our study sites with less than 0.1% of breeding individuals (i.e. ≥ first year of age) among the more than 10 000 adults ringed in the study population during more than 20 yr ever being recorded at another colony in the same or subsequent years (Marzal et al. 2016). Moreover, age at first reproduction occurred at the age of one year for the majority of individuals since all but one of more than 500 individuals marked as nestlings were recruited as firstyear breeding adults. For the analyses of winter habitat use and carryover effects (respectively second and third objectives), we included 197 adult house martins. We categorized as young birds individuals ringed as nestlings/fledglings and recaptured in the next year as yearlings (n = 51; 22 males and 29 females; Online Supplementary Material Appendix 1, Table A1). We categorized as experienced birds Chapter 2 86 house martins ringed as nestlings/fledglings and recaptured two years or more after their first capture (n = 28; 13 males and 15 females). In addition we also included as experienced birds individuals ringed for the first time as adults and recaptured in subsequent years (n = 118; 46 males and 71 females). Hence, the total sample size for experienced birds was 146 individuals (59 males and 86 females; Online Supplementary Material Appendix 1, Table A1). Stable isotope analysis All feathers were cleaned of surface oils in 2:1 chloroform:methanol solvent rinse and prepared for δ2H, δ13C and δ15N analysis at the Stable Isotope Laboratory of Environment Canada, Saskatoon, Canada. Deuterium in the non-exchangeable hydrogen of feathers was determined using the method described by Wassenaar & Hobson (2003) and using three calibrated keratin hydrogen-isotope reference materials (CBS: -197 ‰ SPK: -121.6 ‰, KHS: -54.1 ‰). Hydrogen isotopic measurements were performed on H2gas derived from hightemperature (1350C) flash pyrolysis of 350±10 µg feather subsamples and keratin standards using continuous-flow isotope-ratio mass spectrometry CFIRMS with a Eurovector 3000 (Milan, Italy - www.eurovector.it) elemental analyser interfaced with an Isoprime (Manchester, UK) mass spectrometer. Measurement of the three keratin laboratory reference materials corrected for linear instrumental Wintering areas and breeding phenology 87 drift were both accurate and precise with typical within-run (n=5) SD values of < 2 ‰. All results are reported for non-exchangeable H expressed in the typical delta (δ) notation, in units of per mil (‰), and normalized on the Vienna Standard Mean Ocean Water – Standard Light Antarctic Precipitation (VSMOW-SLAP) standard scale. For δ13C and δ15N analyses, between 0.5 and 1.0 mg of feather material was combusted online using a Eurovector 3000 elemental analyzer. The resulting CO2and N2was separated by Gas Chromatograph (GC) and introduced into a Nu Horizon (Nu Instruments, Wrexham, UK - www.nu-ins.com) triple-collector isotope-ratio mass-spectrometer via an open split and compared to CO2or N2reference gas. Stable nitrogen (15N/14N) and carbon (13C/12C) isotope ratios were expressed in delta (δ) notation, as parts per thousand (‰) deviation from the primary standards: atmospheric nitrogen and VPDB (Vienna Pee Dee Belemnite carbonate) standards, respectively. Using previously calibrated internal laboratory C and N standards, powdered keratin (BWBIII, δ13C: -20‰, δ15N: 14.4‰) and gelatin (PRGEL, δ13C: -13.6‰, δ15N: 4.73‰), within run (n=5), precisions for δ15N and δ13C measurements were ~ ± 0.15 ‰. Chapter 2 88 Defining wintering areas Hobson et al. (2012b) identified four isotopically distinct feather clusters in Africa based on combined δ13C, δ15N and δ2H values (i.e. isoscapes). Briefly, this approach assigns each sample feather to its most-probable cluster origin in Africa, given the regional mean expected values of the three isotopes and the covariance among them. Here, the δ13C and δ15N isoscapes were based on plant physiology and soil models and the δ2H isoscape was based on the measured longterm Global Network for Isotopes in Precipitation (GNIP). These isoscapes were converted to feather isoscape equivalents using assumed isotopic discrimination factors (δ13C and δ15N) or transfer functions (δ2H) based on measured data for known-origin birds. Feather isoscapes were then grouped in different multivariate spaces (i.e. isotopic clusters). This was accomplished using the “clara” algorithm, which is based on the cluster analysis using the partitioning around the medoids (PAM clustering algorithm; see details Hobson et al. 2012b). The isotopic clusters separated well in multivariate space, with relatively low overlap between 90% confidence ellipsoids between clusters (a crossvalidated DFA suggested that 94% of GIS cells within the African continent could be accurately placed into a given isotopic cluster). The optimal number of clusters was determined using a cross-validation criterion known as “prediction strength”: the highest mean prediction strength was found by using four different clusters for δ13C, δ15N and δ2H. Across 50 randomized Wintering areas and breeding phenology 95 Figure 1. Map depicting assigned wintering areas for house martin Delichon urbica breeding at Badajoz (southwestern Spain). The continuous and dotted vertical lines indicate the mean and 95% confidence intervals wintering longitude estimated for the studied population using linear regression analysis (Hobson et al. 2012a). Predicted wintering areas correspond to isotopic clusters 1 and 2 throughout West Africa. Sample size is 300 feathers collected from our two study populations. Chapter 2 96 Second objective: relationships of age and sex on winter habitat use The final model of the analysis of probability to winter either in the isotopic cluster two or the isotopic cluster one showed a significant effect of the interaction term between age and sex (Table 2). The confidence interval of this interaction term excluded zero. This interaction indicated that experienced males (i.e. two years or older) winter in the isotopic cluster two with a greater probability than experienced females (estimated probability: 0.83 ± 0.051 in experienced males, 0.64 ± 0.061 in experienced females), whereas young females winter in the isotopic cluster two with greater probability than young males (estimated probability: 0.55 ± 0.096 in young females; 0.38 ± 0.106 in young males; Fig. 2). Birds differing in body condition did not differ in their probability to winter in either isotopic cluster, as shown by the confidence interval for this predictor including zero (Table 2). The origin of the birds (‘hatched at the study colonies’ or ‘hatched outside the study colonies’) did not affect the probability to winter in either isotopic cluster, as shown by the confidence interval for this predictor including zero (Table 2). Top models selected for averaging are shown in Online Supplementary Material Appendix 2, Table A2. To test if experienced individuals winter in different areas from young ones regardless of sex, we repeated this analysis of probability to winter in either area excluding the interaction term between age and sex. This analysis confirmed that Wintering areas and breeding phenology 97 age showed a significant difference between wintering areas [estimate = –1.57, SE = 0.71, relative importance (RI) = 0.95, 95% CI = (–2.97, –0.18)]. Thus, experienced house martins winter in the isotopic cluster two with a higher probability than young ones (estimated probability: 0.73 ± 0.041 for experienced individuals and 0.49 ± 0.072 for young individuals). In the latter analysis, sex did not correlate with the probability to winter in the isotopic cluster two [estimate = –0.38, SE = 0.36, RI = 0.37, 95% CI = (–1.09, 0.33)]. Table 2. Summary results after model averaging of the relationship between age and sex of house martins on the probability to winter in either isotopic cluster. Sample size is 163 individuals. Parameters estimated for predictors with confidence intervals that do not include zero are highlighted in bold. The reference level of the fixed factor sex was ‘female’, for age was ‘young’ and for origin was ‘hatched at the study colonies’. RI – Relative importance. Parameter Estimate SE 95% CI RI (Intercept) -0.110 0.504 (-1.10, 0.88) Age -1.065 0.822 (-2.68, 0.55) 1.00 Sex 0.540 0.711 (-0.85, 1.93) 0.77 Origin 0.858 0.630 (-0.38, 2.09) 0.48 Scaled body mass index -0.063 0.394 (-0.83, 0.71) 0.25 Age*Sex -1.783 0.773 (-3.30, -0.27) 0.63 Chapter 2 98 Figure 2. Relationship between the interaction term of age (young vs experienced) and sex on the probability to winter in the isotopic cluster two of West Africa for house martins Delichon urbica from Badajoz (southwestern Spain). Sample size is 45 experienced males, 8 young males, 39 experienced females and 15 young females wintering in the isotopic cluster two. Error bars indicate mean percentage ± 1 SE. Wintering areas and breeding phenology 99 Third objective: carry over effects from wintering areas on reproductive performance The final model of the analysis explaining variation in laying date showed a significant effect of the interaction term between age and isotopic cluster. The confidence interval for this predictor excluded zero (Table 3). Specifically, this interaction indicated that there was no difference in laying date between young and experienced birds wintering in the isotopic cluster two (laying date mean ± SE: 113.8 ± 3.2d for young individuals; 114.1 ± 2.9 d for experienced birds), whereas there was a difference in laying date between young and experienced birds wintering in the isotopic cluster one (Fig. 3). Specifically, young birds start to reproduce earlier than experienced ones when wintering in the isotopic cluster one (laying date mean ± SE: 118.4 ± 3.1 d for young birds; 125.8 ± 3.8 d for experienced individuals). Neither sex nor body condition were correlated with laying date, as shown by the confidence interval of these predictors including zero. Top model sets selected for averaging are shown in Online Supplementary material Appendix 2, Table B2. For the model in which age was excluded, we found that birds wintering in the isotopic cluster two bred earlier than birds wintering in the isotopic cluster one (Table 3). Specifically laying date (mean ± SE) for birds wintering in the isotopic cluster two was 114.0 ± 2.2 d, while for birds wintering in the isotopic cluster one was 121.5 ± 2.4 d. In contrast, we did not find any significant effect of winter area on clutch size nor on Chapter 2 100 the number of fledglings, while controlling for laying date and body condition (Table 4). None of the predictors included in these models had a significant effect between the response variables. Mean ± SE clutch and brood size for birds wintering in the isotopic cluster two were respectively, 4.5 ± 0.1 eggs and 3.6 ± 0.1 young; whereas mean ± SE clutch and brood size for birds wintering in the isotopic cluster one were respectively 4.5 ± 0.1 eggs and 3.3 ± 0.1 young. We also introduced age effects in the global model explaining variation in clutch size and the number of fledglings. Again, none of the predictors included in these models had a significant effect with the response variables as all their confidence intervals included zero (Online Supplementary material Appendix 3, Table A3). Wintering areas and breeding phenology 101 Table 3. Summary results after model averaging of the effects of spending the winter in either isotopic cluster on laying date in the house martin. Sample size is 196 individuals in the first model (upper section), which did not include age effects. Results for known age individuals are shown in the lower section (sample size is 112 birds). Parameters estimated for predictors with confidence intervals that do not include zero are highlighted in bold. The reference level of the fixed factor sex was ‘female’, for age was ‘young’ and for isotopic cluster was ‘two’. RI – Relative importance. Parameter Estimate SE 95% CI RI (Intercept) 4.75 0.033 (4.685, 4.816) Isotopic cluster 0.082 0.022 (0.039, 0.125) 1.00 Sex -0.027 0.021 (-0.069, 0.014) 0.56 Scaled body mass index 0.014 0.018 (-0.021, 0.05) 0.32 Sex*Isotopic cluster 0.052 0.033 (-0.014, 0.118) 0.29 (Intercept) 4.744 0.049 (4.648, 4.841) Isotopic cluster 0.052 0.042 ( -0.032, 0.136) 1.00 Sex -0.04 0.027 (-0.095, 0.014) 0.58 Scaled body mass index 0.038 0.026 (-0.014, 0.089) 0.47 Age 0.009 0.059 (-0.106, 0.125) 0.62 Sex* Isotopic cluster 0.054 0.049 (-0.043, 0.151) 0.21 Age* Isotopic cluster 0.114 0.056 (0.003, 0.225) 0.43 Chapter 2 102 Figure 3. Relationship between the interaction term of age (young vs experienced) and isotopic cluster (1 vs 2) on laying date for house martins Delichon urbica from Badajoz (southwestern Spain). Black squares represent mean laying dates for each age category wintering in either isotopic cluster. Within the y-axis, ‘day 75’ corresponds to 16 March and ‘day 150’ to 30 May. Sample size is 112 adult birds. Wintering areas and breeding phenology 103 Table 4. Summary results after model averaging of the effects of spending the winter in either isotopic cluster on clutch size and number of fledglings for house martins. Sample size is 196 individuals. Parameters estimated for predictors with confidence intervals that do not include zero are highlighted in bold. The reference level of the fixed factor sex was ‘female’ and for isotopic cluster was ‘two’. RI – Relative importance. Parameter Estimate SE 95% CI RI Clutch Size: (Intercept) 1.481 0.044 (1.39, 1.57) Isotopic cluster 0.007 0.073 (-0.136, 0.15) 0.27 Sex 0.044 0.069 (-0.091, 0.18) 0.31 Scaled body mass index -0.021 0.068 (-0.155, 0.11) 0.26 Laying date -0.106 0.069 (-0.24, 0.03) 0.54 Sex* Isotopic cluster 0.005 0.14 (-0.27, 0.28) 0.01 Nº of fledglings: (Intercept) 1.229 0.051 (1.13, 1.33) Isotopic cluster -0.04 0.083 (-0.2, 0.123) 0.29 Sex 0.047 0.079 (-0.11, 0.202) 0.30 Scaled body mass index 0.043 0.077 (-0.11, 0.193) 0.28 Laying date -0.081 0.078 (-0.23, 0.073) 0.36 Sex* Isotopic cluster -0.091 0.16 (-0.40, 0.223) 0.02 Chapter 2 104 Discussion Using a multi-isotope (δ13C, δ15N, δ2H) analysis of feathers, we predicted two primary wintering areas in West Africa (i.e. isotopic cluster one and two) for house martins breeding in southwestern Europe. We found that experienced individuals mainly winter in the isotopic cluster two. Furthermore, birds wintering in the isotopic cluster two initiated their clutches sooner than those wintering in the isotopic cluster one. We also showed that experienced males wintered in the isotopic cluster two with a greater probability than experienced females, whereas young females wintered in the isotopic cluster two with a greater probability than young males. Also, young birds from the isotopic cluster one reproduced earlier than experienced birds from the same wintering region. Given the low trans-Sahara recovery ratio for house martins ringed in the European breeding quarters, we provide a correlative approach that provides insights into the winter ecology of this species. Our results can in turn be helpful for understanding the causes of the declining population trend of this species throughout Europe. Previous studies have linked higher winter habitat quality to higher proportion of dominant individuals (Marra & Holmes 2001), to earlier breeding phenology and to improved reproductive success (Marra et al. 1998; Norris et al. 2004). We found that the proportion of experienced males (i.e. the dominant class) is much higher in the Wintering areas and breeding phenology 111 Gelman, A. Scaling regression inputs by dividing by two standard deviations. Stat. Med. 27, 2865–2873 (2008). Grueber, C. E., Nakagawa, S., Laws, R. J. & Jamieson, I. G. Multimodel inference in ecology and evolution: challenges and solutions. J. Evol. Biol. 24, 699–711 (2011). Harrison, X. A., Blount, J. D., Inger, R., Norris, D. R. & Bearhop, S. Carryover effects as drivers of fitness differences in animals. J. Anim. Ecol. 80, 4–18 (2011). Hasselquist, D. Polygyny in great reed warblers: a long-term study of factors contributing to male fitness. Ecology 79, 2376–2390 (1998). Hill, L. A. Trans-Sahara recoveries of house martins Delichon urbica, with discussion on ringing, roosting and sightings in Africa. Safring News 26, 7–12 (1997). Hobson, K. A. & Wassenaar, L. I. Linking breeding and wintering grounds of Neotropical migrant songbirds using stable hydrogen isotopic analysis of feathers. Oecologia 109, 142–148 (1997). Hobson, K. A. & Wassenaar, L. I. Isotopic delineation of North American migratory wildlife populations: loggerhead shrikes. Ecol. Appl. 11, 1545– 1553 (2001). Hobson, K. A. & Wassenaar, L. I. Tracking animal migration using stable isotopes. Academic Press/Elsevier. (2008). Chapter 2 112 Hobson, K. A., Van Wilgenburg, S. L., Piersma, T. & Wassenaar, L. I. Solving a migration riddle using isoscapes: house martins from a Dutch village winter over west Africa. PLoS One 7, e45005 (2012a). Hobson, K. A. et al. A multi-isotope (δ13C, δ15N, δ2H) feather isoscape to assign Afrotropical migrant birds to origins. Ecosphere 3, 44 (2012b). López-Calderón, C. et al. Data from: Wintering areas predict age-related breeding phenology in a migratory passerine bird. Dryad Digital Repository http://dx.doi.org/10.5061/dryad.5sg24 (2016). Lozano, G. A., Perreault, S. & Lemon, R. E. Age, arrival date and reproductive success of male American redstarts Setophaga ruticilla.J. Avian Biol. 27, 164–170 (1996). Marra, P. P. & Holmes, R. T. Consequences of dominance mediated habitat segregation in American redstarts during the nonbreeding season. Auk 118, 92–104 (2001). Marra, P. P., Hobson, K. A. & Holmes, R. T. Linking winter and summer events in a migratory bird by using stable carbon isotopes. Science 282, 1884–1886 (1998). Marzal, A. et al. Malaria infection and feather growth rate predict reproductive success in house martins. Oecologia 171, 853–861 (2013). Marzal, A. et al. A longitudinal study of age-related changes in Haemoproteus infection in a passerine bird. Oikos 125, 1092–1099 (2016). Wintering areas and breeding phenology 113 Mazerolle, M. J. Package ‘AICcmodavg’: model selection and multimodel inference based on (Q) AIC (C). R package ver. 2.0-3 https://cran.rproject.org/web/packages/AICcmodavg/AICcmodavg.pdf (2015). McKinnon, E. A., Stanley, C. Q. & Stutchbury, B. J. Carry-over effects of nonbreeding habitat on start-to-finish spring migration performance of a songbird. PLoS One 10, e0141580 (2015). Michel, N. L., Smith, A. C., Clark, R. G., Morrissey, C. A. & Hobson, K. A. Differences in spatial synchrony and interspecific concordance inform guild-level population trends for aerial insectivorous birds. Ecography 39, 774–786 (2015). Møller, A. P. Sexual selection in the monogamous swallow Hirundo rustica. II. Mechanisms of intersexual selection. J. Evol. Biol. 5, 603–624 (1992). Møller, A. P. Phenotype-dependent arrival time and its consequences in a migratory bird. Behav. Ecol. Sociobiol.35, 115–122 (1994). Nebel, S., Mills, A., McCracken, J. D. & Taylor, P. D. Declines of aerial insectivores in North America follow a geographic gradient. Avian Conserv. Ecol. 5, 1 (2010). Norris, D. R., Marra, P. P., Kyser, T. K., Sherry, T. W. & Ratcliffe, L. M. Tropical winter habitat limits reproductive success on the temperate breeding grounds in a migratory bird. Proc. R. Soc. B 271, 59–64 (2004). Nussey, D. H., Froy, H., Lemaitre, J. F., Gaillard, J. M. & Austad, S. N. Senescence in natural populations of animals: widespread evidence and Chapter 2 114 its implications for bio-gerontology. Ageing Res. Rev. 12, 214–225 (2013). Ockendon, N., Leech, D. & Pearce-Higgins, J. W. Climatic effects on breeding grounds are more important drivers of breeding phenology in migrant birds than carry-over effects from wintering grounds. Biol. Lett. 9, 20130669 (2013). Peig, J. & Green, A. J. The paradigm of body condition: a critical reappraisal of current methods based on mass and length. Funct. Ecol. 24, 1323–1332 (2010). R Core Team. The R Project for Statistical Computing. R Foundation for Statistical Computing http://www.R-project.org (2016). Richards, S. A. Dealing with overdispersed count data in applied ecology. J. Appl. Ecol. 45, 218–227 (2008). Robinson, R. A., Balmer, D. E. & Marchant, J. H. Survival rates of hirundines in relation to British and African rainfall. Ring. Migr. 24, 1–6 (2008). Robinson, W. D., et al. Integrating concepts and technologies to advance the study of bird migration. Front. Ecol. Environ. 8, 354–361 (2009). Rubenstein, D. R. & Hobson, K. A. From birds to butterflies: animal movement patterns and stable isotopes. Trends Ecol. Evol.19, 256–263 2004. Wintering areas and breeding phenology 115 Rubenstein, D. R. et al. Linking breeding and wintering ranges of a migratory songbird using stable isotopes. Science 295, 1062–1065 (2002). Saino, N. et al. Ecological conditions during winter predict arrival date at the breeding quarters in a trans-Saharan migratory bird. Ecol. Lett.7, 21–25 (2004). Sanderson, F. J., Donald, P. F., Pain, D. J., Burfield, I. J. & Van Bommel, F. P. Long-term population declines in Afro-Palearctic migrant birds. Biol.Cons. 131, 93–105 (2006). SEDAC. Population, landscape and climate estimates, v3:biomes, Africa. Socioeconomic Data And Applications Centre (SEDAC) http://sedac.ciesin.columbia.edu/downloads/maps/nagdc/nagdcpopulationlandscape-climate-estimates-v3/place3-biomes-africa.jpg (2012). Senner, N. R., Hochachka, W. M., Fox, J. W. & Afanasyev, V. An Exception to the Rule: Carry-Over Effects Do Not Accumulate in a Long-Distance Migratory Bird. PLoS ONE 9, e86588 (2014). Sergio, F. et al. Individual improvements and selective mortality shape lifelong migratory performance. Nature 515, 410–413 (2014). Smith, R. J. & Moore, F. R. Arrival timing and seasonal reproductive performance in a long-distance migratory landbird. Behav. Ecol. Sociobiol.57, 231–239 (2004). Chapter 2 116 Turner, A. K. Family Hirundinidae (swallows and martins) in Handbook of the Birds of the World (eds del Hoyo, J., Elliott, A. & Christie, D.) 602– 685 (Lynx Edicions, 2004). Wassenaar, L. I. & Hobson, K. A. Comparative equilibration and online technique for determination of non-exchangeable hydrogen of keratins for use in animal migration studies. Isotopes Environ. Health Stud.39, 211– 217 (2003). Webster, M. S., Marra, P. P., Haig, S. M., Bensch, S. & Holmes, R. T. Links between worlds: unraveling migratory connectivity. Trends Ecol. Evol. 17, 76–83 (2002). Winkler, D. W. & Allen, P. E. The seasonal decline in tree swallow clutch size: physiological constraint or strategic adjustment? Ecology 77, 922– 932 (1996). Wintering areas and reproductive success 117 Chapter 3 118 Chapter 3 Wintering areas and reproductive success 119 Environmental conditions during winter predict age- and sex-specific differences in reproductive success of a trans-Saharan migratory bird Cosme López-Calderón1, Keith A. Hobson2, Alfonso Marzal3, Javier Balbontín1, María Isabel Reviriego3, Sergio Magallanes3, Luz García-Longoria3, Florentino de Lope3and Anders P. Møller4 1Department of Zoology, University of Seville (Seville, Spain) 2Environment and Climate Change Canada (Saskatoon, Canada) and University of Western Ontario (London, Canada) 3Department of Anatomy, Cellular Biology and Zoology, University of Extremadura (Badajoz, Spain) 4Laboratoire d'Ecologie, Systématique et Evolution, Université Paris-Sud (Orsay, France) López-Calderón et al. Scientific Reports (2017), doi: 10.1038/s41598-017-18497-2 Chapter 3 120 Abstract Previous studies have linked winter habitat with subsequent breeding phenology and physical condition of migratory birds, but few have found delayed effects of winter habitat on subsequent reproductive success. The aim of this study was to test if African winter habitat is related to subsequent reproductive success of house martins (Delichon urbicum) breeding at a colony in Spain. We measured stable isotope (δ2H, δ13C, δ15N) values from feathers moulted in West Africa and used confirmatory path analysis to test if isotopic values of wintergrown feathers were related to reproductive success through the mediation of breeding phenology and body condition. We conducted separate analyses for males, females and age classes (yearlings vs ≥ 2 years old). Experienced males wintering in habitats of higher rainfall (as inferred from lower feather δ2H values) were in better body condition and produced more offspring during the subsequent breeding season. In contrast, we did not find any effect of winter habitat on reproductive success of young males or females. These findings provide evidence consistent with a complex causal link between winter habitat quality and subsequent breeding success of long-distance migratory songbirds. Wintering areas and reproductive success 127 body condition we used the scaled mass index, which standardizes body mass at a fixed value of a linear body measurement (tarsus) based on the scaling relationship between mass and body length (Peig & Green 2010). Stable isotope analysis All feathers were cleaned of surface oils in 2:1 chloroform:methanol solvent rinse and prepared for δ2H, δ13C and δ15N analysis. Deuterium abundance in the non-exchangeable hydrogen of feathers was determined following standard procedures (Wassenaar & Hobson 2003), and using three calibrated keratin hydrogen-isotope reference materials (CBS = -197‰; SPK = -121.6‰; KHS = -54.1‰). Deuterium measurement was performed on H2gas derived from hightemperature (1350C) flash pyrolysis of 350±10 µg feather subsamples and keratin standards. Measurement of the three keratin laboratory reference materials, corrected for linear instrumental drift, were both accurate and precise with typical within-run (n=5) SD values of < 2‰. For feather δ13C and δ15N analyses, between 0.5 and 1.0 mg of feather material was combusted online using a Eurovector 3000 elemental analyzer (Eurovector, Milan, Italy). The resulting CO2 and N2was separated by Gas Chromatograph (GC) and introduced into a Nu Horizon (Nu Instruments, Wrexham, UK - www.nuins.com) triple-collector isotope-ratio mass-spectrometer via an open Chapter 3 128 split and compared to CO2or N2reference gas. Using previously calibrated internal laboratory C and N standards [powdered keratin (BWBIII; δ13C = -20‰; δ15N = 14.4‰) and gelatin (PUGEL; δ13C = - 13.6‰; δ15N = 4.73‰)], within run (n=5), precisions for δ15N and δ13C measurements were ~ ± 0.15‰. Stable isotope ratios are reported in standard delta (δ) notation relative to VSMOW for δ2H, VPDB for δ13C, and AIR for δ15N analyses. Environmental conditions in the wintering areas of house martins House martins have feather δ2H values reflecting long-term, amountweighted, average δ2H from precipitation prior to moult according to established calibration equations (Bowen et al. 2005; Hobson et al. 2012b), while δ13C and δ15N values are associated more directly from inorganic and organic sources to primary production following isotopic discrimination (Vanderklift & Ponsard 2003; Hobson et al. 2012b). Feather keratin is metabolically inert after synthesis (Hobson 1999) and so isotopic values in feathers reflect the environmental conditions where they were grown (i.e. the wintering areas). Considerable literature has emphasized that foodweb δ2H, δ13C and δ15N values are strongly influenced by climate (reviewed by West et al. 2010). Lower values of δ2H in rain are linked to higher amount of precipitation in tropical latitudes (i.e. the so-called “amount effect”; Dansgaard 1964; Clark and Fritz 1997). Higher values of δ13C are Wintering areas and reproductive success 129 associated with environments dominated by C4 plants as well as C3 plants adapted to hydric stress (Ehleringer 1989). Finally, xeric/cultivated habitats tend to be relatively enriched in 15N relative to mesic/uncultivated habitats (Pardo & Nadelhoffer 2010). The winter areas we previously identified for house martins breeding in southwestern Spain (López-Calderón et al. 2017) were defined by multi-isotopic clusters described for Africa (Hobson et al. 2012b). These isotopic clusters were closely associated with different African biomes, broadleaf forest in the so-called “Cluster 1” and savannah in “Cluster 2” (Fig. 1). Values of feather δ2H and δ13C differed markedly between the two areas of West Africa (López- Calderón et al. 2017), reflecting foodwebs in savannahs relatively enriched in 13C and depleted in 2H. In this wintering area, the savannah receives more rainfall during the rainy season than the broadleaf forest to the south (Sultan & Janicot 2003). Thus, we expect the foodweb to be relatively enriched in 13C here due to expected C4- dominated grasslands (Ehleringer 1989), but depleted in 2H due to greater seasonal rainfall (Dansgaard 1964; Clark and Fritz 1997). Indeed, that pattern was shown previously in the multi-isotopic cluster analysis for Africa (Hobson et al. 2012b). Chapter 3 130 Figure 1. Assigned wintering areas for house martin Delichon urbicum breeding at Badajoz (southwestern Spain). African isotopic clusters were generated in ArcGis 10.2.2., following previously described methods (Hobson et al. 2012b). Land cover classification layer was obtained from freely available images (Allen 2016). Wintering areas and reproductive success 131 Confirmatory path analysis Structural equation models (SEMs) are probabilistic models that hypothesize a causal network with multiple variables that can appear as both predictor and response variables (Lefcheck 2016). SEMs are usually represented as path diagrams in which one-headed arrows represent causal direct relationships from the independent variable to the dependent variable. Path analysis is a type of SEM that only takes into account observed variables; omitting latent variables (i.e. not measured variables; Grace et al. 2012; McDonald & Ho 2002). We used confirmatory path analysis to examine how feather isotope values reflecting winter habitats could explain reproductive success in the subsequent breeding season through mediation of breeding phenology and body condition. We refer to “confirmatory path analysis” as directed acyclic, or piecewise, SEM based on applications from graph theory (Shipley 2016; Lefcheck 2016). In piecewise SEM, the path diagram is translated to a set of linear equations (e.g. linear mixed models), which are then evaluated individually thus allowing a wide range of distributions and sampling designs. Another advantage of this methodology is that it allows fitting models with relatively small sample sizes. In addition, the goodness of fit of the entire causal network can be quantified by a directed separation test (”d-separation test”), which tests the assumption that all variables are conditionally independent (i.e. that there are no missing relationships among unconnected variables) (Shipley 2016). Chapter 3 132 To control the effect of age and sex on breeding success, we developed different path models for each age and sex category. We split our original data set in four subsets: experienced males (n= 38), experienced females (n= 28), young males (n= 52) and young females (n= 77). We included six variables in our models: feather δ2H, δ13C, δ15N, body condition, first-egg laying date and number of fledglings in the first brood. Isotopic values were considered independent (exogenous), while the rest were considered dependent (endogenous) variables. Every observed variable included in our models had the same sample size (i.e. listwise deletion). The structure of the path analysis was designed based on previous knowledge of the migratory ecology of our study population (López-Calderón et al. 2017; Norris et al. 2004), but also on hypotheses to be tested. Exploratory analyses indicated that δ13C and δ15N values were negatively correlated for young males. We found in a previous study that breeding phenology (i.e., first egg laying date) depends on wintering area inferred from stable isotope analysis (López-Calderón et al. 2017). It is also well known that breeding phenology is related to reproductive success in migratory passerines (Norris et al. 2004; Marzal et al. 2013). Different studies have found an effect of wintering habitat (inferred from δ13C values) on body condition before departure to the breeding areas (Marra et al. 1998; Studds & Marra 2005), and during spring migration (Bearhop et al. 2004). Therefore, we included as hypotheses to be tested causal paths Wintering areas and reproductive success 133 from winter-grown feather δ2H, δ13C and δ15N to body condition during the breeding season. We measured body condition when we first captured the individual. Within our original data set, 140 birds were first captured after they had already started breeding, whereas 55 birds were first captured before they started breeding. For this reason, we assumed correlated errors between laying date and body condition, i.e. a relationship that is bidirectional and assumed to be caused by a shared underlying driver (Lefcheck 2016). For instance, arrival time to breeding areas could determine the start of breeding but also the physical condition during the breeding season. The R library piecewiseSEM implements a crude approximation of correlated errors by excluding them from the basis set (since there is no presumed direction of causality), and then running a simple test of significance on the bivariate correlation (Lefcheck 2016). Finally, body condition may influence the potential foraging of adults and hence the quantity of prey delivery they could afford for their progeny, and this, in turn, may influence the number of fledglings reared. In all cases, we captured adults and measured body condition before we recorded the number of fledglings. Each of our path analyses was made up of three linear mixed models, one for each dependent variable. Isotopic values were the independent variables of the linear mixed models built for body condition and first-egg laying date, while these were in turn the independent variables of the linear mixed model built for number of fledglings. In every linear mixed model we included “year” as the random effect, to statistically control for inter-annual effects. Finally, Chapter 3 134 we examined the “qq plots” to visually test for normality of residuals in every individual linear mixed model. We conducted all our analyses with the library piecewiseSEM (Lefcheck 2016) using R version 3.3.1 (R Core Team 2016). Results Winter habitat features carried over to affect reproductive traits more strongly in males than in females. Breeding performance was affected by different winter habitat features depending on age class. For experienced males, confirmatory path analysis showed that rainfall amount as inferred by feather δ2H had a significant direct effect on body condition [estimate (SE) = - 0.33 (0.15), p= 0.04], and also that body condition had a significant direct effect on the number of fledglings produced during the breeding season [estimate (SE) = 0.45 (0.16), p= 0.01]. For young males, confirmatory path analysis showed that vegetation type as indexed by feather δ13C and δ15N had significant direct effects on body condition [δ13C: estimate (SE) = 0.36 (0.14), p= 0.01; δ15N: estimate (SE) = 0.40 (0.14), p< 0.01] and laying date [δ13C: estimate (SE) = - 0.29 (0.13), p= 0.03; δ15N: estimate (SE) = - 0.29 (0.13), p= 0.04]. For females, we only found that feather δ15N had a significant direct effect on the laying date of young females [estimate (SE) = 0.21 (0.09), p= 0.03], possibly related to diet (i.e. trophic position) or foraging microhabitat. Therefore, Wintering areas and reproductive success 135 confirmatory path analyses showed strong differences in path coefficients across age and sex categories (Table 1, Fig. 2). Laying date and body condition were not significantly correlated in any of the confirmatory path analyses [experienced males: estimate = 0.23, p= 0.08; young males: estimate = 0.19, p= 0.09; experienced females: estimate = - 0.13, p= 0.75; young females: estimate = - 0.17, p= 0.93]. Feather δ13C and δ15N values were not significantly correlated for young males (estimate = - 0.37, p= 0.99). We used d-separation test to quantify the goodness of fit of our models, which tests the assumption that all variables are conditionally independent (Shipley 2016). All models provided robust fit to data (experienced males: Fisher’s C= 7.7, df = 6, p= 0.26; young males: Fisher’s C= 3.59, df = 6, p= 0.73; experienced females: Fisher’s C= 7.65, df = 6, p= 0.26; young females: Fisher’s C= 6.66, df = 6, p= 0.35). We did not find any significant association among unconnected variables in any of our models. Thus, we concluded that the hypothesized relationships we examined were consistent with the data. By examining “qq plots”, we determined that every single linear mixed model was well fitted with the only exception of the linear model built for the number of fledglings in experienced females. Since for experienced males we found that feather δ2H had an indirect effect on the number of fledglings mediated through body condition, we used standardized path coefficients from this model to predict how final reproductive success shifts in response to the change Chapter 3 136 in feather δ2H (Fig. 3). We quantified that the increase by one SD of δ2H decreases the number of fledglings by 0.2 SD. Table 1. Summary results from confirmatory path analyses built for each age and sex class. Estimates shown here are standardized path coefficients (i.e. slopes of effects). R2shown here is the conditional R2, based on fixed and random effects. Sample size is 38, 28, 52 and 77 respectively for experienced-males, experienced-females, youngmales and young-females. Significant effects are highlighted in bold. Carry-over effects vary throughout life 239 Smith, R. J. & Moore, F. R. Arrival timing and seasonal reproductive performance in a long-distance migratory landbird. Behav. Ecol. Sociobiol. 57, 231–239 (2004). Studds, C. E. & Marra, P. P. Nonbreeding habitat occupancy and population processes: an upgrade experiment with a migratory bird. Ecology 86, 2380–2385 (2005). Sumner, M. D. graticule: Meridional and Parallel Lines for Maps. R package version 0.1.2 https://CRAN.R-project.org/package=graticule (2016). Van Wilgenburg, S. L. & Hobson, K. A. Combining stable-isotope (δD) and band recovery data to improve probabilistic assignment of migratory birds to origin. Ecol. Appl. 21, 1340-1351 (2011). Wassenaar, L. I. & Hobson, K. A. Comparative equilibration and online technique for determination of non-exchangeable hydrogen of keratins for use in animal migration studies. Isot. Environ. Health S. 39, 211–217 (2003). Winkler, D. W. & Allen, P. E. The seasonal decline in tree swallow clutch size: physiological constraint or strategic adjustment? Ecology 77, 922– 932 (1996). Yosef, R., L. et al. Differential spring migration of adult and juvenile Levant Sparrowhawks (Accipiter brevipes) through Eilat, Israel. J. Raptor Res. 37, 31-36 (2003). General discussion 240 241 General discussion In this thesis we have addressed several topics considered as priority research areas in ornithology, such as the identification of migratory routes, quality and distribution of non-breeding habitats, migratory schedules, carry-over effects, factors influencing migratory behaviour, and migratory connectivity (Webster & Marra 2005; Faaborg et al. 2010; Norris & Marra 2007; Harrison et al. 2011). Furthermore, we have approached other issues of relevance under the current scenario of global change, such as the importance of winter area choice on blood parasite infection, the environmental deterioration of tropical wintering areas, and responses of migratory birds to global change (Gordo et al. 2007; Altizer et al. 2015; Kramer et al. 2018). Our results have provided important insights within theoretical migration ecology that can be in turn helpful for conservation of declining species (Sanderson et al. 2006; Nebel et al. 2010; Patchet al. 2018). We have used both stable isotope analyses and light-level geolocators as our tracking techniques, which are considered the best options to study movements of small-sized birds (Hobson & Wassenaar 2008; McKinnon et al. 2013). In addition, we have incorporated other sources of information such as ringing recoveries (du Feu et al. 2009) and satellite-derive biological indexes (Pettorelli et al. 2005, 2011), completing the toolbox required to meet our goals. Moreover, we have adopted some of the most updated statistical procedures in our research areas; such as novel methods to reconstruct migratory routes General discussion 242 from raw light data (Rakhimberdiev et al. 2015; 2017), the use of Structural Equation Models to detect carry-over effects (Shipley et al. 2016; Lefcheck et al 2016), as well as new concepts for the study of migratory connectivity (Finch et al. 2017). We used light-level geolocators to analyse in detail the migration system of barn swallows breeding in southwestern Spain. Beyond the large amount of descriptive information provided here, we found some interesting results that may help to explain carry-over effects from the winter to the breeding stage. First, we found that migration schedules in spring were more variable among individuals than those of autumn migration. Second, we found that the range of stopover locations and the variance in the number of stopovers within individuals were larger in spring than in autumn migration. Third, we found that departure date from wintering areas significantly predicted arrival date to breeding grounds. On one hand, these results suggest that individuals should experience different environmental conditions during winter and spring migration, and thus some individuals arrive to their breeding areas much earlier than others do, which in turn can have subsequent fitness effects (Saino et al 2017). On the other hand, these results highlight that wintering and breeding stages are linked by means of migration schedules in spring (McKinon et al. 2013; Lemke et al. 2013; Szép et al. 2017; Cooper et al. 2017). We have also found complementary results using stable isotope analyses. For instance, house martins breeding in Spain and 243 wintering in West Africa at the isotopic cluster associated with savannahs, started to breed earlier than those wintering at the isotopic cluster associated with broadleaf forest. In addition, young male house martins wintering in habitats of higher δ13C and δ15N (indicating open habitats dominated by C4 vegetation and C3 plants adapted to hydric stress) initiated reproduction earlier than young males wintering in habitats of lower δ13C and δ15N (indicating closed or moist habitats dominated by C3 vegetation). Furthermore, barn swallows breeding in northern Denmark and wintering in Central Africa, advanced onset of breeding after years of higher NDVI during winter, being this response increasingly stronger as they aged. Unfortunately, we lack estimates of arrival date to breeding areas for our analyses based on stable isotope analyses. Nevertheless, arrival time to breeding areas usually determines the start of reproduction (e.g. Smith & Moore 2004). We also found other mechanisms driving fitness carry-over effects from the winter to the breeding stage, such as body condition and infection by blood parasites. Interestingly, both could be operating with opposite selective pressures, since experienced male house martins improved body condition and breeding success, but at the same time suffered a hig her risk of haemosporidian infection when wintering in habitats of higher rainfall (i.e. lower δ2H). Therefore, we have shown different ways to propagate fitness carry-over effects from the winter to the breeding stage, which may also interact among them. Considering also the different responses that we found across General discussion 244 age and sex classes, this thesis provides some evidence to unveil the complexity of the Palearctic-Afrotropical migration system. In accordance with the objectives of each chapter and body size restriction, we used either stable isotope analyses or light-level geolocators as tracking technique. To the best of our knowledge, house martins have been tagged with geologgers only in one study (Szép et al. 2017), and apparent survival rates of marked birds were much lower than that of control birds (i.e. 12.5% compared to 38.9%). Therefore, due to both ethical and economic reasons, we did not fit house martins with geolocators. Nevertheless, the negative effect of tag deployment on apparent survival rates of barn swallows was proven smaller than that of house martins, yet 25.31% of tagged swallows survived to the next year compared to 38.60% of control ones (Scandolara et al. 2014). The relatively low cost of stable isotope analyses allowed us to work with a much higher sample size for house martins compared to barn swallows breeding in Spain (~300 vs 24 birds, respectively). In addition, we believe that any relationship between wintering habitats and risk of infection by blood parasites would be more easily detected in house martins than in barn swallows, because the former suffers much higher prevalence (von Rönn et al. 2015; Marzal et al. 2016). Thus, in our case, stable isotope analyses were necessary to identify fitness carry-over effects mediated by blood parasite infection. We were also interested in long-term responses of barn swallow populations breeding in Denmark to 245 environmental changes occurring at their wintering areas. In order to maximize sample size to enhance the inference we made for wintering areas, we needed to work with stable isotope analyses rather than light-level geolocators. We suggest that stable isotope analyses have an additional advantage compared to geologgers, which is found in long-term studies. Many ornithologists have been sampling feathers during the last decades for different purposes. Stable isotope measurements from these collected feathers through the years can provide information related to migration ecology before the development of solar geolocation. We count with a large data set of ~ 4,000 individually identified barn swallows monitored over thirty years in Denmark. This large data set, in combination with stable isotope analyses and ringing recoveries, allowed us to highlight the importance of age in shaping the strength of carry-over effects and thus the response of migratory birds to global change. Interestingly, this is in agreement with the previous results we found for house martins, which always revealed the importance of age in migration ecology. References Altizer, S., Bartel, R. & Han, B. A. Animal migration and infectious disease risk. Science 331, 296–302 (2011). General discussion 246 Cooper, N. W., Hallworth, M. T. & Marra, P. P. Light-level geolocation reveals wintering distribution, migration routes, and primary stopover locations of an endangered long-distance migratory songbird. J. Avian Biol. 48, 209–219 (2017). du Feu, C. R., et al. EURING Data Bank geographical index 2009. http://www.euring.org/edb (2009). Faaborg, J. et al. Conserving migratory land birds in the New World: Do we know enough? Ecol. Appl. 20, 398–418 (2010). Finch, T., Butler, S. J., Franco, A. M. A. & Cresswell, W. Low migratory connectivity is common in long-distance migrant birds. J. Anim. Ecol. 86, 662–673 (2017). Gordo, O. Why are bird migration dates shifting? A review of weather and climate effects on avian migratory phenology. Clim. Res. 35, 37–58 (2007). Harrison, X. A., Blount, J. D., Inger, R., Norris, D. R. & Bearhop, S. Carryover effects as drivers of fitness differences in animals. J. Anim. Ecol. 80, 4–18 (2011). Hobson, K. A. & Wassenaar, L. I. Tracking animal migration with stable isotopes (Academic Press, 2008). Kramer, G. R. et al. Population trends in Vermivora warblers are linked to strong migratory connectivity. P. Natl. Acad. Sci. 115, E3192–E3200 (2018). 247 Lefcheck, J. S. piecewiseSEM : Piecewise structural equation modelling in R for ecology, evolution, and systematics. MEE 7, 573–579 (2016). Lemke, H. W. et al. Annual Cycle and Migration Strategies of a Trans- Saharan Migratory Songbird: A Geolocator Study in the Great Reed Warbler. PLoS ONE 8, e79209 (2013). Liechti, F. et al. Timing of migration and residence areas during the nonbreeding period of barn swallows Hirundo rustica in relation to sex and population. J. Avian Biol. 46, 254–265 (2015). Marzal, A. et al. A longitudinal study of age-related changes in Haemoproteus infection in a passerine bird. Oikos 125, 1092–1099 (2016). McKinnon, E. A., Fraser, K. C. & Stutchbury, B. J. M. New discoveries in landbird migration using geolocators, and a flight plan for the future. The Auk 130, 211–222 (2013). Nebel, S., Mills, A., McCracken, J. D. & Taylor, P. D. Declines of aerial insectivores in North America follow a geographic gradient. Avian Conserv. Ecol. 5, (2010). Norris, D. R. & Marra, P. P. Seasonal interactions, habitat quality, and population dynamics in migratory birds. The Condor 109, 535 (2007). Patchett, R., Finch, T. & Cresswell, W. Population consequences of migratory variability differ between flyways. Curr. Biol. 28, R340–R341 (2018). General discussion 248 Pettorelli, N. et al. Using the satellite-derived NDVI to assess ecological responses to environmental change. Trends Ecol. Evol. 20, 503–510 (2005). Pettorelli, N. et al. The Normalized Difference Vegetation Index (NDVI): unforeseen successes in animal ecology. Clim. Res. 46, 15–27 (2011). Rakhimberdiev, E. et al. A hidden Markov model for reconstructing animal paths from solar geolocation loggers using templates for light intensity. Movement Ecology 3, 25 (2015). Rakhimberdiev, E., Saveliev, A., Piersma, T. & Karagicheva, J. FLightR: an r package for reconstructing animal paths from solar geolocation loggers. Methods Ecol. Evol. 8, 1482–1487 (2017). Saino, N. et al. Sex-dependent carry-over effects on timing of reproduction and fecundity of a migratory bird. J. Anim. Ecol. 86, 239–249 (2017). Sanderson, F. J., Donald, P. F., Pain, D. J., Burfield, I. J. & Van Bommel, F. P. Long-term population declines in Afro-Palearctic migrant birds. Biol. Cons. 131, 93–105 (2006). Scandolara, C. et al. Impact of miniaturized geolocators on barn swallow Hirundo rustica fitness traits. J. Avian Biol. 45, 417–423 (2014). Shipley, B. Cause and correlation in biology: a user’s guide to path analysis, structural equations and causal inference with R (Cambridge University Press, 2016). Supplementary Materials 255 Figure 1. Migration track inferred with light-level geolocators for the bird ringed Y66266 (female breeding in “Las Coladas”). Supplementary Materials 256 Figure 2. Migration track inferred with light-level geolocators for the bird ringed Y66029 (female breeding in “Las Coladas”). Supplementary Materials 257 Figure 3. Migration track inferred with light-level geolocators for the bird ringed Y66103 (male breeding in “Las Coladas”). Supplementary Materials 258 Figure 4. Migration track inferred with light-level geolocators for the bird ringed 2L12365 (male breeding in “La Calera”). Supplementary Materials 259 Figure 5. Migration track inferred with light-level geolocators for the bird ringed Y65845 (female breeding in “La Calera”). Supplementary Materials 260 Figure 6. Migration track inferred with light-level geolocators for the bird ringed 1L92850 (male breeding in “Las Coladas”). Supplementary Materials 261 Figure 7. Migration track inferred with light-level geolocators for the bird ringed Y66049 (female breeding in “La Calera”). Supplementary Materials 262 Figure 8. Migration track inferred with light-level geolocators for the bird ringed Y65702 (male breeding in “Las Coladas”). Supplementary Materials 263 Figure 9. Migration track inferred with light-level geolocators for the bird ringed Y65848 (female breeding in “La Calera”). Supplementary Materials 264 Figure 10. Migration track inferred with light-level geolocators for the bird ringed Y66105 (male breeding in “Las Coladas”). Supplementary Materials 271 Figure 17. Migration track inferred with light-level geolocators for the bird ringed 2L12735 (female breeding in “Asesera”). Supplementary Materials 272 Figure 18. Migration track inferred with light-level geolocators for the bird ringed Y65905 (male breeding in “Las Coladas”).* Supplementary Materials 273 Figure 19. Migration track inferred with light-level geolocators for the bird ringed Y66251 (female breeding in “La Alegría”).* Supplementary Materials 274 Figure 20. Migration track inferred with light-level geolocators for the bird ringed Y66220 (male breeding in “La Alegría”). Supplementary Materials 275 Figure 21. Migration track inferred with light-level geolocators for the bird ringed Y65702 (male breeding in “Las Coladas”). Supplementary Materials 276 Figure 22. Migration track inferred with light-level geolocators for the bird ringed 1Y83478 (male breeding in “Las Coladas”). Supplementary Materials 277 Figure 23. Migration track inferred with light-level geolocators for the bird ringed 2L12590 (male breeding in “Asesera”). Supplementary Materials 278 Figure 24. Migration track inferred with light-level geolocators for the bird ringed Y66027 (female breeding in “Las Coladas”). Supplementary Materials 279 Supplementary Materials