Copyright is owned by the Author of the thesis. Permission is given for a copy to be downloaded by an individual for the purpose of research and private study only. The thesis may not be reproduced elsewhere without the permission of the Author. Extreme migration and the annual cycle: individual strategies in New Zealand Bar-tailed Godwits A thesis presented in partial fulfilment of the requirements for the degree of Doctor of Philosophy in Ecology Massey University Palmerston North, New Zealand Jesse Ray Conklin 2011 1 Synopsis Long-distance migration places severe constraints on the annual cycles of birds, as they balance the energetic and scheduling requirement s of breeding, moult, pre-migratory fuelling, and the journey itself. The most extreme migrations, traversing vast, inhospitable areas of the globe in protracted non-stop flights, may push birds to the limits of their capabilities, and would be expected to tolerate little variation in performance. Despite this, Bar-tailed Godwits Limosa lapponia baueri , which are among the world?s greatest endurance migrants, embark on northward migration from New Zealand across a month-long period, and individuals are quite faithful to their particular schedules. Godw its are highly sexually dimorphic in plumage and body size, and there is additionally substantial individual variation within each sex in both traits. These patterns demonstrate a surprising diversity of strategies within a system that should contain little room for error. In this thesis, I sought to identify the roots and consequences of both persistent and ephemeral individual differences in migration and moult of New Zealand Bar-tailed Godwits, and to identify constraints and potential bottlenecks in their annual cycle. To do this, I combined a fine-resolution multi-year focus on individuals and an entire annual-cycle perspective, both of which have generally been impossible in studies of long-distance migratory birds. At a single non-breeding site, I closely monitored moult an d migration of individual Bar-tailed Godwits for three non-breeding seasons, and linked th ese with events outside of New Zealand by tracking a subset of the same in dividuals on their complete migrations to Alaska breeding grounds and back. I supplemented this by trav elling to Alaska myself and describing how godwits are distributed by size and plumage across their vast breeding range. I found that most of the var iation among individual Bar-tailed Godwits was linked to where they nested in Alaska: within each sex, nort herly breeders were smaller, had more extensive breeding plumage, and migrated later on both northbound and southbound migrations. The differences in migration timing can be explained by variation in when tundra breeding sites become snow-free and available across a latitudi nal gradient, but reasons for geographic differences in plumage and size are less clear. Va riation in breeding plumage was associated with different strategies for scheduling moult, both in New Zealand and during northbound migratory stopover in the Yellow Sea. Indi vidual godwits were extraordinarily consistent between years in their timing of departure from New Zealand, and most ?off - schedule? departures were attributable to birds avoiding unfavourable winds for migration. Surprisingly, 2 timing of arrival in New Zealand after the longest recorded non-st op flight did not appear to influence a godwit?s ability to prepare for its next migration, as timing of subsequent migratory departure and extent of breeding plum age on departure were both unaffected and very consistent. Across the entire year, schedu ling of events became more precisely timed as the breeding season approached, but movements were generally much more tightly scheduled than moults. These findings show that Bar-tailed Godwits adopt and enact an array of individualised strategies within an apparently constrained system. The inter-relationships among events in different parts of the globe show that an individual-based, full annual-cycle perspective is required to understand patterns in any partic ular season. The consistent manner in which godwits conduct their annual routines, while s till demonstrating flexibility to address unforeseen circumstances, challenges us to r econsider the view of extreme long-distance migrants as organisms operating at the limits of their capabilities. 3 Preface ?I?m sorry?they do what?? The above quotation is only a dramatisation. I can?t honestly remember how I reacted upon first hearing the idea that an otherwise unremarkable shorebird might spread its wings in Alaska and fly the length of the Pacific Ocean w ithout stopping, only to next touch ground on some mudflat in New Zealand. But over the last seven years, I have explained this concept to a sufficient number of people to see the entire range of possible reactions. Some people can?t get past the simple fact of it, and stand bewildered or move straight to denial?they must have misunderstood what I was saying. Others roll w ith it, their minds moving quickly to the pertinent questions: Why? How? After years of intimate involvement with those exact problems, it sometimes requires the reactions of others to remind me that what I now take for granted is by no means commonplace. Although difficult to maintain on a daily basis, awe is in fact the proper response. Whatever my immed iate reaction was years ago, my ultimate response was profound: I was going to study Bar-tailed Godwits. Since 2005, through the wonder s of satellite telemetry, hypothesis became reality: it was in fact possible for a godwit to sustain powered flight for eight or nine consecutive days without stopping once to eat, drink, or rest, crossing more than 11,000 kilometres of open ocean. Furthermore, there were tens of thousands of godwits performing this astounding migration every year without drawing much attention to themselves, only to head back north six months later on a non-stop flight to Asia that was only slightly shorter than the southward trip. However, each does not go about these flights in precisely the same way. Godwits depart New Zealand across a month- long period from early March to early April, and they don?t all look the same when they leave; some are as dull and gray as mid-winter birds, while others sport the full regalia of summer breeding plumage found in Alaska. It became apparent that individual godwits were remarkably consistent in these qualities across years: certain birds always migrated earlier than others, and the re ddest birds were always the reddest. But why was this true? What makes a part icular godwit an early bird, or a red bird? I had found my research topic, and what follows will describe my four years of attempts to address this ostensibly simple question. 4 For Brian McCaffery, who never ceases to inspire. 5 Contents Synopsis 1 Preface 3 Co-authors 7 Acknowledgements 8 Chapter 1 Introduction: individuality and long-distance migration 11 From populations to individuals 12 Migration as a research challenge 14 Migration as a lifestyle 15 Scheduling of moult 17 Individual quality versus individual strategies 18 Aims of this thesis 20 Study species 20 Study site 24 Overview of research chapters 26 Chapter 2 Breeding latitude drives individual schedules in a trans-hemispheric 33 migrant bird Chapter 3 Geographic variation in morphology of Alaska-breeding Bar-tailed 45 Godwits is not maintained on their non-breeding grounds in New Zealand Chapter 4 Impacts of wind on individual migration schedules of New Zealand 67 Bar-tailed Godwits Chapter 5 Contour feather moult of Bar-tailed Godwits in New Zealand and the 85 Northern Hemisphere reveals multiple strategies by sex and breeding region 6 Chapter 6 Carry-over effects and compensation: late arrival on non-breeding 107 grounds affects wing moult but not plumage or schedules of departing Bar-tailed Godwits Chapter 7 Absolute consistency: individual versus population variation in timing 131 of annual life-history stages of a long-distance migrant bird Chapter 8 Synthesis: an evolving view of long-distance migration 147 Key findings of this thesis 148 Regulation of the migrant annual cycle 149 How are individually-optimised schedules maintained? 153 How ?close to the edge? are long-distance migrants? 154 The role of individual quality 159 Model system or evolutionary outlier? 163 Chapter 9 Future directions 169 What now? 170 Appendix 1 Attachment of geolocators to Bar-tailed Godwits: a tibia-mounted 173 method with no survival effects or loss of units Appendix 2 Analysis of geolocator data 181 Appendix 3 Calculation of wind effect 189 Appendix 4 Supplementary information on primary moult 191 References 193 7 Co-authors I wrote all portions of this thesis, collected near ly all of the data, and performed all analyses. However, four of my collaborators made esse ntial contributions warranting co-authorship of specific sections. Phil F. Battley (Massey University, Ecology Group) As my primary supervisor, Phil collab orated with me to conceive and design every part of this research, and is consequently a co-author on Chapters 2 ? 7 and Appendix 1. He provided logistical and financial support for all trapping, track ing, and travel enterprises, and assisted in the field for local captures and geolocator deployments. He reviewed all manuscripts and helped proof the final publications. Murray A. Potter (Massey University, Ecology Group) As my secondary supervisor, Murray contributed to many aspects of study design and interpretation of results, and reviewed most chap ters. His inclusion as co-author on Chapters 2, 3, and 7 reflects his greater role in developi ng the results and ideas of those chapters for publication. He additionally contri buted to Chapter 3 by accompan ying me to Alaska, helping with nest-searching and trapping, and provid ing many in-field photographs of godwits. James W. Fox (formerly of British Antarctic Survey, UK) For the geolocator portion of this study, Ja mes provided the units and analysis software, advised on matters of data analysis and interpretation, and provided technical assistance. He reviewed and is a co-author on Chapter 2. Dan R. Ruthrauff (U.S. Geological Survey, Alaska Science Center, USA) Dan contributed to Chapter 3 by providing equipm ent and logistical support for all fieldwork in Alaska, and leading the expedition to the North Slope. He additionally provided data on historic captures and tracking data for godwits in Alaska (property of USGS), contributed in- field photographs, and reviewed the manuscript. 8 Acknowledgements First of all, I?d like to thank 23.5?. If the Earth did not hurtle through space tilted on its axis at this precise angle, there would be no regular seasons and no impetus for birds to migrate. Consequently, I would have to find an honest job. More recently, my specific research benefited fro m an array of kind, hard-working individuals spread all over the world. The shorebird world is a close and supportive group, and I am fortunate to have been accepted into their fold on a professional and personal level. Previous decades of godwit work across the flyway created a strong foundation upon which I could build, and I must acknowledge some of the key players (knowing I will also forget some) who contributed both indirectly and very directly to my success. In Alaska: Bob Gill, Brian McCaffery, Dan Mulcahy, Lee Tibbitts, Dan Ruthr auff, and Nils Warnock. In New Zealand: Phil Battley, David Melville, Adrian Riegen, Rob Schuckard, and Keith Woodley. In Australia: Adrian Boyle, Chris Hassell, and Clive Minton. Many people helped trap and band godwits at Foxton. Long before I reached New Zealand, the seed of my project was sown with some key colour-banding by Phil Battley, David Melville, Ralph Powlesland, Adrian Riegen, Hugh Robertson, Rob Schuckard, and Graeme Taylor, which I continue to benefit from today. I bow to either their skill or luck at mist- netting, because I apparently have neither. Duri ng my tenure at Foxton, all of these figures plus legions of Massey Ecology students and ot hers flocked to my aid for trapping attempts (unfortunately, in many cases attempt cannot be stressed enough). I am not sure how many make up a legion, but I trust they will understand when I don?t even attempt to list them all. Phil Battley, Jimmy Choi, Paul Gibson, Craig Steed, Brent Stephenson, and Graeme Taylor contributed photographs of Foxton godwits. All ph otos in this document are my own unless specified otherwise. The geolocator work central to Chapters 2 and 7 ow es its success to many people. In February 2008, Bob Gill tossed me a bag of geolocators, with a casual, ? Here: you might find a use for these.? When they turned out to be the wrong type of units for leg-attachment, James Fox at the British Antarctic Survey replaced them withou t question and, more importantly, without charge. James also generally guid ed me through the struggle of retrieval and analysis of the data, and extracted data from troublesome units, with a very quick turn-around time. Several people helped me scatter ground-truthing geol ocators around the flyway, to help with interpretation of data: Ty Donnelly, Brian McCaffery, and Dan Ruthrauff in Alaska, Andreas 9 Kim and Nial Moores in South Korea, and Adrian Riegen in China. It?s not Ty?s fault (I assume) that the thing got eaten by an Arctic Fox. Many thanks to the superb team at Massey?s Institute of Veterinary, Animal and Biomedical Sciences (particularly Brett Gartrell, Jodi S alinsky, and Janelle Ward), whose tireless work saw me through the undisputed low point of my research. Other contributions to specific chapters are as follows: Chapter 2: Thanks to Lee Tibbitts fo r assistance with the Figure 2.2 map. The manuscript benefited from comments by David Penny on an earlier draft. Comments by Barbara Helm on the final paper helped put these findings into perspective, which greatly benefited my overall thesis. Chapter 3: I thank Bob Gill, David Melville, and Craig Steed for field assistance in Alaska. Todd Buckley, Randy Hill, Sarah Lovibond, Julie Morse, Adrian Riegen, Rob Schuckard, Craig Steed, Pavel Tomkovich, Keith Woodley, and Steve Zack kindly provided photographs of Alaska godwits. I thank Carla Cicero (Museu m of Vertebrate Zoology at the University of California at Berkeley), Bob Gill (U.S. Geological Survey?s Alaska Science Center in Anchorage), and Kevin Winker (University of Alaska at Fairbanks) for access to museum specimens. Adrian Riegen provided New Zealand Wader Study Group capture data. Finally, thanks to Brian McCaffery for pumping out an entire paper just so I could cite it. Chapter 4: I thank Bob Gill, Michael Kemp, and Judy Shamoun-Baranes for thoughts on wind analyses. Chapters 5 and 8: I thank Theunis Piersma for helpfu l comments on early drafts of the Chapter 5 manuscript, and for h elping flesh out the Black-tailed Godwit section of Table 8.2. Chapter 7: I thank Yvonne Verkuil for reviewing th e manuscript and inspiring the title. Chapters 2 ? 6 and Appendix 1 greatly benefited from the editorial and review processes at their respective journals. On a more business-like note, my research wa s supported by a Massey University Doctoral Scholarship. This project was generally su pported by the David and Lucile Packard Foundation through the Pacific Shorebird Migr ation Project, and by a Marsden Fund grant administered by The Royal Society of New Zealan d to Phil Battley. Fieldwork was conducted with Massey University Animal Ethics Comm ittee approval (#07_163 ). The Ornithological Society of New Zealand provided capture data partly obtained under Department of Conservation Research and Develop ment contracts 3739 -01 and 3599. 10 I was already grateful for the re search grant I received from the Manawatu Estuary Trust. Three weeks later, I became even more grateful when my laptop died and had to be quickly replaced during the home stretch of thesis-writing. On a more personal note, I would not be where I am today (alone and penniless at a computer on a remote South Pacific island) if it weren?t for my lucky meeting with a few particular people along the way. I?m not sure why Mark Colwell offered me the job counting shorebirds on Humboldt Bay, a job I almost turned down because I had already decided I wanted to work on any birds except the wet and cold ones. I am just glad he did offer it, because I haven?t thought about anything but shorebirds since. As my Ornithology TA in my very first semester at Humboldt State, Dan Ruthrauff still claims that he taught me nothing, but in fact he taught me the only thing I really needed to know: go to Alaska. And that led me to Brian McCaffery, who for some reason hired me sight-unseen after I cold-called him and deftly dropped the names of Dan and Mark. Strangely, he insisted on repeating this error, and three years later I was marooned at an Alaskan field camp with David Melville. Somehow, amongst repeatedly fleeing for our lives from Biblical floods, David convinced me that I should endeavour to become the PhD student of a then-unemploye d New Zealand biologist who was younger than me. Why wouldn?t I do that? I?d like to thank my parents, Barbara and Les, who never seem to question the wisdom of the many strange things I have considered good id eas in my supposed adulthood. Their trust (as well as the knowledge that I am not completely r eliant on these exploits for financial support!) has given me the confidence to go where I needed to go. I could not have made it through the last four years without being adopted by a pretty special group of friends in New Zealand. Although some have already dispersed abroad, they have been a wonderful and much appreciated support group: Doroth?e Durpoix, Rich Seaton, Jay Gedir, Fleur Maseyk, Robbie Andrew, Pete McGregor, Yvan Richard, Melanie Kiessner, Marco Wenzel, Jay McCartney, and Anne-Marie Emerson, to name a few. I owe a great deal of the success of this thesis to my supervisors, Phil Battley and Murray Potter, who could hardly have been more suppor tive and helpful throughout the entire process. I know I am fortunate to have supervisors wh o are not only so capable and constructively critical, but are also as excited about my research as I am. It has been a great collaboration. However, considering the clear evidence for an inverse relationship between alcohol and the productivity of scientists (Grim 2008) , I have to question Phil?s deliberate role in my development as an aficionado of single malt whisky. Any shortcomings of this thesis should be considered Phil?s sole responsibility. 11 Chapter 1 Introduction: individuality and long-distance migration 12 Chapter 1 From populations to individuals Natural selection and I have something in comm on. We are both focuse d on individuals. For me, the fascination began when I first put radio-transmitters on Dunlin Calidris alpina in 2002 and discovered that each one of these apparently indistinguishable birds had its own way of doing things. As for natural selection, it probably became interested as soon as the first unicellular organisms began to develop idiosyncrasies. Differences among individuals are, of course, the raw material for evolution and therefore the ultimate source of diversity in all higher levels of organisation from population to kingdom. In fact, individuals must differ for natural sele ction to occur (Darwin 18 59). Because not all behavioural or morphological variants are equal in terms of promoting an organism?s success, biologists may view individual differences on a continuum from ?adaptive? to ?maladaptive?, and expect all examples on the latter end to be removed from the population eventually through a failure to survive and reproduce. Op timal behaviour theory (e.g., Stephens and Krebs 1986) places a similarly lo w value on individuality, view ing variation as inherently unimodal: for a given scenario, there is a sing le best solution, and the fringes of variation extending outward from this are increasingly wrong. A radically different view is that individual differences are not just fodder for natural selection, but also the end product of it (Wilson 1998). There is ample evidence that there can be more than one viable solution to an ecological problem. It is well-established that different priorities and capabilities of individuals based on sex or age can foster multiple strategies in the same population (e.g., Ketterson and Nolan 19 83) , but more intrig uing are differences within a class, such as male mating polymorp hisms (e.g., Jukema and Piersma 2006). These show that individual variation need not be discouraged by natural selection, but can be perpetuated when alternative strategies have similar outcomes (Shuster an d Wade 1991) or if each is superior in different contexts (Wilson 1998, Dingemanse et al. 2004). Examples of stable behavioural or morphological polymorphisms come from taxa as diverse as isopods, fish, lizards, and birds, and are almost certainly more common than has been documented. A particularly rich and interesting field of study is the fitness consequences of animal personalities (Dingemanse and R?ale 2005). Much of this research has focused on the propensity of individuals to exhibit shy or bold behaviour when addressing novel circumstances (Wilson et al. 1994, Dingemanse et al. 2004), and has shown that such traits can be heritable and linked to entire suites of other behaviours. Recognising study organisms as individuals with distinct personalities and preferences forces biologists to view them as more than automatons seeking optimality or noise around a mean value. Furthermore, Chapter 1 13 population parameters may be seen as merely the downstream manifestations of individual variation. Science is by nature reductionist: it is fundamental to examine the smallest discernible parts in order to understand larger patterns. Despite this, ecological research was for decades dominated by a focus on populations and species, rather than individual organisms (?omnicki 1 9 9 2 ). One reason for this was philosophical. Prop onents of a more holistic approach rejected the application of reductionism to ecology because it seemed to ignore the consequences of evolutionary theory and downplay the importance of interactions among organisms, artificially simplifying what was inherently complicated and hierarchical (Caplan 1988). It was generally believed that evolutionary and ecological processes could best be understood through describing demographic parameters at the population level. Other obstacles to an individual approach were practical. Intensive study of individuals is difficult and time-consuming, and acquiring large datasets for cryptic, highly mobile, or low- density species can be nearly impossible. Without large samples, an individual approach risks being anecdotal in nature, more naturalistic than empirical. It may bias conclusions by under- representing the range of variation in the population and missing important processes. Another problem is that of analysis; biologists are most accustomed to statistical methods that address average values, and multi-modal (or non-modal) datasets can be difficult to summarise or understand. This is to some extent addressed by the rapidly growing field of individual-based modelling (Judson 1994), but useful models still depend on copious and diverse data that can be very difficult to obtain. Here, methodological questions find commonality with the ecological problems themselves: there is, of course, no single correct answer. Much can be learned from both ?bottom - up? (using individuals to unde rstand how the properties of a larger system emerge) and ?top- down? (using higher -level patterns to understand the diversity of individuals) approaches, but neither tells the whole story on its own (Grimm 1999). In fact, th ese approaches should not be seen as two competing ways to answer the same question. Population-level and individual- level processes must be viewed separately and in conjunction to really see how a system works. The most complete understanding will s urely come from a combination of approaches, as I hope this thesis will help demonstrate. 14 Chapter 1 Migration as a research challenge The field of avian migration research in particular has been historically limited to population- level studies, due to the inherent difficulties of gathering data on individuals that are more mobile than biologists are. Wh erever you try to conduct your research, the birds leave. Depending on the nature of your study site, an y migratory species will be essentially absent for 5 ? 1 1 months each year. So, no matter how much information you gather from one location, you are missing a substantial part of the story. More inconveniently, it has long been recognised that events in one season of a migratory bird?s annual cycle may be influenced by prior conditions or events on the other side of the globe. These cross-seasonal effects operate at evolutionary scales, shaping long-term p atterns within and among populations (Myers 1981), and also at smaller scales, in which envi ronmental conditions in a particular time and place may influence a specific individual?s performance later in that year (Harrison et al. 2011). Consequently, studies that focus (usually by necessity) on a single season are limited in the spectrum of inferences to be drawn from their data even with regard to that season. This is the fundamental challenge of studying migratory species. Except for species travelling through and to extremely inaccessible regions, it is generally possible to capture more of the annual cycle by observing, for example, breeding birds in one location and non-breeding birds in another, a nd then drawing reasonable conclusions regarding connections between the two. Histori cally, most information regarding the annual routines of migratory birds was derived in this manner. This can help describe population parameters, but relies on the central assumpti on that all variation present in the entire population (or at least equivalent segments of it) exists in the respective study samples. In practice, this assumption is almo st always violated to some extent, especially for species that are widely distributed in one or more seasons. It is common for basic attributes of morphology or behaviour to vary geographically across the breeding and/or non-breeding range of a species (Zink and Remsen 1986) , and so spurious conclusions may arise when the entire range of variation is not considered. Without eviden ce of direct connections between studies in different seasons, researchers may unwitti ngly compare apples to oranges. Such direct connections are possible only thr ough the tracking of individual birds across seasons, the difficulty of which remains the si ngle biggest hurdle to understanding annual routines of migrants. One of th e earliest and most elegant approaches to this problem was the attachment of leg bands (Preuss 2001), which allowed individuals or cohorts captured at one site to be identified when they were observed or captured at another site. Still the most widely-used method to study migration after mo re than a century, leg bands continue to provide valuable and surprising insights about migratory routes, timing, and distances, as well Chapter 1 15 as survival and longevity. In the last thirty years, rapid technological advances have allowed more precise and thorough tracking of indivi dual movements by radio telemetry, satellite telemetry, and, most recently, geolocation (F iedler 2009). Each of these technologies experienced exponential increases in both use an d effectiveness for bird research as issues regarding price, miniaturisation, attachmen t, and battery life were resolved. This ongoing renaissance in the field of bird movements, accompanied by concurrent advancements in physiology and endocrinology, has led inevitably to an increased focus on individuals. Traditionally, researchers used po pulation-level data to infer what specific birds did; this practice tends to obscure individual variation and important links between seasons, and thus hinders understanding of the mechanisms and evolutionary processes underlying the phenomena of interest. New methods and attitudes are allowing us to turn the tables on this approach, by instead viewing populations as the sum of many distinct individuals. Migration as a lifestyle To think about how an individual bird manages its migration, it is helpful to first consider what would motivate it to migrate at all. After all, migration is only one of many possible strategies to address the fundamental challenge that faces all organisms, which is to reproduce. In circumstances wher e resources are consistent and appropriate to support both breeding and year-round survival, th ere is no need to migrate. But most environments on earth are seasonal, and so a sedentary animal face s peaks of relatively abundant resources alternating with periods of more restricted resources. Mobile an imals, on the other hand, can benefit from the fact that seasonal resource peaks occur at different times at different locations by moving among them to essentially create a sustained resource peak. This also means they can go places that are completely inappropriate to their needs at other times of the year. For example, a migrant can use a site that offers prim e breeding habitat in the summer but no food during the winter, because it simply leaves before resources disappear. Thus, migration involves the movement of anim als to exploit spatially and temporally variable resources. This is an opportunity, but also a constraint. Because the resources of both breeding and non-breeding locations, and potentially those of locations used in transit between them, operate on different schedules, the migratory lifes tyle places considerable time constraints on the annual routines of individuals (Drent et al. 2006). If maximising lifetime breeding success is an individual?s primary motivation, the breeding season can be considered the focal point of the annual cycle, with all other events th eoretically optimised for maximum reproductive output. For this reason, how the timing of arrival on the breeding grounds relates to a 16 Chapter 1 migrant?s breeding success has received much attention. Successful breeding requires resources (in terms of both habitat and food) fo r both the production of a clutch of eggs as well as the later development of young through fledging, and therefore should be finely tuned to the phenologies of breeding sites. Early breedin g arrival, particularly in species with strong competition among individuals, confers many benefits, including reduced competition for territories and mates, more time for multip le breeding attempts, and appropriate timing of hatch in relation to resources for provisioning young. These benefits are reflected in the seasonal declines in breeding success observed in many species (Daan et al. 1990, Wiggins et al. 1994, Verhulst et al. 1995, Lozano et al. 1996, Hasselquist 1998) . However, ?early? is not necessarily equivalent to ?optimal?; there is also evidence of the reproductive costs of initiating breeding too early (Prop et al. 2003, Bety et al. 2004) . In addition to getting the timing right, a migrant must contend with the direct costs of getting to its destination, which can be considerable. Travelling takes time which could be otherwise spent on self-maintenance activities such as rest, moult, or foraging, and can involve survival threats associated with crossing inhospitable geographic barriers (e.g., oceans, mountain ranges) or exposure to inclement weather or predators. Birds must also spend time and energy preparing for the journey; prerequisites include sufficient fuel stores to fly between successive foraging sites along the migratory route, and f light feathers of adequ ate quality to conduct the trip. For those making protracted non-stop fli ghts of thousands of kilometres, physical preparation includes not only accumulation of sufficient fat stores (in some cases, equal to 50 ? 1 0 0 % of basal non-breeding mass; Zwarts et al. 1990b, Gudmundsson et al. 1991, Piersma and Gill 1998, Battley and Piersma 2005) , but also increases in ma ss and capability of flight muscles, and reduction in mass of body organs (e .g., the digestive system) unnecessary for the flight itself (Piersma and Gill 1998, Battley et al. 2000, Landys-Ciannelli et al. 2003). Temporal and energetic constraints may be particularly challenging for high-latitude breeders. Seasonal changes are most profound at high latitude s, leading to intense resource peaks of shorter duration than those experienced at te mperate latitudes. For example, arctic tundra breeding habitats may be ice- free and available for only 2 ? 4 months per year and the intense annual bloom of insect prey necessary to support fledging of chicks may last just 2 ? 3 weeks (Meltofte et al. 2007b, Tulp and Schekkerman 2008). These brief breeding seasons make timing of migration even more critical, because di fferences of even a week or two may have severe impacts on breeding success or even adult survival. In addition, low food availability and unpredictable conditions at the start of the breeding season may require birds to carry greater nutritional reserves than are needed for the flight itself, to support both survival and immediate investment in breeding after arrival on breeding grounds. Chapter 1 17 So we see that migration is not just an action, but rather a ?syndrome? of behaviours and physiological processes that encompass a bird?s entire annual routine (Piersma et al. 2005, Dingle 2006). These traits have deep evoluti onary roots that reach far beyond the few specific concerns discussed above, influencing morp hology, mating systems, chick development, longevity, and any number of other basic lif e-history attributes. Th is fact profoundly influences not just the birds themselves, but how we must view and study them. Scheduling of moult The most conspicuous and defining attribute of birds is their feathers. However, because feathers are subject to wear and damage through use, they are not permanent, and most birds replace all of their feathers at least once per year (Howell 2010). Feather replacement involves costs, including both direct costs of feather synthesis and indirect effects of decreased performance in thermoregulation or flight during moult (Payne 1972, Murphy and King 1991, Lindstr?m et al. 1993, Swaddle and Witter 1997). How birds schedule their annual moults reflects both these costs and the relative importance of feather quality to performance in different seasons (Holmgren and Hedenstr?m 19 95). Scheduling of moult is a particular challenge for migratory species, because activities such as long-distance flights and pre- migratory mass gain may be energetically incompatible with moult. For example, some species initiate mass gain only after the completion or suspension of moult (Thomas and Dartnall 1971, Serra et al. 1999), but others may overlap the two processes (Marks 1993, Lindstr?m et al. 1994). Compared to breeding and migr atory movements, sc heduling of moult appears to be much more plastic and open to strategic flexibility at both evolutionary and individual scales (Helm and Gwinner 2006, Flinks et al. 2008). Flight feather moult has obvi ous functional consequences with regard to migratory performance and is generally completed to the same degree by all migrating individuals. Intra- and inter-specific differences suggest that timing and duration of wing moult is plastic in evolutionary terms (Serra 2001, Summers et al. 2004), but there is also evidence that individuals can respond to much more immediate stimuli to alter moult schedule. For example, nutrient deprivation may resu lt in a protracted wing moult (Murphy et al. 1988, Marks and Underhill 1994). Also, time-stressed in dividuals may be able to compensate for a late moult by increasing the growth rate of particular feathers or growing more feathers concurrently (Murphy et al. 1988, Dawson 2004, Serra and Underhill 2006). However, this may come at the expense of feather quality, as a hurried moult may produce less durable or more asymmetrical feathers (Se rra 2001, Dawson 2004). 18 Chapter 1 Along with flight feathers, all body (or ?contour?) feathers are typically replaced in the non - breeding season, in what is referred to as the ?pre - basic? moult ( Humphrey and Parkes 1959). Many species that benefit from different plumages in different seasons have evolved a second annual moult, in which all or most of the contour feathers are replace d again. This ?pre - alternate? moult is commonly a transformation into breeding plumage, which, unlike the pre - basic moult, can be a source of substantial individual variation in terms of extent or intensity of colouration. Variation in breeding plumage is traditionally thought to advertise relative individual quality by honestly signalling a bird?s condition ( Hill 1991, Fitzpatrick 1998, Jawor and Breitwisch 2003). In add ition, young or socially subordinat e birds may invest less in breeding plumage, conserving energy and preven ting costly agonistic encounters by signalling their low social status (Flood 1984, Lyon and Montgomerie 1986, Chu 1994). Alternatively, plumage may vary geographically with habitat characteristics (Mumme et al. 2006), and intra- individual variation may additionally reflect environmental stochasticity (Griffith and Sheldon 2001). Whatever the mechanism, variation in inve stment in breeding plumage can lead to substantial differences in timing and duration of moult among individuals in the same population. Individual quality versus individual strategies The observation that individuals differ in some respect often reveals little regarding why they are different. As discussed earlier, there are clear and demonstrable advantages to performing migration ?correctly?; if so, why don?t all individuals do exactly the same thing? One possible answer is that they can?t, due to innate differences in quality. To illustrate, let?s assume that for a given population there exists an optimal time to arrive on breeding grounds, an optimal state of breeding plumage, and an optimal body condition with which to begin breeding activity. The highest quality individuals will a ccomplish all of these optima through proper management of the non-breeding season; perhaps they are highly efficient foragers, or secure the best feeding and roost sites. Other individuals will be forced to trade off some part of the equation, by arriving later, with lesser breeding pl umage, or with suboptimal body condition, and thus suffer the likely consequences of reduced breeding success or survival. Of course, similar disparities can occur without in trinsic differences in quality, if individuals experience very different conditions prior to the breeding season. For example, a bird may arrive late or in poor condition due to unusually low food availability at a particular wintering site, or because it encountered particularly ha rsh weather during the flight. In this case, Chapter 1 19 individual differences in performance may vary annually, as oppo sed to quality-based differences that may be persistent. These scenarios assume that all individuals have the same ?goal? , a concept that has permeated much of optimal migration theory. Alternatively, bi rds might differ simply because they had a different plan to begin with. It has long been recognised that different age and sex classes in a population may have different migration patterns (Myers 1981, Cristol et al. 1999), but strategic differences within these groups have received much less attention. However, several recent studies have emphasised that temporal and energetic trade-offs related to migration must be evaluated in terms of the individual. This approach recognises that the costs and benefits of early arrival may differ among individuals, perhaps due to factors such as morphology, behaviour, social status, or prior experience (M?ller 1994, Forstmeier 2002, Prop et al. 2003). In addition, some facets of migratory behaviour such as breeding arrival date appear to be endogenously programmed and to some extent heritable (Rees 1989, Berthold and Pulido 1994, M?ller 2001, Pulido et al. 2001), and perhaps not strongly influenced by competition and environmental conditions. In dividual repeatability of migration timing (M?ller 2001, Forstmeier 2002, Bety et al. 2004, Battley 2006, Gunnarsson et al. 2006) attests to this, but can be difficult to disentangle from differences in individual quality, which may be quite stable over time (Catry et al. 1999). If one optimal migration strategy exists in a population, with only the highest quality individuals achieving optimality, it follows that migration timing should be highly correlated with individual condition. In two recent studies of long-distance migratory waders, timing of departure from non-breeding grounds was unr elated to individual body condition (Battley et al. 2004) or extent of breeding plum age (Battley 2006). This suggests that individuals behave according to strategies or constraints that are their own. For exampl e, individuals from a single non-breeding site may breed in different re gions, and thus vary their departure dates according to phenologies of their respective breeding areas. In this case, inferring relative quality of individuals based on migration tim ing would be simplistic and in error. Also, differences in initial departure timing do not necessarily result in corresponding differences in breeding arrival, due to poten tial variation in migration speed and length of stay at any stopover site en route (Bety et al. 2004, Warnock et al. 2004). Within one population, there may be multiple viable strategies for maximising lifetime reproductive success, and only by viewing each individual?s annual cycle can we begin to understand the significance of individual variation in any one component. 20 Chapter 1 Aims of this thesis My goal was to study long-distance migration w ith a combination of fine-resolution focus on individuals and broad annual-cycle perspective never achieved in previous studies. By viewing migration in the context of an individual bird?s entire annual routine , I sought to identify the roots and consequences of both pe rsistent and ephemeral individual differences in migration and moult, and to identify constraints and potential bottlenecks in the annual cycle. Study species Godwits (Order Charadriiformes, Family Scolopacidae, Genus Limosa ) are medium-sized, long-legged and long-billed shorebirds that br eed on low tundra and grasslands, and spend non-breeding seasons on coastal mudflats and to a lesser extent on pasture and agricultural lands. This circumpolar genus is particularly well-suited for migration studies, because of the diversity of migration systems found both among and within its four species, from relatively short-distance movements within the northern temp erate zone to epic journeys linking the far reaches of both hemispheres. The wealth of population-level research previously conducted on godwits has described the broad spatial and temporal patterns of their annual routines, so that new information can be immediately placed in a useful comparative context. In addition, the closely-related and ecologically similar Red Knot Calidris canutus is among the most well- studied of migratory birds (see Buehler an d Piersma 2008) , offeri ng additional scope of comparison and inference. The Bar-tailed Godwit Limosa lapponica (Figure 1.1; hereafter, ?godwit?) comprises four recognised subspecies (plus on e small population of uncertain tax onomic status) whose arctic and sub-arctic breeding grounds stretch discon tinuously from Scandinavia east to Alaska (Engelmoer and Roselaar 1998). The Alaska-breeding subspecies L. l. baueri is among the world?s most extreme avian migrants. Post -breeding individuals fuel in southwestern Alaska for several weeks, approximately doubling their mass in preparation for the longest non-stop flight reported in the animal world, a trans- Pacific journey of over 11,000 km to non-breeding grounds in New Zealand and eastern Australia (Gill et al. 2005). In the austral summer, they moult and refuel once more for a 9,000 ?1 0 , 0 0 0 km non-stop flight to the Yellow Sea region of Asia in late March and early April (Battley et al. 2012 ). After a month-long stopover on the productive tidal flats of Korea and China, the go dwits embark on a third non-stop flight across the Bering Sea back to their Alaska breeding grounds. This an nual routine was spectacularly demonstrated when a female godwit (known as ? E 7 ? ) captured in New Zealand was tracked by satellite telemetry on her entire round-trip journey, which in cluded three migratory flights Chapter 1 21 encompassing 29,000 km and 21 total days of flying, including the longes t single flight ever recorded: 11,690 km non-stop from sout hwest Alaska to northern New Zealand (Gill et al. 2009, Battley et al. 2012). Appropriate to these astoundi ng flights, fat content of pre- migratory Bar-tailed Godwits is among the highest recorded in birds: 30 ?4 5 % of body mass prior to northbound departure (Battley and P iersma 2005) and up to 55% upon southbound departure from Alaska (Piersma and Gill 1998). Adult males in New Zealand increase from non-breeding masses of 230 ?2 8 0 g in October ? December to 420 ?5 0 0 g in early March; females increase from 310 ? 3 6 0 g to 520 ? 6 0 0 g (P. Battley and A. Riegen unpubl. data). In general, coastal shorebirds greatly simplify research by congregating in great numbers at a limited number of predictable sites in both winter and on migration, as opposed to dispersed, broad-fronted migrations of, for example, con tinental passerines (Busse 2001). The annual routine of L. l. baueri is particularly elegant, in that indi vidual birds essentially use only three or four sites for most of their lifetimes: one wintering site, one staging site in Asia, and one or two areas of Alaska. From a population perspective, the entire subspecies stops only in small areas of the Yellow Sea and Alaska during exten ded stops on migration, as opposed to species that make numerous shorter stops en route (Warnock et al. 2004, Eichhorn et al. 2006). While this is a potential conservation concern, it greatly simplifies the researcher?s job by limiting the sources of potential variation that must be considered. Figure 1.1 Bar-tailed Godwits at the Manawatu River estuary, New Zealand. 22 Chapter 1 In addition, Bar-tailed Godwits are ideal subjects for individual-based studies of moult and migration in the non-breeding seas on. They are long-lived (~10 ? 25 years) and have very high inter- and intra-annual site fidelity, making it po ssible to monitor individual birds throughout a multi-year study. They are gregarious, use op en habitats, and have predictable daily routines according to tidal cycles, allowing relatively easy capture and very profitable resighting efforts of individually marked birds. They undergo conspicuous and dramatic seasonal changes in plumage, which enables visual observa tion of moult without the need to repeatedly capture birds. Finally, they are large and durab le enough for the use of instrumentation such as satellite transmitters and geolocators to track movements far from the capture site, including to remote and relatively inaccessible breeding sites where direct observation is extremely difficult. Bar-tailed Godwits show delayed maturity; after ma king the southbound flight from Alaska at 3 ? 4 months of age, most young bi rds do not return to the breeding grounds or acquire full adult breeding plumage until their third or fourth summer (McCaffery and Gill 2001). However, some individuals apparently migr ate in their second year (Battley 2007), and godwits on their first northward migration have less extensive breeding plumage than they do in later years (Battley 2006). The bright red vent ral breeding plumage of adult male godwits is assumed to play a role in mate selection by females, although dir ect evidence for this is scant in this species (McCaffery and Gill 2001). It ha s been proposed that godwit plumage signals relative individual quality to po tential mates and rivals (Piersma and Jukema 1993, Drent et al. 2003). However, there is mounting eviden ce for a geographical cline in male breeding plumage within Alaska, with redder males dis proportionally represented in the northern parts of the breeding range (Rynn 1982, McCaffery et al. 2010). Also, males and females share incubation of ground nests, so plumage of both sexes has a crypsis function. Tantalising results from recent studies have shown there is much to learn about how individual godwits optimise their timing of migration. In the Firth of Thames, a major New Zealand non-breeding site, individual godwits ha d highly repeatable departure dates and extent of pre-migratory breeding plumage (B attley 2006). However, plumage at departure, which was highly variable among individuals, was not related to departure date. Also, there was a trend of larger males departing earlier. These findings contradict the simple prediction that high-quality individuals complete their br eeding plumage and depart first, with others departing later upon reaching an optimal breeding plumage. Rather, it suggests that birds operate on individualised schedules, which may re sult from heritable or derived differences in behaviour, endogenously programmed schedules of moult and migration, variation in breeding area and migration route, or carry-over effects from other parts of the annual cycle. Chapter 1 23 Body composition analysis on a cohort of fuelli ng male Bar-tailed Godwits in northern New Zealand (Battley and Piersma 2005) revealed tw o distinct groups. Fatter birds had longer wings, more developed flight muscles, less de veloped breeding plumage, and had suspended body moult. Lighter birds, presumably furt her from readiness for departure, had more developed breeding plumage, but were still in active body moult. This strongly suggests that multiple migration strategies exist in the po pulation, which may include: (1) achieve early New Zealand departure and early Alaska a rrival by compromising extent of breeding plumage; (2) achieve early New Zealand departure in order to spend more time in Asia, allowing the completion of the breeding moult; an d (3) complete the breeding moult in New Zealand, and thus have a shorter stopover in Asia or later arrival in Alaska. Whether these strategies exist on a continuu m in the population, or represent distinct subgroups, perhaps according to breeding area, re quires further investigation. Differential timing of migration to the breedin g grounds occurs in many species; specifically, the sex which benefits most by early occupatio n of breeding territories often migrates earlier (Myers 1981, Cristol et al. 1999). This is not apparent in Bar-tailed Godwits; sexes appear to depart New Zealand across the same time period (Battley 2006). This could be because New Zealand departure and Alaska arrival are not clo sely correlated, once additional variation is introduced by migration speed and length of stay at Asian stopover sites. However, there is some evidence that males and females arrive on breeding grounds approximately simultaneously (McCaffery and Gill 2001) , i ndicating there may simply be no benefit to earlier arrival by one sex. Even so, sex differences in Bar-tailed Godwits are of particular interest, due to substantial sexual dimorphism in the species . Females are approximately 20 ? 4 0 % larger than males, the sexes have minimal overlap in bill length (M cCaffery and Gill 2001) , and no other monogamous scolopacid shorebird shows such extreme plumage dimorphism (Figure 1 in Jukema and Piersma 2000). These differences may have significant implications for how the sexes prepare for migration. For example, cons equences of larger female body size might include greater energetic costs of maintaining body condition throughout the non-breeding season, greater absolute fuel mass required for migration, and gr eater mass of flight feathers to be replaced. Also, females may forage (and thus fuel) quite differently from males due to their longer bills, and may invest less energy grow ing their less extensive breeding plumage. The contribution of sexual dimorphism to patterns of moult and migration has great relevance for understanding individual constraints and strategies within the population. In summary, the annual routines of New Zeal and Bar-tailed Godwits pose some serious challenges for them. Their northerly breeding grounds are profoundly seasonal, with very 24 Chapter 1 short breeding seasons and brief but intense blooms of resources. Their migratory flights are extreme and unforgiving, stretchin g the bounds of physical endurance, while offering little or no opportunity to stop along the way. Their non-breeding season is a busy schedule of physical preparation for the next northbound migration. Thus, Bar-tailed Godwits may be among the most time- and energy-constrained of migratory birds, and therefore represent an excellent model for understanding both the potential extremes of the migratory lifestyle and the basic challenges faced by all migrants. In an annual cycle with so little apparent room for error, we might expect to find the greatest be nefits of strict adherence to individually- optimised schedules, and also the most profound consequences of environmental variation and individual condition. Study site The Manawatu River estuary (hereafter, ?Foxton?; 4 0.47?S, 175.22?E ; Figure 1.2) is a small river mouth on the west coast of New Zealand?s North Island. Before emptying into the Tasman Sea at a sandy beach, the Manawatu Rive r is bordered for approximately 3 km by tidal mudflat and saltmarsh, and accommodates moderate human recreational use from the Figure 1.2 The Manawatu River estuary, New Zealand. Box indicates core study area. Image source: Google Earth. Chapter 1 25 adjacent community of Foxton Beach. One of th e more popular bird-watching spots in New Zealand, the estuary hosts small populations of both local- and arctic-breeding shorebirds during the austral summer months. Although godwits use many sites in New Zealand in larger numbers (Figure 1.3), Foxton is ideal for studies of individual variation for several reasons. Firs t, intra-season movements of individuals to and from the site are rare and negligible; the estuary is isolated by many kilometres from other suitable godwit sites and ap pears to provide for all of their non-breeding requirements. The site is small and easy to acces s, allowing observation of the entire godwit population from several single (a nd dry) vantage points in most circumstances, even when the flock is widely scattered and foraging at low tid es. The resident godwit population is less than 300 birds, so it is much easier to locate and follow specific individuals than at other sites where they may be easily lost among thousands of flock-mates. Despite the small population, Figure 1.3 Numbers of Bar-tailed Godwits at non-breeding sites in New Zealand (from Southey 2009). Arrow indicates location of Manawatu River estuary study site. 26 Chapter 1 variation in plumage and migration timing is similar to that seen at larger sites and in the population as a whole, and so extrapolation of findings to the larger New Zealand context seems appropriate. Most important for my specific study are two final points. Due to the prescience of my primary supervisor, more than 60 godwits were individually colour-banded at the estuary in 2006 ? 2 0 0 7 , 45 of which were still using the site when I began research in November 2007. This allowed the immediate collection of observational data without trapping, and the opportunity to observe any effect of new trapping on the phenomena of interest. Lastly and perhaps most crucially, predictable high-tide roost use by a small population of extremely site- faithful birds allowed the repeated captures necessary for the deployment and retrieval of geolocators. This fostered a previously unheard-of recapture rate of non-breeding shorebirds that is presently the envy of shorebird biologists worldwide. Overview of research chapters The methodological approach for my research was quite simple. Using direct observation and digital photography, I closely monitored in dividually marked Bar-tailed Godwits in New Zealand for multiple years, through the major non -breeding events of migratory arrival, contour and flight feather moults, and u ltimately northbound departure. I linked this information with events outside of New Zealand by tracking a subset of the same individuals on their complete round-trip migrations, through stopover in Asia, breeding in Alaska, and eventual return to New Zealand. I used these data to address the following key questions: 1. How are a godwit?s moult and migration schedules in New Zealand influenced by its ultimate destination in Alaska? 2. Do individual differences in timing of migratory departure from New Zealand persist through the migration and result in differences in timing of arrival at breeding sites? 3. What external local factors affect timing of departure from New Zealand? 4. How do individuals with varying extents of breeding plumage schedule their moults differently? 5. What magnitude of delay in migratory arrival or moult in the non-breeding season will compromise a godwit?s breeding plumage or migration schedule the following season? 6. How do flexibility in moult and migration schedules differ, and what does this tell us about selective forces operating throughout the annual cycle? Chapter 1 27 The heart of this thesis is six research chap ters, each written as stand-alone contributions for international peer-reviewed journals. Therefore, there is a certain amount of redundancy among them that was necessary to make each an independent piece, but I have standardised the formats to make the entire thesis more cohe sive. Furthermore, the chapters appear in the order in which they were written. This is important because I generally analysed, wrote, and submitted each chapter for publication before I fu lly understood what would be revealed in subsequent chapters. So, a chronological reading of this thesis is a good representation of the process of discovery that I experienced across a pe riod of nearly two years. In some cases, I propose hypotheses in one chapter that are rejected in the next one. That is one of the exciting parts of science, and so I have left it all in there. In the following overview, I describe the central challenges of each chapter, with an emphasis on how specific methods addressed my key questions. Chapter 2. Breeding latitude drives individual schedules in a trans-hemispheric migrant bird When I started this project, a new technology wa s revolutionising the study of long-distance migration. Light-level geolocation is an elegant concept: because day length varies with latitude and time of day varies with longitude, a datalogger that records only sunrise and sunset can allow the calculation of an animal?s daily position on the globe. By 2008, geolocators had already taken the seabird world by storm (Shaffer et al. 2006, Gonz?lez-Sol?s et al. 2007), and had recently been miniaturised sufficiently to enable tracking of birds smaller than 200 grams (e.g., Stutchbury et al. 2009). Today, scores of geolocator-equipped studies worldwide are unmasking heretofo re-unknown migratory routes and patterns at an astonishing rate (e.g., Egevang et al. 2010, Klaassen et al. 2011). Geolocators offer some major benefits over satellite transmitters, which have been pr ohibitively expensive and bulky, have limited battery life, and require attachment methods (imp lantation or backpack harness) that have potentially significant impacts on bird behaviour and survival. Geolocators are relatively inexpensive, much smaller, and last up to two years. The drawbacks are reduced precision of locations, including an inability to calculate latitude within about two weeks of either equinox, and the requirement to recapture birds to acces s the data. This last problem had previously limited deployment of geolocators to the breeding season; typically, only site-faithful breeding birds offered much likelihood of r ecapture. The fortuitous nature of my study site and species allowed the first deployment of geolocators on birds from a non-breeding site, and the first on any shorebird. One challenge was to design an attachment method that would withstand extremes of temperature and salt-water immersi on for multiple migrations, and yet be easily removed when the bird was recaptured (see Appendix 1). 28 Chapter 1 Many of the big, fundamental questions rega rding Bar-tailed Godwit migration had already been answered with satellite telemetry (Gill et al. 2009, Battley et al. 2012) and a very successful mark-resight programme (Riegen 1999, Battley et al. 2011), and so I generally knew where New Zealand godwit s would go, and about when th ey would go there. My goal was to go beyond just tracking movements, by linking geolocation data with detailed non- breeding information on the same individuals. I sought to describe the entire range of migration strategies in the population, and so it wa s crucial to deploy geolocators on a fairly representative sample of birds. For the first deployment in March 2008, we captured a small group of godwits whose migratory habits were entirely unknown. Because we made these captures after migration had already begun, we di d not sample early-departing birds, and by chance we also captured few late-departing bird s. For the second deployment in October 2008, I therefore specifically targeted some indi viduals that had previously departed very early or late relative to the population, to tr y to explain what thes e birds were doing differently. This was probably the most important single decision of my entire project. With year-round information on individual go dwits, I could investigate whether patterns observed in New Zealand were strongly influenced by what godwits were doing elsewhere. Where in t he world (literally) should I be looking for the reasons behind a godwit?s plumage and migration timing? Also, I could test whether differences in timing of departure from New Zealand persisted through subsequent stages of migration. Did early-departing godwits actually arrive in Alaska and breed earlier, or did variation in the Asia stopover erase initial timing differences? Chapter 3. Geographic variation in morphology of Alaska-breeding Bar-tailed Godwits is not maintained on their non-breeding grounds in New Zealand Over several decades, a mostly unpublished body of evidence had accumulated, suggesting that plumage and size of godwits might vary geographically across their breeding range in Alaska (Rynn 1982, McCaffery et al. 2010). This was not particularly surprising, as many shorebird species with broad Northern Hemisphere distributions, including Bar-tailed Godwits, demonstrated measurable inter-populati on differences (Engelmoer and Roselaar 1998). However, these differences generally o ccurred along a longitudinal axis, between populations that were isolated by discrete breeding ranges and separate migratory flyways. Geographic variation within Alaskan godwits woul d have to occur along a latitudinal axis and within a population using a single migratory route. Because such a pattern in the breeding range could significantly affect inferences I would make about godwits in New Zealand, I needed to formally describe it. This required me to assemble and add to all previously collected data on Alaskan godwits. First, I ?simply? Chapter 1 29 (an elegant solution masking great expense, effort , and logistical difficulty) went to two Alaskan breeding areas, on the Seward Peninsula and the nort h slope of the Brooks Range, and digitally photographed every godwit I could fi nd. Then I visited three museum collections in North America and examined every av ailable Alaskan Bar-tailed Godwit specimen. I supplemented this by compiling all available go dwit photographs taken by myself and other biologists during previous work in Alaska, an d morphometric data from all previous captures in Alaska and all New Zealand-banded birds that had been tracked to precise breeding locations. Together, these data formed the most comprehensive dataset on Alaskan godwits yet assembled. For comparison, I compiled non-breeding morphometric data from over 20 years of captures across New Zealand (with a span of latitude similar to the breeding range in Alaska). Chapter 4. Impacts of wind on individual migration schedules of New Zealand Bar-tailed Godwits Individual Bar-tailed Godwits are very consistent across years in the dates they depart from New Zealand on migration (Battley 2006). However, godwits are not just automatons with accurate calendars, and a no n-stop flight of 9,000 ? 1 0 , 0 0 0 km is not something to enter into lightly. An individual?s readiness or willingness to embark on migration on a given day may depend on previous conditions (in terms of its ability to moult and fuel sufficiently for migration) or factors experienced on the day in question. For example, a bird otherwise ready to migrate may not do so if current weather would make the flight more difficult, or if no other birds were willing to serve as travelling companions. Such day-to-day influences can be very difficu lt to study, due to the requirement of precise movement data for a large number of individuals simultaneously. Many species migrate at night or from habitats that do not allow direct observation, so that departures can only be detected through remote tracking of individuals , which limits sample size and cannot provide a population context. For gregar ious species that migrate conspicuously from open habitats, such as shorebirds, the challenge is to know the individual composition of departing flocks. Once again, my study site was the star and saviou r of my project, this time allowing me to collect near-perfect departure data on 50 ? 6 0 individual godwits each year (to the exact minute in 76% of cases), using only a spotting scope and a digital camera (Figure 1.4). A worrisome question at the start of my pr oject was whether godwits migr ating from a very small site would behave similarly to those at sites hosting thousands of birds. At two large sites in New Zealand, godwits departed over a four-week period (Battley 1997, 2006), trickling out in flocks of tens to hundreds. What if my entire flock of 250 departed together, or moved to other sites before their actual departures from New Zealand? I would not have much of a study. Fortunately, that was not the case. 30 Chapter 1 Though it required considerable physical and ment al endurance, relentless focus, and more patience than I thought I possesse d, standing out on a mudflat for hours on end for four weeks straight waiting for individual birds to migrate was (all three years) among the most rewarding wildlife experiences I have ever had. It was pa rticularly during these times that the godwits transcended the data they provided and were clearly individuals. It was truly humbling to watch an individual bird that I had followed intimately for the previous six months lift its wings and disappear to the horizon, for I knew it would next touch eart h on a mudflat in China or Korea. Chapter 5. Contour feather moult of Bar-tailed Godwits in New Zealand and the Northern Hemisphere reveals multiple strategies by sex and breeding region In Alaska, male godwits have more extensiv e breeding plumage than females, and plumage within each sex varies substantially. How are these differences accomplished? During the non-breeding season, all godwits undergo a comp lete feather replacement (including flight feathers) into winter plumage and then replac e a portion of their body feathers again to transform into breeding plumage. Godwits with more extensive breeding plumage must either moult faster or spend more time in moult, and this extra investment could occur either before northbound migration in New Zealand or during a month-long stopover in Asia. Describing moult rate and duration at the indivi dual level has rarely been attempted in wild populations, because it usually requires ?in - hand? assessment involving multiple captures. Figure 1.4 Migratory departure of Bar-tailed Godwits from the Manawatu River estuary on 26 March 2010. There are five colour-banded birds in this flock. Chapter 1 31 Because plumage changes in Bar-tailed Godwits are very conspicuous, and individuals at Foxton could be encountered and photographed r eliably throughout the non-breeding season, I decided to try it without captures. Three years and more than 18,000 digital photographs later, I had described the progression of pre-basic an d pre-breeding contour feather moult in New Zealand for every marked godwit in my study. Ho wever, the problem remained that neither of these moults occurred entirely in New Zealand; pre-basic moult began in Alaska before southbound migration, and pre-breeding moult was completed in Asia after the first leg of the northbound trip. Based on ultimat e breeding plumage in Alaska (Chapter 3), I estimated the duration and proportion of moults performed by godwits while they were in the Northern Hemisphere. Chapter 6. Carry-over effects and compensation: late arrival on non-breeding grounds affects wing moult but not plumage or schedules of departing Bar-tailed Godwits The idea that long-distance migrants are high ly constrained by their demanding annual routines contains the assumption that they are operating near the limits of their capabilities. The logical and testable prediction from this is that unfavourable conditions or events during one part of the year will negatively affect a bird?s performance in subsequent stages, by reducing the time or energy available for later activities. In my study, I had the unique opportunity to test whether timing of arrival after the longest non-stop flight ever recorded affected an individual bird?s ability to prepare for its next migration. During six months in New Zealand, godwits recover from their 8-da y flight from Alaska, complete pre-basic contour feather moult, replace their entire set of flight feathers, start moulting back into breeding plumage, and fuel for another 7-day f light to the Yellow Sea. If ever there was a scenario in which to expect carry-over effects, this was it. The non-breeding season of godwits is largely occupied by their 3 ? 4 month full replacement of flight feathers, and to assess the effects of migration timing on this task, I had to do what had never been done before: describe timin g and duration of primary feather moult for individuals in a wild population, for multip le years and without capturing the birds. Fortunately, I possessed 18,000 photographs of individually identifiable godwits, many of which contained birds flying or stretching their wings. From these, I extracted two-year primary moult phenology information for 43 indi vidual godwits. I added corresponding data on migratory arrival, contour feather moult (Cha pter 5), and migratory departure (Chapter 4) to make inter-year comparisons across the entire non-breeding season. 32 Chapter 1 Chapter 7. Absolute consistency: individual versus population variation in timing of annual life-history stages of a long-distance migrant bird Sometimes punctuality is important and at other times it doesn?t matter. Presumably, natural selection has shaped the annual routines of birds such that they schedule life-history stages very precisely when the fitness benefits of correct timing are great, but are more lax when the costs of precise timing outweigh the benefits. Many studies have focused on how migratory birds schedule arrival on their breeding grounds, because it is clear that timing of breeding is important in seasonal habitats. However, the scheduling of this important event is rarely viewed in the context the entire annual cycle, du e to the difficulty of following individual birds through successive seasons. This is a significant omission, because the relative precision with which an event is scheduled, as opposed to absolute precision, may tell us more about a bird?s annual priorities. Using multi-year data on individual godwits observed directly in Ne w Zealand (Chapters 4 ? 6 ) and remotely via geolocators for the rest of the year (Chapter 2), I constructed the most complete picture of year-round schedules availab le for long-distance migrant bird. With these data, I tested two key predictions: (1) that both inter- and intra-individual variation in timing would decrease through successive stages leading up to the breeding season, and (2) that migratory movements would be more precisely scheduled than moults. Chapter 8. Synthesis: an evolving view of long-distance migration In this chapter, I focus on synthesising the disp arate research chapters into a coherent whole, and discuss the significance of the collective findings to our understanding of individuality, constraints, regulation of the annual cycle, and long-distance migration. Chapter 9. Future directions In conclusion, I briefly outline some potenti al avenues of research that logically follow from the findings of this thesis. 33 Chapter 2 Breeding latitude drives individual schedules in a trans-hemispheric migrant bird Conklin, J.R., P.F. Battley, M.A. Potter & J.W. Fox Nature Communications 1: article 67, online (2010) doi: 10.1038/ncomms1072 34 Chapter 2 Abstract Despite clear benefits of optimal arrival time on breeding grounds, migration schedules may vary with an individual?s innate quality, non-breeding habitat, or breeding destination. Here, we show that for the Bar-tailed Godwit Limosa lapponica baueri , a shorebird that makes the longest known non-stop migratory flights of any bird, timing of migration for individuals from a non-breeding site in New Zealand was strongly correlated with their specific breeding latitudes in Alaska, USA, a 16,000?18,000 km journey away. Furthermore, this variation carried over even to the southbound return migration, six months later, with birds returning to New Zealand in approximately the same order in which they departed. These tightly scheduled movements on a global scale strongly suggest endogenously-controlled routines, with breeding site as the primary driver of temporal variation throughout the annual cycle. Introduction Some migratory birds ?winter? vast distances from where they nest, yet are under strong selection pressure to arrive on the breeding grounds at the time that best assures successful reproduction. Despite clear links between th e timing and success of breeding, substantial variation exists in the migration timing of indivi duals, which can largely be explained in terms of three primary mechanisms: (1) variation in individual quality; (2) variation in fuelling resources; and (3) geographic variation among br eeding destinations. In the first case, which has received the most attention in studies of migration timing, individuals possess heritable or acquired differences in moulting or fuelling efficiency, with only the highest-quality individuals achieving an optimal migration schedule (M?ller 1994). In the second, individuals from high-quality non-breeding habitats may migrat e earlier than those in low-quality habitats (Marra et al. 1998, Gunnarsson et al. 2006) and consequently experience increased breeding success (Smith and Moore 2005). The third scenario involves geographic variation in temporal availability of breeding resources, resu lting in individuals having different optimal schedules based on their respective breeding sites. This phenomenon is easily illustrated when races of the same species have clear differences in breeding range and migration timing (Wood 1992, Battley et al. 2005), but within-population demonstr ations are rare. For example, tundra-breeding birds at northerly latitudes generally begin nesting as soon as snow recedes from their breeding sites (Smith et al. 2010). Within Alaska, tundra becomes snow-free 2 ? 4 weeks earlier in the southwest than it does in the far north (NOAA 2010b). As a result, optimal arrival dates for individuals in widely-distributed breeding populations may vary substantially (Holmes 1971). Chapter 2 35 Bar-tailed Godwits Limosa lapponica baueri depart New Zealand during March and early April, and fly approximately 10,000 km to sites on the Yellow Sea coast of Korea and China (McCaffery and Gill 2001). After refuelling, godw its depart Asia from late April to late May, and fly approximately 6,000 km to coastal sites in southwestern Alaska. They then disperse to coastal tundra sites across western and northern Alaska (59 ?7 1 ? N ) for breeding in May ? July. The godwits then congregate in southwestern Alaska to refuel before departing in late August?early October on a non-stop tr ans-Pacific flight of 11,000 ?12,000 km to New Zealand, the longest non-stop flight recorded in birds (Gill et al. 2009). The difficulty of tracking small-bodied bird s across great distances has hampered our understanding of cross-seasonal interactions betw een events on opposite sides of the globe. Studies are often limited to methods (e.g., stable isotopes) that require indirect steps to link the breeding and non-breeding grounds of individu als. The advent of small, lightweight geolocators allows these links to be made directly (Stutchbury et al. 2009), but the need to retrieve units to access data has limited their use to site-faithful species using accessible, high- density breeding sites (Eichhorn et al. 2006, Shaffer et al. 2006). Bar-tailed Godwits breed cryptically in low densities across over 1,800 km of remote coastal tundra within Alaska (McCaffery and Gill 2001) , making breeding stud ies extremely difficult. However, high non- breeding site-fidelity of godwits in New Zealand allows the use of geolocators for inferences across the entire breeding range, while co ntrolling for non-breeding habitat quality. Our study is the first to deploy geolocators on non-breeding birds, and also the first involving shorebirds (suborder Charadrii) , a group famous for extreme long-distance migrations. We attached geolocators to non-breeding Bar-tailed Go dwits (Figure 2.1) at a small estuary in New Zealand, and recaptured them after their return migration. We show that the timing of each step in a >30,000 km round-trip migration (northbound departure from New Zealand, departure from fuelling sites in Asia, arrival on Alaska breeding grounds, and subsequent southbound departure) is primarily dependent upon the latitude of an individual?s breeding site. Methods Geolocation data1 During two non-breed ing seasons (2007 ? 2 0 0 8 and 2008 ?2 0 0 9 ) , we captured godwits from a small (200 ? 2 8 0 birds) population at the Manawatu River estuary, New Zealand (40.47?S, 175.22?E) and attached leg-mounted light-sens itive geolocators (British Antarctic Survey 1 See Appendices 1?2 for more detailed methods. 36 Chapter 2 model MK14; 1.4g; 2-year life). Return rate of these birds was 95%; we recaptured 80% of those available. Some birds carried geolocators for two migrations, and some units failed during deployment; for each bird, we used the first year with the most complete data. We calculated breeding locations for 16 birds (7 males, 9 fem ales); for three of these, subsequent date of Alaska departure was unav ailable due to geolocator failure. One other female was tracked for the entire migration, but did not settle at a breeding site (thus n = 14 in Figure 2.5). The geolocators record sunrise/sunset, allowi ng calculation of latitude and longitude (?130 km error, based on ground-truthing units and re sightings of instrumented godwits), except during ?15 days (d) of the vernal or autumnal equinox, when only longitude is reliable. To derive fuelling sites in Asia and breeding sites in Alaska, we averaged twice-daily locations over periods when birds were relatively stationary, excluding clear outliers caused by weather or bird behaviour. For thre e birds breeding north of the Arctic Circle (thus providing no sunrise/sunset data during this time), we as sumed a breeding latitude of 70.2?N (known breeding range in this region occurs 69.5 ? 7 0.8?N; McCaffery and Gill 2001). Based on resightings of colour-marked godwits at the Mana watu River estuary, timing of migration was similar between years; mean northbound departure ( n = 48 birds) and southbound arrival Figure 2.1 Individually colour-banded male Bar-tailed Godwit with tibia-mounted geolocator, prior to departure from New Zealand. Photo by Phil Battley. Chapter 2 37 ( n = 45 birds) in 2008 and 20 09 differed by 0.48 and 1.0 d, respectively. In addition, snow- free dates for Alaska breeding sites differed by <5 d between the two years (NOAA 2010b), despite annual variation up to 18 d (1998 ? 2 0 0 9 ). Therefore, we combined geolocator data from 2008 (11 birds) and 2009 (6 birds) for analysis. Departure from the Manawatu River estuary was similar for colour-banded males and females in both years (2008: t61 = 1.42, P = 0.16; 2009: t 60 = 0.747, P = 0.46), as was post-breeding return to the site (2008: t 51 = 0.546, P = 0.59; 2009: t5 7 = 0.807, P = 0.42). Among geolocators , there was no apparent difference in migration timing between males and females after controlling for breeding latitude, although sample sizes were quite small. In addition, field observations on the breeding grounds provide no evidence for protandry in breeding arrival (J. Conklin pers. obs.) Therefore, we combined male and female geolocator data for analysis. Hatching success Geolocators also allowed estimation of hatch ing success by indicating periods of nest incubation (Eichhorn et al. 2006). During the breeding season, geolocators registered nights as regular, cleanly demarcated periods of 0 ? 4.5 hours (h) of darkness , depending on latitude. Conversely, days appeared as con tinuous light, irregularly broken by very brief (<1 h) shading events, most likely corresponding to behavi ours such as wading or sitting. Within 6 ? 2 5 d of apparent arrival on breeding grounds, most birds disp layed a conspicuous pattern of incubation, in which semi-reg ular shading events of 4 ? 13 h were overlayed upon the day/night pattern for periods up to 25 d. When data clearly indicated incubation ( n = 14), we designated birds as ?successful? ( 2 1 ?25 consecutive days of incubation; expected incubation duration 20 ? 2 1 d; McCaffery and Gill 2001) or ?failed? (<18 d). Breeding plumage Prior to departure from New Zealand, Bar-tailed Go dwits undergo a breeding plumage moult, during which some portion of non-breeding vent ral contour feathers are replaced with bold rufous-coloured feathers (McCaffery and Gill 2001). Using field observations and digital photographs, we scored breast and belly plumage on a scale of 1 ? 7 (1 = no red feathers; 7 = completely replaced with red feathers; Piersma and Jukema 1993). Males carrying geolocators departed the study site with plumage scores of 4.5 ? 5.5. Results Timing of migration Godwits from our study site used breeding sites spanning 59.7 ?7 0.2?N latitude (Figure 2.2), encompassing most of the known Alaska breeding range for the subspecies. Northerly 38 Chapter 2 breeders migrated later than southerly breeders in each phase of the northward journey: departure date from New Zealand was significantly correlated with breeding latitude ( Figure 2.3a), as were Yellow Sea departure (Figure 2. 3b) and date of arrival at breeding sites (Figure 2.3c). The correlation between New Zealand depart ure date and the location of a breeding site a 16,000 ?1 8 , 0 0 0 km journey away strongly sugges ts endogenously-controlled schedules, as environmental cues indicating tundra conditions in the opposite hemisphere are improbable. The relationship between migration timing and latitude became stronger with each stage of the northbound migration, implying a tightening of these programmed schedules with proximity to the breeding grounds. For th e post-breeding southbound ret urn flight, northerly breeders again migrated later: departure date from Alaska was significantly correlated with breeding latitude (Figure 2.3d). Duration of stopover in the Y ellow Sea was highly variable (30 ? 50 d), which may reflect individual strategies with regard to moult and fuelling, or variation in stopover site quality. Godwits use fuelling sites spanning 600 km of latitu de in the Yellow Sea, and these sites vary greatly in size, level of human impacts, and competition with other migrating shorebirds. There was no apparent relationship between choice of fuelling site and timing of migration, stopover duration, or breeding latitude. However, stopover duration for godwits breeding at Figure 2.2 Breeding locations derived from geolocators for 16 Bar-tailed Godwits tracked from New Zealand. Alaska image courtesy of NASA. C ha pt er 2 3 9 Fi gu re 2 .3 T im in g of m ig ra tio n of B ar -t ai le d G od w its w as c or re la te d w ith b re ed in g la tit ud e in A la sk a. S om e ov er la pp in g po in ts h av e be en n eu tr al ly o ff se t fo r cl ar ity . F ig ur e ax es h av e be en t ra ns po se d to a id v is ua lis at io n of g eo gr ap hi ca l r el at io ns hi ps . ( a) D at e (D ay 1 = 1 Ja nu ar y) o f de pa rt ur e fr om N ew Z ea la nd i n re la tio n to b re ed in g la tit ud e (r 2 = 0. 45 , sl op e = 1. 22 ? 0 .3 6, n = 1 6, P = 0 .0 05 ). (b ) D at e of d ep ar tu re fr om t he Y el lo w S ea in r el at io n to b re ed in g la tit ud e (r 2 = 0. 80 , s lo pe = 1 .6 7 ? 0. 22 , n = 1 6, P < 0 .0 01 ). (c ) D at e of a rr iv al o n br ee di ng g ro un ds i n re la tio n to b re ed in g la tit ud e (r 2 = 0. 91 , sl op e = 2. 16 ? 0 .1 9, n = 1 6, P < 0 .0 01 ). (d ) D at e of de pa rt ur e fr om A la sk a in r el at io n to b re ed in g la tit ud e (r2 = 0 .3 2, s lo pe = 1 .4 3 ? 0. 64 , n = 1 3, P = 0 .0 46 ). Chapter 2 39 40 Chapter 2 higher latitudes was greater than those breeding farther south ( Figure 2.4). Similarly, after arrival in Alaska, northerly breeders remained at coastal sites slightly longer before moving to breeding sites (range: 0 ? 1 3 d; duration vs. breeding latitude: r2 = 0.32, n = 16, P = 0.023). Thus, the variation among individuals increa sed with each successive segment of the migration (range of New Zealand departure = 20 d; Yellow Sea departure = 23 d; breeding arrival = 28 d), and total travel time from New Zealand to breeding sites increased with breeding latitude (range: 48 ? 6 7 d; duration vs. breeding latitude: r2 = 0.40, n = 16, P = 0.009). This is unsurprising, as godwits breedin g in northern Alaska must travel 1,000 ? 1,600 km further than southerly breeders, and may require greater reserves to prepare for colder and less predictable conditions on their breeding grounds. Although breeding arrival has well-understo od time constraints, migration to non-breeding grounds is considered to be much less time-selected (Alerstam and Lindstr?m 1990, McNamara et al. 1998). However, birds departed Alaska in approximately the same order in which they departed New Zealand (Figure 2.5), an d the span of departures from Alaska was 27 d, similar to the northboun d migration. Breeding success cre ates great potential variation in the date an individual may begin fuelling for southbound migration, because birds caring for young through fledging may invest 3 ?6 weeks more than those that fail during incubation. However, there was no evidence that failed bre eders used this ?advantage? to achieve earlier southbound departure: total time spent in Alaska was nearly identical for successful (mean = 125.0 d, n = 4) and failed (mean = 125.6 d, n = 7) breeders. Furthermore, southerly breeders might be expected to migrate south later, because apparently suitable breeding conditions persist later at lower latitudes and offer greater re-nesting opportunity; our findings do not support this. These data suggest rigidity in migration schedule, perhaps evolved to best exploit a predictable peak of fuelling resources or favourable wind conditions (Battley et al. 2005) necessary for the southbound flight. Figure 2.4 Length of stay in the Yellow Sea for Bar-tailed Godwits on northbound migration was positively correlated with breeding latitude in Alaska (r2 = 0.51, n = 16, P = 0.002). Chapter 2 41 Breeding plumage For species in which one sex competes for br eeding partners or territories, variation in plumage may communicate relative individual quality to competitors or potential mates (Jawor and Breitwisch 2003). The extremely var iable breeding plumage of male godwits has led to the hypothesis that redder males may be higher quality individuals, and therefore may migrate earlier than paler males (Drent et al. 2003). In our study, extent of male breeding plumage was unrelated to departure from New Zealand ( r2 = 0.36, n = 9, P = 0.87); this was also true at another non-breeding site in New Zealand (Battley 2006). Howe ver, redder males departed from the Yellow Sea later than paler males ( r2 = 0.46, n = 9, P = 0.044) and arrived later on the breeding grounds ( r2 = 0.79, n = 7, P = 0.007), contrary to the prediction, but similar to observations of L. l. taymyrensis in Europe (Drent et al. 2003). Although supported in other species, the use of breeding plumage as an index of individual quality in Bar-tailed Godwits is far from straightfo rward. First, there is no empirical evidence linking breeding plumage with reproductive succe ss in this species, and links with other parameters such as body condition, parasite lo ads, or survival are equivocal (Piersma and Jukema 1993, Piersma et al. 2001, Drent et al. 2003, Battley and Piersma 2005, Battley 2007). Second, plumage varies geographically among Al askan godwits, with redder males breeding at higher latitudes, on average (Cha pter 3). Accordingly, redder males in this study bred at higher latitudes ( r2 = 0.73, n = 7, P = 0.015). However, due to co ntinued contour moult during northbound stopover (Piersma and Jukema 1993 ) , the degree of which may vary among individuals, it is unknown how closely variatio n in plumage upon New Zealand departure reflects that on the breeding grounds. Figure 2.5 Order of migratory departure from New Zealand and Alaska were similar for individual Bar- tailed Godwits (Spearman rank correlation: rs = 0.74, n = 14, P = 0.003). Some overlapping points have been neutrally offset for clarity. 42 Chapter 2 Discussion Our findings implicate individually-optimised br eeding arrival date as the primary driver of variation in migration schedule of godwits: latitude of an individual?s breeding site (and, by extension, the approximate timing of snow melt from that patch of Alaskan tundra) could be traced back to that bird?s departure from New Zealand 7? 1 0 weeks earlier and more than half a world away. This strongly suggests endogenous programming of migration, and warns against quality-based inferences regarding any single stage of the migration without knowledge of the entire annual cycle. Nonethele ss, competition and individual quality may yet act at a fine scale; i.e., birds at the same latitude vary in quality, arrival date, and breeding success. Our sample is insufficient to evaluate ti ming differences among birds breeding at the same latitude, but the small amou nt of variation in breeding arrival left unexplained by latitude itself (10%) suggests such variation is r elatively minor. Latitude is a coarse index for earliest availability of a site, and does not accou nt for small-scale geographic variation, which may promote differing optimal arrival dates even for sites at similar latitudes (Smith et al. 2010). This makes the extremely tight relationshi p between breeding arrival and latitude all the more surprising. However, given the br ief breeding season at high latitudes, it is nonetheless conceivable that a few days difference in arrival could significantly affect mate acquisition or retention (Gunnarsson et al. 2004), or breeding success. If individual quality or condition were largely dr iving timing of migration, we would expect early-arriving birds to maximise reproductiv e potential by occupying the highest-quality breeding sites. Low-latitude breeding sites could be more desirable, due to a shorter migration distance, milder and more predictable conditi ons, and a longer snow-free season. However, the drawbacks of breeding farther north may be offset by a latitudinal cline in nest predation rates (McKinnon et al. 2010). Given the low breeding density of Bar-tailed Godwits and a general lack of evidence of food- or habitat- limitation among tundra-breeding shorebirds, it is improbable that early-arriving birds, through ear ly occupation of low-latitude breeding sites, force later birds to breed at high latitudes. It is more likely that migration timing and breeding site are linked and heritable, and maintained by either genetic structure within the population or environmentally-mediated regulation of gene expression (Jaenisch and Bird 2003). Some Holarctic shorebird species appear to have radiated across vast regions very recently, such that clear population differences in breeding range and migration are only weakly reflected in genetic structure within the population (Buehler et al. 2006). In some cases, epigenetics, a field of vast potential for ecologists, may be the key mechanism for persistent intra- specific variation. Chapter 2 43 Our study sheds light upon tantalising results from prior Bar-tailed Godwit studies. At the Firth of Thames, New Zealand, individual godw its showed high annual repeatability of both departure date and extent of breeding plumage, but did not conform to expectations of ?high - quality? (redder) birds migrating earlier ( Battley 2006). Likewise, male godwits whose moult and mass indicated readiness for early departure were paler and larger than males not yet in migratory condition (Battley and Piersma 2005). Both of these results are now explained by the links between breeding site, timing, and plumage. The similar later northbound departure of redder males in the European population ( L. l. taymyrensis ; Drent et al. 2003) suggests these patterns may exist across the entire ra nge of the species, but that study found no evidence of geographic variation in plumage within the breeding range. Interestingly, the population structure within Alaska does not appear to persis t in the non-breeding season. Other sites in New Zealand show a range of no rthbound departure dates similar to our study site (Battley 1997, 2006) , suggestin g that godwits from the entire Alaska breeding range mix freely in the non-breeding season. For long-distance migrants, selection for the tim ing of breeding may occur at very different temporal and spatial scales than selection for timing of migration per se. In Bar-tailed Godwits, timing of migration appears quite consis tent, both at the population and individual (Battley 2006) level, despite substantial annual va riation in date of snow melt; this pattern appears common among tundra-b reeding, long-distance migrant shorebirds (Niehaus and Ydenberg 2006, Smith et al. 2010). Timing of long-distance f lights likely evolved in response to long-term global patterns in the timing of fuelling resources and beneficial prevailing winds, while nesting phenology appears very sensitive to local and annual variation in conditions dictating availability of breeding sites (Smith et al. 2010). This may explain why timing of southbound migration was unresponsive to apparent duration of breeding investment in our study. Unfortunately, the potentially conflicting pressures of optimal timing of migration and breeding may make long-distance migrants such as godwits particularly vulnerable to effects of climate change (Both et al. 2006), if rigid flight schedules contribute to a mismatch between breeding arrival and optimal nest initiation (Bot h and Visser 2001) , or preclude adaptation to temporal shifts in resources or weather during migration. The departure of individuals from New Zealand and Alaska (two events separated by six months and over 11,000 km) in approximately the same order and span of days is quite surprising, in light of potential intervening var iation caused by individual differences in flight speed, stopover duration, migration distance, duration of suitable breeding conditions, breeding success, body size, moult speed, fo raging ability, and habitat quality. This relationship between timing of northbound and southbound migration, which lacks a clear 44 Chapter 2 theoretical foundation in migration literature (McNamara et al. 1998) , may reflect constraints operating on the entire annual cycle. Although flexibility in duration and investment in annual activities such as moult (Hall an d Fransson 2000, Dawson 2004) and fuelling (Prop et al. 2003), and even in migration route (Eichhorn et al. 2009) , have been demonstrated in other species, the extreme nature of the godwit?s migration may naturally ensure reduced variation in these parameters. Accordingly, the specificity with whic h breeding site dictates migration timing and the rigid repeatability of individual migration schedules (Battley 2006) seem especially high, compared to that found in sh orter-distance migrant birds. Every New Zealand Bar-tailed Godwit breeding in Alaska must be cap able of a 10,000 km non-stop flight to Asia, followed by a 6,000 km flight across the northe rn Pacific Ocean; this is surely among the most challenging migrations in birds. Low-quality i ndividuals are unlikely to complete this migration, and those in poor condition may not attem pt it at all. It is unlikely that any truly ?low- quality? godwits reach the breeding grounds. In this model of godwit migration, an individual?s breeding site is endogenously controlled, and all other annual events are shifted temporally to optimise arrival on the breeding grounds. Among the consequent predictions is that no rtherly breeders should exhibit delayed completion of feather moult and pre-migratory mass gain in New Zealand, and the duration of these activities may be relatively constant, rega rdless of migration schedule. Using migration timing as a reliable index for br eeding latitude, these predictions are now testable by godwit studies restricted to non-breeding sites. We migh t also expect to find similar patterns in other long-distance migrant species in which indi viduals from one non-breeding site may breed across vast geographic ranges (e.g., Great Knots Calidris tenuirostris ; Battley et al. 2004). 45 Chapter 3 Geographic variation in morphology of Alaska-breeding Bar-tailed Godwits is not maintained on their non-breeding grounds in New Zealand Conklin, J.R., P.F. Battley, M.A. Potter & D.R. Ruthrauff Auk 128: 363?373 (2011) 46 Chapter 3 Abstract Among scolopacid shorebirds, Bar-tailed Godwits Limosa lapponica have unusually high intra- and inter-sexual differences in size and breeding plumage. Despite historical evidence for population structure among Alaska-breeding Bar-tailed Godwits L. l. baueri , no thorough analysis, or comparison with the population?s non-breeding distribution, has been undertaken. We used live captures, field photography, museum specimens, and individuals tracked from New Zealand to describe geographic variation in size and plumage within the Alaska breeding range. We found a north?south cline in body size in Alaska, in which the smallest individuals of each sex occurred at the highest latitudes. Extent of male breeding plumage (proportion of non-breeding contour feathers replaced) also increased with latitude, but female breeding plumage was most extensive at mid-latitudes. This population structure was not maintained in the non-breeding season: morphometrics of captured birds and timing of migratory departures indicated that individuals from a wide range of breeding latitudes occur in each region and site in New Zealand. Links among morphology, phenology, and breeding location suggest the possibility of distinct Alaska breeding populations that mix freely in the non-breeding season, and also imply that the strongest selection for size occurs in the breeding season. Introduction Within the breeding range of many species, in dividuals exhibit geographic variation in morphology, appearance, or behaviour, reflecting either inherited or environmental differences (Zink and Remsen 1986). In migratory species, differential migration patterns within an apparently continuous geographic range (e.g., ?leapfrog? or ?chain? migration systems) may create stable population segreg ation (Lundberg and Alerstam 1986) and, potentially, breeding isolation, promoting popu lation structure and phenotypic diversification (Mayr 1963). Therefore, spatial distribution of in dividuals throughout the entire annual cycle may indicate the strength of population structure, and may also reveal where differential selection for phenotypic traits occurs. The Bar-tailed Godwit Limosa lapponica (hereafter, ?godwit? ) , a long-distance migratory shorebird, breeds in a discontinuous band of arctic and sub-arctic tundra from Scandinavia east to Alaska. There are four reco gnised subspecies (from west to east: L. l. lapponica, taymyrensis, menzbieri, and baueri ; Engelmoer and Roselaar 1998 ) and a small, isolated population in far-eastern Russia of un certain taxonomic status (purported L. l. anadyrensis ; Engelmoer and Roselaar 1998, Tomkovich 2010) . These populations have distinctive Chapter 3 47 migratory routes, timing of migration, an d morphology (Rynn 1982, Engelmoer and Roselaar 1998). Although the sexes are similar in non-breeding plumage, male godwits grow much more striking breeding plumage than females, resu lting in unusually dramatic sexual plumage dimorphism among scolopacid shorebirds (Fig ure 1 in Jukema and Piersma 2000). Size dimorphism in godwits (larger females) is also remarkably high among monogamous shorebirds that share incubation and parental care (McCaffery and Gill 2001). In addition, significant individual variation in both plumage and size occurs wi thin each sex. In particular, individuals undergo substantially different degrees of pre-supplemental contour feather moult (from very little to >90%), which results in conspicuous variation in breeding plumage (Piersma and Jukema 1993). The subspecies L. l. baueri breeds in western and northern Alaska (Figure 3.1) and migrates >10,000 km to non-breeding grounds in New Zealand and eastern Australia (McCaffery and Gill 2001). Field observations suggest geogra phic variation within Alaska: males with the greatest extent of breeding plumage were absent from southern breeding sites, but arrived later than local breeders and passed through these areas en route to northerly breeding areas (McCaffery et al. 2010). This agrees with data from Al aska museum specimens, in which males collected north of 64?N latitude had a greater extent of breeding plumage and were smaller than southern males (Rynn 1982). Distinct strategies in timing of moult and fueling among pre-migratory godwits in New Zealand also support the possibility of multiple breeding populations within L. l. baueri (Battley and Piersma 2005). Recently, godwits tracked on migration with light -sensitive geolocators shed further light on population structure: males depa rting New Zealand with a greater extent of breeding plumage arrived later in Alaska, and later-arriving birds of both sexes bred farther north (Chapter 2). In fact, breeding latitude was linked with timing of every stage of northbound migration, as well as with post-breeding departure from Alaska. It is thus plausible that breeding latitude may influence the distribution of individuals in the non-breeding season, but this hypothesis has yet to be tested. Here, we describe population structure within the breeding range of L. l. baueri and ask whether this structure persists in the non-br eeding season. We examined geographic variation in size and plumage of both sexes within A laska, using museum specimens in conjunction with capture, photography, and tracking of live bi rds. For comparison, we examined historical capture data within New Zealand to describe th e population structure by morphology across a similar range of latitude in the non-breeding season. 48 Chapter 3 Figure 3.1 Alaska breeding locations of Bar-tailed Godwits in this study. Dashed ellipses indicate three main regions used for geographic comparisons (YKD = Yukon-Kuskokwim Delta, SP = Seward Peninsula, and NS = North Slope). Solid ellipses indicate sites of godwit captures and field photography. CB = resights of godwits colour-banded in New Zealand (n = 3). PTT = godwits tracked from New Zealand using satellite telemetry (n = 8). GL = godwits tracked from New Zealand using geolocators (n = 16). MS = sites of museum specimens collected outside the three main regions. Unshaded area indicates known breeding range (McCaffery and Gill 2001). Methods Morphometric and plumage data Morphometrics For live captures, we report culmen (mm; exposed length), wing chord (mm; maximum flattened), and mass (g); no t all measurements were available for all captures. Despite numerous observers, we assume insignif icant systematic observer bias. Godwit mass undergoes drastic seasonal changes; for New Zealand captures, we report mass only for captures during October to mi d-December, when non-breeding mass is relatively stable (Wilson et al. 2007, P. Battley and J. Conklin unpubl. d ata). For Alaska (AK) captures, we Chapter 3 49 pooled masses taken during incubation and brood rearing, although data are lacking on breeding-season mass changes. For museum specimens, we measured length of exposed culmen (mm); all measurements were taken by JRC. On the basis of expected po st mortem shrinkage of 1.69% (Table 6 in Engelmoer and Roselaar 1998) , we corrected culmen lengths of museum specimens for direct comparison with live culmen measurements. Plumage Beginning in January, godwits moult from non- breeding (?basic?) to breeding plumage, in partially overlapping ?pre -alte rnate? and ?pre - supplemental? contour feather moults ( Jukema and Piersma 2000) , the latter of wh ich appears to affect only ventral regions. In general, males undergo much more extensive pre-supplemental moult than females, but there is substantial individual variation in both sexes. Ventral alte rnate plumage typically features lateral barring on a pale background, whereas supplemental f eathers are reddish and lack barring (Jukema and Piersma 2000). Therefore, we characterised ventral breeding plumage as the extent of red supplemental feathers visible against a pale back ground of basic and/or alternate feathers. We visually estimated proportion (in 5% increments) of red feathers in three ventral regions: ?vent? (posterior ventral plumage from leg to vent); ?breast? (anterior ventral plumage from leg to upper breast); and ?throat? (ventral plumage above breast to chin). Dorsal plumage appears to undergo only one pre- breeding moult: basic feathers are plain gray with a dark central stripe, whereas alternate feathe rs are blackish brown with pale or reddish spotting on the edges (McCaffery and Gill 2001) . We estimated ?dorsal? (mantle and scapulars, not including wing coverts) breeding plumage as the proportion (in 5% increments) represented by dark, spotted feathers. The extent of ventral alternate barring varies among individuals, and the amount of barring still evident during the breeding season depends on the extent of pre-su pplemental moult that is subsequently completed. We scored barring in the anterior ventral region as follows: 0 = no barring; 1 = barring on flanks only; 2 = barring on flanks and upper breast; 3 = barring on flanks, breast, and belly. To remove potential observer differences, all pl umage was scored from photographs by JRC. Depending on available photogra phs, not all plumage regions were scored for every individual. To gauge the comparability of different photographic sources (see below), we conducted a blind scoring trial using individuals photographed both free-living and in-hand during the same week ( n = 13); 92% of scores differed by ?10%, and there was no consistent 50 Chapter 3 directional bias in plumage scores. Therefore, we combined plumage scores from all data sources for analysis. Bill colour Bill colour of godwits varies seasonally: non-br eeding birds have predominantly pinkish bills that darken to black at the distal end, while bi lls of breeding birds are mostly black. From photographs of live godwits in Alaska, we scored bill colour as the proportion (in 5% increments) of both mandibles that looked bl ack. We excluded museum specimens because of potential post mortem changes in bill colour. Sources of data Museum specimens From three collections of godwit specimens , we examined breeding individuals ( n = 70; 40 male, 30 female) collected from 18 91 to 2001 in Alaska (60.4 ? 7 1.3?N). To exclude passage birds, we included only birds collected at kn own breeding areas from late May to late July or recorded as exhibiting breeding behaviour. We photographed specimens using standardised lighting and multiple angles, to enable scoring of plumage at a later date. Live captures We compiled morphometric data from adult go dwits captured during incubation or brood rearing at breeding sites in Alaska (61.8 ? 7 0.0?N; Figure 3.1) during May ? July of 2005 ?2 0 1 0 ( n = 57; 30 male, 27 female). We compiled morp hometric data from adult godwits captured at non-breeding sites in New Zealand (34.5 ? 4 6.6?S; Figure 3.2) during late September to early April of 1987 ? 2 0 1 0 ( n = 1,807 ; 932 male, 875 female). Godw its were aged based on plumage (McCaffery and Gill 2001) ; we excluded bird s of unknown age and those aged <3 years. Godwits were sexed by culmen length, plumage, or both. Females are generally larger than males (culmen > 99 mm = female; < 90 mm = male), but intermediate birds (culmen = 90 ? 99 mm) cannot be sexed by size alone. Plumage enables subjective sexing when supplemental plumage is present (January ? October): greater extent and richer red colour indicate male. However, we estimate that 1 ? 2 % of godwits in the New Zealand sample may be incorrectly sexed (P. Battley and J. Conklin unpubl. data). Some godwits captured in New Z ealand were tracked to Alaska breeding sites using satellite telemetry ( n = 8; Battley et al. 2012), geolocators ( n = 16; Chapter 2), or colour-band resightings ( n = 3). Consequently, morphometric data fro m these individuals occur in both Alaska and New Zealand data sets. Field photography To collect plumage data from free-living godwit s in Alaska, we visited known breeding areas Chapter 3 51 Figure 3.2 New Zealand capture sites of non-breeding Bar-tailed Godwits. Dashed ellipses indicate three main regions used for geographic comparisons. near Nome, on the Seward Peninsula (64.5 ? 6 5.2?N, 164.8 ? 1 6 6.7?W), and south of Deadhorse, on the north slope of the Brooks Range (69.7 ? 7 0.1?N, 148.7 ? 1 5 1.5?W; Figure 3.1) during June to early July 2009. We conducted walk ing surveys, digitally photographing all breeding individuals encountered. We used geographic positioning system (GPS) locations, times, and individual characteristics (e.g., bill length, uni que plumage traits) to avoid pseudoreplication of individuals. Biologists involved in prior field work (2003 ?2 0 0 9 ) provided photogra phs of free-living godwits from numerous Alaska breeding sites (58.8 ? 7 0.5?N), plus in-hand photos of eight godwits captured at three sites on the Yuko n-Kuskokwim Delta National Wildlife Refuge (61.1 ?6 1.4?N, 165.4 ?1 6 5.6?W; Figure 3.1). The final data set of live photographs included 123 Alaska godwits (72 male, 51 femal e; median = 11 photos/bird, range = 1 ? 1 3 5 ). Migratory departures from New Zealand At the Manawatu River estuary, New Zealand (40. 47?S, 175.22?E; Figure 3.2), we monitored departures in a small population of godwits (200 ? 2 8 0 individuals; ~25% were individually 52 Chapter 3 colour-banded). Using direct ob servation and digital photography, we recorded exact time and individual membership of departing flocks during three migration periods (4 March ? 5 April, 2008 ? 2 0 1 0 ). We conducted daily high-tide survey s to confirm remaining flock size and presence of marked godwits; daily resightin g probability of marked birds was >95 %. As a result, departures of marked birds were kn own to the day in 84% of cases, and for the remaining 16% we are confiden t of accuracy within ?1 day. Geolocator data (Chapter 2) confirmed that observed departures from the estuary matched departure from New Zealand. We determined departure dates for 76 marked go dwits (36 male, 40 female); for individuals monitored in multiple years, we averaged departure dates across available years. Analysis Although godwits breed in a nearly continuous band of coastal tundra in Alaska from near the Canadian border in the northeast to Bristol Ba y in the southwest (McCaffery and Gill 2001; Figure 3.1), for logistical reasons most field w ork (including all live captures and photographs in this study and most historical collection) ha s been conducted in three discrete regions (Figure 3.1): Yukon- Kuskokwim Delta (?YKD?; 59.7? 6 3.3?N, 161.8 ? 1 6 6.2?W), Seward Peninsula (?SP?; 64.4? 6 5.8?N, 162.3 ?166.7?W), and North Slope (?NS?; 69.6?71.3?N, 148.4 ? 1 6 0.1?W). For comparison, we divided godw it captures in New Zealand into three regions separated by >200 km (Figure 3.2): ?North? (34.5?37.2?S), ?Central? (40.4?41.3?S), and ?South? (43.5?4 6.6?S). Because of differences in size and plumage, we considered the sexes separately in our analyses. For each morphometric variable, we ex amined geographic variation using single- factor analysis of variance (ANOVA), and Tukey?s post -hoc test for between-region differences. For plumage variables and bill co lour, we used Kruskal-Wallis non-parametric ANOVA, and Tamhane?s post -hoc test. Museum specimens offered a more continuous representation of the breeding range than capture and field photography (Figure 3.1). In addition, tw o New Zealand-captured females were tracked to breeding areas outside the thr ee Alaska regions. For these reasons, Alaska totals and sample sizes for some tests exceed the sums for the three regions. We examined the association between breeding latitude and culmen length (pooled live and corrected museum culmen lengths) using linear regression, and compared male and female regression coefficients using Student?s t (Zar 1999). For male plumage va riables, we pooled the three regions with specimens collected elsewhere in Alaska ( n = 13), and examined associations with breeding latitude using linear regression. Chapter 3 53 Results Morphometrics Female godwits are much larger than males, on average (Tables 3.1 and 3.2), despite overlap in bill length, wing chord, and body mass ( t -tests, all measures for both AK and NZ: P < 0.0001). Body proportions also differed by sex: females had longer bills compared with wing chord (wing/culmen) than males ( t -tests, both AK and NZ: P < 0.0001). Alaska Within AK, we found geographic variation in size for both sexes. On average, birds were largest on YKD, smallest on NS, and interm ediate on SP (Table 3. 1). Body proportions also varied geographically, and for both sexes, wing /culmen length was greatest for NS, least for YKD, and intermediate for SP. Geographic variation was strongest in culmen length: the three regions were statistically distinguishable for both sexes. SP could not consistently be distinguished from YKD and NS, but the trend of d ecreasing size with increasing latitude was consistent across nearly all measures. Considering the full range of sizes present in AK, the smallest birds of each sex by culmen, wing, and mass were absent on the YKD, wherea s the largest birds were absent on the NS (Figure 3.3). The single exception was a conspicu ously long-winged male captured on the NS; however, his other measurements were consis tent with other NS males. The pattern was similar for relative wing/culmen length, as birds with extremely long wings in relation to bill did not appear on the YKD, and vice versa. Including samples outside the three main AK regions, culmen lengths demonstrated a continuous north?south cline in both males ( r2 = 0.390, F = 45.35, P < 0.001, n = 73) and females ( r2 = 0.474, F = 64.03, P < 0.001, n = 73; Figure 3.4). In add ition, the slope of the line describing the cline was lower in males (slope = ?0.949 ? SE 0.141) than in females (slope = ?1.523 ? 0.190; t = 11.88, df = 142, P < 0.001), which resulted in a progressive south-to-north reduction in sexual dimorphism of culmen length: females had 30% longer bills than males on YKD, 28% longer on SP, and 24% longer on NS. Dimorphism in wing and mass showed no latitudinal clines. New Zealand In contrast to AK, we detected minimal popu lation structure among regions in NZ, despite much larger samples (Table 3.2). Although southern birds of both sexes were slightly larger in mean culmen and wing, there was no consistent evidence for a north ?south cline in size. Mean differences among regions were much smaller than similar comparisons within AK, with large sample sizes conferring statistical significanc e to differences of much lower magnitude and 54 C ha pt er 3 Ta bl e 3. 1 G eo gr ap hi c va ri at io n in m or ph om et ri cs o f br ee di ng a du lt Ba r- ta ile d G od w its in A la sk a. D at a fr om li ve c ap tu re s on ly , i nc lu di ng N ew Z ea la nd g od w its t ra ck ed t o A la sk a br ee di ng si te s. Si gn ifi ca nt re su lts in am on g- re gi on A N O VA ar e in di ca te d in bo ld . A st er is k in di ca te s si gn ifi ca nt re su lt in be tw ee n- re gi on po st -h oc te st (P < 0 .0 5) . A bb re vi at io ns : Y K = Yu ko n- Ku sk ok w im D el ta , S P = Se w ar d Pe ni ns ul a, a nd N S = N or th S lo pe . A ll A la sk a YK D el ta Se w ar d Pe ni ns ul a N or th S lo pe A N O V A Tu ke y po st -h oc n m ea n ra ng e n m ea n SE n m ea n SE n m ea n SE F df P YK ?S P YK ?N S SP ?N S M al e Cu lm en (m m ) 38 84 .6 71 .3 ?9 4. 0 21 88 .0 0. 94 7 83 .6 0. 96 10 78 .4 1. 32 19 .6 5 2, 35 < 0. 00 1 * * * W in g (m m ) 37 23 2. 3 21 8? 24 3 20 23 5. 2 1. 30 7 23 1. 7 1. 21 10 22 7. 0 2. 29 6. 56 2, 34 0. 00 4 * W in g/ Cu lm en 37 2. 76 2. 45 ?3 .2 0 20 2. 68 0. 03 7 2. 77 0. 03 10 2. 91 0. 06 7. 45 2, 34 0. 00 2 * M as s (g ) 29 25 5. 7 20 5? 32 6 16 26 9. 9 5. 07 5 23 8. 8 9. 90 8 23 7. 9 5. 43 9. 55 2, 26 0. 00 1 * * Fe m al e Cu lm en (m m ) 46 10 8. 0 88 .5 ?1 25 .5 25 11 4. 2 1. 10 4 10 6. 8 1. 25 15 97 .5 1. 58 42 .8 4 2, 41 < 0. 00 1 * * * W in g (m m ) 44 24 4. 7 22 8? 26 1 25 24 9. 0 1. 14 4 24 1. 5 1. 89 14 23 7. 3 1. 69 19 .0 6 2, 40 < 0. 00 1 * W in g/ Cu lm en 44 2. 28 2. 02 ?2 .7 3 25 2. 18 0. 02 4 2. 26 0. 03 14 2. 47 0. 03 28 .7 3 2, 40 < 0. 00 1 * * M as s (g ) 28 31 9. 4 26 5? 38 4 12 33 9. 7 6. 63 4 29 4. 5 9. 98 12 30 7. 3 7. 96 7. 48 2, 40 0. 00 3 * * 54 Chapter 3 C ha pt er 3 5 5 Ta bl e 3. 2 G eo gr ap hi c va ri at io n in m or ph om et ri cs o f n on -b re ed in g ad ul t B ar -t ai le d G od w its in N ew Z ea la nd . S ig ni fic an t r es ul ts in a m on g- re gi on A N O VA a re in di ca te d in b ol d. A st er is k in di ca te s si gn ifi ca nt r es ul t i n be tw ee n- re gi on p os t- ho c te st (P < 0 .0 5) . A bb re vi at io ns : N = N or th , S = S ou th , a nd C = C en tr al . A ll N ew Z ea la nd N or th Ce nt ra l So ut h A N O V A Tu ke y po st -h oc n m ea n ra ng e n m ea n SE n m ea n SE n m ea n SE F df P N ? C N ?S C? S M al e Cu lm en (m m ) 92 0 83 .8 69 .0 ?9 8. 6 59 3 83 .5 0. 24 23 8 84 .1 0. 39 89 84 .8 0. 66 2. 28 2, 91 7 0. 10 W in g (m m ) 47 2 23 0. 5 21 0? 25 6 26 4 22 9. 8 0. 39 15 6 23 1. 8 0. 53 52 23 0. 2 0. 94 4. 73 2, 46 9 0. 00 9 * W in g/ Cu lm en 46 1 2. 73 2. 31 ?3 .3 2 25 3 2. 71 0. 01 15 6 2. 77 0. 01 52 2. 74 0. 03 5. 60 2, 45 8 0. 00 4 * M as s (g ) 35 3 27 7. 4 19 4? 38 4 19 7 27 9. 5 1. 55 12 1 27 6. 4 2. 11 35 26 9. 3 3. 96 3. 24 2, 35 0 0. 04 0 * Fe m al e Cu lm en (m m ) 86 2 10 8. 9 90 .0 ?1 29 .0 64 1 10 8. 5 0. 30 16 8 10 9. 8 0. 58 53 11 1. 1 1. 03 4. 30 2, 85 9 0. 01 4 * W in g (m m ) 54 9 24 3. 7 21 6? 26 4 38 8 24 3. 2 0. 37 11 7 24 4. 3 0. 65 44 24 7. 0 1. 08 5. 90 2, 54 6 0. 00 3 * W in g/ Cu lm en 53 8 2. 23 1. 85 ?2 .6 6 37 7 2. 23 0. 01 11 7 2. 24 0. 01 44 2. 21 0. 02 0. 57 2, 53 5 0. 57 M as s (g ) 34 3 33 3. 2 24 5? 40 0 23 4 33 3. 4 1. 40 81 33 0. 3 2. 41 28 33 9. 8 5. 58 1. 92 2, 34 0 0. 15 Chapter 3 55 56 C ha pt er 3 Fi gu re 3 .3 M or ph om et ri cs o f a du lt Ba r- ta ile d G od w its b y re gi on in A la sk a (Y KD = Y uk on -K us ko kw im D el ta , S P = Se w ar d Pe ni ns ul a, a nd N S = N or th S lo pe ) an d N ew Z ea la nd . ( a) M al e cu lm en ; ( b) m al e w in g ch or d; ( c) m al e w in g/ cu lm en ; ( d) m al e m as s; ( e) f em al e cu lm en ; ( f) f em al e w in g ch or d; (g ) f em al e w in g/ cu lm en ; a nd (h ) f em al e m as s. S ee T ab le s 3. 1 an d 3. 2 fo r sa m pl e si ze s. B ox es in di ca te m ed ia n an d 25 th a nd 7 5t h pe rc en til es . W hi sk er s in di ca te 1 0t h an d 90 th p er ce nt ile s, a nd d ot s in di ca te m or e ex tr em e va lu es . 56 Chapter 3 Chapter 3 57 Figure 3.4 Culmen length was negatively related to breeding latitude of Bar-tailed Godwits in Alaska. Includes live captures and corrected culmen lengths of museum specimens. Filled circles = males; open circles = females. presumably less biological significance. Each NZ region contained the full range of variation in culmen and wing found in AK (Figure 3.3). Alaska vs. New Zealand The grand means for culmen length and wing chord in AK and NZ (Tables 3.1 and 3.2) were similar for both males (culmen: t = 0.86, df = 956, P = 0.39; wing: t = 1.64, df = 507, P = 0.10) and females (culmen: t = 0.82, df = 906, P = 0.41; wing: t = 0.87, df = 591, P = 0.38). Hence, no morphological segment of the AK popu lation appeared to be missing from NZ. For most variables, NZ data contained extremes of distribution not found in AK, as expected given the much larger NZ samples. One excep tion was a northern AK female with a culmen length of 88.5 mm, smaller than the currently recognised minimum for females in NZ (90 mm). This suggests that overlap in male and fe male size, and consequently the number of NZ - captured godwits that are missexed, is gr eater than previously recognised. Migratory departures from New Zealand At the Manawatu River estuary, northbound mi gratory departures occurred from 4 March to 5 April (2008 ?2 0 1 0 ). Among colour-banded individuals of both sexes, larger birds departed earlier than smaller birds (mean departure date vs. culmen length; males: r = 0.562, F = 15.26, P < 0.001, n = 35; females: r = 0.651, F = 28.76, P < 0.001, n = 41; Figure 3.5). 58 Chapter 3 Figure 3.5 Culmen length was negatively correlated with migration departure date (day 1 = 6 March) of colour-banded Bar-tailed Godwits from the Manawatu River estuary, New Zealand (2008?2010). Filled circles = males; open circles = females. Plumage and bill colour in Alaska Plumage On average, male godwits in AK had a greater extent of breeding plumage than females in all body regions (Mann-Whitney tests, all measures: P < 0.0001; Table 3.3). For both sexes, an individual?s breast plumage score was positively correlated with vent (male: r = 0.794, n = 109; female: r = 0.810, n = 78), throat (male: r = 0.837, n = 108; female: r = 0.758, n = 75), and dorsal plumage (male: r = 0.615, n = 95; female: r = 0.587, n = 69; for all tests, P < 0.001). Patterns of geographic variation in plumage differed by sex. For males, extent of breeding plumage was greatest for NS and least for YKD (Table 3.3). For all plumage variables, SP males were more similar to NS than to YKD; with the exception of ve nt plumage, the two northern regions were statistically indistinguishable. Among the reddest males, the great majority were found nort h of 64?N (Figure 3.6). For example, 53% of males from SP and NS had br east scores >90%, compar ed with only 3% of YKD males. Conversely, only one male (1.5%) from SP ? N S had a breast score <70%, whereas 22% of YKD males were in that category. A similar pattern occurred in vent scores, although far fewer males attain extensive re d vent plumage; only one male (from SP) reached C ha pt er 3 5 9 Ta bl e 3. 3 G eo gr ap hi c va ri at io n in p lu m ag e an d bi ll co lo ur o f b re ed in g ad ul t B ar -t ai le d G od w its in A la sk a, in cl ud in g bo th li ve b ir ds a nd m us eu m s pe ci m en s. S ig ni fic an t r es ul ts in a m on g- re gi on K ru sk al -W al lis t es t ar e in di ca te d in b ol d. A st er is k in di ca te s si gn ifi ca nt r es ul t in b et w ee n- re gi on p os t- ho c te st ( P < 0. 05 ). A bb re vi at io ns : YK = Y uk on Ku sk ok w im D el ta , SP = S ew ar d Pe ni ns ul a, a nd N S = N or th S lo pe . D or sa l, ve nt , br ea st , an d th ro at v al ue s in di ca te p ro po rt io n of b re ed in g pl um ag e in e ac h bo dy r eg io n. Ba rr in g va lu es in di ca te a m ou nt o f b ar ri ng in a nt er io r ve nt ra l r eg io n. B ill v al ue s in di ca te p ro po rt io n of b la ck c ol ou r. A ll A la sk a YK D el ta Se w ar d Pe ni ns ul a N or th S lo pe Kr us ka l-W al lis Ta m ha ne p os t- ho c n m ea n ra ng e n m ea n SE n m ea n SE n m ea n SE ?2 df P YK ?S P YK ?N S SP ?N S M al e D or sa l ( % ) 95 92 .3 65 ?1 00 25 88 .2 1. 78 31 93 .7 0. 93 26 94 .8 0. 76 11 .4 4 2 0. 00 3 * * V en t (% ) 10 9 64 .5 5? 10 0 30 49 .2 4. 17 33 68 .2 3. 05 33 77 .0 1. 79 28 .4 9 2 < 0. 00 1 * * * Br ea st (% ) 11 1 85 .7 35 ?1 00 32 74 .8 2. 57 33 89 .9 1. 76 33 92 .4 1. 36 34 .4 0 2 < 0. 00 1 * * Th ro at (% ) 10 8 92 .8 60 ?1 00 31 88 .5 1. 56 33 95 .2 0. 99 31 95 .8 1. 07 17 .8 1 2 < 0. 00 1 * * Ba rr in g (0 ?3 ) 99 1. 00 0? 3 26 1. 19 0. 12 32 0. 69 0. 12 28 0. 93 0. 15 7. 16 2 0. 02 8 * Bi ll (% ) 60 93 .1 75 ?1 00 19 92 .6 1. 85 22 93 .6 1. 62 19 92 .9 1. 23 0. 89 2 0. 64 Fe m al e D or sa l ( % ) 70 78 .8 20 ?9 5 14 72 .5 3. 51 28 86 .4 0. 96 17 72 .7 6. 13 10 .9 7 2 0. 00 4 * V en t (% ) 78 19 .5 0? 50 17 14 .7 2. 37 29 26 .9 2. 08 21 12 .9 2. 82 18 .3 7 2 < 0. 00 1 * * Br ea st (% ) 80 35 .4 0? 80 17 24 .1 2. 85 30 47 .5 2. 96 22 26 .6 4. 47 21 .1 3 2 < 0. 00 1 * * Th ro at (% ) 76 55 .4 0? 90 17 48 .5 4. 94 28 62 .3 2. 99 20 50 .5 4. 64 6. 03 2 0. 04 9 Ba rr in g (0 ?3 ) 74 1. 74 1? 3 16 1. 31 0. 15 28 1. 68 0. 14 19 2. 11 0. 15 11 .3 8 2 0. 00 3 * Bi ll (% ) 46 83 .2 40 ?1 00 16 80 .0 2. 81 19 85 .8 3. 32 10 87 .7 2. 61 4. 09 2 0. 13 Chapter 3 59 60 C ha pt er 3 Fi gu re 3 .6 P lu m ag e of b re ed in g ad ul t Ba r- ta ile d G od w its b y re gi on in A la sk a (Y KD = Y uk on -K us ko kw im D el ta , S P = Se w ar d Pe ni ns ul a, a nd N S = N or th S lo pe ). Va lu es in di ca te e xt en t (% ) o f b re ed in g pl um ag e in e ac h bo dy r eg io n: (a ) m al e do rs al ; ( b) m al e ve nt ; ( c) m al e br ea st ; ( d) m al e th ro at ; (e ) fe m al e do rs al ; (f ) fe m al e ve nt ; (g ) fe m al e br ea st ; an d (h ) fe m al e th ro at . S ee T ab le 3 .3 f or s am pl e si ze s. B ox es in di ca te m ed ia n an d 25 th a nd 7 5t h pe rc en til es . W hi sk er s in di ca te 1 0t h an d 90 th p er ce nt ile s, a nd d ot s in di ca te m or e ex tr em e va lu es . 60 Chapter 3 Chapter 3 61 100%, and 50% of YKD males scored ?50%. Patterns in throat and dorsal plumage were less dramatic because those plumage scores showed relatively little variation; all males reached ?60% in both throat and dorsal scores. Variation in male plumage was consistent with a north?south cline: all measures of breeding plumage demonstrated significant linear increases with latitude (vent: r2 = 0.258, n = 109; breast: r2 = 0.246, n = 111; throat: r2 = 0.116, n = 108; dorsal: r2 = 0.128, n = 95; for all tests, P < 0.001). By contrast, female plumage did not conform to a north?south cline. For all breeding-plumage variables, female scores were greatest for SP, whereas YKD and NS were statistically indistinguishable (Table 3.3). Differences in br east plumage were most conspicuous; 43% of SP females scored >50%, compared with 9% for NS and 0% for YKD. We also found no SP females with dorsal scores <80%, whereas 35% of NS and 57% of YKD females fell in this category. On average, females had more heavily barred unde rparts than males in each AK region (Table 3.3). For females, ventral barring increased with latitude. For males, ba rring was greatest for YKD and least for SP, although neither wa s statistically distinguishable from NS. Bill colour In AK, males had darker bills than females, on average (Table 3.3), and all birds with bills <75% black were female. Bill colour did not vary significantly by geographic region for either sex. Blackness of bill was positively c orrelated with breast score for males ( r = 0.321, P = 0.013, n = 59), but not for females ( r = 0.135, P = 0.37, n = 46). Discussion Our study confirms and clarifies geographic variation within the breeding range of L. l. baueri , indications of which date back more than a century (McCaffery et al. 2010). In the only previous quantitative analysis, Rynn (1982) arbitrarily divided Alaska into two regions (at 64?N) and found regional di fferences in size and plumag e among museum specimens of both sexes. By treating the Seward Peninsula and North Slope separately and examining linear relationships with latitude, we have demonstrated that variation in size of both sexes and in breeding plumage of males are consistent with north?south clines. However, we found no evidence that the Alaska populatio n maintains its structure in the non-breeding season, despite occupying a similar range of latitude in New Zealand. 62 Chapter 3 Where does selection for body size occur? No geographic variation in size was eviden t among New Zealand godwits. By contrast, directional selection appears to occur in the breeding season in Alaska; godwits were smaller at higher latitudes. In >70% of bird species, co lder climates are associated with larger body size, but this pattern appears to be least appl icable to migratory species, whose annual routines are an adaptation to avoid environmental extr emes (Meiri and Dayan 2003). However, the hypothesis that migration distance limits body size is not supported by our data. The additional 1,000 ? 1 , 2 0 0 km traveled to northern Alaska represents <8% of the total migration distance from New Zealand, and because northern breeders stop in southwest Alaska on both northbound and southbound migration (Chapter 2), th ey do not actually perform longer non- stop flights than southerly breeders. On the breeding grounds, male godwits perfor m spectacular, aerobatic displays in their efforts to secure mates and territories (McCaffery an d Gill 2001). These aerial displays may select for smaller males, whose greater maneuverabilit y allows them to outperform larger males (Jehl and Murray 1986, Sz?kely et al. 2000). In turn, if mate comp etition (and, thus, selection for these displays) is stronger at high er latitudes, it could foster the observed size cline in males. However, there is no evidence for geographic variation in mate competition in godwits, and this scenario fails to explain th e equivalent size cline in females. Because bill morphology is related to foraging method in probing shorebirds (Barbosa and Moreno 1999) , both intra- and inter-specific vari ation in bill length is often attributed to partitioning of prey resources (Nebel et al. 2005). Non-breeding godwits forage primarily on mudflats, where their long bills are suited to probing for subsurface prey. By contrast, they spend the breeding season primarily on tundra, often far from mudflats, and forage primarily near or above the surface (McCaffery and G ill 2001). We therefore expect stronger selection for bill length in the non-breeding season. Ho wever, although culmen length did not vary geographically in New Zealand, there was a cline within Alaska beyond that found in wing chord and mass; northern birds were not just shorter-billed but were proportionally shorter- billed for their size. In addition, sexual dimo rphism in culmen length varied geographically, with male and female bills most similar in the north. These findings suggest selection against long bills at high latitudes, consistent with Allen?s Rule (for a given body volume, surface area w ill be minimised in colder climates; Allen 1877). Bird bills can be a significant source of heat loss (Symonds and Tattersall 2010) , and the very long bill of godwits may be a thermoregulation liability during the breeding season, particular at the highest latitudes. Alternatively, habitat differences may also contribute to geographic variation in bill length. Chapter 3 63 Why does breeding plumage vary geographically? Broad geographic patterns within Alaska weak en the hypothesis that variation in godwit breeding plumage is primarily driven by relative individual quality and its honest signaling to rivals and mates (Piersma and Jukema 1993, Piersma et al. 2001, Drent et al. 2003). The distinct geographic patterns in male and female plumage indicate nonparallel selection acting upon the sexes, which is consistent with the as sumption that male plumage plays a greater role in pair formation. However, godwit plumage has yet to be linked to basic fitness components such as reproductive success or quality of territories or mates. One clear function of godwit plumage is nest cr ypsis, because both sexes incubate eggs in open ground nests (McCaffery an d Gill 2001) , relying on the di sruptive pattern of mantle and scapular feathers to blend with the surrounding tundra. Accordingly, dorsal scores of the sexes differed by much less than ventral scores. Furt hermore, dorsal plumage was the least variable plumage region within each sex, indicating si milar selection across the breeding range. However, dorsal and ventral plumage covaried in both sexes, which suggests that geographic variation in dorsal plumage is not driven exclu sively by adaptation to local habitats. The patchy red and white ventral plumage of females and southern males, roughly matching tundra backgrounds (J. Conklin pers. obs.), ma y provide crypsis for non-incubating godwits. However, the striking full-red breasts of no rthern males are very conspicuous, drawing attention to themselves and often away from their more cryptic mates. This suggests trade-offs between crypsis and mate acquisition that va ry geographically, which could occur if competition for mates were more intense in the north, or if the brevity of the northern breeding season increased the importance of rapid mate acquisition. Bill colour may be a component of the breeding ?plumage? of godwits, and thus subject to sexual selection itself, as in some passerines (e.g., Jawor et al. 2003). Alternatively, the seasonal increase in bill blacknes s may be an adaptation to mediate heat loss at high latitudes (Symonds and Tattersall 2010) , given that black pigmentation confers greater absorption of heat. However, we found no geographic var iation in bill colour, despite a correlation with breast plumage in males. Because barred feathers are a component of alternate plumage rather than the subsequent supplemental plumage, interpretation of ventral ba rring in Alaska is complicated, particularly for males. However, it is intriguing that ventr al barring of females was consistent with a north?south cline, whereas their vent, breast, an d throat scores were highest at mid-latitudes. This suggests that alternate plumage is, or was, subject to different selection than supplemental plumage. If alternate plumage represents the ancestral ?breeding? plumage, subsequently replaced (in evolutionary terms ) by the supplemental plumage (Jukema and 64 Chapter 3 Piersma 2000) , the conflicting patterns may reflect s election at different points in evolutionary history. The temporal overlap of pre-alternate and pre-supplemental moults (Piersma and Jukema 1993) warrants further investigation, bu t geographic variation suggests that the extent of barring apparent on the non-breeding gr ounds prior to most pre-supplemental moult (i.e., January ? February) may roughly indicate a godwit?s breeding region. Population structure in the breeding season Geographic variation in size and plumage among Alaskan godwits is similar in magnitude to differences among recognised godwit subspecies that occupy separate migratory flyways (Rynn 1982, Engelmoer and Roselaar 1998). Within L. l. baueri , links between breeding latitude and migration timing (Chapter 2) indicate that morphology is linked with phenology and behaviour as well. In New Zealand, godwit s are extraordinarily site-faithful (P. Battley and J. Conklin unpubl. data) and have high ly repeatable individual migration schedules (Battley 2006). If such behavioural rigidity ex tends to natal philopatry and breeding-site fidelity, segments of the Alaskan population could be reproductively isolated despite having completely overlapping non-breeding ranges. Ho wever, latitudinal clines in size and plumage, with substantial overlap among regions, sugges t no distinct geographic limits to breeding populations. Genetic analyses may elucidate the age and degree of any division within L. l. baueri in relation to recognised godwit subspecies. Geographic variation within Alaska has direct r elevance to the diagnosis of godwits breeding in the Anadyr region of Russia as a separate subspecies ( L. l. anadyrensis ; Engelmoer and Roselaar 1998) . Tomkovich (2010) found L. l. anadyrensis specimens to be intermediate between menzbieri and baueri specimens in both size and plumage, and concluded that anadyrensis was a valid subspecies. However, his Alaska sample ( n = 5) was entirely from the Yukon-Kuskokwim Delta, where baueri godwits are largest. Considering all of Alaska, the measurements of Anadyr specimens fall largely within the range of values we have presented. Geographic variation in plumage of baueri suggests that the separation of anadyrensis on the basis of plumage also warrants further investigation. Therefore, it remains plausible that Anadyr godwits represent a geographically isolated segment of baueri. Is latitudinal variation within Alaska consisten t with patterns in other godwit populations? Among the four recognised subspecies, exten t of male breeding plumage is greater in northerly breeding races ( L. l. taymyrensis and menzbieri ; 63 ? 7 5 ? N ) than in southerly races ( lapponica and baueri ; 58 ? 7 0 ? N ) , and the southernmost male baueri are the palest in the species (Rynn 1982). In body size, the most northerly race ( taymyrensis ) is the smallest and the most southerly ( baueri ) is the largest, but latitudinal tren ds are obscured by a longitudinal pattern, in which western races are smaller than eastern races (Rynn 1982, Engelmoer and Chapter 3 65 Roselaar 1998). Thus, variation within Alaska may reflect more general processes, but because latitude is only an index for a suite of environmental factors (e.g., temperature, habitat type, duration of breeding season), identifying sources of selection will require detailed analysis. A comparison of L. l. baueri and taymyrensis may be instructive, for they breed across similar spans of latitude (~12?) and may co ntain comparable variation (but see Drent et al. 2003). In the ecologically similar Red Knot Calidris canutus , northerly populations also had redder plumage than southerly populations (Buehler and Piersma 2008). In addition, there was a negative relationship between extent of breedin g plumage and migration distance, implying energetic and temporal trade-offs between moult and migration. This latter relationship is not apparent in godwits, because the shortest distance migrant ( L. l. lapponica ) is among the paler races, and redder males in Alaska migrate farther. Reports of intra-population va riation such as we have described are rare among arctic- breeding shorebirds. Engelmoer and Roselaar (1998) identified latitudinal variation for only 1 of 14 shorebird species, the Grey Plover Pluvialis squatarola. Among Alaska-breeding shorebirds, we are aware of only one other example: northern-breeding Dunlin Calidris alpina arcticola are smaller than southern breeders ( C. a. pacifica ) , but these populatio ns follow very different migration patterns (Warnock and Gill 1996). Population structure in the non-breeding season The lack of population structu re in New Zealand shows that L. l. baueri lacks the differential migration patterns (e.g., leap-frog migration) often found to accompany structure in breeding populations (e.g., Swa rth 1920, Kelly et al. 2002). Because a significan t portion of the Alaska population winters in eastern Australia (McCaf fery and Gill 2001), some structure may yet occur across the entire non-breeding range. However, the morphological diversity in each New Zealand region suggests that godw its from across Alaska mix freely at non- breeding sites. This is consistent with geolocator data showing that individuals from the Manawatu River estuary used breeding sites spanning most of the known Alaska breeding range ( 59.7 ?70.2?N; Chapter 2). Links between migratory timing an d breeding location in that study are further supported by morphological data presented here: early-departing (presumably southerly- breeding) godwits were larger (F igure 3.5), in accordance with the size cline that we found within Alaska. This pattern appears general to New Zealand sites, because larger males also departed earlier at the Firth of Thames (although this was reported in error as the opposite relationship; Battley 2006), and the 4-week span of departures implies individuals from a wide range of breeding latitudes. 66 Chapter 3 We did not examine geographic variation in plum age in New Zealand, because the correlation between plumage at departure and ?ultimate? breeding plumage is unclear, as a result of the resumed pre-supplemental mo ult during a stopover of 4 ?7 weeks 1 in Asia (Chapter 5). Also, plumage at departure has been studied at only two New Zealand sites (Battley 2006, this study). However, male plumage was highly var iable at both sites (range of breast scores: 20 ? 1 0 0 % ) , which is consistent with individuals from a wide range of breeding latitudes occurring at each site. The factors that govern non-breeding distri bution of godwits remain mysterious. Although many aspects of godwit life history, such as br eeding site and migration timing, appear to be ?hard - wired? and presumably heritable, non -breeding site does not. Af ter their first migration from Alaska, young godwits (<2 years) appear to freely roam New Zealand and eastern Australia before settling on specific sites, to which th ey are extraordinarily faithful as adults (P. Battley and J. Conklin unpubl. data). Iden tifying the social and ecological factors that govern this site ?choice? may reveal patterns in an apparently random non -breeding distribution of individuals. 1 This was reported in error as ?3?5 weeks? in the published version. 67 Chapter 4 Impacts of wind on individual migration schedules of New Zealand Bar-tailed Godwits Conklin, J.R. & P.F. Battley Behavioral Ecology 22: 854?861 (2011) 68 Chapter 4 Abstract Despite clear links between wind conditions and timing of migration at the population level, no study has examined the contribution of winds to annual variation in the migration timing of individual birds. At a single non-breeding site in New Zealand, we closely monitored three years of departures of Bar-tailed Godwits Limosa lapponica baueri , a long-distance migrant with remarkable annual consistency in individual migration schedules. Although individual godwits showed very little variation in departure date and generally experienced favourable departure conditions, most off-schedule departures were explained by maximising initial wind assistance for the non-stop flight to Asia. Surprisingly, early departures attributable to wind were more common and of greater magnitude than wind-related departure delays, and prolonged weather-related departure lulls did not always result in late-departing individuals. Thus, our results show that knowledge of individual departure decisions with regard to wind can strongly influence interpretation of population patterns. Early departures associated with winds, previously only demonstrated theoretically, may reflect conservative timing and extent of pre-migratory fuelling, a possible adaptation for extreme long-distance migration in variable conditions. Introduction Favourable winds confer signif icant time- and energy-minimis ation benefits for migrating birds (Alerstam and Lindstr?m 1990, Liechti and Bruderer 1998) , promoting timely arriving at breeding sites in appropriate physiological condition (Drent et al. 2003). Accordingly, birds respond to wind conditions with flexibility in departure timing, flight altitude, and migratory route (Liechti 2006). Specifically, headwinds or crosswinds may discourage migratory departure, but, in theory, trade-offs between optimal timing and flight costs may lead to departure in unfavourable winds if such conditions persist for extended periods (Weber et al. 1998a). Numerous studies have demonstrated the in fluence of wind on intensity and timing of migration at the population level, but such data do not directly address departure decisions of individual birds. No multi-year study has ex amined how individuals reconcile migration schedules with annual variation in weather. Optimisation models typically view departure date as flexible, varying with individual differences in body condition and fuelling rate, and with temporal variation in fuelling resources and weather (Hedenstr?m 2008). This view is most appropriate for species with multiple stopovers en route to breeding grounds (Weber et al. 1998b, Clark and Butler 1999) , a strategy allowing potential compensation for ?errors? in initial departure timing or Chapter 4 69 unanticipated conditions encountered after departure (Shamoun-Baranes et al. 2010). By contrast, departure of migrants making extreme long-distance flights (now known to reach >11,000 km; Gill et al. 2009) may require a more rigid concept of ?optimal? departure date. These birds typically cross vast inhospitable barriers offering no refuelling sites (Gill et al. 2005) and little opportunity to correct errors in timing or fuel load. In addition, they initiate these journeys effectively blind to conditions at the destination, as environmental cues are unlikely to correlate at such great distances (but see J?rvinen 1989). Therefore, departure date in long-jump migrants (Piersma 1987 ) is likely to be conservative and ?hard-wired ? by adaptation to long-term and large-scale enviro nmental trends, showing less annual variation than in short-hop migrants. Such rigid migration schedules are exempl ified by New Zealand Bar-tailed Godwits Limosa lapponica baueri. Their round-trip migration to Alaska breeding grounds covers approximately 30,000 km in three trans-oceanic f lights, including the longest non-stop flight recorded in birds (Gill et al. 2009, Battley et al. 2012). Accordingly, godwits in New Zealand and Alaska accumulate among the highest pre-mi gratory fuel loads yet reported (Piersma and Gill 1998, Battley and Piersma 2005). Despite thes e apparent hardships, godwits demonstrate extraordinary inter- annual rigidity in migration timing: individuals? departure dates from New Zealand generally varied by less than one week, am id inter-individual variation of nearly a month (Battley 2006). This latter variation is st rongly linked with phenology of Alaska breeding sites across a wide range of latitude, with southern breeders departing New Zealand earliest (Chapter 2). These findings impl y that migration timing in godwits reflects endogenous, individually-optimised schedules, r ather than putative annual or individual differences in fuelling rate. Weather may therefore be the primary source of intra-individual variation in migration timing in this system, offering the opportunity to ex amine the effects of wind with few confounding factors. In this study, we monitored godwit s departing from a single site in New Zealand for three consecutive years, to examine the associatio n between wind conditions and repeatability of individual departure date, and how indivi dual departure decisions were reflected in migration patterns at the population level. Methods Migratory departures We monitored northbound migratory departures (4 March ? 5 April, 2008 ? 2 0 1 0 ) of Bar-tailed Godwits from the Manawatu River estuary, New Zealand (40.47?S, 175.22?E). At this small 70 Chapter 4 estuary (~1 x 2 km), the sm all godwit population (200 ?2 8 0 birds; ~25% ar e individually colour-banded) is highly approachable, and can usually be observed entirely from one of several vantage points. We monitored expected peak departure hours (13:00 ? 2 1 : 0 0 local time) on 72 of 99 days (d); migratory depa rtures of shorebirds generally occur 1 ? 6 hours (h) before civil twilight (Piersma et al. 1990b, Battley 1997). During surv eys, a single observer watched and listened for migratory behaviour, which included distinct vocalisations and low, circling flights expressing intent to depart (Piersma et al. 1990b, Battley 1997). Typically, flocks engaged in very active calling, preening, and short exploratory flights, for 0.5 ? 4 h before actual departure. During this time, the observe r recorded all marked individuals involved, using a spotting scope and digital camera. All flying flocks were watched and/or photographed until resettling or disappearing from sight. Departures were easily distinguished from local movements by both altitude and direction (always NW/NNW and slowly disappearing to the horizon). After a departure, the observer quickly surveyed the estuary for all remaining marked godwits. In addition, we conducted daily high-tide surveys to confirm size and composition of the remaining flock; d aily resighting probability of marked godwits was >95%. The final dataset included 45 indi viduals (25 female, 20 male) wi th known departure dates in all three years; we excluded birds with <3 y ears of data from analyses. For 102 (75.5%) cases, individuals were directly observed preparing and/ or actually departing with an observed flock. The remaining individuals were assigned a departure date based on the last day they were recorded at the estuary. For 16 (11.9%) cases, this coincided with an observed departing flock of partially or completely unknown individual co mposition; therefore, we considered time and flock size to be known. For 17 (12.6%) cases, this coincided with a decrease in local flock size unexplained by observed departures; we considered these departures ?unobserved? and calculated flock size based on successive high -tide counts. Observed departures occurred 13:30 ? 2 0 : 5 5 , and unobserved departures likely o ccurred outside survey times (Leyrer et al. 2009). By including unobserved departures, annual totals represent a virtually complete accounting of individuals migrating. We are confident that movements out of the es tuary represented migratory departure from New Zealand. Site-fidelity of marked godwits was extremely high for the entire non-breeding season (September ? March), and non-migratory movements in and out of the estuary were extremely rare (J. Conklin and P. Battley unpubl . data). Furthermore, some marked godwits carried geolocators (Chapter 2), and subsequently provided independent support for departures in 2008 ? 2 0 0 9. In all cases, longitude data from geolocators indicated birds were clearly west of New Zealand on the day subsequent to observed ( n = 23) or assigned ( n = 6) departure date, and all continued directly to As ia (J. Conklin and P. Battley unpubl. data). Chapter 4 71 We calculated repeatability (intra-class correlatio n coefficient) of individual departure date according to Lessells and Boag (1987) . We used the median of an individual?s three departures to represent its preferred departure date, and considered departures within ?3 d of this median date to be ?on schedule?. Weather data During surveys, we estimated local surface wi nd direction (in 16 categories; e.g., NNW) and velocity (in 5-km/h increments) hourly and whenever departing flocks were observed. Concurrently, we recorded precipitatio n in three subjective categories: ?heavy? = steady or heavy rain; ?light? = drizzle or intermittent light rain; or ?none?. For days with observed departures, we used winds at the exact time of depa rture to represent conditions. For days with multiple departures ( n = 7 d), we used the mean wind direction and velocity among all departures that day; in all cases, winds were similar among same-day departures. For days without departures and for two days in which departures were apparently missed despite a survey, we used winds at 18:00 local time to represent conditions (median observed departure time = 17:38). For days without surveys, we used local predicted winds and precipitation (MetVUW 2010) , when available ( n = 7), to represent surface conditions. To represent wind conditions at likely migration altitude (Landys et al. 2000, Green 2004) , we obtained data from the NCEP/NCAR Reanalysi s Project (NOAA 2010a) for 40?S, 175?E (50 km NW of the study site) at 850 mb geopo tential height (~1,500 m altitude) at 18:00 local time. From these data, we calculated wind direction and velocity. Calculation of wind effect Because wind effect calculation method ma y affect conclusions, we compared two commonly-used formulae 1 . The first (Tailwind, hereafter ?TW?; see formula in ?kesson and Hedenstr?m 2000) derives relative wind profit ( ?WP ? , in which positive values indicate assisting winds) considering on ly the magnitude of tailwind or headwind along the preferred migratory direction. The second (Crosswind, hereafter ?CW?; see formula in Piersma and Jukema 1990) additionally accou nts for drift from the intended flight path caused by crosswinds; thus, this method always produces an equal or less favo urable assessment of winds. We assumed a migratory direction of NNW (337.5?), consistent with observed migratory departures from the site (315 ? 3 5 0 ? ) and tracks of satellite-tagged godwits after New Zealand departure (Battley et al. 2012). CW requires an assumption of preferred air speed; we used 65 km/h, consistent with northbound tracks of satellite-tagged godwits (Battley et al. 2012). For days when both local surface and 850 mb wind data were available ( n = 87 d), we calculated maximum wind profit ( ? M WP ? ) as the greater of the two WP values. 1 See Appendix 3 for formula details. 72 Chapter 4 We considered MWP values ?5 to +5 km/h to represent calm conditions; lesser or greater values indicated headwinds or tailwinds, respectively. Results Migratory departures ? population level We observed 39 departing flocks on 29 days (1 ?4 flocks/d) across three years. Surveys indicated unobserved departures on seven additional days (Figure 4.1). Observed departing flocks contained 3 ? 3 4 godwits each (mean = 14.1, n = 39 flocks); 3 ? 7 0 godwits departed on each departure day (mean = 18.1, n = 36 d). In 2008, a total of 251 godwits departed on 13 days (6 ?4 8 birds/d) from 10 March ? 2 April (Figure 4.1a). In 2009, 212 godwits departed on 13 days (3 ?3 8 birds/d) from 4 March ? 3 April (Figure 4.1b). In 2010, 189 godwits departed on 10 days (5 ? 7 0 birds/d) from 6 March ? 5 April (Figure 4.1c). Wind profit calculated by Tailwind and Crosswind were nearly identical for local surface winds ( r2 = 0.99, n = 87 d, F = 9966.4, P < 0.001), indicating that significant low-altitude crosswinds were rare. At 850 mb, the two measures were less similar ( r2 = 0.78, n = 99 d, F = 344.7, P < 0.001), indicating more frequent cr osswinds at higher altitudes. However, crosswinds did not appear to deter departures, as four flocks departed when MWP calculated by CW (including crosswinds) was ?7.6 to ?3 2.4 km/h. When crosswinds were not considered (TW), no flocks departed when MWP was +3 MWP and no heavy rain) on 19 days (6 in 2008, 7 in 2009, and 6 in 2010; Figure 4.2), suggesting that no individuals were waiting to depart on these days. On 12 of 19 days (63 %), no remaining marked go dwits were past MDD; 7 of these days were before 10 March, when ve ry few godwits (and only 2 marked birds) ever departed. On four days in 2008, 1 ? 2 individuals had reached MDD +1 ? 6 d, but all had been captured in the previous five days. On three other days, 1 ?2 birds had reached MDD +1 ? 2 d, but waited one more day to depart, still on schedule. Three departing flocks (6, 9, and 13 godwits) contained no marked birds. MDD helps explain why all remaining marked birds skipped two of th ese departures: on 31 March 2008 and 18 March 2009, none had passed MDD yet (Figure 4.4a ?b). In the third case (22 March 2008), one late-running bird had passed MDD (having skipped three de partures within his 7-day departure window), and departed the following day (MDD +6 d). Chapter 4 79 Discussion Variation in individual departure date Several striking results derive from our study, wh ich is the first to examine the effect of winds on intra-individual variation in migration timin g. First, individual departure date was extremely consistent across years, leaving very little variation to explain. Second, unfavourable departure conditions occurred infrequently, so godwits apparently had little difficulty reconciling preferred departure date with weather. Third, most off-schedule departures were attributable to either recent capture or the avoidance of unfavourable winds. Most surprisingly, cases of indi viduals departing earlier than ex pected due to wind conditions were more numerous than wind-related departure delays. Repeatability of individual departure date ( r = 0.836) was nearly identical to that found among godwits at another New Zealand site ( r = 0.83; Battley 2006). To our knowledge, these are by far the highest figures for repeatable migration timing in birds (see Bety et al. 2004 and references in Battley 2006) ; the nearest comparable finding was breeding arrival in Barn Swallows Hirundo rustica ( r = 0.51; M?ller 2001). In both go dwit studies, most individuals departed within a one-week period each year, but the rate was higher in the current study (87% vs. 73% of departures). This difference is probably methodological, as a smaller population and greater intensity of effort at the Manawatu River estuary led to greater precision in departure dates. Regardless, we co nfirm that the bird making the longest known non-stop flights may also have the most rigid migration schedules. This rigidity may be enabled by the generally favourable conditions experienced by godwits: only 14 of 87 days (16%) featured headwinds an d/or heavy rain. A visible setting sun has been proposed as a vital navigational cue for departing shorebirds, explaining their tendency to depart in late afternoon (Moore 1987, Piersma et al. 1990b). However, New Zealand godwits have been shown to depart without a visible sun (Battley 1997), and this was true during the current study, although we did no t specifically measure cloud cover. Therefore, weather rarely presented an obstacle to departure. Furthermore, godwits often departed in calm conditions and did not always choose the most favourable day close to MDD, suggesting that significant wind assistance, at least at the time of departure, was not a requirement for migration. However, most off-schedule departure s (10 of 18) appeared directly attributable to wind: these birds achieved greater wind assistance by departing 1 ? 7 days outside their expected windows. Weather-related departure delays are easily unders tood in terms of optimal migration: a bird reaches migratory condition, and then should depa rt on the first occasion of favourable winds 80 Chapter 4 after its optimal departure date (Weber et al. 1998a). This principle may explain four late departures in our study, in which birds experien ced poor winds at MDD, but departed as soon as conditions improved. Wind conditions causing early departure is less intuitive, and we believe has never before been demonstrated. Weber and Hedenstr?m (2000) showed theoretically that birds may benefit from early departure in favourable conditions, provided that sustained unfavourable winds can be exp ected to follow. Thus, a bird below its optimal fuel load for migration may be ?promoted? to readiness by favourable winds, because wind assistance effectively decreases the fuel required for the flight (Alerstam and Lindstr?m 1990). On 17 March 2010, the greatest single departure day of the study, four flocks comprising 37% of the migratory population departed. Joining many on-schedule birds, five godwits departed 4 ? 1 0 days prior to MDD, avoiding delay by a subsequent 8 -day departure lull and experiencing better conditions than any during their expected departure windows. This is consistent with two different scenarios, both in triguing. In the first, birds anticipated sustained poor conditions, and avoided those by departin g on the last favourable day before the change in weather. The second scenario requires no fore sight, but only recognition of extremely good current conditions: winds that day were among the most favourable in our study (MWP +40.4 km/h), and unusual conditions no rth of New Zealand suggested significant wind assistance for perhaps 2 ? 3 days along the migratory route. The benef its of an unusually easy flight may have outweighed any costs of early departure or early arrival in the Y ellow Sea. Godwits may respond to weather at larger temporal and spatial scales than those addressed here, and we don?t know how well local conditions reflect those encountered later in the 7?8 day flight to Asia. Migration schedules may also be disrupted by trau ma or stress. Four godwits captured near MDD in 2008 appeared to delay departure by 3 ? 9 days, only to resume their normal schedules in following years. Delays may have resulted from capture-related stre ss or mass loss (Warnock et al. 1997), which has been linked with ex tended stopover duration (Warnock and Bishop 1998). However, any capture effect was br ief and did not affect individuals equally: other godwits captured near MDD and all of thos e captured more than eight days prior to MDD departed on schedule. Only six off-schedule departures (33%) were unexpl ained by wind or recent capture, implying that little variation in departure can be attributed to annual variation in fuelling rate or body condition, or to carry-over effects from previo usly disrupted moult schedules or delayed arrival on non-breeding grounds. In addition, we used an admittedly coarse measure of an individual?s preferred departure date (median departure among three years). Although most Chapter 4 81 birds had two or three temporally clumped departure dates during our study, others left more uncertainty regarding their ?intended? departure dates. With additional years of data, and the inclusion of weather variables at greater spatial and temporal scales, even these few unexplained departures may become scrutable. Understanding population patterns with individual data The tendency of individual godwits to depart in a 7 -day period each year (Battley 2006, this study) allows inferences regarding the proporti on of a population available to migrate at a given time (Figure 4.4). For example, after a pr olonged departure lull, it is reasonable to expect that subsequently departing flocks will largely comprise late individuals (e.g., Battley 1997). Indeed, there is a clear case of this in 2009: after a 4-day lull during 25 ? 28 March, approximately 20% of the populati on departed over a 3 -day pe riod (Figure 4.1b). Figure 4.4b reveals a conspicuous accumulation of individuals past MDD during this time, and 3 of 4 wind-related late departures in our study occurred 30 ?3 1 March (Figure 4.3). By contrast, individual data provided an unexp ected view of the departure lulls in 2010. After an 8-day lull amid sustained prohibitive cond itions, there was an intense migration peak 26 ? 2 9 March once conditions improved (Figure 4.1c). In this case, however, the prolonged lull caused no accumulation of late individuals: by 26 March, on ly three godwits had passed MDD (Figure 4.4c), and no late departures resu lted (Figure 4.3). In fact, although 2010 featured two significant departure lulls and the most extreme weather conditions in the study, not a single late departure occurred that year. These insights are impossible from population data alone. Individual data also allow conclusions regarding apparently favourable departure days that were unused by the population. Using our expected departure wi ndows, it was clear that these days were unused primarily due to a lack of r eady and/or late individuals in the population. Likewise, some departures contained no marke d birds, due to a lack of individuals approaching MDD. Shorebirds typically migrate in flocks, which ma y confer navigational, energetic and/or safety benefits to individuals (Alerstam 1990). Depa rting flock sizes in our study were quite small compared to departing Bar-tailed Godwit flocks observed elsewhere (mean flock sizes 40 ? 2 3 5 , and up to 700; Piersma et al. 1990b, Tulp et al. 1994, Battley 1997) , but these studies involved larger populations (1,500 ?8 , 0 0 0 birds). In small popu lations, individuals may encounter few others willing to migrate on a given day, and the smallest departing flocks we observed (3 ?6 birds) may attest to the prioritisation of individual departure date over flock size. But there are apparently limits. In 2009, one female (bird #3 0 in Figure 4.3) attempted to depart during the departure lull of 25 ? 2 8 March (Figure 4.1b). For four days, she 82 Chapter 4 conspicuously called and flew about, but could not rally others for a successful departure. On 29 March, her vocalisations were more vehement , and she attempted to depart alone, circling high above the estuary and several hundred metres NW before turning back in a significant surface headwind. Finally, two godwits joined her in preparation for departure, and this tiny flock departed late on 29 March. The next day, in much improved wind conditions (Figure 4.1b), 19 godwits departed, including several individuals at MDD +3 ? 4 d (Figure 4.3). For those three earlier birds, departure date app eared more important than flock size or wind assistance. The costs of such apparently suboptimal departures are unknown, but departing flock size did not appear to influence likelih ood of returning the following autumn in this study. Interpretation of wind data Migrating birds may experience very different winds at low and high altitudes (e.g., Piersma and Jukema 1990, Piersma et al. 1990a, Schaub et al. 2004), and have been shown to adjust flight altitude to encounter favourable winds (Gauthreaux 1991, Bruderer et al. 1995). Studies evaluating departure conditions using winds at a single altitude have typically indicated some proportion of flocks departing in subo ptimal winds (e.g., Battley 1997, ?kesson and Hedenstr?m 2000) , suggesting that not all bird s can achieve favourable departure conditions. Consistent with this, birds in our study often departed when either surface winds or 850 mb appeared quite unfavourable. However, when bo th altitudes were considered, it was clear that godwits only avoided departing when both were unfavourable (Figure 4.2). Viewed this way, winds were rarely an important obstacle, and birds were not obligated to depart in unfavourable conditions. Flocks often made seve ral aborted flights before ultimate departure, and occasionally circled slowly upward before leaving the site, possibly testing different altitudes for favourable winds. On a 7 ? 8 day flight to Asia, it is likely that godwit flocks change altitude frequently to achieve the course of least resistance. Conclusions regarding the influence of wind on departure decisions can depend upon the method of calculating wind profit. When we included the disp lacement effect of crosswinds, several flocks appeared to depart in very poor conditions. However, this was not the case when we ignored crosswinds and calculated only the tailwind component. This may suggest the energetic cost of compensation for crosswinds is negligible for godwits departing this site, or that they simply allow themselves to drift without compensation, expecting that winds encountered later may compensate for drift from the shortest-distance path. With variable winds, theoretically optimal behaviour of migrants with a distant destination is to allow lateral drift without compensation initially, but to increase compensation upon approaching the destination (Alerstam 1979, Green et al. 2004). This is consistent with tracks of satellite- tagged godwits, which deviated laterally from the sh ortest-distance route by as much as 500 ? Chapter 4 83 6 0 0 km in the first days after departure from New Zealand, only converging upon approaching the Yellow Sea (Battley et al. 2012 ). Deviations of this magnitude add very little distance or time to a flight of approximately 10,000 km. Therefore, we believe crosswinds do not significantly influence departure decisions of New Zealand godwits. Implications for understanding long-distance migration Results of the current study support our view of departure timing in Bar-tailed Godwits as a fundamental, ostensibly fixed window, primar ily governed by overall migration speed and appropriate timing of arrival on distant breeding grounds (Chapter 2). At the population level, mean departure date varied by only two days, suggesting a negligible effect of annual variation in non-breeding resour ces, at least during the period we studied. Individual godwits demonstrated astoundingly repeatable departure when presented with consistently favourable weather, but easily tolerated ?3 days of variation around a preferred date, and responded to more extreme conditions with greater deviations in departure. This implies strong selection for individually-optimised departure date, but also flexibility to respond to unpredictable circumstances. In this study, no bird failed to r eturn the next year after off-schedule departures up to MDD ?9 ? 1 0 days; however, we do not know wh at magnitude of error in departure will reduce survival or reproductive success. Northbound godwits ha ve an extended stopover (30 ? 5 0 d) in Asia, perhaps allowing compensatio n for suboptimal timing of New Zealand departure before the flight to Alaska, which appears much more rigorously scheduled with regard to breeding-site phenology (Chapter 2). The early departure of individuals in response to wind has intriguing implications for optimal migration. Achieving optimal fuel loads well before departure is considered disadvantageous, because an associated decrease in maneuverability may increase mortality risk from predation (Alerstam and Lindstr?m 1990, Dietz et al. 2007). It is unclear wheth er wind benefits could compensate for 10 days of lost fuelling time, as required by the most extreme cases we observed. However, New Zealand godwits face minimal danger from avian predators, currently or historically, and so the cost of ca rrying full fuel loads prior to departure may be negligible. In addition, we ha ve no evidence that departure of godwits is limited by fuel deposition rates in New Zealand; they fuel quite slowly, starting up to three months before departure (J. Conklin and P. Battley unpubl. data). Thus, godwits may regularly reach migratory condition well in advance of departure and may be prone to overloading (Gudmundsson et al. 1991), allowing flex ibility to address unusual weather patterns. The extraordinary migratory flights and high -latitude breeding areas of Bar-tailed Godwits imply extreme time and energy constraints. Howe ver, the maintenance of rigid schedules and high survival rate (>90% individual return rate in this study) do not suggest a bird near the 84 Chapter 4 limit of its capabilities. In evolutionary terms, su ch an unforgiving annual routine may foster a conservative approach to fuelling and migration timing, as opposed to the strategies of short- hop migrants, which may precariously balance crucial trade-offs in time-, energy-, and predation-minimisation (Alerstam and Lindstr?m 1990) with lesser consequences. Thus, safe and predictable fuelling conditions and generally favourable wind patterns may be prerequisites for the evolution and persis tence of such extreme migration systems. 85 Chapter 5 Contour feather moult of Bar-tailed Godwits in New Zealand and the Northern Hemisphere reveals multiple strategies by sex and breeding region Conklin, J.R. & P.F. Battley Emu 111: 330?340 (2011) 86 Chapter 5 Abstract The extreme long-distance migration of Alaskan breeding Bar-tailed Godwits Limosa lapponica baueri may present severe constraints on annual moult, and high individual variation in plumage and migration timing suggests that multiple strategies by sex and breeding region may exist. We used digital photography of free-living Bar-tailed Godwits to describe the timing and extent of pre-basic and pre-breeding contour feather moults in New Zealand, and used plumage of breeding birds in Alaska to infer the proportion of moults occurring in Alaska and Asia. These data demonstrated that: (1) godwits conducted overlapping pre-alternate and pre-supplemental moults; (2) pre-basic and pre-breeding moults were scheduled differently in relation to southbound and northbound migration respectively; (3) northern and southern Alaskan breeding godwits of each sex were distinguishable by plumage differences throughout the non-breeding season; and (4) males and northern breeders achieved more extensive breeding plumage by spending longer in pre- breeding moult in New Zealand, rather than through faster moult rates or greater investment in moult during migratory stopover in Asia. The existence of a ventral pre-supplemental moult implies that contemporary selection for red breeding plumage overrides older selection for barred alternate plumage. Our use of individual-based data revealed a continuum of annual moult strategies within the population, which may reflect individual differences in any combination of sex, size, migration distance, or breeding location. Even within the highly constrained annual cycle of extreme long-distance migrants, differential selection influences how individuals manage trade-offs among non-breeding activities such as moult, fuelling, and migration. Introduction The scheduling of moult is finely tuned by the balancing of direct and indirect costs of feather replacement, the availability of resources, and competing energetic requirements of other annual activities (Payne 1972, Murphy and Ki ng 1991, 1993). Migratory birds face the particular challenge of conducting moult amidst the demands of pre-migratory fuelling, time- critical reproductive efforts on seasonal breeding grounds, and migration itself (Alerstam and Lindstr?m 1990) , any of which may be to some extent incompatible with moult. Thus, moult schedules in relation to migration may indicate the relative fitness consequences of feather quality and the strength of selection in different segments of the annual cycle (Holmgren and Hedenstr?m 1995, Barta et al. 2008). Chapter 5 87 Bar-tailed Godwits Limosa lapponica achieve extraordinarily high pre-migratory fuel loads (Piersma and Gill 1998, Battley and Piersma 2005 ) , undertake the most extreme non-stop migratory flights yet recorded (9,000 ?1 2 , 0 0 0 km; Gill et al. 2009, Battley et al. 2012) , and breed during brief high-latitude summers. Con sequently, godwits may face severe time and energy constraints. In addition, their dram atic seasonal changes in plumage (Figure 1 in Jukema and Piersma 2000) present potentially significant conflicts between investment in moult and other mandatory annual tasks. Bar-tailed Godwits moult from non-breeding to breeding plumage in what appear to be overlapping pre-alternate and pre-supplemental co ntour feather moults (Piersma and Jukema 1993, Jukema and Piersma 2000). European Bar-tailed Godwits L. l. taymyrensis appeared to have three generations of ventral plumage: at the population level, plain (basic) feathers were replaced by barred (alternate) feathers, the extent of which then decreased, particularly in males, as the extent of red (s upplemental) feathers increased (Piersma and Jukema 1993). Data from individuals are required to confirm th at specific feathers are actually replaced twice in the pre-breeding moult(s). However, it ha s been proposed that barred feathers constitute the ancestral breeding plumage, whereas red feathers reflect more current selection processes (Jukema and Piersma 2000). If true, this presents a conundrum: if moult is costly, why would godwits retain a redundant and seemingly wasteful moult? Recently, geographical variation in the Alaskan breeding population of Bar-tailed Godwits L. l. baueri has been demonstrated: northern breeders of both sexes were smaller, had more extensive breeding plumage, and migrated later on both northbound and southbound migrations (Chapters 2 ? 3, McCaffery et al. 2010). It is not yet clear whether the observed variation constitutes discrete evolutionary units or a latitudinal cline within Alaska. However, coupled with dramatic sexual dimorphism in si ze and breeding plumage (Chapter 3), this population structure suggests multiple strategies for scheduling of moult in relation to migratory flights. For example, how do males manage a more extensive moult in the non- breeding season, when both sexes must accomplish full flight feather replacement and prepare for equivalent migratory flights? How do nort hern breeders reconcile a more extensive moult with a longer migration (~1,2 00 km longer each way) between New Zealand and Alaska? If high-quality breeding plumage increases reproductiv e success, birds should schedule moult as late as possible before breeding (Holmgren and He denstr?m 1995). This predicts a significant portion of moult may occur during the godwits? 4?7 week stopover in Asia before arrival in Alaska (Chapter 2), consistent with obse rvations of continued moult on migration in L. l. taymyrensis (Piersma and Jukema 1993). In ad dition, we expect that moults (both pre- basic and pre-breeding) should be temporally shifted according to breeding phenology, as is 88 Chapter 5 migration itself (Chapter 2). However, schedulin g of moult in the Alaskan breeding godwit population has never been described at the individual level (but see McCaffery and Gill 2001). In this study, we use detailed observations of colour-banded Bar-tailed Godwits to describe plumage and moults throughout the non-breeding season in New Zealand, and compare these with plumage of breeding godwits in Alaska (Chapter 3) to es timate the duration and proportion of moults occurring outside New Zealand. We use the resulting data to: (1) present the first individual-based eviden ce for three feather generations in godwits, and describe the temporal overlap of pre-alternate and pre-suppl emental moults; (2) assess whether the timing of moults matches intra-population differen ces in timing of migration; (3) demonstrate the extent to which geographical patterns in Alaska n breeding plumage correspond to differences upon migratory arrival and departure in New Zealand, and in basic plumage; and (4) determine whether differences in the ex tent of breeding plumage by sex and breeding region are achieved through strategic differences in timing, rate, or duration of moult. We then discuss the implications of multiple moult strategies within the population for differential selection for plumage and potential trade-offs with other non-breeding activities. Methods Fieldwork in New Zealand We studied plumage and migration timing in a small population of Bar-tailed Godwits (200 ? 2 8 0 individuals, ~25% of which were colour-banded) at the Manawatu River estuary, New Zealand (40.47 ?S, 175.22?E). During two migratory arrival periods (1 September ? 2 0 October 2008 ? 2 0 0 9 ) we conducted high-tide surveys every 3 ? 4 days (d) to record initial arrival and plumage of marked individu als. During three migratory departure periods (4 March ? 5 April 2008 ? 2 0 1 0 ) we conducted daily surveys to record departure date (details in Chapter 4) and plumage of marked birds. In the in tervening summer months (1 January ?3 March 2008, 20 October 2008 ? 3 March 2009, 20 October 2009 ? 3 March 2010) we conducted surveys every 4 ? 8 d. During surveys, we digitally photog raphed marked godwits to enable detailed scoring of plumage later. This resulted in ~18,00 0 identifiable phot ographs of 78 birds 1 . Ageing and sexing Bar-tailed Godwits <2 ? 3 years old were excluded from the study by considering only migratory individuals (young birds do not migrate). Although plumage at departure from New Zealand may be age-dependent (Battley 2006), we included individuals on their first northbound migration (3 males, 1 female). 1 See Figure 6.2 for more details. The number of days per season each individual was photographed was incorrectly reported here in the published version. Chapter 5 89 Godwits were sexed by culmen length and plumag e. Females are larger and have longer bills than males (culmen length >99 mm for fema les and <90 mm for males), but intermediate birds cannot be sexed by size alone. Howeve r, obvious plumage differences at departure (Table 5.1) allowed unambiguous sexing of all individuals of intermediate size (5 males, 3 females). Moults and plumage scoring Bar-tailed Godwits have three types of ventral feathers (Piersma and Jukema 1993, Jukema and Piersma 2000) : ?basic? (plain whitish), ?alternate? (dark barring on a pale background), and ?supplemental? (plain pale to rusty red). They have two types of dorsal feathers: non - breeding feathers are plain grey with a dark central stripe, whereas breeding feathers are blackish brown with pale or reddish spottin g on the edges (McCaffery and Gill 2001). Following the suggestion by Jukema and Piersma (2000) that barred ventral feathers represent a distinct moult, we include red ventral and spotted dorsal feathers in all references to ?breeding plumage? (BP) or ?pre - breeding moult?, and consider the barred, alternate plumage separately. The non-flight feather component of pre-basic moult begins July ?August in Alaska, after breeding, and continues on the non-breeding grounds (McCaffery and Gill 2001). The flight feather component of pre-basic moult begins af ter southbound migration, and is not addressed in this study. Hereafter, all references to ?pre - basic moult? consider only non -flight feathers. Table 5.1 Evidence for Bar-tailed Godwit contour feather moult outside New Zealand. Breeding data represent godwits in Alaska (Chapter 3). Non-breeding data represent marked godwits at the Manawatu River estuary, New Zealand (arrival = 2008?2009; departure = 2008?2010). Barring scores indicate amount of barring in anterior ventral region. BP scores indicate extent (%) of breeding plumage. NZ departure Breeding NZ arrival mean range n mean range n mean range n Male Barring (0?3) 1.30 0?3 37 1.00 0?3 99 Dorsal BP (%) 71.8 30?95 37 92.3 65?100 95 24.3 0?55 30 Ventral BP (%) 56.9 23?96 37 81.0 33?100 105 19.6 0?43 30 Female Barring (0?3) 1.59 0?3 41 1.74 1?3 74 Dorsal BP (%) 20.9 0?80 41 78.8 20?95 70 18.6 0?50 32 Ventral BP (%) 2.5 0?33 41 37.1 0?68 75 5.3 0?27 32 90 Chapter 5 Using the digital photographs (all scoring done by JRC), we characterised plumage as follows. Barring on the anterior ventral region (belly, flanks, and breast): 0 = no barring visible; 1 = barring on flanks only; 2 = barring on flan ks and upper breast; 3 = barring on flanks, breast, and belly. Ventral supplemental plum age: the proportion (5% increments) of red feathers visible against a pale background of ba sic or alternate feathers on the vent (posterior ventral plumage from leg to vent), breast (anterior ventral, as above) and throat (ventral plumage above breast to chin) regions. We averaged these to create a summary ?ventral? BP score. ?Dorsal? (mantle and scapulars, not including wing coverts) BP: the proportion represented by dark, spotted feathers (5% increments). Alaskan data For comparison with plumage in New Zealand, we summarised population-wide data from Alaskan breeding grounds (60 ? 7 1 ? N ; details in Chapter 3). These data include breeding godwits captured or photographed in the field during May ? July 2003 ?2 0 0 9 or collected from breeding sites from 1891 ? 2 0 0 1 and preserved as museum specimens. Marked birds at the New Zealand site contained similar variation in size (culmen length: males 71 ? 95 mm, females 90 ? 1 2 5 mm) evident in the entire popu lation (Chapter 3), and used breeding sites encompassing most of the known Alaskan breeding range (Chapter 2). Therefore, we assumed that New Zealand and A laskan godwits in this study represent comparable samples. Analysis When a marked bird was first observed after arri val, we assigned the midpoint of the period after the previous survey as its arrival date. Consequently, we are co nfident of individual arrival dates within 2 ?6 d, and recorded initial BP scores ?10 d after arrival for nearly all birds; other individuals were omitted from an alyses of arrival BP. We are confident of individual departure dates within ?1 d (see Cha pter 4) and photographed every individual within 1 ?5 d of departure. We defined the completion of pre-basic moult as the latest calendar day on which any worn, unmoulted contour feathers were visible; as the pre-breeding moult is partial, these were not necessarily BP. We defined the initiation of pr e-breeding moult as the first day on which dorsal or ventral BP was visi ble. For birds departing New Zealand with no BP, we used departure date + 8 d (the earliest possible day after arrival in Asia) to represent initiation date of pre-breeding moult. We considered pre- breeding moult suspended when an individual?s plumage was first equal to its score at depa rture. We excluded individuals with >10 d of uncertainty in completion, in itiation, or suspension of mo ult from relevant analyses. Chapter 5 91 Breeding latitude of New Zealand Bar-tailed Godw its has been strongly linked with migration timing (Chapter 2) and body size (Chapter 3). We assigned th e likely Alaskan breeding region (?north? or ?south? of 64?N) to marked godwits at the Manawatu River estuary (n = 78) by the following process. Sixteen godwits were tracked to breeding sites using geolocators (Chapter 2); among these, all birds departing New Zealand by 24 March bred south of 64?N, and five of six that departed after 24 March bred north of 64?N. We therefore divided the remaining 62 godwits into ?early? and ?late? by average departure date (cutoff: 24 March). Also, northern breeders of both sexes are smaller than southern breeders (Chapter 3), so we divided the birds into ?small? and ?large? classes using culmen length (cutoffs: males 82 mm, females 108 mm). For 46 birds (74%), the two criteria suggested th e same breeding region (small and late departure or large and early departure). When the criteria disagreed, we used the criterion more divergent from the cutoff to assign breeding region (e.g., very large + slightly late = ?south?). This method may misclassify a small number of individuals. To test whether data on the phenology of moult and migration were comparable across years, we first ran analyses of variance (ANOVA) for each parameter (sexes separate), using Alaskan region and year as fixed factors (results not shown); there were no significant region ? year interactions. Therefore, we averaged v alues for each individual across available years (1 ? 3 years, depending upon bird and parameter). Results Contour feather moult in New Zealand Upon arrival at the Manawatu River estuary (September ?early November), all marked godwits had both worn and new contour feathers (e.g., Figure 5.1a). Most individuals (95 ? 9 8 % ) arrived with visible dorsal or ventral BP, th e extent of which then decreased after moult resumed within 1 ?2 weeks of arrival. The last trace of BP was seen 23 September ? 22 December (sexes similar). Wo rn dorsal feathers (basic type, or unknown type owing to extreme wear) were often visible 10 ?35 d after an individual?s last BP disappeared. All birds completed pre-basic moult by late December (Figure 5.2). Basic plumages (Figure 5.1b) of males and females were similar. All birds were strongly barred on the vent and completely lacked barring on the throat, central breast, and belly. All males and most females had some barring on the flanks (score 1); 29% of females had none (score 0). 92 Chapter 5 Figure 5.1 Plumage of one northern breeding male Bar-tailed Godwit through the non-breeding season in New Zealand. (a) 6 October, showing incomplete pre-basic moult; (b) 5 January, in basic plumage; (c) 29 January, showing increases in ventral barring, dorsal spotted feathers, and red ventral feathers; and (d) 1 March, showing further increases in dorsal and ventral breeding plumage, and a decrease in barring. First increases in ventral barring occurred on all males between 5 January and 9 February (Figure 5.3a). Most females showed increased barring between 9 January and 4 March; 18% showed no increase before leaving New Zealand. For males, the first increases in ventral barring , dorsal spotting, and red ventral feathers were approximately simultaneous (generally within 1 ? 2 weeks; see Figure 5.1c); first dorsal BP appeared 12 January ?2 8 February (Figure 5.3b) and ve ntral BP appeared 7 January ?1 7 February (Figure 5.3c). After initial increase s, 74% of male barring scores subsequently dropped as ventral BP scores increased (e.g., Figure 5.1c ?d). All males appeared to suspend moult before migration: breast plum age scores stopped increasing 5 ? 4 0 d (mean = 18.5 d) before departure. (a) (b) (d) (c) Chapter 5 93 Figure 5.2 Timing of pre-basic contour feather moult of male and female Bar-tailed Godwits in New Zealand, 2008?2009; Day 1 = 1 September, Day 120 = 29 December. Values represent mean proportion of individuals with visible, unmoulted feathers. For migratory arrivals, median and span of dates are indicated. By contrast, female BP appeared later and not universally. A total of 79% of females gained dorsal BP, which first appeared 30 January ? 2 3 March (Figure 5.3b); 41% gained ventral BP, which appeared 23 January ? 2 3 March (Figure 5.3c). No fema le barring scores dropped after a pre-departure peak. Among moulting females, suspension was not uniformly evident: some appeared to add BP as late as 1 ? 5 d before departure. Evidence for moult beyond New Zealand Dorsal and ventral BP scores of both sexes were higher in Alaska than at arrival or departure in New Zealand (Table 5.1; Mann ? Whitney tests, all P < 0.001), indicating that portions of both pre-basic and pre-breeding moult occurred outside of New Zealand. Males had lost 75% of total Alaskan BP before southbound migratio n, and females had lost 81% of BP. Upon departure from New Zealand, total BP scores of males and females were 74% and 17%, respectively, of Alaskan scores. Barring scores of males were lower in Alaska than at departure from New Zealand ( U9 9 ,3 7 = 2271.5, Z = 2.34, P = 0.02), but female scores did not differ ( U7 4 , 41 = 1379.0, Z = ?0.87, P = 0.39). 94 Chapter 5 Figure 5.3 Timing of pre-alternate and pre-supplemental contour feather moults of male and female Bar-tailed Godwits in New Zealand, 2008?2010; Day 127 = 5 January, Day 211 = 30 March. Values represent mean proportion of individuals with: (a) alternate ventral barring; (b) dorsal spotted feathers; and (c) ventral red feathers. For migratory departures, median and span of dates are indicated. Chapter 5 95 Plumage differences by breeding region Ventral barring scores of northern and sout hern breeding males were indistinguishable throughout the non-breeding season in New Zealand (Figure 5.4a; all tests P > 0.20). For both groups, the peak barring score in New Zealan d was greater than the score at departure (Wilcoxon paired tests, north: W1 9 = 0.0, Z = ?3.35, P = 0.001; south: W1 6 = 0.0, Z = ?3.30, P = 0.001). In Alaska, northern breeding males had less barring than southern breeding males ( U6 8 , 31 = 1344.5, Z = 2.38, P = 0.017). Northern breeding females had more barring th an southern breeding females at every stage in New Zealand and Alaska (Figure 5.4b; Basic: U 1 1 ,23 = 71.5, Z = ?2.56, P = 0.010; NZ peak and departure: U1 5 , 2 6 = 56.0, Z = ?4.00, P < 0.001; Breed: U5 4 ,20 = 324.0, Z = ?2.85, P = 0.004). All godwits with barring scores of 0 in basic plumage were southern breeding females. Figure 5.4 Change in ventral barring (score 0?3; data are mean score ? 1 SD) of (a) male and (b) female Bar-tailed Godwits from northern and southern Alaskan breeding regions (north or south of 64?N). On x-axis, Basic indicates full basic plumage in mid-December, Peak NZ indicates highest score achieved before departure from New Zealand, Depart NZ indicates plumage at departure, and Breed indicates full breeding plumage in Alaska. Cohorts are offset horizontally for clarity. Asterisks indicate significant north?south differences. 96 Chapter 5 Northern breeders of both sexes always had mo re extensive BP than southern breeders (Figure 5.5). On departure from New Zealand, northe rn and southern breeding males differed in ventral BP ( U 2 1 ,16 = 87.0, Z = ?2.48, P = 0.012) but not dorsal BP ( U2 1 , 16 = 113.5, Z = ?1.67, P = 0.10). In Alaska, male cohorts differed in both dorsal BP ( U6 5 , 30 = 566.5, Z = ?3.42, P = 0.001) and ventral BP ( U7 1 ,34 = 388.5, Z = ?5.61, P < 0.001). On arrival in New Zealand, males differed in dorsal BP ( U1 6 , 14 = 64.5, Z = ?1.99, P = 0.047) but no t ventral BP ( U1 6 , 1 4 = 85.0, Z = ?1.13, P = 0.28). Northern and southern breeding females were always distinguishable by both dorsal BP (NZ departure: U1 5 , 26 = 93.0, Z = ?2.77, P = 0.006; breeding: U5 2 ,18 = 306.5, Z = ?2.21, P = 0.027; NZ arrival: U1 0 , 22 = 47.0, Z = ?2.57, P = 0.010) and ventral BP (NZ departure: U1 5 , 26 = 90.0, Z = ?3.17, P = 0.002; breeding: U54 ,21 = 319.5, Z = ?2.92, P = 0.003; NZ arrival: U1 0 , 22 = 18.0, Z = ?3.78, P < 0.001). Figure 5.5 Change in breeding plumage (BP): (a) male dorsal; (b) male ventral; (c) female dorsal; and (d) female ventral breeding plumage of Bar-tailed Godwits from northern and southern Alaskan breeding regions (north or south of 64?N). Data are mean % ? 1 SD. See Figure 5.4 for explanation of x-axis categories. Cohorts are offset horizontally for clarity. Asterisks indicate significant north?south differences. Chapter 5 97 Moult schedules by breeding region Northern breeding godwits of both sexes arrive d in New Zealand later than southern breeders (Table 5.2). On average, northern breeding females and all males completed pre-basic moult 40 ? 4 5 d after arrival, compared with 3 3.5 d for southern breeding females. Both male cohorts initiated pre-breeding moult ca. 21 January and suspended moult ~18 d before departure; thus, northern breeding males spent longer in moult owing to their later departure (Table 5.2). Northern breeding m ales moulted 68.6% of their contour feathers (dorsal and ventral BP combined; Figure 5.5a ?b) in 45.9 d, for a mean moult rate of 1.49% per day. Southern breeding males moulted 58. 9% of their contour feathers in 36.1 d, a rate of 1.63% per day. These mou lt rates did not differ ( t 1 8 ,16 = 1.52, df = 32, P = 0.07); the overall mean rate for males was 1.57 ? SD 0.42% per day. Southern breeding females initiated pre-breeding mo ult ca. 8 March, just 9 d before departure (Table 5.2). Northern breeding females began mo ult ca. 2 March, 25 d before departure. For females, the brief time in moult and uncertainty regarding suspension of moult precluded the calculation of useful moult rates. Assuming the moult rate observed in males (1.6% per day), northern breeding females spent on average 12 d in moult to r each departure BP, and southern breeding females spent just 5 d (Table 5.2). Projected pre-breeding moult in Asia During staging in Asia, northern breeding ma les must increase BP from 69% (New Zealand departure) to 90% (breeding; Figure 5.5). At the observed New Zealand moult rate (1.6% per day), this would require ~13 d. Southern breed ing males similarly require 13 d to increase from 59 to 79%. Assuming the ma le moult rate, northern breedi ng females require 26 d (from 19 to 61%), compared with 27 d for southern breeding females (from 8 to 51%). Thus, total investment in pre-breeding moult is 59 d for northern breeding males, 49 d for southern breeding males, 38 d for northern breeding fema les, and 32 d for southern breeding females (Figure 5.6). 98 C ha pt er 5 Ta bl e 5. 2 M ou lt an d m ig ra tio n sc he du le s of g od w its d ur in g no n- br ee di ng s ea so n in N ew Z ea la nd . A rr iv al /p re -b as ic m ou lt = 20 08 ?2 00 9; p re -b re ed in g m ou lt/ de pa rt ur e = 20 08 ?2 01 0. F or d at es , d ay 1 = 1 S ep ; d ay 2 11 = 3 0 M ar . N or th = p re su m ed A la sk a br ee di ng r eg io n >6 4? N ; s ou th = < 64 ?N . A st er is ks in di ca te s ig ni fic an t n or th ?s ou th d iff er en ce s. So ut h N or th un it m ea n SD n m ea n SD n t df P M al e N Z ar ri va l da te 22 .6 4. 2 15 29 .9 4. 4 17 4. 80 30 < 0. 00 1 * En d pr e- ba si c m ou lt da te 67 .4 14 .6 14 70 .8 13 .8 16 0. 64 26 0. 53 N Z pr e- ba si c m ou lt da ys 44 .8 15 .2 14 41 .6 13 .1 14 0. 59 26 0. 56 St ar t pr e- br ee di ng m ou lt da te 14 1. 8 7. 9 16 14 4. 1 7. 3 18 0. 88 32 0. 39 Su sp en d pr e- br ee di ng m ou lt da te 17 8. 1 7. 9 16 19 0. 4 8. 0 19 4. 56 33 < 0. 00 1 * N Z de pa rt ur e da te 19 7. 1 4. 1 16 20 8. 3 3. 1 19 9. 07 33 < 0. 00 1 * N Z pr e- br ee di ng m ou lt da ys 36 .1 8. 1 16 45 .9 7. 9 18 3. 56 32 0. 00 1 * Fe m al e N Z ar ri va l da te 21 .0 11 .2 22 33 .4 12 .2 10 2. 81 30 0. 00 9 * En d pr e- ba si c m ou lt da te 52 .6 9. 5 20 73 .4 16 .3 9 4. 35 27 < 0. 00 1 * N Z pr e- ba si c m ou lt da ys 33 .5 8. 2 20 40 .1 6. 5 9 2. 12 27 0. 04 3 * St ar t pr e- br ee di ng m ou lt da te 18 8. 5 15 .8 26 18 2. 1 12 .5 12 1. 24 36 0. 22 N Z de pa rt ur e da te 19 7. 8 4. 4 26 20 7. 9 2. 9 15 8. 03 39 < 0. 00 1 * N Z pr e- br ee di ng m ou lt A da ys 4. 7 5. 9 26 11 .9 9. 2 15 3. 07 39 0. 00 4 * A C al cu la te d us in g m al e m ou lt ra te 1 .6 % /d 98 Chapter 5 Chapter 5 99 Figure 5.6 Estimated average duration of Bar-tailed Godwit pre-breeding moult occurring in New Zealand and Asia by sex and Alaskan breeding region (north or south of 64?N). For reference, expected staging duration in Asia (30?50 days; Chapter 2) is indicated. Discussion Evidence for pre-supplemental moult Our study provides the first direct evidence th at individual godwits replace some ventral feathers twice during pre-breeding moult, conf irming that the barred and plain-red feathers represent distinct plumages (alternate and su pplemental, respectively; Humphrey and Parkes 1959). This was strongly suggested by a populatio n-level decline of (recently grown) barred feathers of male European Bar-tailed Godwits L. l. taymyrensis during northbound stopover in the Netherlands (Piersma and Ju kema 1993) , but po pulation turnover during the study precluded strong inferences regarding individual moult. Furthermore, hormonal changes during moult potentially produce colour differences between feathers grown at different stages of the same moult (Howell 2010 ). Our photographic monitori ng of individuals throughout moult provided many specific examples of red feathers replacing barred feathers that were just 3 ? 5 weeks old (e.g., Figure 5.1b ?d). 100 Chapter 5 This rapid replacement of alternate feathers resulted in pre-departure decreases in barring of most males, and lower scores in Alaska testify to additional loss of alternate feathers in Asia. In females, the extent of pre-supplemental mo ult was insufficient to cause significant decreases in barring before departure from Ne w Zealand or arrival in Alaska. However, some females may conduct additional pre-alternate moult in Asia: 7% of females departed New Zealand with a barring score of 0, whereas all fema les in Alaska had some degree of barring. The red ventral feathers of males are generally darker than those of females (Piersma and Jukema 1993, J. Conklin pers. obs.), and male s perform an earlier and more extensive pre- supplemental moult than females. By contrast , the barred feathers of males and females are indistinguishable, and the extent and timing of pre-alternate moult were also similar. This supports the view that historical selection for alternate plumage was similar for the sexes (promoting equivalent crypsis during shar ed parental duties), whereas pre-supplemental plumage reflects current sexual selection actin g more strongly upon males, as in Ruffs Philomachus pugnax (Jukema and Piersma 2000) . Because the ventral breeding ?aspect? of female Bar-tailed Godwits contains significant c ontributions from two moults, both may be under current selection. However, breeding m ales retain little alternate barring, and so this moult may be an evolutionary artefact which is insufficiently costly to be maladaptive. If barred feathers represent the ancestral breed ing plumage within Scolopacidae, other species with both barred and red ventral feathers (e.g., Curlew Sandpiper Calidris ferruginea , dowitchers Limnodromus spp.) may undergo pre-supplemental moults that have yet to be described. It is intriguing that increases in the extent of ventral barred, dorsal spotted, and ventral red feathers occurred approximately synchronousl y in males, whereas the three types were staggered in females (Figure 5.3). This means that initiation of dorsal and ventral pre-alternate moults are temporally decoupled, by ~3 ? 5 weeks, in females but not in males. Another consequence is that temporal overlap of pre-alternate and pre-supplemental moults differs substantially between the sexes. Because mo st male godwits begin the two moults simultaneously, it appears that ventral pre-sup plemental moult replaces some basic feathers and some alternate feathers. Presumably, selection for early timing of pre-supplemental moult in males has led to the effective loss of pre- alternate moult in some follicles, which ?skip? directly from plain to red feathers without an intervening barred feather. As discussed by Battley et al. (2006) with regard to dorsal pre- supplemental moult in Great Knots Calidris tenuirostris , this creates an inconsistency within the Humphrey and Parkes (1959) nomenclature, in which follicles un dergoing just two moults in the definitive cycle can only produce ?basic? and ?alternate? plumage. Because pre -alternate moult appears partial in both sexes, certain follicles may skip directly from basic to supplemental plumage in females as Chapter 5 101 well, but we have no conclusive evidence of this . Regardless, it seems sensible to refer to the red ventral feathering of godwits as supplemental plumage, irrespective of the number of moults undertaken by individual follicles. Thus far, there is no evidence that godwits perform a dorsal pre-supplemental moult. Are timing of moult and migration linked? Because the timing of both northbound and southbound migration were 2 ? 4 weeks later in northern breeding Bar-tailed Godwits tracked with geolocators (Chapter 2), we hypothesised that other annual events may be shifted temporally to accommodate individually optimised migration schedules (e.g., Buehler and Piersma 2008). This predicts that northern breeders should complete pre-basic moult and initiate pre-breeding moult later than southern breeders. However, evidence here does not uniformly sup port this. On average, northern breeding males completed pre-basic moult and initiated pre-breedin g moult later than southern breeding males by ~3 d, which was less than di fferences in migration timing and not statistically significant. For females, the prediction was upheld for completion of pre- basic moult, but not for initiation of pre-breeding moult. Achieving breeding plumage has an effective deadline (arrival on breeding grounds) and, therefore, timing of pre-breeding moult should be closely linked to migration timing (Holmgren and Hedenstr?m 1995). However, geographical breeding cohorts of godwits differed substantially in extent of BP and, conseque ntly, in the duration of pre-breeding moult, obscuring any simple relationship between moult initiation and migratory departure. Because pre-basic is a complete moult fo r all godwits, we do not expect systematic differences in moult duration by sex or breeding region. Ther efore, variation in completion of pre-basic moult is likely to mirror variation in onset of moult, which may be triggered by hormonal changes associated with cessation of breeding activity (Hahn et al. 1992, Dawson 2006). Accordingly, we found no evidence of active moult on the breeding grounds in Alaska (May ?early August; n = 77 captures or specimens); godwit s appear to initiate pre-basic moult on post-breeding staging grou nds (McCaffery and Gill 2001). Earlier thawing of southern Alaskan breeding sites allows clutch initiation 2 ?4 weeks earlier than northern breeding sites (Chapter 2), and godwits caring for youn g through fledging may spend 3 ? 6 weeks longer on breeding sites than those that fail during incubation. Thus, the timing of pre-basic moult reflects both breeding-site phenology, wh ich may vary little annually, and duration of breeding investment, which should vary subs tantially among individuals and years. This explains why completion of pr e-basic moult varied by 10 ? 1 1 weeks between individuals in our study. 102 Chapter 5 Bar-tailed Godwits tracked with geolocators moved from breeding sites to staging sites 40 ? 8 8 days before departing Alaska (J. Conklin and P. Battley unpubl. data). Birds in this study had completed ~50 ? 9 0 % of pre-basic moult upon arrival in New Zealand, and finished moulting 33 ? 4 5 days after arrival. Godwits suspend moult for several days before migratory flights, perhaps to prioritise fuel accumulation and preparation of flight muscles and internal organs for migratory condition (Piersma et al. 1999, Landys-Ciannelli et al. 2003). Upon arrival, they may invest primarily in recovery from long-d istance flights before resuming moult (Piersma and Jukema 1993). Accounting for these periods of suspension, the esti mated total duration of pre-basic moult is ~70 ?90 days. Are breeding cohorts distinguishable in New Zealand? The plumage of Bar-tailed Godwits varies ac ross the latitude of the Alaskan breeding range (59 ? 7 1 ? N ) : the extent of male breeding plumage in creases with latitude, whereas the reddest females occur ~66?N (Chapter 3) . This population structure is not maintained in the non- breeding season, as godwits from all Alaskan re gions mix freely at New Zealand sites. By assigning individuals to probable breeding region based on size and migration timing (which also vary geographically), we asked whether coho rts were distinguishable by plumage in the non-breeding season. In New Zealand, northern br eeders of both sexes had more dorsal and ventral breeding plumage than southern breeders at both arrival and departure. Because investment in moult in Asia was similar for cohorts within each sex, plumage differences upon departure from New Zealand were si milar to those found in Alaska. In Alaska, alternate barring of females increases with latitude (Chapter 3). Males, however, do not show this pattern, probably owing to their more extensive replacement of barred feathers during pre-supplemental moult. We hypot hesised that the ancestral alternate plumage occurred in a north?south cline, before the evolution of the pre-supplemental moult. Thus, the extent of barring at the onset of pre-suppleme ntal moult in present-day godwits may indicate breeding region. This could not be adequately tested for males, because pre-alternate and pre- supplemental moults were nearly simultaneous rather than sequential. However, the prediction held for females: north ?south differences in barring were greatest in New Zealand, before substantial pre-supplemental moult in Asia. We were surprised to find differences among females in basic plumage: all birds with barring scores of 0 were southern females. The drab ba sic plumage of godwits may provide crypsis in tidal estuaries (Ferns 2003), with barring on the flanks and vent serving to enhance countershading (Rowland 2009). However, be cause the population is not geographically structured year-round, plumage differences among breeding cohorts are unlikely to have Chapter 5 103 functional significance in New Zealand. It is po ssible that variation in basic plumage barring simply represents a non-adaptive carry-over from alternate plumage. Moult strategies by sex and breeding region The more extensive BP of northern breeders was not the result of faster moult rates in New Zealand or greater investment in moult in Asia. Rather, northern birds of both sexes achieved greater BP than southern birds by spending ~7 ? 1 0 days longer in pre-breeding moult in New Zealand. All cohorts required moult in Asia to reach expected Alaskan plumage, but the extent of this additional moult was nearly identical for cohorts within each sex, despite evidence that the duration of stopovers in Asia increases with breeding latitude (Chapter 2). Our findings confirm previous indications of di stinct strategies of moult and migration in New Zealand godwits (Battley and Piersma 2005). In early March, non-moulting males were larger and fatter, and had lower BP scores and larger tes tes than those in active moult. Godwits were collected between the average dates of moult suspension for southern (25 February) and northern (9 March) breeding males in our study. Thus, the inference by Battley and Piersma (2005) that non-moulting birds were preparing for earlier migration was correct: these were clearly southern males, which are larger (Chapter 3) and migrate earlier and with less BP than northern males. Because we calculated moult in Asia from coho rt means, the proportion of individuals conducting additional moult after New Zealand de parture is unclear. In the Netherlands, moulting male godwits were redder and heavier than non-moulting ma les, suggesting that only ?high - quality? birds could afford to invest in moult during migration (Piersma and Jukema 1993). This predicts greater plumage variation in Alaska than at New Zealand departure, because differential moult in Asia shou ld magnify individual differences. Our data contradict this: variation at departure from New Zealand was generally greater than in Alaska. The lowest BP scores in Alaska were greater than the lowest scores at New Zealand departure, indicating that the palest birds moulted in Asia. Conversely, some males departed New Zealand with BP scores very close to the greatest observed in Alaska, and thus could add little to their plumage in Asia. If plumage quality influences reproductive success, godwits sh ould complete pre-breeding moult at the latest opportunity (Asia) to ensure that not all breeding feathers must endure flights of 16,000 ?1 8 , 0 0 0 km and two months of migration wear before serving their ultimate function (Holmgren and Hedenstr?m 1995). Consi stent with this, 75% of female pre-breeding moult occurred in Asia, and some southern fe males did not moult in New Zealand at all. Godwits spend 30 ? 50 days staging in Asia (mean = 40 d, sexes similar; Chapter 2), and presumably suspend moult at both ends of their stay. Thus, 26 ? 2 7 days of moult in Asia, as 104 Chapter 5 we have projected, may approach the maximu m achievable by females, without adopting faster moult rates than observed in New Zealand. Time spent in Asia is insufficient for a male?s entire pre -breeding moult (Figure 5.3), which explains why all males initiated moult in New Zeala nd. However, males performed only half as much moult in Asia as females, and so appa rently did not moult as late as possible. This may suggest that females are better suited to moult during migration, owing to greater flight or fuelling efficiency. Alternatively, if moulting c onditions were more reliable in New Zealand than in Asia, males may simply follow a more conservative moult strategy; that is, males are unwilling to ?risk? leaving a large proportion of moult until the latest opportunity, whereas females face lower costs of this risk. This assumes that godwits respond to poor conditions during the northbound flight or at staging sites by increasing investment in thermoregulation and fuelling at the expense of moult. A conseq uence would be annual variation in plumage in Alaska, potentially population-wide and more ex treme in females; there are currently no data to address this. The existence of a pre-supplemental moult stron gly implies contemporary sexual or natural selection for plumage in Alaska. Presumably, sexu al differences in plumage reflect the greater role of males in competition for mates and territories (McCaffery and Gill 2001) , and north ? south differences result from geographical variation in habitat or competition for mates. It is significant that females invested disproportionately in dorsal BP over ventral BP in both New Zealand and Asia; shared incubation requires both sexes to be cryptic, and so selection for male and female plumage should be more similar for dorsal than ventral plumage. Hypothetically, relaxe d selection for ventral BP may ?free? females and southern breeders to conduct slower, higher quality f light feather moults (Dawson et al. 2000, Serra 2001) or migrate with greater fuel stores, at the ex pense of pre-breeding moult. Conversely, prioritisation of ventral BP by males and northern breeders may constrain their investment in wing moult or fuelling. An alternative view is that plumage differences among godwits arise from energetic constraints on moult imposed by body size (Hed enstr?m 2006). This simple, but not mutually exclusive, hypothesis requires no different ial selection for plumage in Alaska. Ranked smallest to largest (northern males, southern m ales, northern females, southern females), the cohorts showed decreases in both duration and proportion of pre-breeding moult in New Zealand. Because pre-migratory fuel stores scale proportionally with body size (Battley and Piersma 2005) , larger birds must accumulate a gr eater absolute fuel mass, while maintaining a greater non-breeding mass. In addition, they must grow a greater mass of both contour and flight feathers (Hedenstr?m 2006, Rohwer et al. 2009). If larger birds were consequently Chapter 5 105 limited in time or energy available for pre-br eeding moult in New Zealand, it would predict the relative departure plumages that we observed. This hypothesis may be tested by comparing rates of pre-migratory mass gain or duration of primary feather moults by sex and body size, and by examining the extent of ov erlap in the timing of moults and fuelling. Most current knowledge of mo ult strategies derives from population-level studies, in which individuals contribute but one data point, owi ng to the difficulty of capturing or otherwise sampling free-living birds multiple times duri ng a season. Although adequate to describe general patterns of moult, this approach can mask considerable variation of ecological interest. Our use of repeatedly sampled individuals rev ealed a continuum of annual moult strategies within the New Zealand populati on of Bar-tailed Godwits, whic h may reflect individual differences in sex, size, migration distance, or breeding location, or some combination of these. Because successive life-history stages of long-distance migrants may be inextricably linked, and differential selection may occu r at any stage of the annual cycle, true understanding of how individuals manage trade-offs between moult and other non-breeding activities may require a year-round individual approach. 106 Chapter 5 107 Chapter 6 Carry-over effects and compensation: late arrival on non-breeding grounds affects wing moult but not plumage or schedules of departing Bar-tailed Godwits Conklin, J.R. & P.F. Battley Journal of Avian Biology in press (2012) Published online May 2012, doi: 10.1111/j.1600-048X.2012.05606.x 108 Chapter 6 Abstract In the annual cycle of migratory birds, temporal and energetic constraints can lead to carry-over effects, in which performance in one life-history stage affects later stages. Bar-tailed Godwits Limosa lapponica baueri , which achieve remarkably high pre-migratory fuel loads, undertake the longest non-stop migratory flights yet recorded, and breed during brief high-latitude summers, may be particularly vulnerable to persistent effects of disruptions to their rigidly-timed annual routines. Using three years of non-breeding data in New Zealand, we asked how arrival timing after a non-stop flight from Alaska (>11,000 km) affected an individual godwit?s performance in subsequent flight feather moult, contour feather moults, and migratory departure. Late arrival led to later wing moult, but godwits partially compensated for delayed moult initiation by increasing moult rate and decreasing the total duration of moult. Delays in arrival and wing moult up to 34?37 days had no apparent effect on an individual?s migratory departure or extent of breeding plumage at departure, both of which were extraordinarily consistent between years. Thus, ?errors? in timing early in the non-breeding season were essentially corrected in New Zealand prior to spring migration. Variation in migration timing also had no apparent effect on an individual?s likelihood of returning the following season. The Bar-tailed Godwits? rigid maintenance of plumage and spring migration schedules, coupled with high annual survival, imply a surprising degree of flexibility to address unforeseen circumstances in the annual cycle. Introduction The annual cycles of birds are organised into distinct, sequential life-history stages, adapted to promote reproduction and survival in the context of likely environmental conditions (McNamara and Houston 2008). With greater envi ronmental variation, the number of adaptive life-history stages (e.g., migration, distinct breeding and non-breeding plumages) increases, but flexibility in the schedulin g of each stage necessarily decreases (Wingfield 2008). Some successive stages may be temporally distinct, due to energetic or logistical incompatibility (e.g., reproduction and migration) or because they are under related endocrine control (Jacobs and Wingfield 2000). For example, hormonal changes associated with the cessation of breeding appear to have a role in triggering the initiation of pre-basic moult (Dawson 2006). Other stages may show some extent of regular or facultative overlap (e.g ., pre-basic moult and pre-migratory fat deposition; Lindstr?m et al. 1994). With greater time and energy constraints, there is increasing potential for circumstances ex perienced in one life-history stage to affect performance in subsequent stages (Harrison et al. 2011). These carry-over effects can occur Chapter 6 109 within (Earnst 1992) or across seasons (Marra et al. 1998) , and may profoundly influence individual fitness and consequently population dynamics (Norris 2005). The non-breeding season of a migratory bird is comp osed of a series of stages scheduled to facilitate travel to and from its wintering quarters and preparation for the following breeding season. These stages include: (1) replacemen t of breeding plumage with non-breeding plumage; (2) replacement of flight feather s; (3) fuelling and conducting post-breeding (autumn) migration; (4) replacement of non-breedin g plumage with breeding plumage; and (5) fuelling and conducting pre-breeding (spring) migration. Within and among species, there exist many strategies for the scheduling of mo ults relative to migrations, depending on resource availability, costs of moult, and bene fits of feather quality, any of which may vary seasonally (Holmgren and Hedenstr?m 1995, Barta et al. 2008). Long-distance migrants face the two-fold challenge of scheduling annual events according to highly seasonal temperate or arctic environments, while meeting the considerab le demands of travel itself, such as large fuel stores and high-quality flight feathers (Ale rstam and Lindstr?m 1990). These constraints may lead to energetic or temporal bottlenecks in the annual cycle (Buehler and Piersma 2008) , making long-distance migrants particularly su sceptible to carry-over effects among successive life-history stages. For example, if the non-breed ing season is very tightly scheduled, a delay early in the season (e.g., a we ather-related delay in arrival on wintering grounds) may cascade through later stages to affect performance during the return migration to breeding grounds. The annual routine of New Zealand Bar-tailed Godwits Limosa lapponica baueri features the longest and second-longest non-stop migratory flights yet recorded (>11,000 km from Alaska to New Zealand in autumn and 9,000 ?1 0 , 0 0 0 km from New Zealand to staging sites in the Yellow Sea in spring (Gill et al. 2009, Battley et al. 2012). At post-breeding fuelling sites in southwestern Alaska (July ? September), Bar-tailed Godwit s (hereafter, ?godwits?) initiate the contour feather component of pre-basic moult, performing ~50 ? 9 0 % of the transition to winter plumage before suspending moult for the southward migration (Chapter 5). After arrival on non-breeding grounds (September ? October), godwits resume the pre-basic contour moult and initiate flight feather moult, which occupies more than half of the approximately six months godwits spend in New Zealand (M cCaffery and Gill 2001). During January ? March, godwits fuel for spring migration and initiate the moult into breeding plumage (Chapter 5). This moult is also suspended for migration, an d is completed on staging grounds in the Yellow Sea. Despite the challenges presented by their extraord inary flights and the fuel stores required to conduct them (Piersma and Gill 1998, Battley and Piersma 2005), individual godwits departing New Zealand on spring migration show remarkable annual consistency in both date 110 Chapter 6 of departure and extent of breeding plumage (B attley 2006, Chapter 4). This suggests that ?errors? in the timing of prior events that may affect spring departure are either inconsequential or corrected prior to migration. Godwits could compensate for delays through faster moult rates, greater temporal overlap among moults and fuelling, or decreased body condition at departure. In this study, we assess the consequences of timing of arrival in New Zealand for Bar-tailed Godwits throughout the non-breedin g season. To do this, we ma ke the first description of timing and duration of primary feather moult at the individual level for a long-distance migrant, using detailed photographic docume ntation of free-living colour-banded godwits across three non-breeding season s. Combining these data with migration and contour feather moult schedules of the same individuals (Chapters 4 ? 5 ) , we ask whether annual variation in an individual?s arrival in New Zealand affects its subsequent timing of contour or flight feather moults, extent of breeding plumage on departure, or timing of departure on spring migration. In addition, we ask whether delays in migration in one season are related to an individual?s likelihood of survival to the following non -breeding season. Methods Data collection During three non-breeding seasons (Year 1 = January ?April 2008, Year 2 = September 2008 ? April 2009, and Year 3 = September 20 09 ?April 2010), we studied plumage and migration timing in a small, site-faithful population of Bar-tailed Godwits (200 ? 2 8 0 birds) at the Manawatu River estuary, New Zealand (40.47?S , 175.22?E). Data presented in this study are derived from 77 individually colour-banded godwit s (42 female, 35 male); 63 of these were present in Year 1, 62 in Year 2, and 58 in Year 3. We cond ucted high-tide surveys every 3 ? 4 days (d) during migratory arrival (1 September ? 2 0 October), daily during migratory departure (4 March ? 5 April), and every 4 ? 8 d during the intervening summer months (21 October ? 3 March). During surveys, we digitally photograph ed marked godwits, where possible depicting the state of primary feather moult (i.e., flying and wing stretching; Figure 6.1) and extent of breeding plumage; this resulted in 17,535 ph otographs in which one of 77 individuals was identifiable (Figure 6.2). For pu rposes of geolocator deployment and retrieval, we conducted three cannon-net captures during the study; fo r all captured godwits, we scored primary moult in the hand (19 individuals in March 2008, 44 in October 2008, and 39 in November 2009). From a combination of photographs and di rect observation, we determined timing of migratory arrival, completion of pre-basic non- flight feather moult (in this paper, ?pre -basic Chapter 6 111 moult? refers to contour feathers only; flight feather moult is considered separately), initiation of ?pre - breeding moult? (which occurs in overlapping pre-alternate and pre-supplemental contour feather moults; Chapter 5), and migratory departure, as well as extent of breeding plumage at departure, for each individual (not all data were available for every bird each year). Migratory departures were observed directly or estimated to within 1 d (Chapter 4), and other timing parameters were estimated to within approximately 1 ? 6 d (Chapter 5). Extent of breeding plumage (?BP?) was scored as the proportion (%) of ventral and dorsal basic Figure 6.1 Examples of photographs used to evaluate state of primary feather moult. (a) Moult score 2; (b) score 34; and (c) score 47; see text for scoring method. (a) (b) (c) Figure 6.2 Photographic data used for assessment of primary and contour feather moult. (a) Number of days each individual was photographed per season. (b) Total number of photographs of each individual per season. Box plots indicate mean ?1 standard deviation; whiskers indicate entire range of values. 112 Chapter 6 plumage feathers that had been replaced by breeding plumage; this was scored 1 ? 5 d before departure for all individuals (details in Chapter 5). During the followin g non-breeding season (Year 4 = September 2010 ?April 2011), we monitored the site regularly to determine the presence/absence of marked individuals, but moult was not studied. Ageing and sexing For all marked godwits, we measured expose d culmen (mm) and maximum flattened wing chord (mm) at the time of initial capture ( n = 26 in 2006, 21 in 20 07, 28 in 2008, and 2 in 2009). Most godwits ( n = 68) were sexed by culmen length (>99 mm = female; <90 mm = male), but intermediate birds (90 ? 9 9 mm) cannot be sexed by this measure. However, strong sexual dimorphism in plumage before depa rture from New Zealand (Chapter 5) allowed unambiguous sexing of th e remaining individuals ( n = 4 females, 5 males). Non-migrating individuals (aged <2 ? 3 yrs) and those known to be maki ng their first northbound migration were excluded from the study. Scoring of primary feather moult We collected 1069 in-moult records (83 in-han d, 986 from photograph s) of 74 individual godwits across three seasons (44 in Year 1, 64 in Year 2, and 58 in Year 3; Figure 6.3). We scored primary moult according to the method of Newton (1966; illustrated in Ginn and Melville 1983) : each of 10 primaries (denoted P1 through P10, from innermost to outermost) was assigned a score 0 ? 5 (0 = old, unmoulted feather; 5 = fu lly grown new feather), for a total moult score of 0 ? 50. Godwits moult their primaries successi vely from P1 outward to P10. In mid-moult, there are typically three feathers gr owing simultaneously, in predictable relative stages of growth, whereas only 1 ? 2 feathers are growing near the beginning and end of moult. In 439 of 488 (90%) cases where both of an individual?s wings could be evaluated, state of wing moult appeared symmetrical. Therefore, we assumed symmetrical moult when only one wing was scored. In cases of observed asymm etry, we scored each wing separately (mean score difference = 1.94, range = 1 ?6) and averaged these for the individual?s moult score. For approximately 67% of in-moult records, 1 ? 3 growing feathers were obstructed, out of focus, or not visible beyond the length of ad jacent coverts in the photographs, and thus could not be scored directly. In these cases, we recons tructed scores for the unviewed feathers based on the godwits? very predictable progression of moult, using in -hand scores from historical captures of adult godwits in New Zealand (1983 ? 2 0 0 8 , n = 1,434; P. Battley and A. Riegen unpubl. data) as reference 1 . For example, the bird in Fi gure 6.1a was scored ??000000 0 0 (P1 ? 1 0 , respectively) from the photograph. Go dwits typically initiate primary moult by 1 See Appendix 4 for details. Chapter 6 113 dropping P1 and P2 nearly simultaneously, and in every historical case when only P1 and P2 were growing ( n = 11), both feathers were scored as 1. Therefore, this bird?s final score was 2. Usually, however, both growing and unmoulted feat hers were visible. In Figure 6.1b, for example, nine primaries are visible (scored 5555543? 0 0 ). In this case, the hidden P7 was assigned score 2, for a final score of 34. Wh en reconstruction of feather scores introduced more than minimal potential scoring error, we excluded the record. When fully grown, the distal tip of P10 generally extends slightly beyo nd that of P9, in both extended and folded wing positions. Consequent ly, photographs of birds roosting with wings folded were useful to evaluate state of moult after score 49 was reached (Figure 6.4). However, the final length of P10 relative to P9 varied among individuals; we considered moult completed when the length of P10 matched its apparent length in photographs of that individual just prior to migratory departure. Figure 6.3 Number of in-moult records available per individual per season for the assessment of primary moult. (a) Year 1, n = 44 individuals; (b) Year 2, n = 64; (c) Year 3, n = 58. 114 Chapter 6 Figure 6.4 Two photographs of one male godwit taken on (a) 16 January and (b) 26 January of the same year. On the left wing, note that P10 is not yet visible beyond the tip of P9 in (a), but has grown to its full length in (b). Thus, the bird completed primary moult between these two observations. Conversion of primary moult scores to feather mass Because moult scores are not linearly related to the proportion of new feather mass grown, we converted individual feather scores to proportional mass values (Summers et al. 1983). Using three godwit specimens (with short, medium, and long wing chords relative to the population) from New Zealand, we derived population-specif ic mean proportional feather masses (methods in Underhill and Summers 1993) 2 . Also, feather mass is not distributed evenly along the length of a primary feather. Therefore, we derived proportional masses for each of 12 equal-length segments along each feather (method s in Dawson 2003) , and used mean values among all primaries to calculate population- specific conversions for moult score to proportional feather mass. Thus, a feather at score 1 has grown on average 0.10 of its ultimate mass, score 2 = 0.25, score 3 = 0.50, and score 4 = 0.80; these values vary substantially from those recommended by Underhill & Zucchini (1988), which did not acc ount for variable mass along feather length. By summing the proportion al masses of all 10 primaries, we converted moult scores to total Percentage Feather Mass Grown (hereafter, ?PFMG?; Underhill and Summers 1993). 2 See Appendix 4 for details. Chapter 6 115 Compared with inner primaries, outer primaries take longer to grow, have greater mass, and are more visible in photographs, making it both appropriate and feasible to score their growth with greater precision than the 0 ?5 scoring system allows. For all moult scores ?46, we estimated the length of P9 and P10 in 12 equal- length increments (effectively splitting score 3 and 4 into four sub-scores each), and calculated pr oportional mass accordingly. In Figure 6.1c, for example, P9 is approximately 11/12 grow n and P10 is 8/12 grown, resulting in total PFMG of 0.90 (compared with 0.88 if score 47 were simply converted to PFMG). Individual primary moult regressions We analysed primary moult duration individually for each godwit with ?4 in -moult scores spanning ?50% of total moult (mean = 7.4 scores, range = 4? 1 5 ). This sample included 1 godwit in Year 1, 46 in Year 2, and 49 in Year 3 (56 tota l individuals); Year 1 fieldwork began late in moult, and so only one very late -moulting female fit the criteria (Figure 6.3a). Dawson (2003) demonstrated th at feather mass accumulates at a linear rate over most of primary moult, but slower near the beginning and end, when fewer feathers are growing simultaneously. This was true in our study: in all 96 cases, PFMG increased across moult scores 4 ? 4 7 (accounting for 86% of total feather ma ss) at a linear rate (linear regressions; mean r2 = 0.987, range = 0.908 ? 1.00, all P < 0.001), but the rate was generally slower outside that range. Consequently, using observed mo ult rate (PFMG/day) derived from scores 4 ? 4 7 to calculate total moult duration will produce an underestimate (illustrated in Figure 6.5a). Unfortunately, high variation in moult rates late in moult indicated that rates derived from scores 4 ? 4 7 could not simply be scaled proporti onally to produce reliable estimates of total moult duration (see below). Ther efore, we derived two primary moult parameters for analysis: moult rate across scores 4 ?47 (PFMG/d; hereafter, ?primary moult rate? or ?PMR?) and total estimated moult duration from scores 0 ?50 (d; hereafter, ?primary moult duration? or ?PMD?). For estimation of rates early and late in moult, we examined cases in which birds were scored twice within scores 1 ? 4 or 47 ? 4 9. On average, moult rates during scores 1 ? 4 were 0.47 ? SD 0.29 ( n = 21) of an individual?s PMR. Late in moult, mean rates slowed from 0.75 ? SD 0.25 ( n = 43) of the PMR during scores 47 ? 4 8 to 0.54 ? SD 0.24 ( n = 39) within score 49. From these data, we constructed an idealised moult pr ogression for the early and late segments of moult. For each bird, we used the individual?s PMR regression equation to calculate dates of scores 4 and 47, and then applied the idealised early-moult segment to calculate the date of moult initiation (last day of score 0) backwa rd from score 4. All godwits in our study appeared to progress from score 0 to 4 in a 5 ? 7 d period; therefore, this method introduced only minor potential errors (<2 d) in date of mo ult initiation. However, du ration of moult after score 47 was extremely variable (~10 ? 3 0 d). When scores of 48 ? 4 9 were available (85 of 96 116 Chapter 6 cases), we proportionally scaled the idealised late-moult progression to fit observed data, and calculated the days from score 47 to score 50 to derive the date of moult completion; Figure 6.5b illustrates the application of this method to the data in Figure 6.5a. For the remaining cases ( n = 11), we could not confidently estimate da te of moult completion, and so these are excluded from analyses of moult completion and PMD, but are included in those of moult initiation and PMR. For some godwits with insufficient data for a full moult regression analysis, there was sufficient information to estimate date of moult completion. For birds sc ored once at 47 and at least once during score 48 ? 4 9 , we scaled the idealised late-mou lt progression to observed data (as above) to estimate th e first date of score 50 ( n = 15 in Year 1, 1 in Year 2, and 4 in Year 3). Figure 6.5 Illustration of individual primary moult progression analysis. Day 1 = 1 September 2009; 150 = 28 January 2010. ?PFMG? = Percentage Feather Mass Grown. (a) Raw data. Open circles = observed moult scores 4?47, with linear trendline (r2 = 0.995, n = 9, P < 0.0001). Filled circles = moult scores outside 4?47. This female godwit was observed in moult on days 16?134, and was not in moult on day 10 (score 0) and day 144 (score 50); thus, moult duration must be between 119 and 134 d. The daily moult rate calculated from scores 4?47 (0.922 PFMG/day) predicts a total moult duration of 108 d, a demonstrable underestimate. (b) Idealised moult progression, using observed linear rate to predict dates of scores 4 and 47, and modified rates (see text) to predict scores 0, 48, 49, and 50. The resulting estimate of moult duration is 121 d. Chapter 6 117 In addition, some birds were photog raphed at score 49 and then score 50 in a period of just 3 ? 8 d ( n = 18 in Year 1 and 3 in Year 3); for these birds, we used the midpoint between the two dates to represent the date of moult completion. Analysis For analysis, we limited our sample to individuals observed in 2 ? 3 seasons ( n = 58; 31 females, 27 males); each individual contributed 0 ?3 observations per parameter. To illustrate the degree of individual variation in each migration and moult parameter, we calculated the difference between the greatest and least values for each individual and summarised these across all birds of each sex (Table 6.1). To examine differences by sex and year in each parameter, we used mixed-model regression (PASW Statistics 18.0, SPSS Inc.) with sex and year as fixed effects and individual as a random effect. To examine within-season carry-over effects of the timing of non-breeding events, we looked for within-individual correlations between each predictor and subsequent parameter using within-subject centering (van de Po l and Wright 2009) and mixed-model regression. Predictor variables included dates of migratory arrival, completion of pre-basic moult, initiation and completion of primary moult, and initiation of pre-breeding moult; within- and between- individual variation in the predictor variable were included separately as fixed effects. Table 6.1 Within-individual between-year variation in non-breeding Bar-tailed Godwits in New Zealand. Values represent the difference between the greatest and least values for each individual across 2?3 years, summarised for all birds of each sex. Within-individual variation Males Females n mean SD range n mean SD range Arrive (date) 19 5.8 4.3 1?15 25 7.1 7.3 0?34 End pre-basic moult (date) 20 15.8 11.6 2?43 25 12.1 9.0 1?31 Start primary moult (date) 16 5.4 2.9 0?11 24 10.0 8.7 0?37 End primary moult (date) 18 7.8 5.8 0?28 25 10.8 7.9 1?27 Primary moult rate (PFMG/d) 16 0.07 0.05 0.0?0.2 24 0.09 0.07 0.0?0.3 Primary moult duration (d) 12 4.3 3.7 0?11 19 6.4 4.7 0?17 Start pre-breeding moult (date) 24 12.5 5.7 1?28 27 8.5 7.1 1?25 Departure BP (%) 25 2.4 2.0 0.0?7.5 29 2.3 2.5 0.0?7.5 Depart (date) 27 4.8 3.3 0?14 31 4.6 3.2 0?13 118 Chapter 6 Dependent variables included completion of pr e-basic moult, initiation of primary and pre- breeding moults, PMR, PMD, extent of BP at departure, and departure date. Thus, significant within-subject correlations represent between-year differences (carry-over effects) in an individual?s moult or behaviour, after controlling for persistent between -subject variation. To explore the effect of non-breeding schedu les on subsequent survival outside of New Zealand, we examined return probability after earlier and later migratory arrival and departure the previous year. For each godwit with two or more years of migration data in Years 1 ? 3 , we classified each year as ?early? or ?late? based on the extreme values of arrival or departure for that individual; we excluded birds when migr ation timing was identical between years and discarded the intermediate year for birds with three values. We then calculated the return probability of early- and late-migrating godwits ba sed on their presence/absence at the site in the subsequent year (Years 2 ? 4 ). Results Non-breeding schedule differences by sex Male and female godwits both spent an average of 177 d (range = 140 ? 1 9 5 d) in New Zealand per non-breeding season (Figure 6.6); timing of migratory arrival and departure were similar for the sexes (Table 6.2), although the earliest and latest arrivals were typically females. On average, females completed pre-basic contou r moult 9 d earlier than males and initiated pre- breeding moult 43 d later than male s; 19% of females did not init iate pre-breeding moult until after departing New Zealand. Accordingly, male s departed New Zealand with much greater mean BP scores (females 11%, males 64%) and we re more likely to overlap primary and pre- breeding moults: 5% of females and 88% of ma les overlapped these moults (mean overlap; females = ? 3 0.9 d ? SD 20.0 d, range = ? 6 6 to +30 d; males = 5.9 d ? SD 11.1 d, range = ? 1 0 to +33 d). Overlap of pre-basic and primary moul ts was also greater for males (females = 19.5 d ? SD 20.0 d, range = 0 ? 3 8 d; males = 27.1 d ? SD 15.3 d, range = 2 ? 77 d). All godwits initiated primary moult 3 ?2 9 d (mean = 15.7 d) afte r arrival in New Zealand (Figure 6.6); date of initiation di d not differ by sex (Table 6.2). Females completed primary moult 6 d later than males, reflecting their greater moult duration (females = 114.7 ? SD 7.5 d, range = 97 ? 1 2 8 d; males = 107.4 ? SD 5.8 d, range = 96 ? 1 2 0 d). Males had a slightly greater primary moult rate than females (females = 1.00 8 ? SD 0.094 PFMG/d; males = 1.044 ? SD 0.080 PFMG/d); this non-significant differenc e equates to about 4 d across scores 4 ? 4 7. Because females are larger than males (mean wi ng chord; females = 246.9 mm, range = 236 ? 2 6 1 mm; males = 231. 5 mm, range = 216 ? 2 4 3 mm), differences in primary moult rate or Chapter 6 119 Figure 6.6 Timing of migratory arrival and moults in relation to departure from New Zealand. For illustration, dates are standardised against an individual?s timing of departure to control for known migration timing differences based on breeding location. For each sex, upper bar = arrival and departure (Day 0; mean = 20 March, range = 4 March?5 April); middle bar = start and finish of primary moult; lower bar = finish pre-basic contour moult and start pre-breeding moult. Numbers indicate total observations (left) and years (right) of combined data; individual godwits (n = 58) contributed 0? 3 values in each parameter. Boxplots indicate median and 25th and 75th percentiles; whiskers indicate total range of values. Filled circles represent one late-moulting female in Year 1 (see text; start of pre- breeding moult and departure were within population norms). duration could result from size or sex differe nces (Figure 6.7). In mixed-model regression (wing chord as a fixed effect and individual as a random effect), wi ng chord was strongly positively associated with primary moult duration ( F 1,45 = 13.67, n = 83, P = 0.001) but not with primary moult rate ( F 1 , 51 = 3.22, n = 94, P = 0.078). However, with sex as an additional fixed effect, neither sex nor wing chord was significant in either test (all P > 0.07); therefore, we cannot distinguish the effects of wing chord and sex per se on primary moult. Within-individual variation in timing of migratory arrival was greater for females than males (Table 6.1); variation in departure was generally less and similar for the sexes. Variation in timing of contour feather moults (both pre-ba sic and pre-breeding) was generally greater for males than for females, but females showed greater intra-individual variation in timing and duration of primary moult. Exten t of BP at departure was extremely consistent for individuals of both sexes, varying by only 0 ? 7.5% between years. Non-breeding schedule differences by year At the population level, timing of migratory arri val did not vary among years (Table 6.2), but average migratory departure was slightly earlier (<2 d) in Year 3 than in Years 1 ? 2. Completion of pre-basic moult did not vary by year, but initiation of pre-breeding moult was earliest (by 2 ? 6 d) in Year 3. Average initiation of primary moult was 4 d later in Year 2 than 12 0 C ha pt er 6 Ta bl e 6. 2 Re su lts o f m ix ed -m od el r eg re ss io n an al ys is o f t he e ff ec t of s ex a nd y ea r on n on -b re ed in g pa ra m et er s fo r Ba r- ta ile d G od w its in N ew Z ea la nd . Fo r si gn ifi ca nt r es ul ts ( in b ol d) , e st im at es r ep re se nt m ea n ef fe ct o f se x (f em al e sh ow n re la tiv e to m al e) o r ye ar ( Ye ar 1 o r 2 sh ow n re la tiv e to Y ea r 3) on y (? SE ). Se x Ye ar y- va ri ab le n y ea rs n o bs df F P Es ti m at e SE df F P Es ti m at e SE A rr iv e (d at e) 2 10 1 1, 54 0. 04 0. 85 0 1, 46 0. 68 0. 41 5 En d pr e- ba si c m ou lt (d at e) 2 98 1, 53 4. 86 0. 03 2 ?9 .0 0 4. 08 1, 49 3. 04 0. 08 7 St ar t pr im ar y m ou lt (d at e) 2 94 1, 52 0. 04 0. 84 4 1, 45 5. 15 0. 02 8 3. 59 (Y 2) 1. 58 En d pr im ar y m ou lt (d at e) 3 12 0 1, 55 6. 83 0. 01 2 6. 03 2. 31 2, 72 1. 35 0. 26 6 Pr im ar y m ou lt r at e (P FM G /d ) 2 94 1, 51 3. 30 0. 07 5 1, 47 0. 60 0. 44 2 Pr im ar y m ou lt d ur at io n (d ) 2 83 1, 46 16 .5 5 0. 00 0 7. 29 1. 79 1, 37 0. 86 0. 35 9 St ar t pr e- br ee di ng m ou lt (d at e) 3 14 2 1, 55 18 9. 05 0. 00 0 42 .8 7 3. 12 2, 85 9. 68 0. 00 0 2. 06 (Y 1) 1. 47 5. 66 (Y 2) 1. 30 D ep ar tu re B P (% ) 3 14 6 1, 56 23 9. 14 0. 00 0 ?5 3. 60 3. 47 2, 86 3. 41 0. 03 8 0. 63 (Y 1) 0. 39 ?0 .3 8 (Y 2) 0. 38 D ep ar t (d at e) 3 16 1 1, 56 0. 75 0. 39 1 2, 10 2 4. 53 0. 01 3 1. 76 (Y 1) 0. 62 1. 24 (Y 2) 0. 57 120 Chapter 6 Chapter 6 121 Figure 6.7 (a) Primary moult rate (daily percentage of total feather mass gained across moult scores 4?47) and (b) estimated total primary moult duration (scores 0?50) of individual Bar-tailed Godwits in relation to maximum flattened wing chord. Values are individual means (n = 54) across 1?2 years. Lines indicate range of values for individuals with two years of data. in Year 3, but rate, duration, and completion of primary moult did not differ among years. Because variation in plumage at departure was minimal, annual differences of <1% were statistically significant; such differences are un likely to have biological significance, and are within expected measurement error. Within-season individual carry-over effects The timing of arrival in New Zealand is mo re variable than timing of departure: at the population level, marked go dwits arrived across 54 ? 5 5 d each year (Year 2 ? 3 ) , compared to departure spans of 23 ? 30 d (Years 1 ? 3 ). Within individuals, later a rrival had no influence on 122 Chapter 6 dates of departure, pre-basic moult completion, or pre-breeding moult initiation (Table 6.3). However, later arrival led to later initiation of primary moult; this effect was exactly 1:1 d, on average. Although timing of arrival and pre-basic mou lt were unrelated, later completion of pre-basic moult was associated with slightly later primary moult initiation (Table 6.3). Later pre-basic moult completion had no effect on primary moult rate or duration, or on timing of pre- breeding moult initiation. Individuals performed a faster wing moult in years when they started moult later: timing of primary moult initiation was significantly associated with both primary moult rate and total duration (Table 6.3). However, the de gree of the effect on moult duration ( ?0.32 d) indicates that the decrease in moult duration did not fully compensate for the late start, and so moult was completed later as well. Later completion of primary moult had no effect on timing of pre-breeding moult initiation or departure. Although pre-breeding moult ap peared unaffected by any previous parameters, within- individual variation in initiation of pre-breeding moult was surpri singly high (up to 28 d; Table 6.1). However, variation in initiation date had no influence on extent of BP at departure (Table 6.3). For all significant within-season carry-over effect s (Table 6.3), there was great variation in the degree and uniformity of response by individuals (Figure 6.8). The effect of arrival date on primary moult initiation was relatively strong and uniform (Figure 6.8a). For other carry-over effects involving primary moult, the significant results indicate general trends, but some individuals demonstrated no effect or even the opposite pattern (Figure 6.8b ?d). Cross-seasonal carry-over effects Non- breeding delays had no apparent effect on an individual?s likelihood of undertaking spring migration: in all three years, every marke d adult godwit departed the site on migration during March ?early April. Return rates of marked godwits in Years 2 ? 4 (after departures in Years 1 ? 3 ) were 0.85 (53 of 62), 0.93 (56 of 60), and 0.90 (52 of 58), respectively. Among individuals with two years of New Zealand arrival data ( n = 44), return rate was 0.93 in the year following ?early? arrivals and 0.95 after ?late? arrivals. For departures from New Zealand ( n = 58), return rate was 0.96 after ?early? departures and 0.95 after ?late? departures. When we limited data to the most extreme delays (?10 d intra -individual difference between years), there was still no discernible effect: 8 of 9 late-arriving godw its and 5 of 5 late-departing godwits returned the following year. C ha pt er 6 1 23 Ta bl e 6. 3 Ca rr y- ov er e ff ec ts w ith in t he n on -b re ed in g se as on o f Ba r- ta ile d G od w its i n N ew Z ea la nd , as s ho w n by m ix ed -m od el r eg re ss io n an al ys is u si ng w ith in -s ub je ct c en te ri ng . U ni ts a re d at es e xc ep t fo r pr im ar y m ou lt ra te (P FM G /d ), pr im ar y m ou lt du ra tio n (d ) a nd d ep ar tu re B P (% b re ed in g fe at he rs ). Si gn ifi ca nt w ith in -s ub je ct r es ul ts ( in b ol d) in di ca te a c or re la tio n be tw ee n in di vi du al t im in g of t he p re di ct or v ar ia bl e (x ) an d su bs eq ue nt p er fo rm an ce in th e de pe nd en t v ar ia bl e (y ). Fo r si gn ifi ca nt r es ul ts , e st im at es r ep re se nt m ea n ef fe ct o n y pe r un it x (? SE ). Be tw ee n- su bj ec t W it hi n- su bj ec t x- va ri ab le y- va ri ab le n o bs df F P Es ti m at e SE df F P Es ti m at e SE A rr iv e En d pr e- ba si c 96 1, 49 28 .9 5 0. 00 0 0. 98 0. 18 1, 45 2. 61 0. 11 3 St ar t p ri m ar y 91 1, 50 16 6. 70 0. 00 0 0. 84 0. 06 1, 39 92 .9 1 0. 00 0 1. 00 0. 10 St ar t p re -b re ed in g 10 0 1, 55 0. 22 0. 64 1 1, 42 0. 00 4 0. 94 8 D ep ar t 10 1 1, 56 18 .7 0 0. 00 0 0. 36 0. 08 1, 44 0. 37 0. 54 4 En d pr e- ba si c St ar t p ri m ar y 92 1, 48 52 .6 8 0. 00 0 0. 44 0. 06 1, 45 5. 70 0. 02 1 0. 23 0. 10 Pr im ar y ra te 92 1, 47 1. 71 0. 19 8 1, 44 2. 13 0. 15 2 Pr im ar y du ra tio n 81 1, 43 6. 91 0. 01 2 ?0 .1 6 0. 06 1, 35 0. 13 0. 71 1 St ar t p re -b re ed in g 97 1, 50 3. 20 0. 07 9 1, 43 1. 95 0. 17 0 St ar t pr im ar y Pr im ar y ra te 94 1, 50 3. 59 0. 06 4 1, 38 37 .2 1 0. 00 0 0. 00 7 0. 00 1 Pr im ar y du ra tio n 83 1, 43 11 .0 2 0. 00 2 ?0 .3 0 0. 09 1, 31 9. 11 0. 00 5 ?0 .3 2 0. 10 En d pr im ar y St ar t p re -b re ed in g 11 9 1, 56 4. 31 0. 04 2 0. 71 0. 34 1, 60 0. 92 0. 34 1 D ep ar t 12 0 1, 58 4. 40 0. 04 0 0. 20 0. 09 1, 61 1. 82 0. 18 2 St ar t BP D ep ar tu re B P 13 0 1, 55 28 1. 01 0. 00 0 ?1 .1 2 0. 07 1, 72 0. 07 2 0. 79 0 Chapter 6 123 124 Chapter 6 Figure 6.8 Significant within-individual carry-over effects indicated in Table 3. For individuals with two years of data, difference (absolute value in days) between dates of the predictor variable in Year 2 and 3 is indicated on x-axis. Y-axis indicates degree and direction of change in the dependent variable during the year of later x. For reference, the dotted line represents no effect of x on y. Chapter 6 125 Discussion This study presents the most detailed examinatio n of within-individual carry-over effects of the timing of post-breeding migration on the su bsequent non-breeding schedules of a long- distance migratory bird. Surprisingly, annual variatio n (up to 34 d) in a rrival in New Zealand after the longest known non-stop migratory flight had no apparent effect on an individual Bar- tailed Godwit ?s extent of breeding plumage or timing of departure from New Zealand on the following spring migration. In general, godw its compensated for later autumn arrival by conducting a faster wing moult and through a greater overlap of wing and pre-breeding moults. Thus, ?errors? in timing at the start of the non -breeding season were essentially corrected in New Zealand prior to spring migration. Seasonal carry-over effects can arise from many causes and take countless forms (Harrison et al. 2011). In migratory birds, the most commonly examined carry-over effect is that of spring migration timing on breeding success (e.g., Bety et al. 2004). Also, the relationship between breeding investment and timing of pre-basic mo ult has been evaluated for species that breed and initiate moult at the same site (e.g., Earnst 1992 ). Carry-over effects in the non-breeding season have received less attention, largely due to the difficul ties of tracking moult, fuelling, and migration of individual birds. Our study, the first to examine intra-individual variation in duration of primary moult in a wild migratory population, repre sents a significant advance in understanding how individuals manage potential carry-over effects to maintain time-critical migration schedules. Individuality in primary moult Very little is known about inter- or intra-individu al variation in flight feather moult in wild populations, because most information comp rises average values from population-level studies (e.g., Summers et al. 1983, Underhill 2003) or is derive d from controlled experiments (e.g., Dawson 2004). Estimatin g wing moult parameters for individuals typically requires multiple captures (but see Bensch and Grahn 19 93) , which generally provides small samples, except for colonial species that mou lt during breeding (e.g., Emslie et al. 1990). Although wing moult can be conspicuous in the field (Howell 2010), we know of only one previous study using repeated photographs of free-liv ing individuals to measure primary moult progression (California Condor Gymnogyps californianus ; Snyder et al. 1987). Our study was possible because godwits in New Zealand have high non-breeding site-fidelity, make predictable daily movements to open-habitat high -tide roosts, are relatively approachable, and are large enough for discernment of both colour-bands and state of moult even in photographs of relatively low quality. Such an approach re mains impractical for most systems, and our 126 Chapter 6 labourious methods probably cannot be transferred to other studies without a similar investment of time and effort. We found substantial between-indivi dual variation in moult duration (Figure 6.7b): the longest duration we observed (128 d) exceeded the shor test (96 d) by 33%. Th is illustrates the magnitude of individual variation that may be masked by average values calculated from traditional population-level methods (e.g., Underhill et al. 1990). Across species, duration of primary feather moult generally increases allometrically with wing length and body size (Rohwer et al. 2009); therefore, some of the variatio n we observed may result from individual differences in size. In Bar-tailed Godwits, wi ng chord varies substantially between sexes (on average, females 15 mm greater than ma les) and within each sex (variation up to 25 ? 2 7 mm). However, our analysis lacked the power to sepa rate confounding effects of size and sex, and so the greater moult duration of females may be an effect of sex per se. As we observed no sex differences in timing of migration or initiation of primary moult (Table 6.2), it is probably unrelated to systematic differences in reproductive effort or parental care in Alaska. However, males conduct a much more extensive pre-breed ing moult than do females (Chapter 5), which could result in a more time-constrained flight feather moult for males. It is intriguing that the rate of new feather mass gain throughout 86% of moult (PMR), though clearly linear, was not a reliable pr edictor of total moult duration. On average, moult durations (PMD) were 13% longer than a stra ight regression based on scores 4 ? 4 7 would predict (Figure 6.5), similar to the findings of Da wson (2003), but there was substantial individual variation in the degree to which mass gain slowed when only P9 and 10 were growing (scores 47 ? 4 9 ). This variation constituted another sex disparity: although PMR was statistically similar for males and females, PMD was significantly longer for females (Table 6.2). For the average male, PMD was 9% longer than PMR wo uld predict, compared to 16% for females, indicating that the final phase of primary moult was generally more protracted in females. Again, this may indicate that completion of prima ry moult is more time-constrained for males than females in this system. The lack of a direct correlation between PMR and PMD also raises the question of whether total duration is th e variable of greatest interest for studies of flight feather moult. Arguably, variation at the fringes of primary moult may be of less biological significance than the linear rate across the great majority of total feather mass gain. We know of only one previous study that examined intra-individual variation in timing of primary moult. For Barnacle Geese Branta leucopsis , date of initiation of primary moult appeared quite consistent annually and to some extent heritable (Larsson 1996) ; however, initiation dates were generally estimated from single captures during moult and did not account for potential individual variation in moult rates. In our study, timing of primary moult Chapter 6 127 was largely dependent upon date of arrival in New Zealand, with intra-individual variation up to 37 d in date of moult in itiation and up to 17 d in moult duration. Such flexibility does not exclude a heritable component to moult schedules , but does demonstrate short-term ability to respond to conditions that vary annually. According to population-level studies, initiation of primary moult on wintering grounds may vary with breeding success (Barshep et al. 2011b) and later-moulting individuals appear to compensa te by moulting faster (Johnson and Minton 1980, Barter 1989). Starlings Sturnus vulgaris responded to experimentally delayed moult by decreasing moult duration as much as 20 ? 3 0 % (Dawson 2004). Our study, in which an individual?s primary moult duration varied up to 17% annually, explicitly demonstrates this in a wild population. Influence of conditions prior to arrival in New Zealand In this population, between-individual variation in migration timing, and to a certain extent scheduling of the entire annual cycle, is linked w ith breeding latitude in Alaska (Chapters 2 and 5); in general, southern breeders migrate ear lier in both spring and autumn. Additional intra-individual variation in timing and condi tion upon arrival in New Zealand may result from fuelling conditions in Alaska, timing of weather systems conducive to southbound departure, and conditions experienced en route (Gill et al. 2009). Although breeding success (i.e., duration of parental investment) has been linked with autumn migration timing in other studies (e.g., Kjell?n et al. 2001), it does not appear to affect timing of southward departure from Alaska by Bar-tailed Godwits (Chapter 2). However, cessation of breeding activity very likely affects the timing of pre-basic moult (Hahn et al. 1992, Dawson 2006) ; godwits initiate this moult at post-breeding staging areas (McCa ffery and Gill 2001, Chapter 5). This probably explains why intra-individual variation in timin g of arrival in New Zealand and pre-basic moult completion were not strongly correlated (Table 6.3). The direct flight from Alaska to New Zealand takes 8 ? 9 d in favourable conditions, but godwits may make opportunistic stops or detours in adverse weather (Gill et al. 2009). Variable trans-Pacific conditions therefore likely result in some birds arriving in New Zealand in good condition with unused fuel accumulated in Alaska, some arriving in poor condition with completely depleted fuel stores, and some arriving later after recuperating and refuelling in locations such as New Caledonia and sout heast Australia. Godwits reaching New Zealand in poor condition may be unable to initiate moult and replenish body stores simultaneously; this may explain why some bird s initiated primary moult just 3 ? 6 d after arrival, while others delayed up to 29 d. Population-wide patterns of moult initiation after arrival may therefore indicate whether godwits experienced a relatively easy or difficu lt autumn migration that year. For example, 128 Chapter 6 average timing of arrival was similar in Years 2 ? 3 , but mean primary moult initiation was 3.6 d later in Year 2 (Table 6.2). In Year 2, the first four females to arrive (3 ? 1 4 September) delayed the start of primary moult for 19 ? 2 9 d after arrival, on aver age 10 d longer than they did after similar arrivals in Year 3. Geolocator data from on e of these birds revealed a dramatic detour to the vicinity of New Caledonia (800 ? 9 0 0 km west of her prior course toward New Zealand), resulting in a 3-day delay in arrival (J. Conklin and P. Battley unpubl. data). Other possible indications of a difficult autumn migration in Year 2 included a lower return rate (85% vs. 90 ? 9 3 % in Years 3 ? 4 ) and a greater number of birds stopping elsewhere in New Zealand rather than flying directly to our study site (3 of 10 geolocators in Year 2, and 0 of 9 in Year 3; J. Conklin and P. Battley unpubl. data). Within-season trade-offs and carry-over effects Pre-basic moult is typically viewed as a sing le stage, comprising both flight feather replacement and the transition into non-br eeding plumage (Humphrey and Parkes 1959). However, these two components are to some extent modular (Piersma et al. 2008) ; many long-distance migrants initiate pre-basic contour mo ult prior to autumn migration but delay flight feather moult until arrival on wintering grounds (e.g., Thomas and Dartnall 1971, this study). This strategy avoids potential flight pe rformance costs of migrating with incompletely moulted wings (Swaddle and Witter 1997) , and also raises the question of energetic trade-offs between the two components of pre-basic mou lt, which only partially overlap in the non- breeding season. We found that intr a-individual variation in date of pre-basic contour moult completion had a small but significant effect on the initiation of primary moult, beyond the greater effect of arrival (Table 6.3). This sugges ts that the state of contour moult (which could reflect annual variation in breeding success or conditions during fuelling or migration) presents some limitation on the timing of primary moult. However, timing of pre-basic moult completion did not affect rate or duration of primary moult, so we have no evidence that overlap of the two moults (0 ? 3 8 d) affected moult performance. Adaptive responses to experimentally delayed primary moult have varied among studies. In Bluethroats Luscinia svecica , neither timing nor duration of primary moult changed in response to a photoperiod-simulated delay of one month (Lindstr?m et al. 1994). By contrast, starlings responded to a 3-week testosterone- induced delay in primary moult by decreasing moult duration by about 12 days (from 104 to 94 d) and the response to a photoperiod-induced delay was even more dramatic (from 119 to 92 d; Dawson 2004). In our study, godwits responded to later primary moult initiation by moulting faster, but decrea ses in moult duration compensated for only 32% of the delay in initiation, on average (Table 6.3). This could indicate that flexibility in moult rate has an abso lute limit, or perhaps th at only individuals in the best condition have the option of accelerating moult. Also, godwits may face a trade-off Chapter 6 129 between the costs and benefits of timely moult completion. It has been demonstrated that faster moult rates can lead to reduced feather qu ality, in terms of strength , mass, and durability (Dawson et al. 2000, Serra 2001) , and that condition of primary feathers can significantly affect flight performance (Swaddle et al. 1996). New Zealand godwits travel more than 30,000 km between flight feather replacements; this annual routine may tolerate minimal compromise in flight feather quality withou t jeopardising survival or breeding success. If flight feather quality is a top priority for long-distance migrants, we may expect primary moult to temporally preclude other potentially costly activities, such as pre-migratory fuelling and pre-breeding moult. For example, Grey Plovers Pluvialis squatarola in South Africa gained mass only after completing or suspending wing moult (Serra et al. 1999). However, it is clear that fuelling and moult are not mutually exclusive in New Zealand godwits. Godwits appear to fuel quite slowly, starting up to three months before departure (J. Conklin and P. Battley unpubl. data), a period encompassing a significant portion of primary moult and all pre-breeding moult conducted in New Zealand. In fact, due to two closely-timed ventral feather generations (Chapter 5), some godwits in late January/early February are fuelling while conducting portions of three moults (pre -basic flight feathers, pre-alternate and pre- supplemental body feathers) simultaneously. Mo st males and some females overlapped the primary and pre-breeding moults, and annual varia tion in the former did not influence timing of the latter (Table 6.3). Overlap of the moults almost always (49 of 51 cases) occurred when only the last two primaries were growing (score 45 ? 4 9 , >80% total moult completed); two late-running females initiated pre-breeding mo ult while still growing P8. However, some females finished primary moult >50 d before th eir first breeding feathers appeared; therefore, there is no evidence that completion of wing moult frees birds to start pre-breeding moult. Accordingly, differences in duration or timing of wing moult do not appear to drive individual or sex differences in breeding plumage in Alaska (as hypothesised in Chapter 5). It is intriguing that variation in an individual?s initiation of pre -breeding moult had no effect on its extent of breeding plumage at de parture (Table 6.3). Males conduct 12 ? 5 8 d of pre- breeding moult while in New Zealand (Chapter 5) , which corresponds to variation in plumage at departure. However, moult rates varied subs tantially among individuals, as did the timing of moult suspension prior to departure. This suggests that individuals may respond to variation in moult initiation by adjusting moult rate and/or date of pre-departure suspension. However, there was no evidence that birds ahead of schedule seized the opportunit y to improve breeding plumage. Most godwits resume pre-breedin g moult during northbound stopover in Asia (Chapter 5); it may be beneficial to complete mo ult at the last possible opportunity (Holmgren and Hedenstr?m 1995) , even if earlier opportunities present themselves. 130 Chapter 6 In a concurrent study of the same individual go dwits, we showed that most unexpectedly early or late spring departures (67%) were attributable to avoidance of departure in headwinds or delay due to recent capture (Chapter 4); none of the unexplained of f-schedule departures are attributable to variation in arrival or moult shown in this paper. Still, there must be a point at which a poorly scheduled non-breeding season affects when or whether an individual migrates. Each breeding season, a small number of adult godwits do not migrate from New Zealand in spring (P. Battley unpubl. data), presum ably due to poor condition or advanced age. In our study, every bird known to be >3 y ears old migrated each year, so we could not evaluate circumstances leading to a failure to migrate. Potential downstream carry-over effects The demanding annual routines of long-dis tance migratory birds suggest lives spent precariously pressed against time and energy constraints. With their extreme migrations and tight annual schedules, New Zealand Bar-tailed Go dwits may seem particularly vulnerable to ?errors? in timing that may cascade through subsequent life -history stages. Despite this, we found no evidence that delays in the non-breeding season carried over to the timing of spring migration, suggesting that flexibility exists to compensate for unforeseen circumstances. Even the most late-moulting individual in our study (Figure 6.6) departed New Zealand within a normal time frame, despite completing primary mo ult just one day earlier. This bird also did not return the next season, which raises the question: do Bar-tailed Godwits prioritise rigid migration schedules over all other concerns? In this paper, we used timing of non-breeding events and extent of breeding plumage at depart ure to evaluate carry-over effects. We did not measure other potential qualitative indicators su ch as feather condition or extent of pre- migratory fuelling, either of which could be re duced by delays in the non-breeding season and lead to downstream effects on breeding success or survival. Migration timing in New Zealand had no apparent effect on survival, as return r ates were similar after early and late arrivals and departures. However, it is plausible that a go dwit with reduced feather quality or fuel stores could complete migration but fail to breed after arriving in Alaska late or in poor condition. As Bar-tailed Godwits are long-lived and may have 10 ? 2 0 breeding opportunities in a lifetime, we should expect them to prioritise surv ival over breeding success in any particular year. Therefore, the ultimate carry-over effects of non-breeding schedules may be subtle, and cannot be truly assessed without measures of individual fitness. 131 Chapter 7 Absolute consistency: individual versus population variation in timing of annual life-history stages of a long-distance migrant bird Conklin, J.R., P.F. Battley & M.A. Potter Submitted manuscript 132 Chapter 7 Abstract Flexibility in scheduling varies throughout an organism?s annual cycle, reflecting relative constraints and fitness consequences among specific life-history stages. Using light-level geolocators and direct observation, we tracked individual Bar-tailed Godwits Limosa lapponica baueri for two full years (including non-breeding seasons in New Zealand and round-trip migrations to Alaska) to present the most complete annual-cycle view of moult, breeding, and migration schedules yet available for a long-distance migratory bird. At both population and individual scales, our data support two hypotheses: schedules tightened with proximity to the breeding season, and migratory movements were more precisely scheduled than moults. In general, individual godwits were remarkably consistent in timing of events throughout the year, and individual repeatability (r) of pre-breeding movements in particular was extraordinarily high (0.82?0.92). However, we demonstrate that r values misrepresent absolute consistency by confounding two parameters (inter- and intra-individual variation) containing very different information; the biological significance of r values can only be understood when these are considered separately. To evaluate potential flexibility to address stochastic or directional environmental change, temporal variation in any single season must be viewed with an annual-cycle perspective and recognition of the different mechanisms and implications of individual and population variation. Introduction The typical vertebrate annual routine is composed of distinct, sequential life-history stages (e.g., breeding, moult) shaped by natural selection to maximise fitness (McNamara and Houston 2008). Performance in any particular s tage depends on both the costs of sub-optimal performance (selection pressure) and the individual?s ability to behave optimally, which may vary with control mechanisms, resource availability , and individual quality. Species generally respond to substantial and predictable environmental variation by expressing additional life- history stages (e.g., migration), but increasing annual-cycle complexity results in greater temporal or energetic constraints on specific stages (Wingfield 2008) and potential trade-offs among stages. It follows that individuals shou ld prioritise optimal performance in events most critical to fitness (survival or re productive output) and display m ore variation in stages with lesser impacts on fitness. For any life-history parameter, the total range of values observed in a population contains two primary components: (1) inter-individual variatio n resulting from persistent differences in strategy or performance; and (2) intra-individu al variation resulting from annual differences in Chapter 7 133 performance. These two components (which we will refer to as ?population variation? and ?individual variation?, respectively) may reveal very different things about constraints on a population, and the relative contribution of each can only be evaluated in multi-year individual-level studies. For example, migrat ory departures occurring across an extended period of time could result from all individuals having a similar strategy but being inconsistent in performance, or from individuals performing a range of strategies with very high precision; these scenarios lead to very different conclusions regarding constraints and flexibility in the system. There may be only one ?correct? way to accomplish certain annual tasks, whereas other stages tolerate a number of strategies or a range of individual performance with equivalent fitness consequences. For migratory birds, arrival on breeding grounds is generally thought to impose the greatest temporal constraint in the annual routine (McNamara et al. 1998), due to clear associations between the timing and success of breeding (Verhulst and Tinbergen 1991, Bety et al. 2004). At high latitudes, extreme seasonality and shor t breeding seasons make timing of arrival especially important (Meltofte et al. 2007b). Thus, annual cycles of long-distance migrants may be expected to have one primary time-critic al focal point (breeding arrival), with other stages subjected to increasing time-selection as birds conduct the moults, pre-migratory fuelling, and movements that facilitate timely breed ing arrival. We may then expect both population and individual variation in timing to decrease with proximity to the breeding season. This idea is supported by observations of tighter temporal windows for pre-breeding (spring) than post-breeding (autumn) migration (Alerstam et al. 2006, Egevang et al. 2010), and increasingly precise migration timing in successive stages of spring migration (Farmer and Wiens 1999, Chapter 2). Long-distance migratory movements are primar ily governed by endogenous programmes (Gwinner 1996) that are largely unaffected by annually variable social and environmental cues, and may be more rigidly scheduled than breeding itself (Both and Visser 2001, Smith et al. 2010). By contrast, scheduling of moult appe ars more flexible at both evolutionary (Holmgren and Hedenstr?m 1995, Helm an d Gwinner 2006) and annual (Dawson 2004, Chapter 6) time scales, presumably due to lesse r fitness consequences of timing of moult. Therefore, we may expect greater population and in dividual variation in timing of moults than in migratory movements, but this has never been explicitly examined in a wild migratory population. The annual routine of New Zealand Bar-tailed Godwits Limosa lapponica baueri (Figure 7.1) includes a short, high-latitude breedin g season, a complex moult (Chapters 5 ? 6 ) , and the two longest non-stop migratory flights yet recorded (Gill et al. 2009, Battley et al. 2012), and thus 134 Chapter 7 may feature considerable time constraints and trade-offs among life-history stages. In this study, we combine two years of detailed observ ations of colour-banded Bar-tailed Godwits in New Zealand with geolocator-tracking of a subset of the same individuals to present the most complete picture of temporal variation throughout the annual cycle yet available for a long- distance migratory bird. We describe po pulation-level variation and individual-level consistency in moults, movements, and init iation of breeding to test two non-mutually exclusive hypotheses: (1) that variation in timing of key annual events decreases with increasing proximity to breeding; and (2) that ti ming of migratory movements is more rigidly maintained than that of moults. Figure 7.1 Generalised cycle of annual life-history stages of New Zealand Bar-tailed Godwits (typical adult male). Arrows indicate major migratory flights. Chapter 7 135 Methods Individually marked birds During three non-breeding seasons (January ?April 2008, September 2008 ?April 2009, and September 2009 ?April 2010), we studied moult and mi gration timing in a small population of Bar-tailed Godwits (200 ? 2 8 0 godwits; approximately 25% were individually colour-banded) at the Manawatu River estuary, New Zealand (40.47?S, 175.22?E). We conducted high-tide surveys every 3 ? 4 days (d) during migratory arrival (1 September ? 2 0 October), daily during migratory departure (4 March ? 5 April), and every 4 ? 8 d during the intervening summer months (21 October ? 3 March). During surveys, we us ed direct observation and digital photography to monitor primary feather moult, contour feather moult, and presence/absence of all marked individuals (7 7 total; 35 male, 42 female; n = 58 ? 6 3 per season). With these data, we determined for each individual the timing of migratory arrival, completion of pre- basic contour feather moult, initiation and comp letion of primary feather moult, initiation of pre-breeding contour feather moult, and migr atory departure. Departures were generally observed directly, and other parameters were estimated to within approximately 1 ?6 d. We have described details of data collection and analysis specific to each parameter elsewhere: arrival and contour moult (Chapter 5), primar y moult (Chapter 6), and departure (Chapter 4). Light-level geolocators1 A subset of colour-banded godwits at the site we re additionally equipped with leg-mounted light-level geolocators (British Antarctic Survey model MK14; 1.4 g; 2 -year life) to track movements outside of New Zealand. Twenty in strumented individuals (9 male, 11 female) provided data for this study; eight of these were tracked for two entire annual cycles. Derived breeding locations of these godwits spanned most of the known breeding range in Alaska (59 ? 7 0 ? N (Chapter 2), indicating that most variation in migration schedules present in the greater New Zealand population was encompassed by our sample. The geolocators recorded sunrise and sunset, allowing daily calculation of latitude and longitude (?130 km error, base d on ground-truthing units a nd resightings of instrumented godwits), except during ?15 d of the vernal or autumnal equinox, when only longitude is reliable (Fox 2010). To derive fuelling and breed ing sites outside New Zealand, we averaged twice-daily locations over periods when bird s were relatively stationary, excluding clear outliers likely resulting from weather- or behaviou r-related shading events near dawn or dusk. Clear shifts in latitude or longitude indicated the initiation of major mi gratory movements, which we considered concluded when a bird?s location once again stabilised. Sample sizes decreased throughout May ? September due to unit failures. Additionally, we could not 1 See Appendices 1?2 for more detailed methods. 136 Chapter 7 determine timing of departure from the breeding area for two birds, because their breeding and post-breeding staging locations differed by le ss than the location error of the geolocators. Geolocators also indicated periods of nest incubation (Eichhorn et al. 2006, Chapter 2). During the breeding season, ge olocators registered nights as regular, clearly demarcated periods of darkness <4.5 hours in length; thes e did not appear at all if birds bred north of 64?N. Days appeared as continuous light, irregu larly broken by very brief (<1 hour) shading events, most likely corresponding to behaviours such as wading or sitting. Within 6 ?2 5 d of apparent arrival on breeding grounds, most bi rds (14 of 16 cases) displayed a conspicuous pattern of incubation, in which semi-regular shading events of 4 ?1 3 hours were overlaid on the day/night pattern for periods up to 25 d. We co nsidered the first day of this period to be the start of incubation. Analysis We present data for two complete annual cy cles, from New Zealand departure in 2008 through the initiation of pre-breeding contour mo ult in 2010 (the period for which geolocator data are available); this includes geolocator data for events outside New Zealand, and direct observations of colour-banded godwits (includi ng the geolocator-tagged birds) for events within New Zealand. To describe population var iation in timing of each stage, we pooled all observed dates across two annual cycles (1 ? 2 observations per bird); population spans are the difference between the earliest and latest individuals. To describe individual variation in timing for each stage, we calculated the diffe rence (d) between the two values for each individual observed in both years. To ask wh ether population (spans across all birds) or individual (geolocators only) variation decreased with proximity to the breeding season, we ranked data across all stages and tested fo r differences from hypothesised ranks (stages ranked chronologically from post-breeding disper sal to first incubation) using Spearman-rank correlation. To see how well our geolocator sample represented variation in the larger population, we tested for statistical differences from colour-ba nded samples when both types of data were available (six stages within New Zealand). The larger colour-bande d samples naturally contained more extreme values (Figure 7.2b), bu t medians and distributions of values in every stage were similar to geolocator samples for both population (Mann-Whitney tests, all P = 0.25 ? 0.88) and individual data (all tests P = 0.18 ? 0.96). Therefore, we combined the two data sources for the best representation of population variation throughout the year. However, we limited analyses of individual variation to geolocator birds tracked for two entire annual cycles ( n = 8), to ensure that samples for all s tages were as comparable as possible. Chapter 7 137 Figure 7.2 Population and individual variation in timing of events throughout the annual cycle of Bar- tailed Godwits. Stages (x-axis) are in chronological order, starting with dispersal from breeding grounds (see Figure 7.1). (a) Distribution of all observations during March 2008?March 2010, standardised by date (day 1 = earliest observation for each stage). Data are derived from colour- banded (New Zealand only) or geolocator-tracked (outside New Zealand) godwits. Boxplots indicate median and 25 th and 75 th percentiles; whiskers indicate 5 th and 95 th percentiles; filled circles indicate more extreme values; overlapping points are offset for clarity. (b) Individual consistency in timing (boxplots; difference between dates in successive years) for geolocator-tracked godwits, compared to total population spans (bars) shown in (a). Boxplots indicate median and 25th and 75th percentiles; whiskers indicate entire range of values. Circles indicate greatest (open) and least (filled) within- individual differences in the larger sample of colour-banded (CB) birds (New Zealand only; sample sizes shown). For stages in New Zealand, light grey bars indicate the proportion of variation observed among colour-banded birds that was not present in the smaller geolocator sample (dark grey bars). Geolocator (GL) sample sizes = individuals observed in two successive years (boxplots)/total individuals (dark grey bars). 138 Chapter 7 For each stage, we calculated repeatabili ty (intra-class correlation coefficient, r; Lessells and Boag 1987) ? SE (Becker 1984) of individual timing for all go dwits with two years of data. For comparison, we calculated r separately for all colour-banded godwits and the subset of geolocator-tagged birds. Results Variation with proximity to breeding Population variation was least in events leading up to the breeding season ( Figure 7.2a): during stages from spring departure from Ne w Zealand through the on set of breeding, the earliest and latest individuals differed by only 23 ? 3 1 d, compared to spans of 30 ? 78 d for other post-breeding events in Alaska and New Zealand. Population spans decreased chronologically from post-breeding departure to start of incubation (Spearman-rank correlation, one-tailed: rs = 0.71, n = 13, P = 0.003). When considered separately, movement spans decreased chronologically ( rs = 0.77, n = 8, P = 0.013) but moult spans did not ( rs = ?0.45, n = 4, P = 0.28). In general, individual godwits were extremely consistent in timing ( Figure 7.2b): in every stage, some individuals differed by only 0 ? 2 d between years. Individual variation was least for movements toward the breeding grounds, with median differences of only 2.0 ? 3.5 d between years. The greatest individual differences were only 4 ? 1 0 d in pre-breeding movements, compared to 9 ? 4 3 d in post-breeding events in Alaska and New Zealand; accordingly, variances differe d significantly among annual events (Brown-Forsythe test: F 1 2 ,29 = 4.44, P = 0.001). Individual variation (mean, standard deviation [SD] , and maximum values) decreased chronologi cally from post-breeding departure to start of incubation (Spearman-rank correlations, all tests P ? 0.016); considered separately, this was true for movements (all tests P ? 0.005) but not for moults (all tests P ? 0.20). Moults versus movements Among stages, population and individu al variation were strongly correlated ( r = 0.778, n = 13, F = 16.92, P = 0.002) and were greater for moults than for movements (Figure 7.3). Mean population spans were 68.3 d for moult parameters and 34.8 d for movements (Mann-Whitney test: U = 2.39, n = 12, P = 0.017). Individual variation was also greater for moults than for movements (Mann-Whitney tests of mean, SD, and max; all P = 0.033 ? 0.041); these differences were more profound when post-b reeding departure was excluded (all tests P = 0.006). Chapter 7 139 Figure 7.3 Moults were more loosely scheduled than movements at both the population and individual levels. For each stage, population variation represents the difference (in days) between the earliest and latest individuals (from Figure 7.2b). Individual variation indicates the mean difference between years for all individuals (2-year geolocators only; n = 6?8; see Figure 7.2b). Individual repeatability Among geolocator-tagged godwits with two y ears of data, individual repeatability was uniformly very high for timing of spring movements and somewhat lower for autumn migration and initiation of pre-breeding moult ( Table 7.1); repeatability was not significant for other moult parameters, incubation, or depart ure from breeding sites. The larger sample of colour-banded godwits demonstr ated significant repeatability in all stages in New Zealand (Table 7.1), and again this was higher for s pring departure than for autumn arrival. Repeatability of pre-breeding moult initiation wa s very high, and substantially higher than other moult parameters. Discussion Despite the importance of considering specific s tages in the context of the entire annual cycle (McNamara and Houston 2008) , few migration studies have empirically addressed the timing of movements in more than one season, and none has additionally included scheduling of 140 Chapter 7 Table 7.1 Individual repeatability of timing of key events throughout the annual cycle of Bar-tailed Godwits. Significant results are indicated in bold. Geolocator-tagged only All colour-banded n r SE F P n r SE F P Depart breeding grounds 6 0.47 0.26 2.8 0.12 Depart Alaska 8 0.77 0.11 7.8 0.005 Arrive New Zealand 8 0.77 0.11 7.7 0.005 44 0.66 0.06 4.9 <0.001 Start wing moult 7 0.44 0.25 2.6 0.12 40 0.53 0.08 3.3 <0.001 Finish pre-basic moult 8 0.37 0.25 2.2 0.15 45 0.53 0.08 3.3 <0.001 Finish wing moult 7 ?0.23 0.52 0.6 0.71 34 0.55 0.09 3.5 <0.001 Start pre-breeding moult 8 0.76 0.11 7.3 0.006 47 0.91 0.02 20.3 <0.001 Depart New Zealand 8 0.86 0.07 13.2 0.001 49 0.82 0.03 10.3 <0.001 Arrive Yellow Sea 8 0.90 0.05 19.3 <0.001 Depart Yellow Sea 8 0.92 0.04 24.7 <0.001 Arrive Alaska 8 0.89 0.05 17.4 <0.001 Arrive breeding grounds 8 0.91 0.05 21.4 <0.001 First incubation 7 0.35 0.28 2.1 0.18 moults and breeding for a comprehensive view of temporal variation throughout the year. Using a combination of direct observation and ge olocator data, we present a unique view of year-round temporal variation for a set of migr atory individuals, placed within the context of population-level patterns. In our study, Bar-tailed Godwits were remarkably consistent in timing of spring movements toward the breeding grounds, but showed greater flexibility in timing of post-breeding movements and moults . These results supported two hypotheses: at both population and individual levels, schedules ge nerally tightened as the breeding season approached, and migratory movements were more precisely scheduled than moults. Variation with proximity to breeding The idea that time constraints for migratory birds increase with proximity to the breeding season has a wealth of theoretical support (Alerstam and Lindstr?m 1990, McNamara et al. 1998), but surprisingly little empirical evidence derived from year-round data from a single population. Previous studies have shown that individual timing is more consistent for spring migration than autumn migration (Rees 1989, Alerstam et al. 2006, Egevang et al. 2010, Vardanis et al. 2011), and that timing becomes increasingl y precise in successive stages of spring migration, at either the individual (Bety et al. 2004) or population level (Warnock and Bishop 1998, Farmer and Wiens 1999, Battley et al. 2004, Warnock et al. 2004). Our data Chapter 7 141 confirm that stages of spring migration are the most tightly scheduled of the entire year, viewed at both the individual and population scales. We ha ve previously reported extraordinarily high individual consistency of departure from New Zealand for this population (Battley 2006, Chapter 4), but schedules beca me even more finely tuned later: all eight geolocator-tracked godwits departed th e Yellow Sea in windows of only 1 ? 4 days across two years (Figure 7.2b). Population spans also decreas ed across stages of spring migration, from 28 ? 3 1 days for the initial flight from New Zealand to just 25 days for the second flight to Alaska. It is interesting that individual var iation was least for Yellow Sea departure, rather than for breeding arrival or incubation. This supports the view of long-distance flights as ?hard - wired? and endogenously programmed according to long -term environmental patterns (Piersma et al. 1990a), as opposed to breeding phenol ogies, which are more dependent on annually variable local conditions (Meltofte et al. 2007a, Smith et al. 2010). Remarkably, start of incubation differed between years by just 0 ?2 days for five of seven godwits; differences of 8 ? 18 days for two others may represent delays in mate acquisition or loss of a clutch prior to incubation. Moults and post-breeding movement s generally had population spans 2 ? 3 times those observed for spring migration (Figure 7.2). Di fferences in individual variation between autumn and spring migration were perhaps not as great as expected, given the unclear fitness benefits of arrival timing on wintering grounds. Each year of our study, 92 ? 9 6 % of colour- banded godwits arrived in a 32 -day span endi ng by 4 October, and the remainder trickled in over the subsequent four weeks. Because ou r relatively small geolocator sample failed to capture this tail of variation represented by late-arriving birds (Figure 7.2b), we do not know how much variation in New Zealand arrival is explained by l ate departures from Alaska, as opposed to delays or detours en route (Gill et al. 2009). Moults versus movements Temporal variation in moults was generally greater than in movements at both the population and individual levels (Figure 7.2); this pres umably reflects differences in both regulatory mechanisms and selection consequences of moults and movements. Migration timing has theoretical and demonstrated consequences for subsequent activities such as breeding (Alerstam and Lindstr?m 1990, M?ller 1994) an d moults (Holmgren and Hedenstr?m 1995, Barshep et al. 2011b), and timing of protracte d trans-oceanic flights (9,000 ? 1 2 , 0 0 0 km non- stop for New Zealand Bar-tailed Godwits; Gill et al. 2009, Battley et al. 2012) may have direct survival consequences, particularly if wi nd assistance is necessary for successful migration (Liechti 2006). Fitness consequences of moult timing are less clear (but see Dawson et al. 2000); moult is typically seen as flexible, with birds adjusting both timing and duration to keep to their annual schedules (Noskov et al. 1999, Helm and Gwinner 2006). 142 Chapter 7 In this godwit population specif ically, within-individual variation in timing of autumn migration carried over to moult schedules in New Zealand, but timing of spring migration was unresponsive to variation in preceding life-history stages (Chapter 6). It is surprising that timing of moults in New Zealand did not tighten with proximity to the breeding season. High individual variation in completion of pre- basic moult was as expected; like departure from the breeding grounds (by far the most flexible annu al movement in our study), initiation of this moult is strongly as sociated with cessation of breeding activity (Hahn et al. 1992, Dawson 2006, Mitchell et al. 2012), which may vary by more than eight weeks in this population, due to differences in laying d ate and breeding success (Chapter 2). However, individual variation in pre-breeding moult in itiation was greater than expected, given its presumably strict photoperiod control (Noskov et al. 1999) and proximity to the very tightly scheduled spring migration. The cause of this variation is unclear, but it had no apparent effect on subsequent plumage or migration (Chapter 6). Geolocator data included all major annual moveme nts, but we could not evaluate two moult stages that occur in the Northern Hemispher e: initiation of pre-basic moult in Alaska and completion of pre-breeding moult in Asia (Chap ter 5). If these are re spectively the most and least variable moult parameters of the year (which is plau sible), moult schedules could show significant tightening toward the breeding season with their inclusion. Problems with repeatability With increasing interest in how flexible indi viduals are in variable circumstances (e.g., climate change), there is a growing desire to de scribe the consistency with which individuals perform certain annual tasks, and to make di rect comparisons within and among studies. In bird migration literature, repeatability (intra-class correlation coefficient, r; Nakagawa and Schielzeth 2010) is becoming standard for representing the consistency with which individuals perform, but it is not precisely a measure of that; rather, it indicates how consistently individuals differ from each other. This is the variable of interest in some studies (for instance, as an indication of the upper bound of heritability; Nakagawa and Schielzeth 2010), but it unfortunately tells us little about absolute consistency, which may be of greater interest to many migration studies. The problem is that r combines population and individual variation to arrive at one value, and thus obscures two parameters that may signify fundamentally different things about the constraints and selection forces in play. This renders r values essentially incomparable, even am ong parameters within a single study. In our data, the problem is exemplified by th e repeatability of pre-breeding moult initiation among all colour-banded godwit s (Table 7.1). The very high r (0.91) implies extraordinary individual consistency in timing, but in fact the individual variation in this parameter is higher Chapter 7 143 than in migratory arrival and departure, which counter-intuitively show lower repeatability values. This apparent paradox results from the relatively high population variation in pre- breeding moult initiation; due to a wide range of moult strategies in the population (Chapter 5), individuals are more consistently differe nt from each other while being less consistent individually than in other parameters. For th is reason, we propose that authors should not report repeatability values without presenting accompanying descriptions of absolute variation at the population and individual levels, as we have in Figure 7.2. This will be a step toward understanding the biological significance of within- and between-study differences in repeatability. Our data also demonstrate the problem of comparing r values derived from different sample sizes. Both our geolocator-only and larger colo ur-banded samples indicate that moults in New Zealand are generally less repeatable than movement s (Table 7.1), but thr ee parameters that had strongly non-significant r values in the former dataset appeared highly repeatable in the latter dataset. Beyond the obvious ly greater statistical power afforded by larger samples even when effect sizes are identical, r values also naturally increase as more of the total population variation is described in the sample (because r is driven by the ratio of population to individual variation). This is a particular problem for comparing r values currently available in migration literature, because multi-year individu al studies made possible by recent advances in geolocation and satellite-telemetry ty pically contain small samples (Alerstam et al. 2006, Vardanis et al. 2011) that cannot describe total populatio n variation as completely as studies based on large samples of marked indi viduals (Rees 1989, Battley 2006). Individual versus population variation The greatest understanding of variation, and th e selection forces implied by it, comes from considering the two components of repeatability (population and individual variation) separately. Because the fitness of migratory animals can be particularly sensitive to fine-scale timing of certain life-history stages, migratio n research has been especially focused on describing ?optimal? timing and assessing the propensity of individuals to achieve it. Variation in an individual?s performance results from the interaction of a number of factors, including: (1) environmental variation; (2) fitness conseque nces of this variation; (3) availability of relevant cues; and (4) the degree of flexibility afforded by control mechanisms and available resources. Thus, individual variation comprises both what an individual should do and what it can do. Concluding that individual consistency implies strong selection for that trait is somewhat simplistic, as consistency may also result from failing to change when circumstances call for it; this may persist unless it is strongly maladaptive. For example, despite annual variation in breeding phenologies, spring movements of long-distance migrants may be very consistent largely because relevant cues are not available from their non-breeding 144 Chapter 7 grounds. Thus, rigidity, although not precisely optimal, is the best long-term strategy available to them. As we have defined it, population variation necessarily contains all within-individual variation, but additionally includes between-i ndividual differences. Thus, the difference between population and individual variation essentially represents the system?s tolerance for persistent differences in performance (e.g., due to available resources or intrinsic ?quality?) and strategy (e.g., by sex, age, or geograph ic region). For New Zeala nd Bar-tailed Godwits, timing differences among individuals contain 3 ? 4 weeks of variation related to breeding site phenology: due to relative timing of spring thaws, godwits breeding in northern Alaska migrate later than southern breeders in both spring and autumn (Chapter 2), and timing of moults show roughly corresponding differences (Chapters 5 ? 6 ). Accordingly, minimum population spans of nearly a month are expected for all stages, even before additional variation based on individual performance is considered. For spring movements to the breeding grounds, population spans of 25 ? 3 1 days approached this minimum expected window, reflecting very precise scheduling at the individual level. By contrast, some population spans found in moults (up to 78 d) are greater than accounted for by breeding latitude and individual variation combined, and th us suggest greater tolerance for strategic differences within the population. For godwits, profound sexual dimorphism and within-sex variation in size and plumage (Chapter 3) give rise to a wide range of moult schedules in terms of duration and extent (Chapters 5 ? 6 ) , whereas no significant sex differences in migration timing have been observed in this population (Battley 2006, Chapters 4 ? 5 ). Future directions Although we expect that progressiv ely tightening schedules leading up to breeding are general to migratory birds, some of the patterns we demonstrate may be peculiar to extreme long- distance migrants, or to this particular godw it population. For instance, the autumn migration of Alaskan Bar-tailed Godwits is uniq ue in that it typically covers 11,000 ? 1 2 , 0 0 0 km in a single non-stop flight of 8 ? 1 0 days (Gill et al. 2009, Battley et al. 2012) , the timing of which depends primarily on breeding latitude (Chapter 2) and the occurrence of favourable weather systems (Gill et al. 2009). These factors contribute to a s urprisingly rigid autumn migration: the median individual between-year difference was just 4 days and most of the population arrived in New Zealand in a span similar to that of spring departure ( Figure 7.2b). We expect differences between spring and autumn to be much more profound in systems in which timing or duration of autumn migration is strongly influenced by extent of prior breeding investment (Barshep et al. 2011a) or annual variation in conditi ons at staging or stopover sites (Weber et al. 1998b). In terms of moults, birds wintering in the northern temperate zone may be more time-constrained than godwits in New Zealand, because mid-winter periods of inclement Chapter 7 145 weather and low prey availability may be incompatible with moult (Holmgren and Hedenstr?m 1995). Year-round comparative st udies among species along a continuum of time and energy constraints will greatly enhance our understanding of both evolutionary adaptations to the migratory lifestyle at the species level and smaller-scale variation that may result from differences in individual personality and experience. The aim of most contemporary studies of temporal variation is to ascertain levels of flexibility to address stochastic or directional environmental change. Cross-seasonal and cross-species comparisons of temporal variation are certainly a step toward understanding flexibility inherent in systems and identifying specific stages prone to critical time-constraints and seasonal carry-over effects. Fo r migratory birds, however, there has been a general lack of relevant long-term studies; due to the logistical difficulties of tracking individual migrants, few studies have contained individual data spanning more than 2 ? 3 years. Unfortunately, because both individual behaviour and environmental conditions are more likely to be consistent in consecutive years than over longer periods (Catry et al. 1999), short-term studies may misrepresent the flexibility of a system. For our study, individual and population spans of variation would necessarily increase with additional years of data, but the extent of increase should vary among stages. For instance, we wo uld predict start of incubation, which varies directly with local conditions on the breeding grounds, should show a greater long-term increase in overall variation than long-distan ce movements, which vary little with annual conditions. In a long-term study, temporal var iation in certain annual events will respond to persistent environmental change while others may not, and this will help identify the life- history stages most prone to critical timing mismatches. 146 Chapter 7 147 Chapter 8 Synthesis: an evolving view of long-distance migration 148 Chapter 8 Key findings of this thesis The aims of this research were to determine the factors leading to both persistent and ephemeral differences in plumage and migration timing among individual Bar-tailed Godwits, and to identify possible constraints or bottleneck s in the annual cycle. Each of six research chapters addressed different key questions arising from these central goals. In Chapter 2 , I found that an individual?s migration schedule was linked to the location of its breeding site in Alaska, with northern breed ers migrating later than southern breeders on each stage of both northbound and southbound migrations. In a ddition to explaining the 4 ?5 weeks of variation in timing of migratory depa rture, this showed that schedule differences observed in New Zealand persist throughout th e six months that godwits spend in the Northern Hemisphere. Adopting an Alaskan perspective for Chapter 3 , I found that the population was structured geographically in body size and plumage duri ng the breeding season. With increasing breeding-site latitude, godwits of both sexes we re smaller, and extent of male breeding plumage increased; however, female plum age was most extensive at mid-latitudes. Interestingly, this population stru cture did not carry over to the non-breeding season, as individuals from all regions of Alaska ap pear to mix at New Zealand sites. In Chapter 4 , I showed that godwits have extraordina rily consistent individual schedules when departing New Zealand on migration and that most of the exceptions to this pattern can be explained by birds avoiding unfavourable winds at the start of the northbound journey. A surprising finding was that some godwits departed several days earlier than expected when particularly favourable winds occurred, im plying that they were physically ready for migration well before their customary departure dates. Chapter 5 revealed that individuals with more exte nsive breeding plumage achieved it by spending more time moulting in New Zealand pr ior to northbound migration, rather than through faster moult rates or greater investment in moult during their stopover in the Yellow Sea. However, males and females had very different strategies: despite their more extensive breeding plumage, males performed a much high er proportion of their moult in New Zealand, so that females were required to perform twice as much additional moult in Asia as males. Perhaps the most surprising revelation in my study derived from Chapter 6 : variation in timing of arrival after the epic flight from Alaska had no apparent influence on an individual?s ability to prepare for the following northbound migration. Late-arriving birds (by 7 ? 34 days) partially compensated for this by moulting their wing feathers faster, but the delays had no Chapter 8 149 effect on an individual?s timing of pre -breeding moult, extent of breeding plumage at departure, or timing of departure from New Zealand. In Chapter 7 , I showed that temporal variation at bo th the individual and population levels generally decreased through successive life-hist ory stages leading up to arrival on the breeding grounds. Scheduling of migratory mo vements was increasingly rigid from post- breeding events in Alaska through arrival on breeding grounds the following year, but this was not true among successive moults, wh ich were much more loosely scheduled. Northbound migration was even more precisely timed than breeding, consistent with the former?s presumed greater reliance on endogenous programming. Overall, these findings show that Bar-tailed Godwits are very much individuals, and their behaviour, morphology, and schedules cannot be viewed simply as deviation from optimum values. Most of the inter-individual variation observed in New Zealand had its roots in the Alaskan breeding season, indicating that a full annual-cycle perspective is required to understand patterns in any particular season. Furthermore, the precision with which godwits conduct their annual routines, while still demo nstrating flexibility to address unpredictable circumstances, challenges us to reconsider the view of extreme long-distance migrants as severely constrained organisms operating at the limits of their capabilities. Regulation of the migrant annual cycle My methodological approach to this resear ch was decidedly observational: I sought to understand long-distance migration by simply view ing individual behaviour and moult with a resolution that had never previously been achieved. This approach yielded many surprising and important revelations, but my data cannot di rectly address the mechanisms underlying the observed patterns. To understand the significance of the godwit?s extraordinary annual routine, we must consider how bird s in general manage their schedules. This rich subject is replete with mysteries, but it is clear that regulation of annual cycles involves cooperation between internal programmes and external information gleaned from the environment. Virtually every animal with a lif espan longer than a year displays a cyclical pattern of annual activities (such as breeding an d moult) occurring repeatedly in a certain order (Jacobs and Wingfield 2000) . Unexpectedly, an individual?s innate circannual rhythm does not automatically occur in a regular 365 -day cycle, but must be constantly corrected to stay in line with external seasonal patterns (Gwinner 1990). Conven iently, day length, or ?photoperiod?, is an entirely reliable external cue, providing a perfect calendar with which to 150 Chapter 8 synchronise endogenous rhythms with the natural year. Photoperiodic control of the annual cycle of birds has been firmly established through experimentally demonstrated links with gonadal development, moult, and migratory be haviour (reviewed in Da wson 2008) , but it remains unclear whether absolute day length or change in day length is the more important cue. Photoperiod regulates physiological processes through an intricate relationship between photoreceptors in the avian brain (which have yet to be described) and the endocrine system (Dawson et al. 2001). Of course, resources and environmental condi tions are not completely predictable, and so an absolutely rigid annual schedule is not always beneficial. For example, if spring arrives earlier or later than expected, a bird should adjust its breeding schedule to best exploit the peak of resources necessary for brood development and survival. In this case, photoperiod information can be supplemented with locally relevant short- term cues, such as temperature, rainfall, or food abundance (Dawson 2008) , to refine the timing of specific events. In general, it appears that such fine-tuning must still occur in a framework ultimately co ntrolled by photoperiod (Gwinner 1996) ; i.e., endogenous programming pr ovides a window in which the event can occur, and other cues influence the exact timing within that window. Plasticity in timing varies among annual events and within and among species in ways dictated by natural selection. This will influence the extent to which rigid photoperiodic control is tempered by other information. For example, a short-distance migrant may be able to roughly predict the phenology of its breeding site while still at its winter site, due to a correlation between climatic conditions at the two locations (e.g., J?rvinen 1989). This bird should accept local information to inform its decision about when to migrate, and we may expect significant annual differences in timing of both pre-migratory fuelling and migratory departure. By contrast, a bird that winters in the opposite hemisphere from its breeding site cannot expect reliable cues about upcoming breed ing phenology from its non-breeding site. This bird should ignore local environmental information and prepare for migration according to long-term patterns in breeding-site phenolo gy, and thus may rely entirely on photoperiod (Gwinner 1996) and show less flexibility in sc hedules (Both and Visser 2001). In both cases, the exact timing of migratory departure may resp ond to cues at yet another temporal scale, which is daily or hourly conditions. For exampl e, initiating migration in a severe headwind is a poor tactic (Liechti 2006), and so even the rigidly-prescribed departure plan of the long- distance migrant should show small-scale plasticity in this regard. Because each life-history stage has its own set of available cues, mechanistic constraints, and selection processes, flexibility in timing varies dramatically throughout the year (Chapter 7). Most of our knowledge regarding regulation of specific stages in birds comes from passerines Chapter 8 151 ( Dawson 2008), which perform w ell in captive experiments, but some relevant work has been conducted with shorebirds (Cadee et al. 1996, Piersma 2002, Piersma et al. 2008). There has been no formal study of how specific stages are regulated in Bar-tailed Godwits. Based on general knowledge of regulation in birds ge nerally and biology of Bar-tailed Godwits specifically, Table 8.1 summarises, in broad ter ms, the factors potentially involved in the timing of key events throughout the year. Post-breeding events (movement to staging grounds and initiation of pre-basic moult) are expected to be the most te mporally variable, as they are associated with the cessation of breeding, which contains substantial variation based on breeding success. However, there may be very different regulatory mechanisms at work, as post- breeding dispersal is a behavioural ?decision? based on the lack of young to care for, and pre-basic moult is a physiological response related to the endocrinology of gonadal regression (Dawson 2006). Although wing moult is generally considered an integral part of pre-basic moult (Humphrey and Parkes 1959, Howell 2010 ) , godwits and other long-distance migrants illustrate that the two are somewhat modular in both timing and control mechanisms (Piersma et al. 2008). Initiation of wing moult ap peared to be primarily governed by timing of arrival on non-breeding grounds (Chapter 6), and so may be related to hormonal changes associated with the cessation of migratory flight. Arguably, initiation of migratory flights should be the least flexible, as they are under more strict ph otoperiodic control and may vary primarily with daily weather conditions. We saw in Chapter 4, however, that even the most rigidly-scheduled Table 8.1 Factors thought to influence the timing of key events throughout the year for New Zealand Bar-tailed Godwits. Seasonal local cues include weekly or monthly weather, food availability, etc. Daily local cues include hourly or daily weather, behaviour of conspecifics, etc. Regulation of start Regulation of end Primary cue Secondary cue Primary cue Depart breeding end breeding ? ? Pre-basic moult end breeding ? photoperiod Flight Alaska?New Zealand photoperiod daily local weather en route Wing moult arrive NZ ? photoperiod Pre-breeding moult photoperiod ? photoperiod Flight New Zealand?Yellow Sea photoperiod daily local weather en route Flight Yellow Sea?Alaska photoperiod daily local weather en route Arrive breeding seasonal local ? ? Breeding seasonal local daily local breeding success 152 Chapter 8 movements may show surprising flexibility in extreme weather. On northbound migration, final movements to breeding sites after brief stopovers in southwestern Alaska should vary more than the major migratory flights, because godwits may receive local cues to breeding- site phenology, which may vary annually by days or even weeks (Meltofte et al. 2007a, Tulp and Schekkerman 2008). The in itiation of a clutch of eggs may vary still more, due to the additional influence of unpredictable social factors such as acquiring a mate and performance of pre-copulatory rituals. Photoperiod also has a role in the way some stages are performed after their initiation, and thus influences the timing of termination as well. Experimental changes in photoperiod have been linked changes in moult rate (Hall and Fransson 2000, Dawson 2004) ; it appears that photoperiod is the cue late-mou lting birds use to adjust mou lt rate to make up time (as godwits did in Chapter 6). In some species, individuals that are running late on migration are similarly urged by photoperiod to ?hurry up? during later stages of migration (e.g., Fransson 1995). Increasing the speed of migration may be possible for short-hop migrants that incorporate several staging episodes into their journey, but appears unfeasible for those, like godwits, that migrate in only one or two major flights. Once a godwit embarks across the Pacific from Alaska, it may have little control of its own speed. After initiation of such flights, weather encountered during the flight is probably the greatest source of additional variation before arrival (Gill et al. 2009, Shamoun-Baranes et al. 2010). Most of this generally agrees with the patterns of temporal variation I found across the annual cycle of godwits (Chapter 7). One stage is puzzling, however: pre-breeding moult. Presumably, initiation of this moult is associated with gonadal maturation in advance of the breeding season (Peters et al. 2000). In some passerines, non-photoperiodic cues have been implicated in gonadal maturation (Dawson 2008) , suggesting that not just the act of breeding but the readiness to breed can be influenced by environmental variation. However, I have suggested that relevant phenological cues are unavailable to non-breeding godwits, and have shown that pre-breeding moult is unresponsive to timing of both migratory arrival and wing moult (Chapter 6). Despite this, intra-individu al variation in initiation of pre-breeding moult was much greater than that of northbound migration or breeding itself (Chapter 7). Furthermore, this temporal variation (up to 28 days) had no apparent consequence, as it did not affect extent of breeding plumage at departure from New Zealand (Chapter 6). The mechanisms and functional significance of this await explanation. It is possible that the first evidence of pre-breeding moult (visible emergenc e of new breeding feathers) is not a perfect indicator of the initiation of the moult process. Chapter 8 153 How are individually-optimised schedules maintained? A separate question from how annual routines are ge nerally regulated is how distinct routines within a population are maintained. External cues may provide a perfect calendar and roughly predict future resource phenologies, but an individual must still determine how this information applies to its own specific strategies. For simplicity, let?s assume that photoperiod alone informs migration timing and that all godwits breeding at a given latitude in Alaska migrate at precisely the same time. Obviously, all the birds at the same New Zealand site are exposed to the same external information (e.g., ?It is March 13?), but respond differently; those intending to migrate on March 13 do so, while later migrants remain and await their turn. In this example, all birds receive the sa me cue, but somehow either perceive the cue differently or employ different thresholds for translating the information into action. Now imagine the thousands of ?March 13? godwits that are scattered at different New Zealand sites across 12? of latitude. Birds at different sites ex perience different day lengths, but are able to judge date correctly by somehow accounting for their own geographic position. That they do this is supported by the very similar population-level departure windows observed at different sites across New Zealand (Battley 1997, 2006, Chapter 4). These same issues occur in Alaska as well: during the breeding season, birds ar e scattered across a range of latitudes with different breeding phenologies and experience ve ry different day lengths (including 24-hour daylight in the northern half of the range). Af ter breeding, the entire population congregates in a small geographical area before southbound migration, but retains individual departure schedules despite receiving similar photoperiod cues. Setting aside how a bird knows its own position on the globe, how are godwits ?programmed? with their individual schedules? Certainly we expect there to be a genetic component. Heritability of migratory traits such as timi ng and direction has been demonstrated in a number of bird species (Pulido 2007) , although the genetic basis for individual variation in migration remains somewhat poorly studied (van Noordwijk et al. 2006). In Bar-tailed Godwits, migration timing, body size, and plum age all varied geographically within Alaska, and this suite of potentially co-evolved traits co uld be inherited together, or each could result from independent underlying mechanisms. Ge ographic variation in genotype can be maintained by ?isolation -by-dista nce? in highly structured populations or by geographic clines in polymorphic allele frequencies even when gene flow is high (Endler 1977). Particularly intriguing is the discovery of ?clock genes?, providing a molecular basis for the timing of annual events (Tauber and Kyriacou 2005). Polymorphism in clock genes has been preliminarily identified in Bar-tailed Godwits (A . Fidler pers. comm.) and could hold the key to understanding individual schedules in this population. Latitudi nal clines in clock gene polymorphisms have been demonstrated in birds (Johnsen et al. 2007), fish (O ? Malley 154 Chapter 8 and Banks 2008) , and flies (Costa et al. 1992), and there is eviden ce that stable clines can be maintained by selection for local adaptation, as opposed to genetic drift (O ? Malley et al. 2010). Alternatively, persistent individual differences need not be inherited in the traditional sense. The expression of a single ge notype may be altered by epigenetic factors during cellular development (Jaenisch and Bird 2003) or by later facultative phenotypic responses to the environment (Piersma and Drent 2003) . Thus, a godwit?s early environment may permanently imprint upon it certain behavioural or physiological routines and the sensory mechanisms required to maintain them. Experiments with mice have shown that photoperiods experienced perinatally can permanently entrain an individual?s responses to seasonal photoperiod changes (Ciarleglio et al. 2011). This provides a potential mechanism for godwits hatched at a given latitude to return there at the appropriate time every year. It is also possible that some aspects of migratio n are learned. Particularly in the case of specific migration routes and stopover sites, a bird may engage in some level of experimentation and generally repeat what has worked in the past. However, stable social associations did not appear to influence timing of Bar-tailed Godwit migration, despite the gregarious nature of the birds. Scolopacid fa mily groups do not persist in the non-breeding season; members of breeding pairs may winter hu ndreds or thousands of kilometres from each other (Nebel et al. 2002, Gunnarsson et al. 2004), and godwit adults migrate separately from juveniles (McCaffery and Gill 2001). Subg roups in non-breeding sandpiper populations generally appear to represent ephemeral and fluid daily associations, rather than a persistent social organisation (Myers 1983, Conklin and Col well 2008). Accordingly, godwits in my study did not migrate consistently with the same other birds each year, despite each being very consistent in departure date (Chapter 4). The d ecision to depart on migration was very much an individual action. How ?close to the edge? are long-distance migrants? One of the key tenets of ecology is that organi sms are constrained. Existence is full of trade- offs, and there is always an array of obstacles standing between an individual and its optimal performance. Constraints, which may be tempor al, environmental, physiological, or social, have been the cornerstone of wildlife research and conservation because of their central role in natural selection and the success of both individuals and populations. Migratory birds are often portrayed as models of constrained organisms, because their lifestyle of travelling to exploit seasonal resources naturally impose s substantial temporal and energetic demands Chapter 8 155 (Alerstam and Lindstr?m 1990) that require a finely-tuned suite of behavioural and physiological adaptations to overcome (Hedenstr? m 2008). These demands only become more profound as migration distance increases. Lo ng-distance migrants typically breed at high latitudes, where brief, intense resource blooms make timing of migration and breeding even more critical (Tulp 2007). To make their epic flights, they ac cumulate large fuel stores and undergo other physiological changes to optimise flight performance (Piersma and Gill 1998, Battley et al. 2000, Landys-Ciannelli et al. 2003). They often cross vast, inhospitable barriers such as oceans, deserts, or mountain ranges without the option of resting or refuelling along the way (Henningsson and Alerstam 2005) , and may enter circumstances of unknown predation danger and resource availability. These factors combine to create an annual routine with seemingly little room for error. At the extreme end of this continuum, New Zealand Bar-tailed Godwits seem to approach the limits of physical performance (Hedenstr?m 20 10). But just how constrained are they? By considering their behaviour and performance in certain stages throughout the year, we can ask if and when godwits conform to expectations for birds operating close to their limitations. Annual survival First of all, they have very high annual s urvival. Mark-recapture studies have recorded individuals of the Alaska-breeding Bar-tailed Godwit population at minimum ages of 21 years in New Zealand (A. Riegen pers. comm.) and 20 ? 2 4 years in southeastern Australia (C. Minton pers. comm.). There is no reason to think this approaches the limit. Bar-tailed Godwits in the United Kingdom, where banding has occurred for a much longer period, have reached 33 years of age (WWRG 2011) , although this population ( L. l. lapponica ) has a much shorter migration. At my study site, the aver age annual return rate of individually marked godwits was 90% (86% in 2008, 93% in 2009, and 90 % in 2010). When you co nsider that this figure includes all possible causes of disappearance, such as accidents, hunting, depredation, and simply choosing another non-breeding site, it is clear that very little mortality results from godwits being physically overcome by their migrations. Three times per year, these birds fling themselves over vast spans of open water, with only their wings and fuel stores to fend off whatever meteorological challenges the Pacific Ocean can throw at them. And yet they seem to do just fine. Flight distance The longest non-stop avian flight ever recorded was that of the famous ?E7?, a satellite -tagged female godwit from the Firth of Thames that f lew 11,6 90 km on a r easonably direct course from southwestern Alaska back to the s ite of her capture in New Zealand (Gill et al. 2009, Battley et al. 2012). This is impressive, but actually is fairly close to the minimum necessary 156 Chapter 8 to travel from Alaska to New Zeala nd in a single flight. There is plenty of indirect evidence that tens of thousands of godwits make longer flights than this every year, and in fact fly farther than they have to. Most geolocator-track ed godwits in my study (7 of 10 in 2008, 9 of 9 in 2009) appeared to fly directly to Foxton from Alaska, a trip approximately 400 km longer than E7?s flight. Even godwits that stop somewhere else en route do not always do so in order to shorten the trip: one bird in my stud y routinely travels hundreds of kilometres past the North Island on southbound migration and spends a month or more at the Avon-Heathcote estuary near Christchurch, before returning to Foxton. If the southbound flight approached the maximum capabilities of godwits, we would expect many or all South Island birds to stop for rest and recovery at North Island sites before continuing south. However, despite enthusiastic colour-band resighting efforts every year at major northern sites (e.g., Manukau Harbour and the Firth of Thames) during migratory arrival, only a handful of godwits colour-banded on the South Island are ever encountered at these sites (Battley et al. 2011). I have never seen one at my study site in four years. It is also wort h noting the location of the start of the southbound flight. As most of the population does every year, E7 departed Alaska from the southwestern coast of the Yukon-Kuskokwim Delta, rather than from apparently suitable sites farther south on the Alaska Peninsula, approximately 400 km alo ng her flight path. None of this suggests that godwits are attempting to minimise the length of this ex traordinary flight. Carry-over effects Surprisingly, godwits also rarely choose to not to migrate. When I first conceived my study, I thought that close monitoring of godwits throughout the non-breeding season would help me explain why certain individuals did not migrate each year, thus identifying the system?s effective limits. For example, I might be able to trace a departure ?failure? back to a very late arrival or moult, and this would suggest a bi rd had insufficient time or energy to prepare for migration. Four migrations later, not a single adult godwit ever failed to depart my study site. As detailed in Chapter 6, even go dwits arriving or moulting 30 ?4 5 days later than expected still made timely departures the following season without any compromise to their breeding plumage. Surely, there must be limits. Every y ear, small numbers of apparently adult godwits remain in New Zealand during the northern su mmer (P. Battley unpubl. data), and these are presumably birds whose health or advanced age has caused them to forego breeding. But these are clearly in a small minority?most godwits simply get the job done every year. The precision with which they perform it (Chapters 4 and 7) additionally suggests that they have little trouble doing it well. Fuelling Other signs that the performance of godwits may no t be tightly constrained are more subtle, but perhaps no less instructive. A substantial body of literature has focused on the idea that Chapter 8 157 birds accumulating fuel for migration must balance this task with the costs of carrying the extra weight (Hedenstr?m and Alerstam 1992, Witter and Cuthill 1993, Weber et al. 1998b), which include reductions in flight performance and ability to escape avian predators. This predicts that birds should fuel as quickly and as late as possible, to re duce the amount of time spent in heightened danger. An even larger body of literature has focused on the difficulties of attaining optimal mass for migration, due to ephe meral peaks of fuelling resources (Atkinson et al. 2007), direct competition with conspecifics or other species (Schneider and Harrington 1981, Moore and Yong 1991) , and physiological co nstraints such as digestive bottlenecks (McWilliams and Karasov 2005). These seem lik e particularly serious concerns for birds fuelling for a 9,000 ? 1 0 ,0 0 0 km non-stop flight. However, no ne of them appeared to seriously trouble godwits in my study. New Zealand lacks an y important avian predators for shorebirds, and so there may be little cost to carrying ex tra mass well before migration. Perhaps as a result, godwits fuel very slowly, starting as ear ly as late December (Anderson 2003), which is approximately three months before northb ound migration. I did not specifically measure foraging effort in my study, but increase s in foraging leading up to migration (Zwarts et al. 1990a) were not conspicuous at the estuary. Throughout March, many godwits routinely stayed at high-tide roosts even as the mudflat was exposed by the falling tide, and could often be found roosting even at low tide, suggesting they were not under great pressure to maximise their foraging opportunities just before mi gration. Consistent with this, some godwits appeared to depart 1 ? 2 weeks earlier than usual to tak e advantage of unusually strong tailwinds (Chapter 4); this implies they were sufficiently fat well before their expected departure dates. At this estuary, a one-year stud y of the godwits? main polychaete prey Nicon aestuariensis revealed a moderate decline in biomass from January ? March (Powell 2011), but natural fluctuations or reduction by foraging shorebirds did not appear sufficient to threaten a godwit?s ability to fuel for migration. The situation may be different during fuelling for southbound migration, as southwestern Alaska features many avian predators, such as falcons Falco spp. and jaegers Stercorarius spp., which routinely attack shorebirds co nspicuously assembled at coastal mudflats. Regardless, godwits may fuel quite slowly in A laska as well, as geolocator-tracked birds moved to post-breeding staging grounds as early as 88 days before they departed Alaska. However, those that appeared to incubate eggs until hatching moved to staging grounds only 40 ? 6 5 days before departure. Successful breeders may begin fuelling later and be required to fuel at faster rates than failed breeders, or they may begin gaining mass while still attending broods on their breeding grounds. In any case, breeding success did not appear to significantly influence a godwit?s ability to fuel sufficiently, as departure timing was unaffected (Chapter 2 ). Furthermore, the demands of fuelling fo r the >11,00 0 km sout hbound flight do not 158 Chapter 8 preclude a concurrent moult, as all godwits arri ving in New Zealand ha d already finished 50 ? 9 0 % of their pre-basic contour feather moult (Chapter 5). These observations may testify to very high productivity of the mudflats they use for fuelling; it appears that nearly all Alaska- breeding Bar-tailed Godwits (this study and Battley et al. 2012) congregate before migration near Kuskokwim Shoals (59.83? N, 164.13? W ) along a stretch of coastline <200 km long. Migratory birds routinely perform feats pr eviously thought beyond their capabilities (Hedenstr?m 2010) , and force theoretical models to change to fit empirical observations they couldn?t otherwise explain (e.g., Pennycuick and Battley 2003). We may never find a more extreme example of endurance flight than New Zealand Bar-tailed Godwits, simply because the geography of the earth doesn?t present many scenarios in which such a flight is required ( Hedenstr?m 2010). This does not necessarily suggest that birds are demonstrating the greatest performance that physics and physiology will allow. Viewing their annual routines as I have in this study, I failed to identify a po int at which godwits were clearly pushed to the limits of their capabilities. Of course, it wo uld require long-term data on survival and reproductive success to honestly evaluate how their performances translate to overall fitness. But I believe Bar-tailed Godwits encourage us to revisit the traditional view of long-distance migrants as severely constrained creatures perpetually operating at maximum capacity. By monitoring changes in any of the life-hist ory parameters discussed above, we might be alerted if changing circumstances actually push godwits to their limits. Unfortunately, we may have many future opportunities to assess how close migratory birds are to critical tipping points, as we witness their responses to co ntinuing climate change (Crick 2004) and human- induced modification of habitats. Bar-tailed Godw its may be particularly vulnerable to negative effects of climate change if rigid, endogenously programmed migration schedules preclude their ability to adapt to changing phenologies at breeding or fuelling sites (e.g., Both and Visser 2001). Also, predictable long-term weath er conditions may have contributed to the evolution and persistence of certain migratory strategies (Liechti 2006). If changes in prevailing wind patterns reduce the tailwind assistance available to godwits during their flights, they may require more fuel for migration or suffer reductions in flight range. Although the remote and largely inaccessible breeding range of this godwit population is relatively free from direct human impacts, critical staging areas in the Yellow Sea are under great threat of development and habitat loss (van de Kam et al. 2010), which could severely constrain the migrations of many shorebird species in the East Asian-Australasian flyway. Chapter 8 159 The role of individual quality The relative ?quality? of individuals is a topic closely associated with constraints on optimal performance. An individual?s condition, whether this is an innate or acquired state, will naturally influence its ability to survive and reproduce in whatever circumstances befall it. In birds, two relevant phenomena are so commonly observed that they have become axiomatic: higher-quality individuals have better breeding plumage, and they also migrate earlier. In mate competition, individuals of the competing sex typically display behavioural or physical attributes that signal their quality to potential mates and ri vals. In birds, the size and intensity of plumage ornaments have been demonstrated to honestly signal individual quality, through demonstrated correlations with immunocompetence (Folstad and Karter 1992, Dufva and Allander 1995), parasite loads (Figuerola et al. 2003), parental investment (Keyser and Hill 2000), social status (Lyon and Montgomeri e 1986) , and foraging ability (Hill and Montgomerie 1994, Senar and Escobar 2002, McGraw et al. 2005). In migratory birds, these ornaments may advertise that an individual is sufficiently fit to invest in energetically expensive plumage even after the considerable costs of migration itself. Such individuals will presumably be desirable reproductive partners, either by passing on good genes to the young, or by excelling at defense of territories, nests, or broods. Another way birds demonstrate their quality is by breeding at the right time. High-qu ality individuals (variously defined in terms of age, breeding plumage, parasite loads, or imm unocompetence) typically arrive earlier on the breeding grounds (Flood 1984, Francis and Co oke 1986, Hill 1988, M ?ller 1994, Potti 1998). Plumage and migration timing have been correlated explicitly with breeding success or survival frequently enough that they are often cons idered reliable indices of relative individual quality or condition. This may be appropr iate when such relationships are empirically established for the particular study system, and can allow powerful inferences across seasons (see next section for one such example). By contrast, my study illustrates that inferring individual quality from plumage or migratio n timing on non-breeding grounds can be completely spurious. In New Zealand, Bar-tailed Godwits of each sex varied substantially in departure date (Chapter 4) and plumage at depart ure (Chapter 5), but bot h of these attributes were strongly correlated with latitude of their individual breeding sites in Alaska (Chapters 2 ? 3 ). There is no reason to be lieve that quality or condition of individuals varies systematically across more than 12? of latitude in Alaska, part icularly when this pattern cannot be linked to differences in non-breeding habitat quality (e.g., Gunnarsson et al. 2005, Studds and Marra 2007). Furthermore, quality-based predictions based on plumage and migration timing would be contradictory, as southern breeding birds migrate earlier whereas northern breeding birds 160 Chapter 8 have more extensive breeding plumage. It s eems that plumage and migration timing in New Zealand tell us very little about the relative quality or condition of individual godwits. An intensive study of individuals in Alaska shou ld be more instructive. At particular breeding locations, individuals vary in both arrival time (McCaffery et al. 2010) and plumage (Chapter 3), and this may translate to differences in breeding success. However, these presumed relationships await verification in this species. In particular, the function of ventral breeding plumage of Bar-tailed Godwits is unclear. It seems intuitively obvious that the striking red ventral plumage of most male godwits must function to advertise status to mates and rivals, and the evolution of a pre-supplemental moult (Chapter 5) implies strong selection for this trait. To date, however, there is no direct evidence that females use plumage to choose mates, or that plumage in either sex is related to social status, quality of breeding site, or reproductive success. Furthermore, attempts to link Bar-tailed Godwit plumage to other fitness measures such as body condition, parasite loads, or survival have been largely inconclusive or contradictory (Piersma and Jukema 1993, Piersma et al. 2001, Drent et al. 2003, Battley and Piersma 2005, Battley 2007). Understanding the significance of godwit plumage is complicated by geographic variation in Alaska (Chapter 3), the reasons for which remain obscure, but which presumably results from a grad ient in habitat or sexual selection, or both. It is intriguing that in Black-tailed Godwits L. limosa in the Netherlands, males with less plumage ornamentation appeared to have higher fitness than more colourful males (Schroeder et al. 2009). This illustrates that relationships b etween plumage and fitness can be more complicated than, or even opposite to, our expectations. One argument against plumage as a reliable indi cator of quality in godwit s is the extraordinary consistency of individual plumage on departure from New Zealand (Figure 8.1; Chapter 6, Battley 2006). In fact, individual godwits were not just consistent in their plumage scores, but in the specific feathers that were moulted each year; many birds could be identified by distinct patterns of both ventral and dorsal pre-breedin g moult prior to departure (Figure 8.2). If individuals were competing to display the most impressive plumage, they should take every opportunity to perform more ex tensive pre-breeding moult. Surp risingly, annual variation in moult initiation (up to 28 days) did not result in differences in plumage on departure; each bird appeared to have a prescribed extent of moult to perform in New Zealand, regardless of the precise scheduling of it. Sin ce godwits resume moult at staging sites in the Yellow Sea, it is possible that the extent of additional moult cond ucted there is condition-dependent, and thus reflects an individual?s migratory quality (Piersma and Jukema 1993). Currently, there are no data from Alaskan breeding grounds to address annual variation in the ultimate breeding plumage of individuals. Chapter 8 161 Figure 8.1 Plumage (% breeding plumage in all body regions; see Chapter 5 for scoring details) at departure from New Zealand for 34 godwits with 3 years of data. Individual repeatability (Lessells and Boag 1987) of plumage score = 0.98. Birds 1?18, 21, and 22 are females, and the rest are males. One might expect the impacts of variation in indi vidual quality or condition to be particularly profound in long-distance migrants, due to the un forgiving nature of their annual routines. I did not attempt to directly assess condition in this study, but I found that most inter-individual variation appeared attributable to individually optimised strategies based on sex, size, and breeding location. Furthermore, my data alo ne do not support any condition-dependence hypotheses to explain godwit behaviour or mo ult. Why would this be true? Are Bar-tailed Godwits a race of unstoppable super-migrants? Where are all the low-quality godwits? 162 Chapter 8 Fig 8.2 Plumage of one male Bar-tailed Godwit (bird #26 in Figure 8.1) before migratory departure in three successive years: (a) 11 March 2008; (b) 13 March 2009; (c) 1 March 2010. The bird?s median departure date was 15 March. Photo in (c) by Phil Battley. My hypothesis is that the system simply does not tolerate low-quality individuals, which are weeded from the population very qu ickly. This first occurred to me when I realised that the juvenile godwits arriving in New Zealand around early Octo ber (Figure 8.3), having just accomplished an epic trans-Pacific flight with no experience or safety net, had all still been folded up inside their eggs, 12,000 ?1 4 , 0 0 0 km away, just three and half months previously. Consider what had to occur in that short period. First, they hatched and began foraging for themselves within hours. Within about 30 days, they grew to ne ar adult size (an increase in mass of ~1,000% ) while manufacturing a complete set of flight feathers capable of travelling non-stop to Australasia. Next, they found th eir way to staging grounds in southwestern Alaska, up to ~1,200 km away, without any help from the parents who had already abandoned them. Once there, they nearly doubled their w eight in about two months, approximately the same time it takes for an adult with an alrea dy fully-grown bill and at least two years of foraging experience. Finally, they set off from Alaska, and flew 7 ? 9 days straight before touching the earth again. A more efficient method for immediately removing low-quality individuals from a population could hardly be conceived. For most migratory bi rds, the greatest mortality rate is certainly in the first few months of life (Sullivan 1989, Anders et al. 1997, Thomson et al. 1999) , but most (a) (a) (b) (c) Chapter 8 163 Figure 8.3 Bar-tailed Godwits at the Manawatu River estuary on 2 October 2008. The juvenile on the left had arrived on southbound migration from Alaska within the previous 3 days. systems allow some flexibility in the way yo ung birds perform their first migration (Woodrey and Moore 1997, Hake et al. 2003). So, low-quality individual s have the chance to remain in the population and subsequen tly be out-performed by high- quality individuals in someone?s research project. In the case of godwits, the low- quality individuals probably never make it to New Zealand. Model system or evolutionary outlier? I have discussed how godwits in my study a ppear to defy expectations derived from traditional migration theory and sometimes common sense. In general, Alaskan Bar-tailed Godwits occupy an extreme position along a well-understood continuum, and in some respects the evolutionary context of godwit migr ation may be quite singular. It is fair to ask what such an unusual system can teach us about, for example, the short-distance migration of a temperate-breeding songbird. For discussion, I have chosen two well-studied birds with migrations and life-histories that offer inst ructive contrasts to Bar-tailed Godwits: the American Redstart Setophaga ruticilla and the Black-tailed Godwit L. limosa limosa (see Table 8.2 for key characteristics and in formation sources for each species). Ta bl e 8. 2 Li fe -h is to ry c ha ra ct er is tic s in flu en ci ng n or th bo un d m ig ra tio n st ra te gi es o f t hr ee s pe ci es . A m er ic an R ed st ar t Bl ac k- ta ile d G od w it Ba r- ta ile d G od w it Se to ph ag a r ut ici lla L. lim os a l im os a L. lap po nic a b au er i N on -b re ed in g se as on La tit ud e 5? S? 25 ?N 5? S? 30 ?N 46 ?1 5? S Te rr ito ri al ity ye s no no Fu el lin g re so ur ce s un pr ed ic ta bl e un pr ed ic ta bl e pr ed ic ta bl e N or th bo un d m ig ra ti on M ig ra ti on d is ta nc e (k m ) 2, 00 0? 4, 00 0 3, 00 0? 5, 00 0 15 ,0 00 ?1 8, 00 0 M ax im um s in gl e fli gh t d is ta nc e (k m ) 1, 20 0 1, 80 0 10 ,0 00 Fl ex ib ili ty in s to po ve r lo ca tio n hi gh m od er at e lo w N um be r of s to po ve rs m ul tip le fe w on e M ig ra ti on b ou t d ur at io n (d ) ?1 1? 3 4? 8 Br ee di ng p he no lo gy c ue s A va ila bl e at n on -b re ed in g si te no ? no no A va ila bl e at s to po ve r si te ye s ye s no Br ee di ng s ea so n La tit ud e 20 ?6 0? N 45 ?6 2? N 58 ?7 1? N A rr iv al w in do w , 1 s ite (w k) 2? 4 6? 7 1? 2 Cl ut ch in iti at io n w in do w , 1 s ite (w k) 6? 7 6? 7 3? 4 In cu ba tio n pe ri od (d ) 10 ?1 3 22 ?2 4 20 ?2 3 D ur at io n to c hi ck in de pe nd en ce (d ) ?3 0 25 ?3 0 28 ?3 0 A nn ua l b re ed in g at te m pt s 1? 4 1? 2 1? 2 In di vi du al r ep ea ta bi lit y N or th bo un d de pa rt ur e tim in g 0. 38 0. 30 ?0 .4 2 0. 84 So ur ce s of in fo rm at io n pr es en te d he re a nd in t ex t: A m er ic an R ed st ar t: S he rr y an d H ol m es 1 99 7, M ar ra 2 00 0, N or ri s e t a l. 2 00 4, S tu dd s an d M ar ra 2 00 7, R eu di nk et a l. 2 00 9, A ng el ie r e t a l. 2 01 1, S tu dd s an d M ar ra 2 01 1; B la ck -t ai le d G od w it: G ro en a nd H em er ik 2 00 2, G ill et al. 2 00 8, L ou re n? o an d Pi er sm a 20 08 , Lo ur en ?o e t a l. 20 10 , 20 11 , M as er o et a l. 20 11 , T. P ie rs m a pe rs . co m m .; Ba r- ta ile d G od w it: M cC af fe ry a nd G ill 2 00 1, H us se ll 20 04 , G ill et al . 2 00 9, M cC af fe ry et al . 2 01 0, B at tle y et al . 2 01 2, th is s tu dy . 164 Chapter 8 Chapter 8 165 American Redstarts are small (6 ? 1 0 g), insectivorous wood-warblers (Family Parulidae) that winter in the Neotropics and breed in the northern temperate zone. Like many woodland passerines, their migration is broad-fronted, with individuals breeding roughly north of their non-breeding sites and using multiple stopover s ites in between. Migration bouts are generally a few hundred kilometres and o ccur within a single night, but some birds probably cross the Gulf of Mexico in one flight; for most, crossi ng such extensive barriers is unnecessary. For redstarts, annually variable environmental condi tions and habitat quality in the non-breeding season have demonstrable effects on individual condition, migration timing, the quality of sexually-selected plumage ornaments, and ultimately breeding success. Let?s consider the aspects of redstart life history that might foster migratory strategies so different from those of Bar-tailed Godwits. I grou p these fundamental drivers into four main categories (Table 8.2): non-breeding resour ces, the geography of the journey itself, availability of cues for breeding phenology, and factors specific to the breeding season. For redstarts, non-breeding food resources are unpr edictable, both annually and among individuals via territorial defense. The largely overland migratory route is relatively short, contains no especially daunting barriers, and birds may be quite flexible in stopover locations, number of stops, and overall migration speed. Their trop ical non-breeding grounds may offer few cues for phenology of the upcoming breeding season, but such cues are available when birds reach stopover sites in North America, still hundreds or thousands of kilometres from breeding sites. Their breeding season is relatively long and chick development is rapid, allowing multiple attempts should earlier nests fail. Another possible source of differences between the species is longevity: redstarts are short-lived (~3 ? 1 0 years; Sherry and Holmes 1997) and thus s hould place a higher priority on breeding success in any one year, as opposed to godwits, which may have 10 ? 2 0 breeding opportunities in a lifetime and therefore should prioritise annual survival. All of these characteristics of redstarts predict a plastic migration system, in which flexibility in response to seasonal and daily conditions takes precedence over rigid endogenous programming (Studds and Marra 2011). Not only do unpredictable circumstances make rigid timing unfeasible for redstarts, but opportunities exist to correct ?errors? in timing or pre- migratory fuelling during the northbound journey (Marra et al. 2005, Calvert et al. 2012). Furthermore, timing of arriv al on the breeding grounds might be somewhat forgiving, with birds arriving even 2 ? 3 weeks later than normal having so me chance to breed successfully. Little of this is true for Bar-tailed Godwits. For these birds, vast barriers between successive stops mean that cues for future resource phenologies are unavailable, and departing with insufficient fuel stores can result in a complete loss of breeding opportunity or perhaps death. The short arctic summer means that even brief delays in arrival can severely decrease a 166 Chapter 8 godwit?s chances of breeding. These factors have resulted in a relatively structured, yet conservative, migration strategy. Stable and predictable non-breeding resources enable unhurried fuelling and moults that allow room for flexibility in timing but are reliable enough to ensure the performance of rigid, endogenous ly programmed movements. For both redstarts and godwits, natural selection has provided the t ools and the flexibility to succeed in their respective circumstances. How are the various influences on the sc hedules of these populations reflected in the migration behaviour of individual birds? Rep eatability (or intra-class correlation coefficient; Nakagawa and Schielzeth 2010 ) is a measure of how consistently different individuals are through time, on a scale of 0 (not at all) to 1 ( each individual is completely distinguishable from all others). In the case of Bar-tailed Godwits, repeatability of timing of northbound departure from New Zealand was 0.84 (Chapter 4; Table 8.2), which is among the highest values yet reported for migration timing in any species. For comparison, repeatability of departure by American Redstarts from Jamaica was 0.38 (Studds and Marra 2011) , showing that birds were less faithful to individualised schedules. It should be kept in mind that repeatability reflects the ?individuality? of a parameter in the context of population variation, rather than absolute individual consistency (Chapter 7). However, these relative values strongly suggest that individual Bar-tailed Go dwits place a high value on consistency, whereas redstarts keep their options open. It is quite possible that stable and safe non-br eeding conditions are necessary for a migration such as the Bar-tailed Godwits ? to evolve in the first place. It is also possible that the same behaviours and physiological processes that equip them for this lifestyle will render them unable to adapt to future circumstances dictated by climate change and habitat loss. Are birds as different as redstarts and godwits trapped by their evolutionary pasts, no longer able to converge near the centre of the flexible ?rigid spectrum? One only has to look at the Genus Limosa to find striking examples to the contra ry. This group includes several members strongly in the camp of New Zealand Bar-tailed Godwits, breeding at sub-arctic or higher latitudes and migrating in two or three jumps to the southernmost regions available to them (Hudsonian Godwit L. haemastica and two other Bar-tailed Godwit subspecies, L. l. taymyrensis and L. l. menzbieri ). However, godwits also include relatively short-distance migrants ( L. l. lapponica and Marbled Godwit L. fedoa ) and another congener, the western European Black-tailed Godwit (Table 8.2), whos e annual routine contrasts about as sharply as imaginable from the Bar-tailed Godwits in my st udy. This race of Black-tailed Godwits is primarily terrestrial rather than estuarine throughout the year, using mostly human-altered agricultural and freshwater habitats for breeding and foraging. They winter in tropical sub- Saharan Africa, but depart as early as late Decemb er on a long, direct f light to the Iberian Chapter 8 167 Peninsula, where the population is highly concentr ated while staging for several weeks on rice fields. Subsequently, they move to breeding s ites in one or a few shorter continental flights. They nest in grasslands of temperate Europe in a very loosely-sche duled breeding season: variation in arrival is extensive and some do not start breeding until 3 ? 5 weeks after arrival. As may be expected from these patterns, indi vidual repeatability of northbound migration timing in this Black-tailed Godwit population more closely resembles that of redstarts than of the godwits in my study (Table 8.2). This is cle arly a bird with a different concept of time constraints than the Bar-tailed Godwit has. In fact, the lack of correlation between timing of arrival on the breeding grounds and the start of breeding (Louren?o et al. 2011) goes against everything we know about migrating to exploit resources at just the right time. The unusual schedule of the Black-tailed Godwit is generally explained by the conversion of wetlands throughout its range (Gill et al. 2008); the human-altered habitats it now uses throughout the year are no longer on natural schedules, and the birds? schedules have changed accordingly. Godwits begin to depart winter sites in Africa when the rice fields they use for foraging are drained and harvested (Gill et al. 2008). On the European breeding grounds, warmer spring temperatures over the last century have led to earlier mowing schedules on commercial grasslands, making early breeding by Black -tailed Godwits increasingly advantageous (Schroeder et al. 2012). In response, the godwits advanced their egg-laying dates by about 2 weeks during 1930 ? 1 9 7 6 , but then stopped advancing breedin g despite continuing change in climate and agricultural practices. These patterns may tell us quite a bit about re gulation mechanisms and the potential in godwits to respond to change. Under natur al conditions, we would expect the northbound flight from tropical Africa to be under primarily endogenous control and perhaps very consistent, as it is unlikely to be informed by temporal cues regarding resources in Europe and involves a non-stop crossing of the Sahara Dese rt and Mediterranean Sea. Yet, departure from Africa (in December and January, unusually early for a long-distance migrant) appears to have shifted earlier in response to regular annual declines in local food availability. Apparently, rice fields are profitable enough to support fuelling for this flight in an abbreviated season. It would be interesting to know whether this po pulation has altered its wing moult schedule to accommodate this earlier migration, and how quickly any such adaptation occurred. There is growing concern that long-distance migra nts may be unable to keep pace with earlier breeding phenologies resulting from climate change, because their regulation mechanisms or fuelling resources may not accommodate appropriate timing of arrival at breeding sites (Both and Visser 2001, Ahola et al. 2004, but see Jonz?n et al. 2006). The Black-tailed Godwit is an intriguing case of arrival advancing in response to earlier conditions but breeding not 168 Chapter 8 responding appropriately. It is possible that further advancement of breeding is constrained by conditions for nesting itself becoming temporally decoupled with food resources required to raise chicks to fledging (Schroeder et al. 2012). Another possibility is that endogenous programmes regulating the timing of breeding have limits to their flexibility. How do these patterns influence the view of New Zealand Bar-tailed Godwits as inflexible and trapped by the endogenous programmes necessary for their extreme migration? The Black-tailed Godwit, a congeneric long-distance mi grant, has shown startling shifts in timing of annual events in response to extrinsic factors. Why shouldn?t Bar -tailed Godwits be capable of similar flexibility? It is important to remember that what animals do is not always an indication of their capabilities in other circumstances. We don?t know the evolutionary history or age of the New Zealand godwit population, an d therefore whether it has withstood great challenges posed by habitat and climate changes in the past. Perhaps this system is an evolutionary blip, a noble experiment that will not stand the test of time due to the perfect suite of circumstances that must collude for it to persist. Alternatively, perhaps conditions have not yet persuaded it to display its inherent flexibility and perseverance. Perhaps Bar-tailed Godwits migrate this way simply because they can. 169 Chapter 9 Future directions 170 Chapter 9 What now? As with any ecological research that is worth doing, my project answered far fewer questions than it raised or left unresolved. Some of these latter mysteries are general to migratory birds, but have never been satisfactorily answered in any system. These include: How does a bird ?know? when it is fat enough to start or continue a migration? How exactly does a bird assess the weather conditions prior to departure? Particularly fascinating to me is the formation of departure flocks. In my study, I had departing flocks as small as three birds. Did this tiny flock really fly all the way to the Yellow Sea by itself, or do flocks somehow join up w ith others after initial departure? Departure flock size had no apparent impact on an individual?s likelihood of returning to Foxton the following season. Godwits from my study site used many s taging areas in China and Korea, and appear to have at least regional fidelity between years. Do departing godwits flock together based on when or where they are going? Is it a ?carpool? paradigm, in which all birds going to a particular destination about the same time depart and arrive together? Or is it a ?city bus? paradigm, in which all birds leaving on a particu lar day travel most of the way together, but then drop out one by one as the flock passes their intended destinations? Other big questions specifically raised by my findings are central to understanding the evolution and maintenance of this particular system, and provide a foundation for some rich and rewarding future research. I conclude with a brief overview of just three of these. How did this migration evolve? ?Why do they go so far?? Regarding Bar -tailed Godwits, this is the question I hear most often, from both scientists and non-scientists. Of course, the answer to every evolutionary ?why?? question is: ?because it worked.? The question we really want to ask is: ?how did this evolve?? Although evolution can sometimes work in une xpectedly brief time scales, it is hard to envision epic trans-oceanic flights suddenly resulting from the actions of na?ve but lucky birds, because many associated behaviours and physiological processes must have pre-existed to enable such flights (Alerstam et al. 2003). It seems safe to conclude that the migration developed somewhat incrementally. As described by Hedenstr?m (2010) , the two most plausible scenarios are: (1) a Siberian-breed ing population gradually shifted both its summer and winter quarters eastward until Alaska breed ers were flying directly to New Zealand and Australia without an overland Asia route; and (2 ) an Alaskan-breeding population wintering in Asia gradually shifted its wintering quarters to the southeast toward New Zealand. In fact, the original population from the first scenario may still exist today in the form of L. l. menzbieri , which migrates between Siberia and Australia (Battley et al. 2012). There is also the Chapter 9 171 possibility that menzbieri is a later offshoot from an Alaska-breeding population that persisted in ice-free refugia during the last Pleistocene glacial extreme (25,00 0 ?1 0 , 0 0 0 years ago) and then spread its breeding range westward when glaciers retreated. Because no extensive genetic analysis of Bar-tailed Godwits has yet been conducted, we know very little about the evolutionary relationships among the four or five extant subspecies. Tundra-breeding shorebirds are thought to have radiated across vast areas of the arctic and sub-arctic during the most recent inter-gl acial period (Kraaijeveld and Nieboer 2000) ; consistent with this, genetic analyses have f ailed to uphold some recognised taxonomic divisions based on clear differences in morphology and/or migration routes (e.g., Dunlin Calidris alpina : Wenink et al. 1993, Red Knot C. canutus : Buehler et al. 2006). If diversification of godwits is similarly recent, gen etic analysis may raise as many questions as it answers. But it is clearly the best first step. What is the adaptive significance of geographic variation in Alaska? More relevant for elucidating the specific resu lts of my study is identifying any genetic structure within the Alaska population. For instance, how does the s cale of genetic variation in Alaska compare to that found within and among the other subspecies? The present, continuous breeding range in Alaska was at one time inte rrupted (between the Seward Peninsula and the North Slope) by glaciers, so any historical Al askan godwits could have persisted in one or both of these refugia. Thus, Bar-tailed Godwits may have expanded either north or south of the Brooks Range relatively recently, or tw o historical populations may have united (or reunited) to form the present distribution. Th is is relevant for understanding the selective forces maintaining size and plum age clines within Alaska, particularly if they are shown to be correlated with genetic structure in the population. What did the ancest ral Alaskan godwits look like? Small, red godwits in northern Alas ka could have spread south and met with relaxed selection for those traits, or vice versa. Of course, a more direct way of understanding plumage and size in Alaska is through intensive breeding studies, which have yet to be conducted in any Bar-tailed Godwit population. Is there asso rtative mating by plumage or size within a particular breeding site? Do members of either sex honestly advertise th eir quality as reproductive partners with these traits? Do latitudinal clines in breeding plumage reflect subtle habitat differences or variation in intensity of sexual selection? 172 Chapter 9 How do young godwits ?learn? their routines? Endogenous programming, social behaviour, and learning may all play roles in the development and maintenance of individual routines in Bar-tailed Godwits. Adult godwits are extremely habitual, with unwavering non-breed ing site-fidelity and remarkably consistent migration schedules, but young bird s spend some time growing into their roles. It seems from recoveries of godwits banded in eastern Australia as juveniles (Minton et al. 2011) that many birds on their first southbound migration end up in Australia, only to later become faithful New Zealand residents. Whether this reflects un skilled navigation or a different strategy on the first migration is an intriguing subject by itself. After arrival from their first southbound migration, young godwits typically skip two br eeding seasons before they begin migrating annually to Alaska. They spend the first year or more freely roaming New Zealand and eastern Australia with other young birds, before settlin g at their permanent non-breeding sites. Also, scheduling of departure on the first northbound migration appears to be less precise than in subsequent years (Battley 2006) . These patterns suggest that experience and ?choice? play some role in the development of repeatable adult routines, as opposed to strict and complete control by endogenous programming. As yet, no juvenile from a known Alaska breeding site has been tracked to adulthood, and there is no information on natal-site fidelity, but presumably this must exist at least at a re gional scale for geographic variation in the population to be maintained. If breeding location and migration timing are found to have a genetic component, it would suggest that progra mming for adult routines essentially lies dormant until somehow being ?switched on? at about three years of age. Combined with genetic studies, trace element and stable isot ope analyses have potential for identifying the natal regions of young godwits in New Zealand. If these birds can then be followed until they recruit into the migratory population (a diff icult proposition), it may elucidate how young godwits choose their ultimate non-breeding sites and the roles of endogenous programmes and experience in shaping their first northward migrations. 173 Appendix 1 Attachment of geolocators to Bar-tailed Godwits: a tibia-mounted method with no survival effects or loss of units Conklin, J.R. & P.F. Battley Wader Study Group Bulletin 117: 56?58 (2010) 174 Appendix 1 Abstract We describe a method for tibia-mounted geolocator attachment successfully used on Bar- tailed Godwits Limosa lapponica baueri in New Zealand. The return rate of 95% for instrumented godwits was higher than for other colour-banded individuals in the same study, and we observed no negative physical or behavioural effects of attachment. There was no loss of units, even after 24 months and two return migrations to Alaska. We believe this method is appropriate for a wide variety of migratory waders. Introduction With recent advancements in radio telemetry and datalogger technology, the need for long- term instrument attachment to medium- and small-s ized birds is increasing. Glue applied to the skin and/or feathers is appropriate only for short-term deployment (Warnock and Warnock 1993, Mong and Sandercock 2009) , with the expectation of natur al shedding after instrument batteries are expended. Harness attachment, whether backpack or leg-loop design (e.g., Rappole and Tipton 1991) , offers longer retention, but can negatively affect behaviour (Sykes et al. 1990), breeding success (Rotella et al. 1993), flight speed (Irvine et al. 2009), or survival (Mong and Sandercock 20 09). Long-distance migrants may be particularly ill-suited to back-mounted instrumentation, because dras tic body mass changes make a proper harness fit problematic (Gill et al. 2009), and increased wind drag may significantly decrease flight range (Obrecht et al. 1988). Instrument attachment to the tibia is a viable option for sufficiently large birds, and is made practical by the availability of leg bands and flags used for individual identification. Instruments have typically been attached to metal or plastic leg bands with epoxy (e.g., Morris and Burness 1992, Haig et al. 2002) or cable ties (e.g., Phillips et al. 2009). With the advent of small-sized dataloggers that require recapt ure for data retrieval, and also have sufficient longevity to warrant redeployment of units, there is a growing need for durable attachment that is simple to both apply and remove. Here , we describe a leg-mounted attachment method successfully used on the Bar-tailed Godwit Limosa lapponica baueri , which performs the longest known non-stop flight (>11,00 0 km) of any migratory bird (Gill et al. 2009). Appendix 1 175 Methods We captured Bar-tailed Godwits at the Mana watu River estuary, New Zealand (40.47?S, 175.22?E) , from a local non- breeding population of 200 ?2 8 0 individuals. Approximately 25% of the godwits in this highly site-faithful population are individu ally colour-marked as part of an ongoing non-breeding study. We attached ge olocators (British Antarctic Survey model MK14; 9 mm x 21 mm; 1.4 g; 2-year life) to double-wraparound Darvi c leg bands (external diameter 9 mm; height 8 mm) by the following procedure: Step 1: Using a flame-heated large sewing needle , we pierced two holes approximately 5 mm apart on an axis perpendicular to the length of the band, equidistant from either end, and opposite the band opening. Only holes through the outer layer were required, but it was easiest to pierce through both layers. We then filed of f excess plastic so that the inner and outer band surfaces were smooth again. Step 2: Using a small sewing needle, we passed a length of mist-net repair thread through both holes of the outer layer of the band three times, so th at a double loop was formed, plus two ends free. Step 3: After roughening the back surface of the geol ocator slightly with sandpaper, we used a drop of cyanoacrylate glue to attach the geolocator to the band, within the loops of thread and parallel to the length of the band. The MK14 units were designed with a ?waist? to accommodate cable-tie attachment, and we centred this waist on the length of the band (and thus with the holes), with the light sensor below the waist. Step 4: We tightened the double loop around the unit and tied a simple knot with the loose ends, so that the unit was encircled three time s. We snipped off excess thread and placed a single drop of cyanoacrylate glue to secure the knot. Step 5: We spread a thin layer of Araldite ? two-part epoxy over the centre of the unit, just covering the thread to where it reached the band on either side of the unit (carefully avoiding the light sensor) and filling the crevice betwee n the edges of the unit and the band. We removed excess epoxy and let the assembly dry . Some epoxy flowed through the holes and glued the two layers of the band together; af ter drying, we gently snapped this bond by sending a small knife between the band layers, and filed off the excess epoxy so the band layers sat closely together again. Step 6: Because some of the natural curvature of th e band is held fast by the mounting, care must be taken to open the band as little as possible to avoid breaking the band or the epoxy bond during application to the tibia. With the MK14, we positione d the unit with the light- 176 Appendix 1 sensor end toward the ground, to ensure maximu m exposure of the sensor. Because the units were longer than our colour-bands, we placed an additional half-length colour-band below the geolocator assembly, to stop the downward- pointing contact pins from irritating the bird?s leg joint (Figure A.1a). Step 7: Upon recapture, we removed the assembly from the bird by simply cutting the band with small scissors. Removal of the unit from the band was simple: with two pairs of pliers, we twisted the band and unit in opposite directions, breaking the epoxy bond with the band. By pulling the length of thread, the epoxy th en peeled off easily without damaging the geolocator sheath. Very little add itional cleaning of epoxy was required, as it generally came off in a single strip. After data retrieval and testing, the unit could then be remounted to another band for redeployment. In our study, the mounted units weighed ap proximately 1.8 g. The additional half-band weighed 0.14 g. Godwits in our study carry an individual metal band, a region-specific plastic flag, and an individual combination of four colo ur-bands. Thus, a normal bird carried 2.1 g of markings, while an instrumented bird carried a total of 4.1 g. Due to extreme temporal changes and substantial individual variation, mass of Bar-tailed Godwits varies from 210 ?5 7 0 g for males, and 240 ?6 6 0 g for females (J. Conklin an d P. Battley unpubl. data); the combined bands and geolocator represented 0.6 ? 1.9% of body mass. Results We deployed 17 geolocators on godwits (7 males , 10 females) on 12 March 2008, when birds were very fat and preparing for northbound departure from New Zealand. Despite the disruption of capture by cannon-net and ge olocator attachment, all instrumented birds departed on migration within 16 days of capture , during the same period as local birds that were not recently captured. In fact, three inst rumented females departed New Zealand on 13 March, just one day after capture. The following non-breeding season, 15 of 17 (88% ) instrumented birds returned to the site, compared with 38 of 45 (84%) other individually- marked godwits. One of two females that did not return was last resighted on 23 April 2008 at Yalu Jiang National Nature Reserve in China (J. Conklin pers. obs.), and so successfully completed at least the first leg of migration (approximately 10,30 0 km) after departing New Zealand. On 30 October 2008, we captured godwits to retrieve geolocators and make a second deployment. We retrieved 11 of 15 units, and deployed 19 new units (11 on the original Appendix 1 177 cohort, and 8 on new birds). On 1 March 2009, we redeployed one of the original units on a new bird, captured by mist-net. All 24 inst rumented birds (20 newly-deployed and 4 not recaptured since original deployment in March 2008; 12 males, 12 females) departed on migration 4 ?3 1 March 2009, within the departure period of non-instrumented birds. The following non-breeding season , all 24 (100%) instrumented godwits returned to the site, compared with 32 of 36 (89%) other individuall y-marked godwits. On 3 November 2009, we captured godwits by cannon-net, retrieving 17 of 24 geolocators. Seven instrumented godwits were not recaptured, and departed on northbound migration in March 2010. Figure A.1 Geolocators attached to Bar-tailed Godwits: (a) 12 March 2008; newly-attached geolocator on female godwit; (b) 22 March 2010; male godwit with geolocator, 17 months after deployment Photos by Phil Battley. (a) (b) 178 Appendix 1 We retrieved 26 units after one return mi gration (minimum 30,000 km travelled), 7 ? 1 2 months after deployment. We retrieved two un its after two return migrations (minimum 60,000 km travelled), 19 months after deployment . All units remained firmly attached to the band upon retrieval, and the attachment showed no signs of significant deterioration. With birds in the hand, we observed no physical effects of the attachment method. We also observed no behavioural effects of the geolocators, during detailed observations of individuals at the deployment site from September th rough March in both non-breeding seasons (Figure A.1b). In addition, instrumentation did not appear to preclude normal breeding activity. Because tibia-mounted units are shaded during incubation of nests, periods of nest attendance were roughly indicated by light-sensor data. In 20 08, all 11 birds with available breeding data apparently incubated nests. In 2009, 12 of 15 birds apparently incubated nests. The longest periods of nest attendance (24 ? 2 5 days) were sufficient to indicate probable hatching. Discussion Direct effects of instrument attachment are ofte n difficult to ascertain in migratory species, due to confounding natural influences on return rates. However, high annual survival and site- fidelity of Bar-tailed Godwits allowed an oppo rtunity to test both attachment retention and effects on the birds themselves. We observe d no behavioural or survival effects in instrumented godwits, and no loss of units even after two return migrations to Alaska. The Bar-tailed Godwit performs one of the most extrem e long-distance migrations of any bird, and the attached units were exposed to regular s alt water immersion, as well as extreme UV-radiation and low temperatures. Therefore, we conclude that this attachment method is appropriate for a wide range of long-distance mi gratory wading birds, and probably for other species as well. Body size of smaller species naturally represent s a limit to the use of any attachment method, depending upon the size of the units in question. Our attachment , including other individual markings, represented <2% of godwit mass. Th e generally accepted limit of appropriate tag weight for birds is 3 ? 5 % of body mass, but this should be ev aluated with regard to the species in question (Caccamise and Hedin 1985). Also , appropriate limits have not been fully investigated specifically for leg-attachment, which is naturally asymmetrical, unlike other methods; this could have specific importan ce for perambulatory species. As the size of available geolocators decreases, researchers w ill target increasingly smaller species; we are aware of current projects involving Pacific Golden-Plovers Pluvialis fulva , Ruddy Turnstones Appendix 1 179 Arenaria interpres , Dunlin Calidris alpina , Red Knots C. canutus , and Hudsonian Godwits L. haemastica. It is critical for researchers to docume nt the short- and long-term effects of leg-attachment. Many waders disp lay strong size dimorphism, and thus may show greater negative effects of instrumentation in the smaller sex. However, in our study, instrumented birds of both sexes returned at slightly higher rates than colour-banded individuals at the same site. In various large-bodied seabird species, geol ocators have been attached using cable-ties, requiring little pre-deployment preparation. Ho wever, for godwits, which are smaller and spend a significant amount of their time walkin g, we sought to reduce the additional mass and bulk, and minimise uneven surfaces that might cau se irritation to the legs. Some researchers have attached similar units to the flanges of plastic leg-flags, rather than to the leg-band itself. However, placement of the unit aw ay from the axis of the leg creates torque that may irritate the bird, and we recommend av oiding this when possible. Most studies reporting the use of epoxy alone to attach leg-mounted radio transmitters have done so on permanent aluminium bands, resultin g in indefinite retention far exceeding the life of the units (Warnock and Takekawa 2003). Our aim was to create a secure yet easily- removed attachment to aid both retrieval and redeployment of units. The combination of thread and epoxy was well-suited to this: af ter the initial time investment in mounting the units, deployment or removal of th e attachment in the field requir ed less than one minute per bird, hardly more than a conventional colour -band. The assembly showed no significant weakening after 19 months, and two units rem ained attached after 24 months, and so this method may be appropriate for considerably longer deployments. 180 Appendix 1 181 Appendix 2 Analysis of geolocator data 182 Appendix 2 Geolocation basics The geolocator I used (British Antarctic Survey mo del MK14; Fox 2010) is essentially a light- sensor attached to a memory chip. The unit is not a transmitter, and therefore access to the data requires manual download after recapture of the bird. For periods up to two years, the unit measures light every minute, and then records the maximum light level for each 10-minute period. When this information is compared to a known time and location of deployment, it allows the rough calculation of latitude and longitude twice per day; the former is derived from the length of day or night, a nd the latter from absolute time of local noon or midnight. The exceptions to this are periods close to the vernal and autumnal equinoxes (approximately ?15 d), during which day lengt hs are roughly equal at all locations, and so only longitude is reliable. Processing light-level data In general, I followed data processing inst ructions provided by British Antarctic Survey (Fox 2010). Communication with geolocators was con ducted through a custom USB interface box provided by BAS and the program Hypertermi nal in Microsoft Windows. For each retrieved unit, I copied raw downloaded ge olocator data from Hyperterminal and saved it as a .txt file in the program Notepad. I opened each .txt file through the BASTrak data decompression software, and entered the time the unit was originally activated. With this information, BASTrak compensated for any clock drift that occurred since activation, and automatically created a .lig file compatible with the program TransEdit. Using TransEdit, I applied a lin ear interpolation to smooth the light-level data, and identified sunrises and sunsets based on transitions through a user-prescribed light threshold level of 32 (on a scale of 0 ? 6 4 ). I then visually inspected all derive d transitions for quality, removing false and indistinct transitions that clearly resulted from shading events or other interference. At this point, I applied an arbitrary 10 -minu te retardation of sunset transitions; this corrects for the way the geolocators record light levels (as the maximum value across a 10-minute period). TransEdit then created a transition file (.trn) for the next stage of analysis. I opened each .trn file in the program BirdTracker and specified a sun angle corresponding to sunrise/sunset (see below for derivation of th ese sun angles). BirdTracker then calculated a geographical location for each noon and midnight bracketed by valid sunrise and sunset transitions, thus producing two locations per day in the best of circumstances. Determination of sun angles For calculation of locations, BirdTracker requ ires a user-prescribed sun angle, which represents the sun?s position below the horizon at the time the sunrise/sunset threshold light Appendix 2 183 level prescribed in TransEdit is reached. Because the sun angle is applied to sunrise and sunset equally, it profoundly affects derivation of latitude , but has no influen ce on longitude. The appropriate sun angle must be determined for each unit individually, because the light- sensitivity of the units is not entirely uniform at the time of manufacture. BAS recommends a pre-deployment calibration period, in which the un its are placed outdoors to collect light-level data for a period of days or weeks at a known location. Upon retrieval, this information can be used to determine the sun angle appropriate to identify the known location, and then this sun angle can be applied to the data of interest. In my case, I knew that my birds were at my study site for weeks or months after geolocator deployment and prior to retrieval. Therefore, I used these periods as my calibration data. In ever y case, BirdTracker correctly identified the longitude of the study site within 0.23 ? 1.15? (mean = 0.49?), representing mean linear- distance errors of 20 ? 9 8 km. Calibration-derived sun angles for correctly identifying the latitude of the study site ranged from ? 2.8? to ? 4.0?. Outside of the breeding season, Bar-tailed Godwits are restricted to coastlines, and this fact greatly improved my ability to assess the plausibility of derived locations during migratory stopovers in the Yellow Sea and Alaska. Duri ng northbound migration, approximately 70% of New Zealand godwits use an east-west coastline in northwest China (the vicinity of the Yalu Jiang Nature Reserve). When staging for sout hbound migration, nearly all of this population uses a roughly east-west coastline in southwest Alaska (the vicinity of Kuskokwim Shoals). These two locations allowed me to fine-tune my calibration-derived sun angles for the greatest confidence in the latitude of locations during the breeding season. In some cases, derived locations were clearly implausible, because th ey were either too far inland (northward error) or off the coast in open sea (s outhward error). For each bird, I adjusted the sun angle until the locations for both northward and southward migrations were plausible. I then used the final adjusted sun angle for the entire tracking period . In every case, locations during the breeding season appeared to fall within the known breed ing range, giving me confidence in this technique. The corrected sun angles ranged 3.0 ?3.8? and differed from the initial Foxton- derived sun angles by 0.0 ? 0.6?. The reason for these discre pancies is unknown, but they demonstrate that use of calibration-derived sun an gles without year-round knowledge of the biology of the study species can lead to unrecognised errors. Estimation of location error To help assess the reliability of geolocator-de rived locations outside of Foxton, I deployed ground- truthing units to various locations along the godwits? migratory route in 2009, during the times of year the birds would be using these regions. I deployed two units to staging sites in the Yellow Sea, and five to staging and breed ing areas in Alaska (Tab le A2.1). In each case, 18 4 A pp en di x 2 Ta bl e A 2. 1 Gr ou nd -tr ut hi ng ge ol oc at or s d ep lo ye d in 2 00 9 at lo ca tio ns al on g t he B ar -ta ile d Go dw its ? m igr at or y r ou te , t o aid in te rp re ta tio n of d at a pr ov id ed b y i ns tru m en te d bi rd s. M ea n er ro rs sh ow d iff er en ce b et we en tr ue an d de riv ed lo ca tio ns , u sin g s un an gle o f ? 3. 5? . D ep lo ym en t p er io d M ea n er ro r (? ) Re gi on Lo ca ti on D ep lo ym en t s it e La ti tu de Lo ng it ud e St ar t En d La ti tu de Lo ng it ud e Ye llo w Se a Ko re a M ok po 34 .8 1? N 12 6. 42 ?E 14 A pr 25 M ay ?0 .3 5 ?0 .1 6 Ye llo w Se a Ch in a Ya lu Ji an g N at ur e R es er ve 39 .8 2? N 12 4. 06 ?E 4 A pr 17 A pr - - Al as ka Al as ka Pe nin su la Co ld B ay 55 .2 0? N 16 2. 72 ?W 3 A ug 3 O ct - - Al as ka Yu ko n- Ku sk ok wi m D elt a Be th el 60 .7 9? N 16 1. 76 ?W 3 M ay 8 S ep ?0 .3 1 ?0 .2 5 Al as ka Yu ko n- Ku sk ok wi m D elt a Pu no ar at Po in t 61 .3 1? N 16 5. 81 ?W 21 A ug 13 Se p +0 .6 4 ?0 .2 2 Al as ka Se wa rd Pe ni ns ul a No m e 64 .5 0? N 16 5. 41 ?W 12 Ju n 28 Ju n - - Al as ka No rth Sl op e Ga lb ra ith La ke 68 .4 8? N 14 9. 49 ?W 1 J ul 7 J ul - - 184 Appendix 2 Appendix 2 185 the unit was placed for 1 ?17 weeks in an open location free from obstructions along the sun?s daily path. Two ground-truthing units failed to provide data: the Yalu Jiang unit failed for unknown reasons, and the Cold Bay unit was rendered useless after apparently being chewed by an Arctic Fox Vulpes lagopus. As expected from their high latitude s, the North Slope and Seward Peninsula units failed to record nights at all during the summer. In the latter case, which is south of the Arctic Circle, there were technically sunsets during the deployment period, but the sun did not quite reach the requisite angle belo w the horizon for the light sensor to register darkness; the latitude threshold for registering ni ghts during late June is approximately 64?N. For the other three units, the average error from the true location was 0.31 ? 0.64? in latitude (at sun angle ? 3.5?) and 0.16 ? 0.25? in longitude (Table A2.1); these represent location errors of approximately 50 ?130 km. In two fortuitous cases, I could compare ge olocator-derived bird locations outside New Zealand to actual locations obtained through dir ect observation of the instrumented godwits. Colleagues working along the flyway observe d female godwit 6YBBY at the Geum River estuary in South Korea (36.01?N, 126.74?E ) on 5 April 2008 (G. Stiles pers. comm.) and female 6YBYY at Cape Avinof, Alaska (5 9.83?N, 164.08?W) on 5 September 2008 (D. Ruthrauff pers. comm.). My independent geolocato r-derived locations for these birds differed from the true locations by 91 km and 55 km, respectively. Breeding locations After departing the Yellow Sea, godwits ty pically (23 of 25 cases) arrived in coastal southwestern Alaska, and then clearly moved 2 ? 13 d later to a different Alaska location, in most cases further north. This was consistent with birds arriving at coastal staging sites, and then dispersing to breeding sites after a brief period of refuelling and assessment of likely breeding conditions. In one other case, there wa s no discernible movement after initial arrival in Alaska; this bird presumably either flew str aight to its breeding site, or bred <130 km from the spot of its initial arrival. All individuals were apparently stationary (within location error) for 32 ? 8 5 d after arrival at breeding sites. For birds breeding south of 64?N ( n = 16), I determined breeding location by calculating mean latitude and longitude among twice- daily locations (from the BirdTracker program) duri ng this stationary period, after removal of clear outliers. North of 64?N, determining breeding location is problematic, because geolocators do not register regular nights during most of the breeding season. On th e Seward Peninsula (~65 ? 6 6 ? N ) , nights were discernible in the geolocator data until about 24 ? 2 6 May, and then were 186 Appendix 2 absent until 4 ? 1 5 July. For two birds, the last reliable information in May and the first reliable information in July indicated the same ap proximate location on the Seward Peninsula. Therefore, I used the final available locations in May to re present these birds? breeding sites. In five cases, birds were apparently still mo ving north when the final discernible night occurred 24 May ?2 June, suggesting that they continued to the North Slope of Alaska. In three of these cases, the bird was still stationa ry above the Arctic Circle when discernible nights returned 28 ? 3 0 July. The derived longitudes for these locations spanned 154.9 ? 1 5 8.1?W, also indicating the North Slope se gment of the breeding range. Latitude for locations derived from such brief nights (<1 hour) is not very reliable, but in each case was clearly >68?N. In two other cases, the bird was never stationary above the Arctic Circle during a period of discernible night, but the traject ories of northbound (May) and southbound (July) tracks were consistent with North Slope locations 151.9 ? 1 5 8.5?W. At such easterly longitudes, the known breedin g range is a narrow east-west band spanning just 69.5 ? 70.8?N (McCaffery and Gill 2001). Therefore, I assumed a breeding latitude of 70.2?N (the midpoint of this range) for all apparent North Slope br eeders. Longitudes for these breeding locations must also be viewed as approximate, as I ca nnot rule out eastward or westward movements after arrival on the North Slope. In every case in which I had two years of breeding data for an individual ( n = 8), the between- year differences in derived breeding locations were within the expected location error (24 ? 124 km), and so were consistent with breeding site-fidelity. Nest incubation As previously demonstrated with Barnacle Geese Branta leucopsis (Eichhorn et al. 2006), light-level geolocators have an additional benef it of indicating periods of nest incubation. In my study, geolocators recorded nights duri ng the breeding season as regular, clearly demarcated periods of darkness <4.5 hours in lengt h; these did not appear at all if birds bred north of 64?N. Days appeared as continuous light, irregularly broken by very brief (<1 hour) shading events, most likely corresponding to be haviours such as wading or sitting. Within 6 ? 2 5 d of apparent arrival on breeding grounds, most birds (21 of 23 cases) displayed a conspicuous pattern of incubation, in wh ich semi-regular shading events of 4 ? 1 3 hours were overlaid on the day/night pattern for 4 ? 2 5 d. Three birds that showed no such pattern (including one that never appeared to settle at a breeding site) were considered non-breeders, but this could have resulted from either failure to find a mate or loss of a clutch prior to incubation. The full-time incubation period for Bar-tailed Godwits is thought to be 20 ? 23 d (McCaffery and Gill 2001, Hussell 2004) , a nd godwits typically begin part-time incubation when clutches are not yet complete (J. Conkl in pers. obs.). Therefore, periods of 21 ?25 Appendix 2 187 consecutive days of incubation indicated a high likelihood of hatching success, and I considered these cases to be successful breeding attempts. Periods of 4 ? 1 8 consecutive days of incubation were clearly insufficient for eggs to have hatched (either by loss of a clutch or nest abandonment), and I considered these attempts unsuccessful. In four cases, two periods of incubation were separated by several days without shading events; these appeared to indicate renesting attempts after initial clutch losses. Of five apparent renesting attempts, one was successful and four were unsuccessful. 188 Appendix 2 189 Appendix 3 Calculation of wind effect 190 Appendix 3 Wind profit equations In Chapter 4, I used two different methods to calc ulate how a bird gained or lost speed as a result of winds experienced at the time of migratory departure from my study site. The first (Tailwind, or ?TW?) simply represents the magnitude of tailwind (positive values) or headwind (negative values) along the axis of the bird?s chosen flight path (as described in ?kesson and Hedenstr?m 2000) and ignores the effect of crosswinds. TW is calculated as:  = ??? ( ? )  where:  = wind profit experienced by the bird (km/h)  = wind velocity (km/h)  = bird flight direction (radians)  = wind direction (radians) The second formula (Crosswind, or ?CW?) additionally accounts for the work required by the bird to stay on its preferred flight path in the presence of crosswinds (as described in Piersma and Jukema 1990). In this method, wind profit is incalculable when the velocity of crosswinds exceeds the bird?s preferred flight speed. CW is calculated as:  =  ? where:  = bird ground speed (km/h)  = bird air speed (km/h) For  , I assumed a preferred flight speed of 65 km/h.  is calculated as: ? ? ?? ??? ?? ? ?????? ??? ??? where: ? ?  ??? ???? (the difference between bird direction and wind direction) 191 Appendix 4 Supplementary information on primary moult 192 Appendix 4 Table A4.1 Use of historical New Zealand godwit capture data to assign moult scores when only the number of unmoulted primary feathers was known. Source data includes 1,434 captures of adult godwits in New Zealand during 1983?2008 (P. Battley and A. Riegen unpubl. data). ?Moult? represents primaries 1?10, left to right; 0 = old, unmoulted feather. Moult Mean score Range n Assigned score ?000000000 1 1 3 1 ??00000000 2 2 11 2 ???0000000 4.2 3?7 36 4 ????000000 8.1 4?16 62 8 ?????00000 15.4 9?21 81 15 ??????0000 22.6 14?26 61 23 ???????000 29.8 28?33 36 30 ????????00 35.1 27?39 39 35 ?????????0 40.1 36?44 45 40 Table A4.2 Masses of individual godwit primary feathers (n = 3 individuals), as a proportion of total mass of primaries 1?10. Feather Mean length (mm) Mean mass (g) Mean prop. mass P01 92.0 0.044 0.035 P02 100.7 0.055 0.044 P03 110.7 0.071 0.056 P04 124.3 0.090 0.072 P05 136.3 0.110 0.087 P06 148.7 0.132 0.105 P07 160.3 0.154 0.122 P08 173.0 0.178 0.141 P09 181.3 0.201 0.160 P10 179.7 0.220 0.176 193 References 194 References Ahola, M., T. Laaksonen, K. Sippola, T. Eeva, K. Rainio & E. Lehikoinen. 2004. Variation in climate warming along the migration route uncouples arrival and breeding dates. Global Change Biology 10: 1610 ? 1617. ?kesson, S. & A. Hedenstr?m. 2000. Wind selectiv ity of migratory flight departures in birds. Behavioral Ecology and Sociobiology 47: 140 ? 1 4 4. Alerstam, T. 1979. Wind as selective agent in bird migration. Ornis Scandinavica 10: 76 ?9 3. Alerstam, T. 1990. Bird Migration. Cambridge University Press, Cambridge. Alerstam, T., M. Hake & N. Kjell?n. 20 06. Temporal and spatial patterns of repeated migratory journeys by ospreys. Animal Behaviour 71: 555 ?5 6 6. Alerstam, T., A. Hedenstr?m & S. ?kesson. 2003. Long-distance migration: evolution and determinants. Oikos 103: 247 ? 260. Alerstam, T. & ?. Lindstr?m. 1990. Optimal bird migration: the relative importance of time, energy and safety. Pages 331 ?3 5 1 in Bird Migration: Physiology and Ecophysiology (E. Gwinner, Ed.). Spri nger-Verlag, Berlin. Allen, J.A. 1877. The influence of physical conditions in the genesis of species. Radical review 1: 108 ? 1 4 0. Anders, A.D., D.C. Dearborn, J. Faaborg & F .R. Thompson. 1997. Juvenile survival in a population of neotropical migrant birds. Conservation Biology 11: 698 ? 707. Anderson, M.G. 2003. Investigations into shorebird community ecology: interrelations between morphology, behaviour, habitat and abiotic factors. MSc thesis, University of Auckland. Angelier, F., C.M. Tonra, R.L. Holberton & P. P. Marra. 2011. Short-term changes in body condition in relation to habitat and rainfall abundance in American redstarts Setophaga ruticilla during the non-breeding season. Journal of Avian Biology 42: 335 ?3 4 1. Atkinson, P.W., A.J. Baker, K.A. Bennett, N.A. Clark, J.A. Clark, K.B. Cole, A. Dekinga, A. Dey, S. Gillings, P.M. Gonz?lez, K. Kalasz, C.D.T. Minton, J. Newton, L.J. Niles, T. Piersma, R.A. Robinson & H.P. Sitters. 2007. Rates of mass gain and energy deposition in red knot on their final spri ng staging site is both time- and condition- dependent. Journal of Applied Ecology 44: 885 ? 8 9 5. Barbosa, A. & E. Moreno. 19 99. Evolution of foraging strategies in shorebirds: an ecomorphological approach. Auk 116: 712 ? 725. Barshep, Y., A. Hedenstr?m & L.G. Underhill. 2011a. Impact of climate and predation on autumn migration of the Curlew Sandpiper. Waterbirds 34: 1 ?9. Barshep, Y., C. Minton, L.G. Underhill & M. Remisiewicz. 2011b. The primary moult of Curlew Sandpipers Calidris ferruginea in north-western Australia shifts according to breeding success. Ardea 99: 43 ?5 1. Barta, Z., J.M. McNamara, A.I. Houston, T.P. Weber, A. Hedenstr?m & O. Fer?. 2008. Optimal moult strategies in migratory birds. Philosophical Transactions of the Royal Society of London, Series B 363: 211 ? 229. Barter, M. 1989. Bar-tailed Godwit Limosa lapponica in Australia, Part 2: weight, moult and breeding success. Stilt 14: 49 ? 53. Battley, P.F. 1997. The northward migration of arctic waders in New Zealand: departure behaviour, timing and possible migration ro utes of red knots and bar-tailed godwits from Farewell Spit, north-west Nelson. Emu 97: 108 ? 1 2 0. References 195 Battley, P.F. 2006. Consistent annual sc hedules in a migratory shorebird. Biology Letters 2: 517 ?5 2 0. Battley, P.F. 2007. Plumage and timing of mi gration in bar-tailed godwits: a comment on Drent et al. (2003). Oikos 116: 349 ? 3 5 2. Battley, P.F. & T. Piersma. 2005. Body compositio n and flight ranges of bar-tailed godwits ( Limosa lapponica baueri ) from New Zealand. Auk 122: 922 ?9 3 7. Battley, P.F., T. Piersma, M.W. Dietz, S. X. Tang, A. Dekinga & K. Hulsman. 2000. Empirical evidence for differential organ reductio ns during trans-oceanic bird flight. Proceedings of the Royal Society of London, Series B 267: 191 ? 195. Battley, P.F., T. Piersma, D.I. Rogers, A. Deki nga, B. Spaans & J.A. van Gils. 2004. Do body condition and plumage during fuelling predict northwards departure dates of great knots Calidris tenuirostris from north-west Australia? Ibis 146: 46 ?6 0. Battley, P.F., D.I. Rogers & C.J. Ha ssell. 2006. Prebreeding moult, plumage and evidence for a presupplemental moult in the great knot Calidris tenuirostris. Ibis 148: 27 ?3 8. Battley, P.F., D.I. Rogers, J.A. van Gils, T. Piersma, C.J. Hassell, A. Boyle & Y. Hong-Yan. 2005. How do red knots Calidris canutus leave Northwest Australia in May and reach the breeding grounds in Jun e? Predictions of stopover times, fuelling rates and prey quality in the Yellow Sea. Journal of Avian Biology 36: 494 ? 500. Battley, P.F., R. Schuckard & D.S. Melville. 2011. Movements of bar-tailed godwits and red knots within New Zealand. In Science for Conservation 315 , 56 pp. New Zealand Department of Conservation. Battley, P.F., N. Warnock, T.L. Tibbitts, R.E. Gill, Jr., T. Piersma, C.J. Hassell, D.C. Douglas, D.M. Mulcahy, B.D. Gartrell, R. Schuckard, D. Melville & A. Riegen. 2012. Contrasting extreme long-distance migration patterns in bar-tailed godwits Limosa lapponica. Journal of Avian Biology 43: 21 ?3 2. Becker, W.A. 1984. A Manual of Quantitative Genetics. Academic Enterprises, Pullman, Washington. Bensch, S. & M. Grahn. 1993. A new meth od for estimating individual speed of molt. Condor 95: 305 ?3 1 5. Berthold, P. & F. Pulido. 1994. Heritability of migratory activity in a natural bird population. Proceedings of the Royal Society of London, Series B 257: 311 ? 315. Bety, J., J.F. Giroux & G. Gauthier. 2004. In dividual variation in timing of migration: causes and reproductive consequences in greater snow geese ( Anser caerulescens atlanticus ). Behavioral Ecology and Sociobiology 57: 1 ?8. Both, C., S. Bouwhuis, C.M. Lessells & M.E. Visser. 2006. Climate change and population declines in a long-distance migratory bird. Nature 441: 81 ?8 3. Both, C. & M.E. Visser. 2001. Adju stment to climate change is constrained by arrival date in a long-distance migrant bird. Nature 411: 296 ? 298. Bruderer, B., L.G. Underhill & F. Liechti. 1995. A ltitude choice by night migrants in a desert area predicted by meteorological factors. Ibis 137: 44 ? 5 5. Buehler, D.M., A.J. Baker & T. Piersma. 20 06. Reconstructing palaeoflyways of the late Pleistocene and early Holocene red knot Calidris canutus. Ardea 94: 485 ? 498. Buehler, D.M. & T. Piersma. 2008. Travelli ng on a budget: pred ictions and ecological evidence for bottlenecks in the annu al cycle of long-distance migrants. Philosophical Transactions of the Royal Society of London, Series B 363: 247 ? 266. 196 References Busse, P. 2001. European passerine migration system ? what is known and what is lacking. The Ring 23: 3 ? 36. Caccamise, D.F. & R.S. Hedin. 1985. An aerodyn amic basis for selecting transmitter loads in birds. Wilson Bulletin 97: 306 ? 318. Cadee, N., T. Piersma & S. Daan. 1996. En dogenous circannual rhythmicity in a non- passerine migrant, the knot Calidris canutus. Ardea 84: 75 ?8 4. Calvert, A.M., S.A. Mackenzie, J.M. Flemming, P.D. Taylor & S.J. Walde. 201 2. Variation in songbird migratory behavior offers clues about adaptability to environmental change. Oecologia 168: 849 ? 861. Caplan, A.L. 1988. Rehabilitating reductionism. American Zoologist 28: 193 ?203. Catry, P., G.D. Ruxton, N. Ratcliffe, K .C. Hamer & R.W. Furness. 1999. Short-lived repeatabilities in long-lived great skuas: implications for the study of individual quality. Oikos 84: 473 ? 479. Chu, P.C. 1994. Historical examination of dela yed plumage maturation in the shorebirds (Aves, Charadriiformes). Evolution 48: 327 ? 350. Ciarleglio, C.M., J.C. Axley, B.R. Strauss, K.L. Gamble & D.G. McMahon. 2011. Perinatal photoperiod imprints the circadian clock. Nature Neuroscience 14: 25 ? 27. Clark, C.W. & R.W. Butler. 1999. Fitness compon ents of avian migration: a dynamic model of Western Sandpiper migration. Evolutionary Ecology Research 1: 443 ? 457. Conklin, J.R. & M.A. Colwell. 2008. Individual associations in a wintering shorebird population: do dunlin have friends? Journal of Field Ornithology 79: 32 ? 40. Costa, R., A.A. Peixoto, G. Barbujani & C.P. Ky riacou. 1992. A lati tudinal cline in a Drosophila clock gene. Proceedings of the Royal Society of London, Series B 250: 43 ? 49. Crick, H.Q.P. 2004. The impact of climate change on birds. Ibis 146: 48 ? 5 6. Cristol, D.A., M.B. Baker & C. Carbone. 1999. Differential migration revisited: latitudinal segregation by age and sex class. Current Ornithology 15: 33 ? 88. Daan, S., C. Dijkstra & J.M. Tinbergen. 1990. Family-planning in the kestrel ( Falco tinnunculus ) ? the ultimate control of covariation of laying date and clutch size. Behaviour 114: 83 ? 116. Darwin, C. 1859. On the Origin of the Species by Natural Selection. Murray, London. Dawson, A. 2003. A detailed analysis of primary feather moult in the Common Starling Sturnus vulgaris ? new feather mass increases at a constant rate. Ibis 145 (online): E69 ?E76. Dawson, A. 2004. The effects of delaying the star t of moult on the duration of moult, primary feather growth rates and feather mass in common starlings Sturnus vulgaris. Ibis 146: 493 ?5 0 0. Dawson, A. 2006. Control of mo lt in birds: association with prolactin and gonadal regression in starlings. General and Comparative Endocrinology 147: 314 ?322. Dawson, A. 2008. Control of the annual cycle in birds: endocrine constraints and plasticity in response to ecological variability. Philosophical Transactions of the Royal Society of London, Series B 363: 1621 ? 1633. Dawson, A., S.A. Hinsley, P.N. Ferns, R.H.C. Bonser & L. Eccleston. 2000. Rate of moult affects feather quality: a me chanism linking current repro ductive effort to future survival. Proceedings of the Royal Society of London, Series B 267: 2093 ? 2098. Dawson, A., V.M. King, G.E. Bentley & G.F. Ball. 2001. Photop eriodic control of seasonality in birds. Journal of Biological Rhythms 16: 365 ? 380. References 197 Dietz, M.W., T. Piersma, A. Hedenstr?m & M. Brugge. 2007. Intraspecific variation in avian pectoral muscle mass: constraints on main taining manoeuvrability with increasing body mass. Functional Ecology 21: 317 ?3 2 6. Dingemanse, N.J., C. Both, P.J. Drent & J.M. Tinbergen. 2004. Fitness consequences of avian personalities in a fluctuating environment. Proceedings of the Royal Society of London, Series B 271: 847 ? 852. Dingemanse, N.J. & D. R?ale. 2005. Natural selection and animal personality. Behaviour 142: 1165 ? 1190. Dingle, H. 2006. Animal migration: is there a common migratory syndrome? Journal of Ornithology 147: 212 ? 220. Drent, R., C. Both, M. Green, J. Madsen & T. Piersma. 2003. Pay-offs and penalties of competing migratory schedules. Oikos 103: 274 ?2 9 2. Drent, R.J., A.D. Fox & J. Stah l. 2006. Travel ling to breed. Journal of Ornithology 147: 122 ?134. Dufva, R. & K. Allander. 1995. Intraspecific variation in plumage coloration reflects immune-response in great tit ( Parus major ) males. Functional Ecology 9: 785 ?789. Earnst, S.L. 1992. The timing of wing molt in Tundra Swans: energetic and non-energetic constraints. Condor 94: 847 ? 856. Egevang, C., I.J. Stenhouse, R.A. Phillips, A. Petersen, J.W. Fox & J.R.D. Silk. 2010. Tracking of Arctic terns Sterna paradisaea reveals longest animal migration. Proceedings of the National Academy of Sciences 107: 2078 ?2 0 8 1. Eichhorn, G., V. Afanasyev, R.H. Drent & H. P. van der Jeugd. 2006. Spring stopover routines in Russian barnacle geese Branta leucopsis tracked by resightings and geolocation. Ardea 94: 667 ? 678. Eichhorn, G., R.H. Drent, J. Stahl, A. Leito & T. Alerstam. 2009. Skipping the Baltic: the emergence of a dichotomy of alternative spring migration strategies in Russian barnacle geese. Journal of Animal Ecology 78: 63 ? 72. Emslie, S.D., R.P. Henderson & D.G. Ainley. 1990. Annual variation of primary molt with age and sex in Cassin's auklet. Auk 107: 689 ? 695. Endler, J.A. 1977. Geographic Variation, Speciation, and Clines. Princeton University Press, Princeton, New Jersey. Engelmoer, M. & C.S. Roselaar. 1998. Geographic Variation in Waders. Kluwer Academic Publishers, Dordrecht. Farmer, A.H. & J.A. Wiens. 1999. Models an d reality: time-energy trade-offs in pectoral sandpiper ( Calidris melanotos ) migration. Ecology 80: 2566 ?2 5 8 0. Ferns, P.N. 2003. Plumage colour and pattern in waders. Wader Study Group Bulletin 100: 122 ?1 2 9. Fiedler, W. 2009. New techn ologies for monitoring bird migration and behaviour. Ringing & Migration 24: 175 ? 179. Figuerola, J., J. Domenech & J.C. Senar. 2003. Plumage colour is related to ectosymbiont load during moult in the serin, Serinus serinus : an experimental study. Animal Behaviour 65: 551 ? 557. Fitzpatrick, S. 1998. Colour schemes for birds: st ructural coloration and signals of quality in feathers. Annales Zoologici Fennici 35: 67 ? 77. Flinks, H., B. Helm & P. Rothery. 2008. Plas ticity of moult and breeding schedules in migratory European Stonechats Saxicola rubicola. Ibis 150: 687 ?697. 198 References Flood, N.J. 1984. Adaptive significance of delay ed plumage maturation in male northern orioles. Evolution 38: 267 ?2 7 9. Folstad, I. & A.J. Karter. 1992. Parasites, br ight males, and the immunocompetence handicap. American Naturalist 139: 603 ? 622. Forstmeier, W. 2002. Benefits of early arri val at breeding grounds vary between males. Journal of Animal Ecology 71: 1 ? 9. Fox, J.W. 2010. Geolocator Manual , version 8. British Antarctic Survey. Francis, C.M. & F. Cooke. 1986. Differential timi ng of spring migration in wood warblers (Parulinae). Auk 103: 548 ?5 5 6. Fransson, T. 1995. Timing and speed of migratio n in North and West European populations of Sylvia warblers. Journal of Avian Biology 26: 39 ?48. Gauthreaux, S.A., Jr. 1991. The flight behavior of migrating birds in changing wind fields: radar and visual analyses. American Zoologist 31: 187 ? 204. Gill, J.A., R.H.W. Langston, J.A. Alves, P.W. Atkinson, P. Bocher, N.C. Vieira, N.J. Crockford, G. G?linaud, N. Groen, T. Gunnarsson, B. Hayhow, J. Hooijmeijer, R. Kentie, D. Kleijn, P.M. Louren?o, J.A. Masero, F. Meunier, P.M. Potts, S.P. Roodbergen, H. Schekkerman, J. Schr?d er, E. Wymenga & T. Piersma. 2008. Contrasting trends in two Black-tailed Godw it populations: a review of causes and recommendations. Wader Study Group Bulletin 114: 43 ? 50. Gill, R.E., Jr., T. Piersma, G. Hufford, R. Servranckx & A. Riegen. 2005. Crossing the ultimate ecological barrier: evidence for an 11000 -km-long nonstop flight from Alaska to New Zealand and eastern Australia by bar-tailed godwits. Condor 107: 1 ?20. Gill, R.E., Jr., T.L. Tibbitts, D.C. Douglas, C.M . Handel, D.M. Mulcahy, J.C. Gottschalck, N. Warnock, B.J. McCaffery, P.F. Battley & T. Piersma. 2009. Extreme endurance flights by landbirds crossing the Pacific Ocean: ecological corridor rather than barrier? Proceedings of the Royal Society of London, Series B 276: 447 ?4 5 7. Ginn, H.B. & D.S. Melville. 1983. Moult in Birds ? BTO Guide 19. British Trust for Ornithology, Hertfordshire. Gonz?lez-Sol?s, J., J.P. Croxall, D. Oro & X. Ruiz. 2007. Trans-equatorial migration and mixing in the wintering areas of a pelagic seabird. Frontiers in Ecology and the Environment 5: 297 ? 301. Green, M. 2004. Flyi ng with the wind ? spring migration of arctic-breeding waders and geese over south Sweden. Ardea 92: 145 ?1 6 0. Green, M., T. Alerstam, G.A. Gudmundsson, A. Hedenstr?m & T. Piersma. 2004. Do Arctic waders use adaptive wind drift? Journal of Avian Biology 35: 305 ?3 1 5. Griffith, S.C. & B.C. Sheldon. 2001. Phenotypic plasticity in the expression of sexually selected traits: neglected components of variation. Animal Behaviour 61: 987 ? 993. Grim, T. 2008. A possible role of social activity to explain differences in publication output among ecologists. Oikos 117: 484 ?4 8 7. Grimm, V. 1999. Ten years of individual-ba sed modelling in ecology: what have we learned and what could we learn in the future? Ecological Modelling 115: 129 ? 148. Groen, N.M. & L. Hemerik. 2002. Reproductive success and surv ival of black-tailed godwits Limosa limosa in a declining local population in The Netherlands. Ardea 90: 239 ?248. Gudmundsson, G.A., ?. Lindstr?m & T. Alers tam. 1991. Optimal fat loads and long-distance flights by migrating knots Calidris canutus , sanderlings C. alba and turnstones Arenaria interpres. Ibis 133: 140 ? 152. References 199 Gunnarsson, T.G., J.A. Gill, P.W. Atkinson, G. Gelinaud, P.M. Potts, R.E. Croger, G.A. Gudmundsson, G.F. Appleton & W.J. Suth erland. 2006. Population-scale drivers of individual arrival times in migratory birds. Journal of Animal Ecology 75: 1119 ?1127. Gunnarsson, T.G., J.A. Gill, J. Newton, P. M. Potts & W.J. Sutherland. 2005. Seasonal matching of habitat quality and fitness in a migratory bird. Proceedings of the Royal Society of London, Series B 272: 2319 ?2 3 2 3. Gunnarsson, T.G., J.A. Gill, T. Sigurbjorn sson & W.J. Sutherland. 2004. Arrival synchrony in migratory birds. Nature 431: 646 ? 6 4 6. Gwinner, E. 1990. Circannual rhythms in bird migration: control of temporal patterns and interactions with photoperiod. Pages 257 ?2 6 8 in Bird migration: Physiology and Ecophysiology (E. Gwinner, Ed.). Springer-Verlag, Berlin. Gwinner, E. 1996. Circannual clocks in avian reproduction and migration. Ibis 138: 47 ? 63. Hahn, T.P., J. Swingle, J.C. Wingfield & M. Ra menofsky. 1992. Adjustme nts of the prebasic molt schedule in birds. Ornis Scandinavica 23: 314 ?321. Haig, S.M., L.W. Oring, P.M. Sanzenbacher & O.W. Taft. 2002. Space use, migratory connectivity, and population segregation among willets breeding in the western Great Basin. Condor 104: 620 ?630. Hake, M., N. Kjell?n & T. Alerstam. 2003 . Age-dependent migration strategy in honey buzzards Pernis apivorus tracked by satellite. Oikos 103: 385 ?3 9 6. Hall, K.S.S. & T. Fransson. 2000. Lesser wh itethroats under time-constraint moult more rapidly and grow shorter wing feathers. Journal of Avian Biology 31: 583 ? 587. Harrison, X.A., J.D. Blount, R. Inger, D.R. Norris & S. Bearhop. 2011. Carry-over effects as drivers of fitness differences in animals. Journal of Animal Ecology 80: 4 ? 1 8. Hasselquist, D. 1998. Polygyny in great reed warblers: a long-term study of factors contributing to male fitness. Ecology 79: 2376 ? 2390. Hedenstr?m, A. 2006. Scaling of migration and the annual cycle of birds. Ardea 94: 399 ?4 0 8. Hedenstr?m, A. 2008. Adaptations to migration in birds: behavioural strategies, morphology and scaling effects. Philosophical Transactions of the Royal Society of London, Series B 363: 287 ? 299. Hedenstr?m, A. 2010. Extreme endurance migratio n: What is the limit to non-stop flight? PLoS Biology 8: 1 ?6. Hedenstr?m, A. & T. Alerstam. 1992. Climbing performance of migrating birds as a basis for estimating limits for fuel-carrying capacity and muscle work. Journal of Experimental Biology 164: 19 ? 38. Helm, B. & E. Gwinner. 2006. Timing of molt as a buffer in the avian annual cycle. Acta Zoologica Sinica 52 (suppl.): 703 ? 7 0 6. Henningsson, S.S. & T. Alerstam. 2005. Barr iers and distances as determinants for the evolution of bird migration links: the arctic shorebird system. Proceedings of the Royal Society of London, Series B 272: 2251 ?2 2 5 8. Hill, G.E. 1988. The function of delayed plumage maturation in male black-headed grosbeaks. Auk 105: 1 ? 10. Hill, G.E. 1991. Plumage coloration is a se xually selected indicator of male quality. Nature 350: 337 ? 339. Hill, G.E. & R. Montgomerie. 19 94. Plumage color signals nutr itional condition in the house finch. Proceedings of the Royal Society of London, Series B 258: 47 ? 52. 200 References Holmes, R.T. 1971. Latitudinal differences in the breeding and molt schedules of Alaskan red-backed sandpipers ( Calidris alpina ). Condor 73: 93 ? 99. Holmgren, N. & A. Hedenstr?m. 1995. The scheduling of molt in migratory birds. Evolutionary Ecology 9: 354 ? 368. Howell, S.N.G. 2010. Molt in North American Birds. Houghton Mifflin Harcourt, New York. Humphrey, P.S. & K.C. Parkes. 1959. An appr oach to the study of molts and plumages. Auk 76: 1 ? 3 1. Hussell, D.J.T. 2004. Incubation period an d behavior at a bar-tailed godwit nest. Wilson Bulletin 116: 177 ?1 7 8. Irvine, R.J., F. Leckie & S.M. Redpath. 2009. Cost of carrying radio transmitters: a test with racing pigeons Columba livia. Wildlife Biology 13: 238 ? 2 4 3. Jacobs, J.D. & J.C. Wingfield. 2000. Endocrine co ntrol of life-cycle stages: a constraint on response to the environment? Condor 102: 35 ? 5 1. Jaenisch, R. & A. Bird. 2003. Epigenetic regu lation of gene expression: how the genome integrates intrinsic and environmental signals. Nature Genetics 33 (suppl.): 245 ?2 5 4. J?rvinen, A. 1989. Geographical variation in tem perature variability and predictability and their implications for the breeding strategy of the pied flycatcher Ficedula hypoleuca. Oikos 54: 331 ? 336. Jawor, J.M. & R. Breitwisch. 2003. Melanin ornaments, honesty, and sexual selection. Auk 120: 249 ? 265. Jawor, J.M., S.U. Linville, S.M. Beall & R. Breitwisch. 2003. Assortative mating by multiple ornaments in northern cardinals ( Cardinalis cardinalis). Behavioral Ecology 14: 515 ?520. Jehl, J.R., Jr. & B.G. Murray, Jr. 1986. The ev olution of normal and reverse sexual size dimorphism in shorebirds and other birds. Current Ornithology 3: 1 ? 86. Johnsen, A., A.E. Fidler, S. Kuhn, K.L. Carter, A. Hoffmann, I.R. Barr, C. Biard, A. Charmantier, M. Eens, P. Korsten, H. Siit ari, J. Tomiuk & B. Kempenaers. 2007. Avian clock gene polymorphism: eviden ce for a latitudinal cline in allele frequencies. Molecular Ecology 16: 4867 ? 4880. Johnson, C. & C.D.T. Minton. 1980. The primary moult of the Dunlin Calidris alpina at the Wash. Ornis Scandinavica 11: 190 ?1 9 5. Jonz?n, N., A. Lind?n, T. Ergon, E. Knudsen, J.O. Vik, D. Rubolini, D. Piacentini, C. Brinch, F. Spina, L. Karlsson, M. Stervander, A. Andersson, J. Waldenstr?m, A. Lehikoinen, E. Edvardsen, R. Solvang & N.C. Stenseth. 2006. Rapid advance of spring arrival dates in long-distance migratory birds. Science 312: 1959 ?1 9 6 1. Judson, O.P. 1994. The rise of the in dividual-based model in ecology. Trends in Ecology and Evolution 9: 9 ? 14. Jukema, J. & T. Piersma. 2000. Contour feather moult of ruffs Philomachus pugnax during northward migration, with no tes on homology of nuptial plumages in scolopacid waders. Ibis 142: 289 ?2 9 6. Jukema, J. & T. Piersma. 2006. Permanent female mimics in a lekking shorebird. Biology Letters 2: 161 ? 1 6 4. Kelly, J., V. Atudorei, Z. Sharp & D. Finch. 2002. Insights into Wilson?s Warbler migration from analyses of hydrogen stable-isotope ratios. Oecologia 130: 216 ? 2 2 1. Ketterson, E.D. & V. No lan, Jr. 1983. The evolution of diff erential bird migration. Pages 357 ? 4 0 2 in Current Ornithology, vol. 1 (R.F. Johnston, Ed.). Plenum Press, New York. References 201 Keyser, A.J. & G.E. Hill. 2000. Structurally ba sed plumage coloration is an honest signal of quality in male blue grosbeaks. Behavioral Ecology 11: 202 ? 209. Kjell?n, N., M. Hake & T. Alerstam. 2001. Timi ng and speed of migration in male, female and juvenile Ospreys Pandion haliaetus between Sweden and Africa as revealed by field observations, radar and satellite tracking. Journal of Avian Biology 32: 57 ?67. Klaassen, R.H.G., T. Alerstam, P. Carlsson, J. W. Fox & ?. Lindstr?m. 2011. Great flights by great snipes: long and fast non-stop migration over benign habitats. Biology Letters 7: 833 ? 835. Kraaijeveld, K. & E.N. Nieboer. 2000. Late Quat ernary paleogeography and evolution of arctic breeding waders. Ardea 88: 193 ?205. Landys-Ciannelli, M.M., T. Piersma & J. Juke ma. 2003. Strategic size changes of internal organs and muscle tissue in the bar-tailed godwit during fat storage on a spring stopover site. Functional Ecology 17: 151 ? 159. Landys, M.M., T. Piersma, G.H. Visser, J. Ju kema & A. Wijker. 2000. Water balance during real and simulated long-distance migratory flight in the bar-tailed godwit. Condor 102: 645 ? 652. Larsson, K. 1996. Genetic and environmental effe cts on the timing of wing moult in the barnacle goose. Heredity 76: 100 ? 1 0 7. Lessells, C.M. & P.T. Boag. 1987. Unrepeatable repeatabilities: a common mistake. Auk 104: 116 ?1 2 1. Leyrer, J., S. Pruiksma & T. Piersma. 2009. On 4 June 2008 Siberian Red Knots at Elbe Mouth kissed the canonical evening mi gration departure rule goodbye. Ardea 97: 71 ?79. Liechti, F. 2006. Birds: blowin? by the wind? Journal of Ornithology 147: 202 ?211. Liechti, F. & B. Bruderer. 1998. The relevance of wind for optimal migration theory. Journal of Avian Biology 29: 561 ?5 6 8. Lindstr?m, ?., S. Daan & G.H. Visser. 1994 . The conflict between molt and migratory fat deposition: a photoperiodic experiment with bluethroats. Animal Behaviour 48: 1173 ? 1 1 8 1. Lindstr?m, ?., G.H. Visser & S. Daan. 1993. The energetic cost of feather synthesis is proportional to basal metabolic rate. Physiological Zoology 66: 490 ? 5 1 0. ?omnicki, A. 1992. Population ecology from the individual perspective. Pages 3?1 7 in Individual-based Models and Approaches in Ecology: Populations, Communities, and Ecosystems. (D.L. DeAngelis & L.J. Gross, Eds.). Chapman and Hall, New York. Louren?o, P.M., R. Kentie, J. Schroeder, J.A. Alves, N.M. Groen, J.C.E.W. Hooijmeijer & T. Piersma. 2010. Phenology, stopover dyna mics and population size of migrating Black-tailed Godwits Limosa limosa limosa in Portuguese rice plantations. Ardea 98: 35 ? 42. Louren?o, P.M., R. Kentie, J. Schroeder, N.M. Groen, J.C.E.W. Hooijmeijer & T. Piersma. 2011. Repeatable timing of northward depart ure, arrival and breeding in Black -tailed Godwits Limosa l. limosa , but no domino effects. Journal of Ornithology 152: 1023 ? 1 0 3 2. Louren?o, P.M. & T. Piersma. 2008. Changes in the non-breeding distribution of continental black-tailed godwits Limosa limosa limosa over 50 years: a synthesis of surveys. Wader Study Group Bulletin 115: 91 ? 97. Lozano, G.A., S. Perreault & R.E. Lemon. 1996 . Age, arrival date and reproductive success of male American redstarts Setophaga ruticilla. Journal of Avian Biology 27: 164 ?170. 202 References Lundberg, S. & T. Alerstam. 1986. Bird migratio n patterns: conditions for stable geographical population segregation. Journal of Theoretical Biology 123: 403 ?414. Lyon, B.E. & R.D. Montgomerie. 1986. Delayed plumage matu ration in passerine birds ? reliable signaling by subordinate males. Evolution 40: 605 ? 615. Marks, J.S. 1993. Molt of bristle-thighed curlews in the northwestern Hawaiian Islands. Auk 110: 573 ? 587. Marks, J.S. & L.G. Underhill. 1994. Molt, mi gration and mass of a handicapped bristle- thighed curlew Numenius tahitiensis. Ardea 82: 153 ?1 5 5. Marra, P.P. 2000. The role of behavioral domi nance in structuring patterns of habitat occupancy in a migrant bird during the nonbreeding season. Behavioral Ecology 11: 299 ?3 0 8. Marra, P.P., C.M. Francis, R.S. Mulvihill & F.R. Moore. 2005. The infl uence of climate on the timing and rate of spring bird migration. Oecologia 142: 307 ? 315. Marra, P.P., K.A. Hobson & R.T. Holmes. 1998 . Linking winter and summer events in a migratory bird by using stable-carbon isotopes. Science 282: 1884 ? 1 8 8 6. Masero, J.A., F. Santiago-Quesada, J.M. S?nchez-Guzm?n, A. Villegas, J. Abad-G?mez, R.J. Lopes, V. Encarna??o, C. Corbacho & R. Mor? n. 2011. Long lengths of stay, large numbers, and trends of the Black-tailed Godwit Limosa limosa in rice fields during spring migration. Bird Conservation International 21: 12 ? 24. Mayr, E. 1963. Animal Species and Evolution. Oxford University Press, London. McCaffery, B.J. & R.E. Gill, Jr. 2001. Bar-tailed Godwit ( Limosa lapponica ). In The Birds of North America , vol. 581 (A. Poole & F. Gill, Eds.). The Birds of North America, Inc., Philadelphia. McCaffery, B.J., R.E. Gill, Jr., D. Melville, A. Riegen, P. Tomkovich, M. Dementyev, M. Sexson, R. Schuckard & S. Lovibond. 2010 . Variation in timing, behavior, and plumage of spring migrant Bar-tailed Godwits on the Yukon-Kuskokwim Delta, Alaska. Wader Study Group Bulletin 117: 179 ? 1 8 5. McGraw, K.J., G.E. Hill & R.S. Parker. 2005. The physiological costs of being colourful: nutritional control of carotenoid utilization in the American goldfinch, Carduelis tristis. Animal Behaviour 69: 653 ? 660. McKinnon, L., P.A. Smith, E. Nol, J.L. Martin, F.I. Doyle, K.F. Abraham, H.G. Gilchrist, R.I.G. Morrison & J. Bety. 2010. Lower predatio n risk for migratory birds at high latitudes. Science 327: 326 ? 3 2 7. McNamara, J.M. & A.I. Houston. 2008. Optimal annual routines: behaviour in the context of physiology and ecology. Philosophical Transactions of the Royal Society of London, Series B 363: 301 ?3 1 9. McNamara, J.M., R.K. Welham & A.I. Houston. 1998. The timing of migration within the context of an annual routine. Journal of Avian Biology 29: 416 ? 423. McWilliams, S.R. & W.H. Karasov. 2005. Mi gration takes guts: digestive physiology of migratory birds and its ecological significance. Pages 67 ? 7 8 in Birds of Two Worlds (R. Greenberg & P.P. Marra, Eds.). Smiths onian Institution Press, Washington, D.C. Meiri, S. & T. Dayan. 2003. On the validity of Bergmann?s rule. Journal of Biogeography 30: 331 ?3 5 1. Meltofte, H., T. H?ye, N. Schmidt & M. Forchhammer. 2007a. Differences in food abundance cause inter-annual variation in the breeding phenology of High Arctic waders. Polar Biology 30: 601 ?6 0 6. References 203 Meltofte, H., T. Piersma, H. Boyd, B. McCa ffery, B. Ganter, V.V. Golovnyuk, K. Graham, C.L. Gratto-Trevor, R.I.G. Morrison, E. Nol, H. R?sner, D. Schamel, H. Schekkerman, M.Y. Soloviev, P.S. To mkovich, D.M. Tracy, I. Tulp & L. Wennerberg. 2007b. Effects of climate var iation on the breeding ecology of Arctic shorebirds. Bioscience 59: 1 ? 48. MetVUW Weather and Climate Service. 2010. Available at: http://www.metvuw.com/ forecast/ Minton, C., J. Wahl, H. Gibbs, R. Jessop, C. Hassell & A. Boyle. 2011. Recoveries and flag sightings of waders which spend the non-breeding season in Australia. Stilt 59: 17 ? 43. Mitchell, G.W., N.T. Wheelwright, C.G. Gugliel mo & D.R. Norris. 2012. Short- and long- term costs of reproduction in a migratory songbird. Ibis 154: 325 ?3 3 7. M?ller, A.P. 1994. Phenotype-dependent arrival time and its consequences in a migratory bird. Behavioral Ecology and Sociobiology 35: 115 ?122. M?ller, A.P. 2001. Heritability of a rrival date in a migratory bird. Proceedings of the Royal Society of London, Series B 268: 203 ? 2 0 6. Mong, T.W. & B.K. Sandercock. 2009. Optim izing radio retention and minimizing radio impacts in a field study of upland sandpipers. Journal of Wildlife Management 71: 971 ?9 8 0. Moore, F.R. 1987. Sunset and the or ientation behaviour of migrating birds. Biological Reviews 62: 65 ? 86. Moore, F.R. & W. Yong. 1991 . Evidence of food-based competition among passerine migrants during stopover. Behavioral Ecology and Sociobiology 28: 85 ? 9 0. Morris, R.D. & G.P. Burness. 19 92. A new procedure for trans mitter attachment: effects on brood attendance and chick feedi ng rates by male common terns. Condor 94: 239 ?243. Mumme, R.L., M.L. Galatowitsch, P.G. Jablo?ski, T.M. Stawarczyk & J.P. Cygan. 2006. Evolutionary significance of geographic variation in a plumage-based foraging adaptation: an experimental test in the slate-throated redstart ( Myioborus miniatus ). Evolution 60: 1086 ? 1097. Murphy, M.E. & J.R. King. 1991. Nutritional aspects of avian molt. Pages 2186 ? 2 1 9 3 in Acta XX Congressus Internationalis Ornithologici (B.D. Bell, Ed.). New Zealand Ornithological Congress Tr ust Board, Wellington. Murphy, M.E., J.R. King & J. Lu. 1988. Malnut rition during the postnuptial molt of white- crowned sparrows ? feather growth and quality. Canadian Journal of Zoology 66: 140 3 ? 1413. Myers, J.P. 1981. Cross-seasonal interactions in the evolution of sandpiper social systems. Behavioral Ecology and Sociobiology 8: 195 ? 202. Myers, J.P. 1983. Space, time and the pattern of individual associations in a group-living species ? sanderlings have no friends. Behavioral Ecology and Sociobiology 12: 129 ?134. National Oceanic and Atmospheric Administ ration (NOAA), Earth System Research Laboratory. 2010a. Available at: http://www. esrl.noaa.gov/psd/data/gridded/data. ncep.reanalysis2.html National Oceanic and Atmospheric Administratio n (NOAA), National Climatic Data Center. 2010b. Available at: http://lwf.ncdc.no aa.gov/snow-and-ice/snow-cover.php Nakagawa, S. & H. Schielzeth. 2010. Repeatab ility for Gaussian and non-Gaussian data: a practical guide for biologists. Biological Reviews 85: 935 ?9 5 6. 204 References Nebel, S., D.L. Jackson & R.W. Elner. 2005. F unctional association of bill morphology and foraging behaviour in calidrid sandpipers. Animal Biology 55: 235 ?2 4 3. Nebel, S., D.B. Lank, P.D. O ? Hara, G. Fernandez, B. Haase, F. Delgado, F.A. Estela, L.J.E. Ogden, B. Harrington, B.E. Kus, J.E. Lyon s, F. Mercier, B. Ortego, J.Y. Takekawa, N. Warnock & S.E. Warnock. 2002. Western sandpipers ( Calidris mauri ) during the nonbreeding season: spatial segregation on a hemispheric scale. Auk 119: 922 ?928. Newton, I. 1966. The mo ult of the Bullfinch Pyrrhula pyrrhula. Ibis 108: 41 ?6 7. Niehaus, A. & R. Ydenberg. 2006. Ecological factors associated with the breeding and migratory phenology of high-latitude breeding western sandpipers. Polar Biology 30: 11 ? 17. Norris, D.R. 2005. Carry-over effects and habitat quality in migratory populations. Oikos 109: 178 ?1 8 6. Norris, D.R., P.P. Marra, T.K. Kyser, T.W. Sh erry & L.M. Ratcliffe. 2004. Tropical winter habitat limits reproductive success on the temperate breeding grounds in a migratory bird. Proceedings of the Royal Society of London, Series B 271: 59 ?6 4. Noskov, G., T. Rymkevich & N. Iovchenko. 1999. Intraspecific variation of moult: adaptive significance and ways of realisation. Pages 544 ? 5 6 3 in Proceedings of the 22nd International Ornithological Congress, Durban (N. Adams & R. Slotow, Eds.). BirdLife South Africa, Johannesburg. O ? Malley, K.G. & M.A. Banks. 2008. A latit udinal cline in the Chinook salmon ( Oncorhynchus tshawytscha ) clock gene: evidence for selection on PolyQ length variants. Proceedings of the Royal Society of London, Series B 275: 2813 ? 2821. O ? Malley, K.G., M.J. Ford & J.J. Hard. 2010. Clock polymorphism in Pacific salmon: evidence for variable selection along a latitudinal gradient. Proceedings of the Royal Society of London, Series B 277: 3703 ?3 7 1 4. Obrecht, H.H., C.J. Pennycuick & M.R. Fuller. 19 88. Wind-tunnel experiments to assess the effect of back-mounted radio transmitters on bird body drag. Journal of Experimental Biology 135: 265 ?2 7 3. Payne, R.B. 1972. Mechanisms and control of molt. Pages 103 ? 1 5 5 in Avian Biology , vol. 2 (D.S. Farner & J.R. King, Eds.). Academic Press, New York. Pennycuick, C.J. & P.F. Battley. 2003. Burning th e engine: a time-marching computation of fat and protein consumption in a 5420 -km non-stop flight by great knots, Calidris tenuirostris. Oikos 103: 323 ? 332. Peters, A., L.B. Astheimer, C.R.J. Boland & A. Coc kburn. 2000. Testosteron e is involved in acquisition and maintenance of sexually selected male plumage in superb fairy-wrens, Malurus cyaneus. Behavioral Ecology and Sociobiology 47: 438 ? 445. Phillips, R.A., J.C. Xavier, J.P. Croxall & A.E. Burger. 2009. Effects of satellite transmitters on albatrosses and petrels. Auk 120: 1082 ? 1090. Piersma, T. 1987. Hop, skip or jump? Constr aints on migration of arctic waders by feeding, fattening, and flight speed. Limosa 60: 185 ? 194. Piersma, T. 2002. When a year takes 18 mont hs: evidence for a strong circannual clock in a shorebird. Naturwissenschaften 89: 278 ? 2 7 9. Piersma, T., M. Brugge, B. Spaans & P.F. Ba ttley. 2008. Endogenous circannual rhythmicity in body mass, molt, and plumage of great knots ( Calidris tenuirostris). Auk 125: 140 ?148. Piersma, T. & J. Drent. 2003. Phenotypic flexib ility and the evolution of organismal design. Trends in Ecology & Evolution 18: 228 ? 2 3 3. References 205 Piersma, T. & R.E. Gill, Jr. 1998. Guts don't fly: small digestive organs in obese bar-tailed godwits. Auk 115: 19 6 ? 203. Piersma, T., G.A. Gudmundsson & K. Lilliendahl. 1999. Rapid changes in the size of different functional organ and muscle groups during refueling in a long-distance migrating shorebird. Physiological and Biochemical Zoology 72: 405 ? 415. Piersma, T. & J. Jukema. 1990. Budgeting the fli ght of a long-distance migrant: changes in nutrient reserve levels of bar-tailed godwit s at successive spring staging sites. Ardea 78: 315 ?3 3 7. Piersma, T. & J. Jukema. 1993. Red breasts as honest signals of migratory quality in a long- distance migrant, the bar-tailed godwit. Condor 95: 163 ? 177. Piersma, T., M. Klaassen, J.H. Bruggemann, A.M. Blomert, A. Gueye, Y. Ntiamoa-Baidu & N.E. van Brederode. 1990a. Seasonal timing of the spring departure of waders from the Banc d?Arguin, Mauritania. Ardea 78: 123 ? 134. Piersma, T., L. Mendes, J. Hennekens, S. Ratiarison, S. Groenewold & J. Jukema. 2001. Breeding plumage honestly signals likelihood of tapeworm infestation in females of a long-distance migrating shorebird, the bar-tailed godwit. Zoology 104: 41 ?4 8. Piersma, T., J. P?rez-Tris, H. Mouritsen, U. Bauchinger & F. Bair lein. 2005. Is there a ?migratory syndrome? common to all migrant birds? Annals of the New York Academy of Sciences 1046: 282 ? 293. Piersma, T., L. Zwarts & J.H. Bruggemann. 19 90b. Behavioral aspects of the departure of waders before long-distance flights ? flocking, vocalizations, flight paths and diurnal timing. Ardea 78: 157 ?1 8 4. Potti, J. 1998. Arrival time from spring migr ation in male pied flycatchers: individual consistency and familial resemblance. Condor 100: 702 ? 708. Powell, M. 2011. Seasonal and spatial variation of Nicon aestuariensis (Polychaeta: Nereididae) at the Manawatu Estuary, New Zealand. Honours thesis, Massey University. Preuss, N.O. 2001. Hans Chris tian Cornelius Mortensen: aspects of his life and of the history of bird ringing. Ardea 89: 1 ? 6. Prop, J., J.M. Black & P. Shimmings. 2003 . Travel schedules to the high arctic: barnacle geese trade-off the timing of migration with accumulation of fat deposits. Oikos 103: 403 ?4 1 4. Pulido, F. 2007. The genetics and evolution of avian migration. Bioscience 57: 165 ?1 7 4. Pulido, F., P. Berthold, G. Mo hr & U. Querner. 2001. Heritability of the timing of autumn migration in a natural bird population. Proceedings of the Royal Society of London, Series B 268: 953 ?9 5 9. Rappole, J.H. & A.R. Tipton. 1991. New harnes s design for attachment of radio transmitters to small passerines. Journal of Field Ornithology 62: 335 ?3 3 7. Rees, E.C. 1989. Consistency in the timing of migration for individual Bewick?s swans. Animal Behaviour 38: 384 ? 3 9 3. Reudink, M.W., C.E. Studds, P.P. Marra, T. Kurt Kyser & L.M. Ratcliffe. 2009. Plumage brightness predicts non-breedin g season territory quality in a long-distance migratory songbird, the American redstart Setophaga ruticilla. Journal of Avian Biology 40: 34 ?41. Riegen, A.C. 1999. Movements of banded Arctic waders to and from New Zealand. Notornis 46: 123 ?1 4 2. 206 References Rohwer, S., R.E. Ricklefs, V.G. Rohwer & M.M. Copple. 2009. Allometry of the duration of flight feather molt in birds. PLoS Biology 7: e100013 2. Rotella, J.J., D.W. Howerter, T.P. Sankowski & J.H. Devries. 1993. Ne sting effort by wild mallards with 3 types of radio transmitters. Wildlife Society Bulletin 57: 690 ? 695. Rowland, H.M. 2009. From Abbott Thayer to the present day: what have we learned about the function of countershading? Philosophical Transactions of the Royal Society of London, Series B 364: 519 ? 5 2 7. Rynn, S. 1982. A revision of the taxonomy of the genus Limosa. PhD thesis, Liverpool Polytechnic. Schaub, M., F. Liechti & L. Jenni. 2004. Departure of migrating European robins, Erithacus rubecula , from a stopover site in r elation to wind and rain. Animal Behaviour 67: 229 ?2 3 7. Schneider, D.C. & B. Harrington. 1981. Timing of shorebird migration in relation to prey depletion. Auk 98: 801 ? 811. Schroeder, J., P.M. Louren?o, J.C.E.W. Hoo ijmeijer, C. Both & T. Piersma. 2009. A possible case of contemporary selection leading to a decrease in sexual plumage dimorphism in a grassland-breeding shorebird. Behavioral Ecology 20: 797 ?8 0 7. Schroeder, J., T. Piersma, N. Groen, J. Hooijmeijer, R. Kentie, P. Louren?o, H. Schekkerman & C. Both. 201 2. Reproductive timing and inve stment in relation to spring warming and advancing agricultural schedules. Journal of Ornithology 153: 327 ?3 3 6. Senar, J.C. & D. Escobar. 2002. Carotenoid derived plumage coloration in the siskin Carduelis spinus is related to foraging ability. Avian Science 2: 19 ? 2 4. Serra, L. 2001. Duration of primary moult a ffects primary quality in grey plovers Pluvialis squatarola. Journal of Avian Biology 32: 377 ? 3 8 0. Serra, L. & L.G. Underhill. 2006. The regulatio n of primary molt speed in the grey plover, Pluvialis squatarola. Acta Zoologica Sinica 52: 451 ?4 5 5. Serra, L., D.A. Whitelaw, A.J. Tree & L.G. Un derhill. 1999. Moult, mass and migration of grey plovers Pluvialis squatarola wintering in South Africa. Ardea 87: 71 ? 81. Shaffer, S.A., Y. Tremblay, H. Weimerskirch, D. Scott, D.R. Thompson, P.M. Sagar, H. Moller, G.A. Taylor, D.G. Foley, B.A. Block & D.P. Costa. 2006. Migratory shearwaters integrate oceanic resources across the Pacific Ocean in an endless summer. Proceedings of the National Academy of Sciences 103: 12799 ? 12802. Shamoun-Baranes, J., J. Leyrer, E. van Loon, P. Bocher, F. Robin, F. Meunier & T. Piersma. 2010. Stochastic atmospheric assistance and the use of emergency staging sites by migrants. Proceedings of the Royal Society of London, Series B 277: 1505 ? 1511. Sherry, T.W. & R.T. Holmes. 1997. American Redstart ( Setophaga ruticilla ). In The Birds of North America , vol. 277 (A. Poole & F. Gill, Eds.). The Birds of North America, Inc., Philadelphia. Shuster, S.M. & M.J. Wade. 1991. Equal m ating success among male reproductive strategies in a marine isopod. Nature 350: 608 ? 610. Smith, P.A., H.G. Gilchrist, M.R. Forbes, J. Ma rtin & K. Allard. 2010. Inter-annual variation in the breeding chronology of arctic shorebirds: effects of weather, snow melt and predators. Journal of Avian Biology 41: 292 ? 3 0 4. Smith, R.J. & F.R. Moore. 2005 . Arrival timing and seasonal reproductive performance in a long-distance migratory landbird. Behavioral Ecology and Sociobiology 57: 231 ? 239. References 207 Snyder, N.F.R., E.V. Johnson & D.A. Clendenen. 1987. Primary molt of California condors. Condor 89: 468 ? 485. Southey, I. 2009. Numbers of waders in New Zealand 1994 ? 2003. In Research & Development Series 308 , 70 pp. New Zealand Department of Conservation. Stephens, D.W. & J.R. Krebs. 1986. Foraging Theory. Princeton University Press, Princeton. Studds, C.E. & P.P. Marra. 2007. Linking flu ctuations in rainfall to nonbreeding season performance in a long-distance migratory bird, Setophaga ruticilla. Climate Research 35: 115 ?1 2 2. Studds, C.E. & P.P. Marra. 2011. Rainfall-indu ced changes in food availability modify the spring departure programme of a migratory bird. Proceedings of the Royal Society of London, Series B 278: 3437 ? 3443. Stutchbury, B.J.M., S.A. Tarof, T. Done, E. Gow, P.M. Kramer, J. Tautin, J.W. Fox & V. Afanasyev. 2009. Tracking long-distance songbird migration by using geolocators. Science 323: 896. Sullivan, K.A. 1989. Predation a nd starvation: age-specific mortality in juvenile juncos ( Junco phaenotus ). Journal of Animal Ecology 58: 275 ? 286. Summers, R.W., R.W. Swann & M. Nicoll. 1983. The effects of methods on estimates of primary moult duration in the Redshank Tringa totanus. Bird Study 30: 149 ?1 5 6. Summers, R.W., L.G. Underhill, M. Nicoll, K.-B. Strann & S.?. Nilsen. 2004. Timing and duration of moult in three populations of Purple Sandpipers Calidris maritima with different moult/migration patterns. Ibis 146: 394 ?4 0 3. Swaddle, J.P. & M.S. Witter. 1997 . The effects of molt on the flight performance, body mass, and behavior of European starlings ( Sturnus vulgaris ) : an experimental approach. Canadian Journal of Zoology 75: 1135 ? 1 1 4 6. Swaddle, J.P., M.S. Witter, I.C. Cuthill, A. Bud den & P. McCowen. 1996. Plumage condition affects flight performance in Common Starlings: implications for developmental homeostasis, abrasion and moult. Journal of Avian Biology 27: 103 ? 1 1 1. Swarth, H.S. 1920. Revision of the avian genus Passerella with special reference to the distribution and migration of the races in California. University of California Publications in Zoology 21: 75 ?2 2 4. Sykes, P.W., Jr., J.W. Carpenter, S. Holzma n & P.H. Geissler. 1990. Evaluation of three miniature radio transmitter attachment methods for small passerines. Wildlife Society Bulletin 18: 41 ? 48. Symonds, M.R.E. & G.J. Tattersall. 2010. Geogra phical variation in bill size across bird species provides evidence for Allen?s rule. American Naturalist 176: 188 ? 197. Sz?kely, T., J.D. Reynolds & J. Figuerola. 2000 . Sexual size dimorphism in shorebirds, gulls, and alcids: the influence of sexual and natural selection. Evolution 54: 1404 ? 1413. Tauber, E. & C.P. Kyriacou. 2005. Molecular ev olution and population genetics of circadian clock genes. Methods in Enzymology 393: 797 ? 8 1 7. Thomas, D.G. & A.J. Dartnall. 1971. Moult of the curlew sandpiper in relation to its annual cycle. Emu 71: 153 ? 158. Thomson, D.L., S.R. Baillie & W.J. Peach. 19 99. A method for studying post-fledging survival rates using data from ringing recoveries. Bird Study 46 (suppl.): 104 ? 111. Tomkovich, P.S. 2010. Assessment of the Anad yr Lowland subspecies of bar-tailed godwit Limosa lapponica anadyrensis. Bulletin of the British Ornithologists Club 130: 88 ?95. 208 References Tulp, I. 2007. The arctic pulse: timing of breeding in long-distance migrant shorebirds. PhD thesis, University of Groningen. Tulp, I., S. McChesney & P. De Goeij. 1994. Migratory departures of waders from north- western Australia ? behavior, timing and possible migration routes. Ardea 82: 201 ?221. Tulp, I. & H. Schekkerman. 2008. Has prey availability for Arctic birds advanced with climate change? Hindcasting the abundance of tundra arthropods using weather and seasonal variation. Arctic 61: 48 ? 60. Underhill, L.G. 2003. Within ten feathers: pr imary moult strategies of migratory waders (Charadrii). Pages 187 ? 197 in Avian Migration (P. Berthold, E. Gwinner & E. Sonnenschein, Eds.). Springer-Verlag, Berlin. Underhill, L.G. & R.W. Summers. 1993. Relativ e masses of primary feathers in waders. Wader Study Group Bulletin 71: 29 ?3 1. Underhill, L.G. & W. Zucchini. 1988. A model for avian primary moult. Ibis 130: 358 ?3 7 2. Underhill, L.G., W. Zucchini & R.W. Summers. 1990. A model for avian primary moult-data types based on migration strategies and an example using the Redshank Tringa totanus. Ibis 132: 118 ? 123. van de Kam, J., P.F. Battley, B.J. McCaffery, D. Rogers, J.-S. Hong, N. Moores, J. Yung-Ki, J. Lewis & T. Piersma. 2010. Invisible Connections: Why Migrating Shorebirds Need the Yellow Sea. CSIRO Publishing, Co llingwood, Australia. van de Pol, M. & J. Wright. 2009. A simple method for distinguishing within- versus between-subject effects using mixed models. Animal Behaviour 77: 753 ? 758. van Noordwijk, A.J., F. Pulido, B. Helm, T. Copp ack, J. Delingat, H. Dingle, A. Hedenstr?m, H. van der Jeugd, C. Marchetti, A. Nilsso n & J. P?rez-Tris. 2006. A framework for the study of genetic variation in migratory behaviour. Journal of Ornithology 147: 221 ?2 3 3. Vardanis, Y., R.H.G. Klaassen, R. Strandberg & T. Alerstam. 2011. Individuality in bird migration: routes and timing. Biology Letters 7: 502 ?5 0 5. Verhulst, S. & J.M. Tinbergen. 1991. Expe rimental evidence for a causal relationship between timing and success of reproduction in the great tit Parus m. major. Journal of Animal Ecology 60: 269 ?282. Verhulst, S., J.H. van Balen & J.M. Tinbe rgen. 1995. Seasonal decline in reproductive success of the great tit ? variation in time or quality? Ecology 76: 2392 ?2 4 0 3. Warnock, N. & M.A. Bishop. 1998. Spring stop over ecology of migrant western sandpipers. Condor 100: 456 ?4 6 7. Warnock, N. & R.E. Gill, Jr. 1996. Dunlin ( Calidris alpina ). In The Birds of North America , vol. 203 (A. Poole & F. Gill, Eds.). The Bi rds of North America, Inc., Philadelphia. Warnock, N., G.W. Page & B.K. Sandercock. 19 97. Local survival of du nlin wintering in California. Condor 99: 906 ? 915. Warnock, N. & J.Y. Takekawa. 2003. Use of ra dio telemetry in studies of shorebirds: past contributions and future directions. Wader Study Group Bulletin 100: 138 ?1 5 0. Warnock, N., J.Y. Takekawa & M.A. Bishop. 2004. Migration and stopover strategies of individual dunlin along the Pacific coast of North America. Canadian Journal of Zoology 82: 1687 ?1 6 9 7. Warnock, N. & S. Warnock. 1993. Attachment of radio-transmitters to sandpipers: review and methods. Wader Study Group Bulletin 70: 28 ?3 0. References 209 Wash Wader Ringing Group (WWRG). 2011. Available at: http://freespace.virgin.net/holme. vale/WaderLongevity.htm Weber, T.P., T. Alerstam & A. Hedenstr?m. 1998 a. Stopover decisions under wind influence. Journal of Avian Biology 29: 552 ? 5 6 0. Weber, T.P., B.J. Ens & A.I. Houston. 1998b. Op timal avian migration: a dynamic model of fuel stores and site use. Evolutionary Ecology 12: 377 ? 4 0 1. Weber, T.P. & A. Hedenstr?m. 2000. Optimal st opover decisions under wind influence: the effects of correlated winds. Journal of Theoretical Biology 205: 95 ? 1 0 4. Wenink, P.W., A.J. Baker & M.G. Tilanus. 19 93. Hypervariable-control-region sequences reveal global population structuring in a long-distance migrant shorebird, the Dunlin ( Calidris alpina ). Proceedings of the National Academy of Sciences 90: 94 ? 98. Wiggins, D.A., T. Part & L. Gustafsson. 19 94. Seasonal decline in collared flycatcher Ficedula albicollis reproductive success: an experimental approach. Oikos 70: 359 ?364. Wilson, D.S. 1998. Adaptive individual di fferences within single populations. Philosophical Transactions of the Royal Society of London, Series B 353: 199 ? 205. Wilson, D.S., A.B. Clark, K. Coleman & T. Dearstyne. 1994. Shyness and boldness in humans and other animals. Trends in Ecology and Evolution 9: 442 ? 446. Wilson, J.R., S. Nebel & C.D.T. Minton. 2007. Migration ecology and morphometrics of two bar-tailed godwit populations in Australia. Emu 107: 262 ? 274. Wingfield, J.C. 2008. Organization of vertebrate annual cycles: implications for control mechanisms. Philosophical Transactions of the Royal Society of London, Series B 363: 425 ? 441. Witter, M.S. & I.C. Cuthill. 1993. The eco logical costs of avian fat storage. Philosophical Transactions of the Royal Society of London, Series B 340: 73 ?9 2. Wood, B. 1992. Yellow Wagtail Motacilla flava migration from West Africa to Europe: pointers towards a conservation strategy for migrants on passage. Ibis 134: 66 ?76. Woodrey, M.S. & F.R. Moore. 1997. Age-related differences in the stopover of fall landbird migrants on the coast of Alabama. Auk 114: 695 ?7 0 7. Zar, J.H. 1999. Biostatistical Analysis, 4th ed. Prentice Hall, Englewood Cliffs, New Jersey. Zink, R.M. & J.V. Remsen, Jr. 1986. Evolutionary processes and patterns of geographic variation in birds. Current Ornithology 4: 1 ?6 9. Zwarts, L., A.M. Blomert & R. Hupkes. 1990a. In crease of feeding time in waders preparing for spring migration from the Banc d?Arguin, Mauritania. Ardea 78: 237 ? 256. Zwarts, L., B.J. Ens, M. Kersten & T. Piersma. 1990b. Molt, mass and flight range of waders ready to take off for long-distance migrations. Ardea 78: 339 ?3 6 4.