Spatiotemporal Group Dynamics in a Long-Distance Migratory Bird

Spatiotemporal group dynamics in a long-distance migratory bird 1 Kiran L. Dhanjal-Adams 1,2,*, Silke Bauer 1, Tamara Emmenegger 1,3, Steffen Hahn 1, Simeon 2 Lisovski 1, Felix Liechti 1 3 1 Department of Bird Migration, Swiss Ornithological Institute, Seerose 1, 6204 Sempach, 4 Switzerland; 5 2 Lead contact; 6 3 Institute of Integrative Biology, ETH Zurich, Universitätstrasse 16, 8092 Zürich, Switzerland 7 * Correspondence: [email protected] 8 Key words: Animal migration; Social interactions; Group decision-making; Pressure sensor; PAM 9 logger; European bee-eater, Merops 10 11 Summary 12 Thousands of species migrate [1]. Though we have some understanding of where and when they 13 travel, we still have very little insight into who migrates with whom and for how long. Group 14 formation is pivotal in allowing individuals to interact, transfer information and adapt to changing 15 conditions [2]. Yet it is remarkably difficult to infer group membership in migrating animals without 16 being able to directly observe them. Here, we use novel lightweight atmospheric pressure loggers to 17 monitor group dynamics in a small migratory bird, the European bee-eater (Merops apiaster). We 18 present the first evidence of a migratory bird flying together with non-kin of different ages and sexes 19 at all stages of the life cycle. In fact, 49% stay together throughout the annual cycle, never separating 20 longer than 5 days at a time despite the ~14,000 km journey. Of those that separated for longer, 89% 21 reunited within less than a month with individuals they had previously spent time with, having flown 22 up to 5,000 km apart. These birds were not only using the same non-breeding sites but displayed 23 coordinated foraging behaviours – these are unlikely to result from chance encounters in response to 24 the same environmental conditions alone. Better understanding of migratory group dynamics, using 25 the presented methods, could help improve our understanding of collective decision-making during 26 large scale movements. 27 28 Results 29 From zebras [3] to monarch butterflies [4], migratory species undertake some of the most extreme 30 feats of endurance known in the animal kingdom. With the advent of novel tracking technologies, we 31 are gradually completing the picture of where and when they travel [5]. However, without being able 32 to directly observe migration [e.g. 2], we have very little knowledge of who might migrate with whom. 33 Migratory species are notable for their propensity to aggregate in large numbers. The stability of 34 migratory groups over time can be important in determining survival [6], navigational accuracy [7], 35 migratory speed [8], transfer of information [7] and new migratory behaviours [2]. However, 36 migrating with others is not without risk, as it can increase both disease prevalence [9] and resource 37 competition [10]. Group size typically fluctuates over time and space, with individuals coming 38 together and separating [11; hereafter termed “fission-fusion dynamics”] as they trade-off the different 39 benefits and costs of cooperation [11,12]. Indeed, resource patches are distant, seasonal and often 40 unpredictable. One slow individual could, for instance, force the entire group to slow down and miss 41 peaks in resource availability, creating conflict [11]. Groups can therefore either compromise to 42 remain together, or spilt into sub-groups, for example of different migratory speeds. 43 Fission-fusion can occur without individuals being able to “recognise” each other per se [11]. The 44 same individuals could encounter each other again and again at the same site as a result of migratory 45 connectivity, simply because it is the only one available to them at a particular period [13–15]. Under 46 such circumstances, resource bottlenecks are likely driving group fusion, not social relationships [14]. 47 On the other hand, where resources occur broadly over a large area, animals must coordinate decisions 48 to fuse into a long-term group, especially if they regularly fissure and must find each other again 49 [13,16]. Only species with high social cognition, such as elephants [17], dolphins [18] and bats [19] 50 have been found to form long-term social bonds by coordinating decisions, despite separations 51 imposed by migration. In birds, long-term social bonds despite fission-fusion dynamics have been 52 observed between non-migratory non-kin [20,21], migratory kin [22], or migratory bonded-pairs [23]. 53 Long-term social bonds, despite fission-fusion dynamics are poorly understood in non-kin migratory 54 birds. 55 Here, we use novel lightweight (~1.4g) multi-sensor loggers to track the spatiotemporal pattern of 56 group cohesion between 29 European bee-eaters (Figure S1A; Merops apiaster) over the annual cycle. 57 Indeed, European bee-eater are gregarious. They can breed cooperatively, making complex decisions 58 on whether to help another breeding pair, which pair to help and how much [24]. The species also 59 forages socially [25], and can cooperate with other bee-eater species to mob predators, preen and 60 forage [26] In the non-breeding grounds, they form vocal flocks of 8 individuals (on average with a 61 range of 5-40 individuals based on e-bird data from the non-breeding grounds [27]), 8-39 during 62 stopover [28] ands of 30-100 during migration [29,30]. However, what is less well established is how 63 gregariousness might change over time More specifically, we aim to determine (i) whether birds from 64 the same colony have similar migratory routes, (ii) whether they remain together during migration, 65 (iii) what the stability and (iv) composition of these groups might be. 66 We used novel multi-sensor loggers to measure both light for geolocation, and ambient air pressure for 67 altitudinal changes during the annual cycle (2015 to 2016 and 2016 to 2017). To confirm potential 68 groups suggested by geolocation overlap (Figures 1 vs S1B), we applied a hidden Markov model 69 (HMM) to ambient air pressure measurements and identified periods of synchronisation in altitudinal 70 changes between birds (Figures 2 and S2). Indeed, altitudinal changes can easily be identified from the 71 pressure measurements: background variations in pressure driven by weather are less than 2 hPa per 72 hour, while those caused by flight range from 2-205 hPa per hour (equivalent to a change in altitude of 73 16.89-1934.97m assuming a starting pressure of 1000 hPa at 15°C; Figures 2 and S2E-F). We assume 74 that if these highly dynamic altitudinal changes are synchronous, then the decision to fly/not fly, to go 75 up/down, how high/low is coordinated between individuals. Thus, if some individuals made the same 76 decision at the same time repeatedly, especially over weeks or months, the decision must have been 77 shared between individuals flying within the same flock. 78 To test the method, we then compared birds within the same breeding colony (Figure 3A) and found 79 that even birds that were nesting within 500m from each were not always classified as having similar 80 pressure signatures (Figures 3A and 3E). Thus, the observed patterns are likely driven by behaviour, 81 not overestimated dur to geographic proximity or weather fronts (Figure S2E and F). 82 Even within a relatively small sample size of 29 tagged individuals recaptured between 2016 and 83 2017, 89% formed long-term groups with one or more other tagged individuals outside the breeding 84 grounds (Figures 3, 4, S3 and S4). Many groups formed in the breeding grounds prior to migration 85 (Figures 3A and 4) with none of the recaptured individuals having bred together before (Table S1). In 86 total, we identified one group of five individuals (group 1), one group of four (group 5), and six 87 groups of two (groups 2, 3, 4, 6, 7 and 8; Figures 1, 3 and 4). The group of four (group 5 i.e. 19% of 88 grouped birds) persisted throughout the annual cycle, covering 14,000 km together (Figures 1, 3 and 89 4). HMMs never classified these individuals as having separated during migration (Figure 4). Only, 90 during the non-breeding residency period did we observe individuals breaking into subgroups for short 91 periods of no longer than 5 days (e.g. 2-6th November 2016; Figure S3K). 92 For two groups (1 and 6 i.e. 33% of grouped birds), fission occurred during southward migration for 5 93 and 4 days respectively (Figures S4A, B, M and N). Group 1 fissured into two sub-groups while 94 crossing France, while group 6 fissured while crossing Algeria (Figures S4A and M). Both groups 95 fused again to remain stable during the rest of migration, crossing the Sahara and spending their non-96 breeding residency repeatedly coming together and separating (Figures 4A and F). Group 1 97 occasionally formed subgroups for a maximum of 9 days before fusing again (Figure S3G). Group 6 98 only separated for 1 or 2 days at a time (Figure S3L). Group 1 then migrated north to the breeding 99 grounds as a stable group, without separating (Figures 3 and 4). For group 6, fission-fusion dynamics 100 remain unknown because the pressure logger on individual OF failed during the non-breeding season 101 (Table S1). 102 All other groups (2, 3, 4, 7 and 8 i.e. 48% of grouped birds) started migration from their breeding 103 grounds to their non-breeding grounds together (Figure 4). Of these bonded birds, 80% (groups 2, 3, 4 104 and 7) parted from their flight partner while crossing the Sahara (Figures 4 and S4). Of these separated 105 birds, 80% (groups 2, 3 and 4) then came back together, having migrated up to 5,000 km over one 106 month separately, in their non-breeding grounds spread across Cameroon, Equatorial Guinea, Gabon, 107 Congo, Democratic Republic of Congo and Angola (Figures 3, 4, S3 and S4D, E and I). Pressure 108 loggers failed on both individuals in group 4, however, groups 2 and 3 then started migrating north to 109 the breeding grounds together, but separated after crossing the Sahara, only meeting again in the 110 breeding colony (Figures 4 and S4H and L). 111 Also, 17% of birds did not migrate with any tagged birds, but repeatedly joined a group in the non-112 breeding grounds (Figures 3C and S3A-F; UT and TV joined TY-UK, AA sometimes joined TY-UK 113 and sometimes TZ-UG, and SH and RZ joined BL-RH-TQ-TW-UO). In fact, two groups (3 and 4) 114 occasionally foraged together in the non-breeding grounds, particularly with UG foraging more often 115 with AA than its migratory partner TZ (Figures 3 and S3F), and AA in turn foraging with TY-UK-UT 116 (Figures 3C, S3B and S3E). Most of these birds were already classed as having foraged together in 117 the breeding grounds prior to migration (Figure 3). 118 Two breeding pairs formed after migration together: TQ-UO from group 1 and OO-OI from group 5 119 (Table S1). In fact, UO switched colonies from 2016 to 2017 to breed with TQ, though both birds 120 already foraged together in the breeding grounds in 2016, as did OO and OI in 2015 (Table S1 and 121 Figure 3). Neither pair bred together in the year before they were tagged, suggesting that these 122 migratory groups formed independently from pair formation the previous year (Table S1). In total, 8 123 birds switched breeding colonies, 5 of which moved to the colony of their travel companion (i.e. UO-124 TW to the breeding colony of BL-TQ-RH, SJ to SO, TY to UK, and TZ to UG; Table S1 and Figure 125 3A vs 3E). All in all, group formation was not consistent with age or sex, and no birds were ever 126 ringed or tagged within the same burrow before this study, indicating they were not likely kin or 127 previously bonded pairs (Table S1). Indeed, roughly 80% of the juveniles from these colonies have 128 been ringed since 2003 and over 95% since 2007 [31]. 129 130 Discussion 131 Without physically following birds with an ultralight aircraft [e.g. 7], it has previously been 132 impossible to monitor spatiotemporal group dynamics in small migrating birds. Here, we show how 133 novel lightweight multi-sensor loggers can be used to better understand who migrates with whom at 134 all stages of the annual cycle. Indeed, our analyses provide strong evidence for long-term group 135 formation in a small migratory bird both during migration and in the non-breeding grounds, between 136 non-kin of mixed age and sex. Though our results do not exclude the possibility of tagged birds 137 forming groups with non-tagged kin, our sample size only included non-kin. This is particularly rare 138 between non-kin, as there is no direct genetic benefit to be gained from remaining together over long 139 periods. In fact, this is some of the first evidence of migratory birds remaining in long-term non-kin 140 groups throughout all stages of the annual cycle. Despite evidence of waterbirds migrating in non-kin 141 groups, most research indicates that these groups still separate into family or same sex and age sub-142 groups in the non-breeding grounds, most frequently unpaired juvenile [2,7,32,33]. 143 During migration, theory suggests that stable groups may arise as a result of environmental 144 bottlenecks or social interactions [34], with the importance of sociality increasing with decreasing 145 group size [34,35]. Given that hundreds of bee-eaters migrate simultaneously in flocks of 5-39 146 individuals [27–30] and that they encounter difficult flight conditions [36], we expected high fission-147 fusion [11,34]. Indeed, soar-gliding requires birds to identify suitable thermal updrafts, adjust their 148 speed to navigate within the updraft and then find the right moment to leave with enough momentum 149 to get to the next updraft [37]. Older individuals are therefore better at navigating this challenge than 150 younger individuals [37] and species such as storks rarely remain together long-term despite short-151 term coordination [38,39]. It is therefore surprising that all birds remained together during these 152 periods of rapid altitudinal changes for a minimum of 3 weeks, and 45% during the entire migratory 153 period, hinting at some social aspects to group stability [34]. Though our data cannot directly measure 154 sociability, it is well documented in the species at different stages of the annual cycle [24–26,36], 155 Surprisingly, of the separated migratory groups 89% reformed again in the Congo Basin [40], an area 156 of roughly 4 million km2 with individuals they had previously interacted with in the breeding grounds 157 or on migration Figures 3 and 4). To some degree, non-breeding range can be genetically driven [41], 158 forcing birds into the same region where they form groups due to proximity. For this population 159 however, the non-breeding ranges are not necessarily overlapping (Figure S1C-J), and sparsely spread 160 out over thousands of kilometres between Gabon and Angola [40]. Given (i) the lack of resource 161 bottlenecks in the region which might force all birds into the same tree or waterhole[11,34], (ii) the 162 fact that non-breeding flocks are relatively small (average size of 8 [27]) [35], and (iii) that separated 163 individuals primarily reunited with individuals they had previously spent time with – suggests these 164 reunions may not have occurred by chance. Indeed, the only individual which was tagged over two 165 years (OO in 2016 and TO in 2017) returned to the same breeding site both years, suggesting that 166 individuals could be returning to sites that they had used with other flock members in the past. 167 However, the mechanisms by which separated individuals reunited despite long separations remains to 168 be elucidated. 169 The benefits of cooperation, both in the non-breeding grounds and during migration, may explain the 170 need to reach consensus decisions by maintain long-term groups with non-kin. Indeed, within the non-171 breeding grounds, grouping can help with predator detection and competition for prime feeding areas, 172 thus increasing fitness and reducing stress levels [33]. Not only can this increase survival, but it can 173 also help maintain a better body condition during migration and increase later reproductive success. 174 During migration, flocking can increase navigational accuracy [7,42] either through social learning, 175 where experienced individuals guide less experienced individuals [2], or through collective learning, 176 where groups pool their knowledge to generate better migratory decisions than solitary individuals 177 [43]. 178 Whether through collective or social learning, being able to transfer information within a group to 179 identify new non-breeding sites, allows birds to respond to environmental changes [2,7]. This could 180 potentially be the case for our study population whose migratory range has rapidly expanded, with 181 new breeding and non-breeding sites appearing in Europe and the Congo Basin respectively ([40]when 182 birds were previously only known to migrate to Western and South Eastern Africa [41]). Given the 183 stability of these non-kin groups, and the rapid emergence of new migratory routes, it is possible that 184 social transfer of information could, in combination with phylogenetic plasticity, be affecting this 185 change. Indeed, though phylogenetic plasticity can allow populations to change migratory routes over 186 generations, behavioural plasticity can allow these changes to occur within the lifespan of an 187 individual. 188 Overall however, migratory birds are declining more severely than non-migratory birds [44]. Given 189 the current rate and extent of anthropogenic driven changes, adaptability could be key in averting 190 population declines. Disentangling the relative roles of genetic, social and environmental factors in 191 migration could help understand how collective decision-making affects large-scale movements, and 192 how new migratory routes might (or might not) arise from social transfer of information, and thus how 193 adaptable a species might be to a changing environment. 194 Conclusions 195 In conclusion, we find that (i) birds from the same colony do not always follow the same migratory 196 routes but will in fact join with birds from nearby colonies post-breeding to (ii) form groups which 197 migrate together. Groups are generally (iii) stable during migration. However, if groups separate, they 198 will reunite in the non-breeding grounds to form dynamic groups which repeatedly forage together, 199 sometimes separating for 1-5 days at a time before migrating back to the breeding grounds together. 200 Most surprisingly, these groups showed (iv) no age or sex structure and consisted of non-kin. Our 201 research is the first to show such behaviour between migratory non-breeding non-kin bird groups, 202 displaying rare spatiotemporal group dynamics more often observed in mammals [17,19]. 203 204 Acknowledgements 205 Funding was provided by Swiss National Science Foundation 31003A_160265 to SB and SH. The 206 Swiss federal office for the environment (FOEN) contributed financial support for the development of 207 the tags (grant UTF 400.34.11). We thank M. Schulz and P. Tamm for field assistance and long-term 208 monitoring. SOI-GDL3pam loggers were fitted under licence LAU 43.17-22480-58/2015. 209 210 Author contributions 211 FL came up with the initial concept of the paper, KLDA performed the analysis and wrote the first 212 draft of the manuscript, TE and SH collected data, SL and KLDA wrote code. All authors developed 213 the concepts of the paper and edited final version of manuscript. 214 215 Declaration of Interests 216 The authors declare no competing interests. 217 218 References 219 1. Robinson, R.A., Crick, H.Q., Learmonth, J.A., Maclean, I., Thomas, C.D., Bairlein, F., Forchhammer, 220 M.C., Francis, C.M., Gill, J.A., and Godley, B.J. (2009). Travelling through a warming world: 221 climate change and migratory species. Endanger. Species Res. 7, 87–99. 222 2. Teitelbaum, C.S., Converse, S.J., Fagan, W.F., Böhning-Gaese, K., O’Hara, R.B., Lacy, A.E., and 223 Mueller, T. (2016). Experience drives innovation of new migration patterns of whooping cranes 224 in response to global change. Nat. Commun. 7, 12793. 225 3. Naidoo, R., Chase, M.J., Beytell, P., Du Preez, P., Landen, K., Stuart-Hill, G., and Taylor, R. (2016). 226 A newly discovered wildlife migration in Namibia and Botswana is the longest in Africa. Oryx 50, 227 138–146. 228 4. Urquhart, F., and Urquhart, N. (1978). Autumnal migration routes of the eastern population of 229 the monarch butterfly (Danaus p. plexippus L.; Danaidae; Lepidoptera) in North America to the 230 overwintering site in the Neovolcanic Plateau of Mexico. Can. J. Zool. 56, 1759–1764. 231 5. Block, B.A., Jonsen, I.D., Jorgensen, S.J., Winship, A.J., Shaffer, S.A., Bograd, S.J., Hazen, E.L., 232 Foley, D.G., Breed, G.A., Harrison, A.L., et al. (2011). Tracking apex marine predator movements 233 in a dynamic ocean. Nature 475, 86–90. 234 6. Caro, T. (2005). Antipredator defenses in birds and mammals (University of Chicago Press). 235 7. Mueller, T., O’Hara, R.B., Converse, S.J., Urbanek, R.P., and Fagan, W.F. (2013). Social Learning of 236 Migratory Performance. Science 341, 999. 237 8. Portugal, S.J., Hubel, T.Y., Fritz, J., Heese, S., Trobe, D., Voelkl, B., Hailes, S., Wilson, A.M., and 238 Usherwood, J.R. (2014). Upwash exploitation and downwash avoidance by flap phasing in ibis 239 formation flight. Nature 505, 399–402. 240 9. Altizer, S., Bartel, R., and Han, B.A. (2011). Animal Migration and Infectious Disease Risk. Science 241 331, 296–302. 242 10. Hopcraft, J.G.C., Morales, J.M., Beyer, H.L., Borner, M., Mwangomo, E., Sinclair, A.R.E., Olff, H., 243 and Haydon, D.T. (2014). Competition, predation, and migration: individual choice patterns of 244 Serengeti migrants captured by hierarchical models. Ecol. Monogr. 84, 355–372. 245 11. Sueur, C., King, A.J., Conradt, L., Kerth, G., Lusseau, D., Mettke-Hofmann, C., Schaffner, C.M., 246 Williams, L., Zinner, D., and Aureli, F. (2011). Collective decision-making and fission–fusion 247 dynamics: a conceptual framework. Oikos 120, 1608–1617. 248 12. Silk, M.J., Croft, D.P., Tregenza, T., and Bearhop, S. (2014). The importance of fission–fusion 249 social group dynamics in birds. Ibis 156, 701–715. 250 13. Krause, J., and Ruxton, G.D. (2002). Living in groups. 251 14. Conradt, L., and Roper, T.J. Consensus decision making in animals. Trends Ecol. Evol. 20, 449–252 456. 253 15. Myers, J.P. (1983). Space, time and the pattern of individual associations in a group-living 254 species: Sanderlings have no friends. Behav. Ecol. Sociobiol. 12, 129–134. 255 16. Krause, J., James, R., and Croft, D.P. (2010). Personality in the context of social networks. Philos. 256 Trans. R. Soc. B Biol. Sci. 365, 4099. 257 17. Wittemyer, G., and Getz, W.M. (2007). Hierarchical dominance structure and social organization 258 in African elephants, Loxodonta africana. Anim. Behav. 73, 671–681. 259 18. Fox, K.C.R., Muthukrishna, M., and Shultz, S. (2017). The social and cultural roots of whale and 260 dolphin brains. Nat. Ecol. Evol. Available at: https://doi.org/10.1038/s41559-017-0336-y. 261 19. Kerth, G. (2008). Causes and Consequences of Sociality in Bats. BioScience 58, 737–746. 262 20. Loretto, M.-C., Schuster, R., Itty, C., Marchand, P., Genero, F., and Bugnyar, T. (2017). Fission-263 fusion dynamics over large distances in raven non-breeders. Sci. Rep. 7, 380. 264 21. Aplin, L.M., Farine, D.R., Morand-Ferron, J., and Sheldon, B.C. (2012). Social networks predict 265 patch discovery in a wild population of songbirds. Proc. R. Soc. B Biol. Sci. 279, 4199. 266 22. Scheiber, I.B., Kotrschal, K., Weiß, B.M., and Hemetsberger, J. (2013). The social life of Greylag 267 Geese (Cambridge University Press). 268 23. Stutchbury, B.J.M., Fraser, K.C., Silverio, C., Kramer, P., Aeppli, B., Mickle, N., Pearman, M., 269 Savage, A., and Mejeur, J. (2016). Tracking mated pairs in a long-distance migratory songbird: 270 migration schedules are not synchronized within pairs. Anim. Behav. 114, 63–68. 271 24. Jones, C.S., Lessells, C.M., and Krebs, J.R. (1991). Helpers-at-the-nest in European Bee-eaters 272 (Merops apiaster): a Genetic Analysis. In DNA Fingerprinting: Approaches and Applications, T. 273 Burke, G. Dolf, A. J. Jeffreys, and R. Wolff, eds. (Basel: Birkhäuser Basel), pp. 169–192. Available 274 at: http://dx.doi.org/10.1007/978-3-0348-7312-3_13. 275 25. Campobello, D., and Hare, J.F. (2007). Information transfer determined by association of 276 neighbours in European bee-eater (Merops apiaster) colonies. Ethol. Ecol. Evol. 19, 237–243. 277 26. Kossenko, S.M., and Fry, C.H. (1998). Competition and coexistence of the European Bee-eater 278 Merops apiaster and the Blue-cheeked Bee-eater Merops persicus in Asia. Ibis 140, 2–13. 279 27. ebird (2018). eBird: An online database of bird distribution and abundance [web application]. 280 EBird Ithaca N. Y. Available at: http://www.ebird.org [Accessed May 20, 2017]. 281 28. Yosef, R. (2010). Thermoregulatory behavior in migratory European bee-eaters (Merops 282 apiaster). Wilson J. Ornithol. 122, 378–380. 283 29. Sapir, N., Wikelski, M., Avissar, R., and Nathan, R. (2011). Timing and flight mode of departure in 284 migrating European bee-eaters in relation to multi-scale meteorological processes. Behav. Ecol. 285 Sociobiol. 65, 1353–1365. 286 30. López-Gordo, J. (1975). Sobre la migración postnupcial del Abejaruco (Merops apiaster) en el 287 Estrecho de Gibraltar. Ardeola 21, 615–625. 288 31. Arbeiter, S., Schulze, M., Tamm, P., and Hahn, S. (2016). Strong cascading effect of weather 289 conditions on prey availability and annual breeding performance in European bee-eaters Merops 290 apiaster. J. Ornithol. 157, 155–163. 291 32. Frigerio, D., Weiss, B., and Kotrschal, K. (2001). Spatial proximity among adult siblings in greylag 292 geese (Anser anser): evidence for female bonding? Acta Ethologica 3, 121–125. 293 33. Scheiber, I.B.R., Kotrschal, K., and Weiß, B.M. (2009). Benefits of family reunions: Social support 294 in secondary greylag goose families. Horm. Behav. 55, 133–138. 295 34. Guttal, V., and Couzin, I.D. (2010). Social interactions, information use, and the evolution of 296 collective migration. Proc. Natl. Acad. Sci. 107, 16172. 297 35. Kao, A.B., and Couzin, I.D. (2014). Decision accuracy in complex environments is often maximized 298 by small group sizes. Proc. R. Soc. B Biol. Sci. 281. Available at: 299 http://rspb.royalsocietypublishing.org/content/281/1784/20133305.abstract. 300 36. Sapir, N., Horvitz, N., Wikelski, M., Avissar, R., Mahrer, Y., and Nathan, R. (2011). Migration by 301 soaring or flapping: numerical atmospheric simulations reveal that turbulence kinetic energy 302 dictates bee-eater flight mode. Proc. R. Soc. B Biol. Sci. 278, 3380. 303 37. Harel, R., Horvitz, N., and Nathan, R. (2016). Adult vultures outperform juveniles in challenging 304 thermal soaring conditions. Sci. Rep. 6, 27865. 305 38. Nagy, M., Couzin, I.D., Fiedler, W., Wikelski, M., and Flack, A. (2018). Synchronization, 306 coordination and collective sensing during thermalling flight of freely migrating white storks. 307 Philos. Trans. R. Soc. B Biol. Sci. 373. Available at: 308 http://rstb.royalsocietypublishing.org/content/373/1746/20170011.abstract. 309 39. Flack, A., Nagy, M., Fiedler, W., Couzin, I.D., and Wikelski, M. (2018). From local collective 310 behavior to global migratory patterns in white storks. Science 360, 911. 311 40. Arbeiter, S., Schulze, M., Todte, I., and Hahn, S. (2012). Das Zugverhalten und die Ausbreitung 312 von in Sachsen-Anhalt brütenden Bienenfressern (Merops apiaster). Ber Vogelwarte Hiddensee 313 21, 33–40. 314 41. Ramos, R., Song, G., Navarro, J., Zhang, R., Symes, C.T., Forero, M.G., and Lei, F. (2016). 315 Population genetic structure and long-distance dispersal of a recently expanding migratory bird. 316 Mol. Phylogenet. Evol. 99, 194–203. 317 479 Figure 2. Examples of raw air pressure measurements (P) in hectopascals (hPa) for bird (A) TQ in 480 September 1st -13th 2016 compared with pressure for (B) TW, (C) TV and (D) UH on September 481 10th. Grey shading represents nightime periods for TQ derived from geolocation. We only consider 482 pressure during daytime because this is the period when birds are actively changing altitude and 483 therefore less likely to be similar when birds are not together. Indeed, looking at raw atmospheric 484 temperature measurements in panel a for TQ (in black), we can see that they follow the same 485 background pressure variations recorded at the local weather station (in red) until the start of 486 migration. The change in atmospheric pressure during flight bouts is much higher than that of 487 background fluctuations in atmospheric pressure driven by weather (Sep 1st -9th in panel a) or by 488 geography and topography (Sep 10-11th). It is therefore possible to distinguish between pressure 489 changes caused by flight (during daytime), weather (during nightime) and geography (from one night 490 to the next). Birds classified as migrating together (B) are birds whose raw pressure measurements are 491 highly correlated [r(P)], whose direction and amplitude of altitudinal changes is correlated [r(dir)] and 492 whose difference in pressure measurements between birds is low [diff]. High synchronisation occurs 493 between individuals migrating together (B). However, birds can experience similar background 494 pressure conditions while following similar migratory routes, without having synchronised behaviour 495 (C). Finally, some birds record completely different atmospheric pressure, indicating their migratory 496 behaviours are different (D). See also Figure S2. 497 498 499 Figure 3. Network representation of social interactions between all tagged birds. Nodes represent 500 individuals and edges represent pairs of birds that were classified as together by a hidden Markov 501 model during (A) pre-migration breeding (i.e. capture), (B) southward migration, (C) non-breeding 502 residency and (D) northward migration and (E) post-migration breeding (i.e. recapture). In all 503 networks, the thickness of the edges indicates the proportion of time within the season where these 504 bird pairs were classified as together. Warm colours (red/orange/yellow) represent birds tagged in 505 2015 and recaptured in 2016, while cold colours (blue/green/black) represent birds tagged in 2016 and 506 recaptured in 2017. All nodes are coloured according to group, node shapes represent the breeding 507 colony the birds were caught at. Note that the air pressure loggers on TW, UG, AA and OF stopped 508 working before north migration and are therefore not represented as nodes in the network in sector d, 509 as were TZ, SJ, OF, PR and OZ in sector e. See also Figures S3 and S4. 510 511 512 Figure 4. Raw air pressure measurements in hectopascal for all social groups across the annual cycle. 513 These illustrate fission-fusion for (A) group1: BL-TH-TQ-TW-OU, (B) group 2: SJ-SO, (C) group 3: 514 TY-UK, (D) group 4: TZ-UG, (E) group 5: OI-OO-OZ-SA, (F) group 6: OF-PR, (G) group 7: GA-NR 515 and (H) group 8: PL-QK, where the grey background represents periods when the birds were classified 516 as “together”. For A and E, there are 5 and 4 birds respectively within the groups, and darker grey 517 represents days when all birds are classed as together, and lighter grey when only some birds within 518 the group are classed as together. Black bars represent migratory periods, with the left bar indicating 519 south (post-breeding) migration, and the right bar north (pre-breeding) migration. Note that A, B, C 520 and D were tagged in 2016-2017, and E, F, G and H in 2015-2016. See also Figures S3 and S4. 521 522 Supporting information 523 524 Figure S1. European bee-eater migratory patterns and non-breeding ranges as revealed by 525 geolocation, Related to Figure 1 and STAR Methods. A Two European bee-eaters (Merops 526 apiaster) from this study. Individual on the right has been fitted with a 1.4g multi-sensor logger (SOI-527 GDL3pam, Swiss Ornithological Institute) Copyright: Bernd Sekka, image used with permission. B 528 Median geolocation estimates for birds which did not show groups with another tagged bird: AA, BM, 529 RZ, SH, TO, TV, UH and UT. All were tagged between July 2016 and July 2017. Note that the 95% 530 confidence intervals of tracks are not presented. Note also that equinox occurs just as birds start 531 migrating northwards, pulling the track south in the non-breeding grounds. C. Non-breeding 532 distribution of group 1: BL-RH-TQ-TW-UO; D Non-breeding distribution of group 2: SJ-SO. E Non-533 breeding distribution of group 3: TY-UK. F Non-breeding distribution of group 4: TZ-UG. G Non-534 breeding distribution of group 5: OI-OO-OZ-QK. H Non-breeding distribution of group 6: OF-PR. I 535 Non-breeding distribution of group 7: GA-NR and J non-breeding distribution of group 8: PL-QK. 536 537 538 539 Figure S2. Classification output from the hidden Markov model, Related to STAR Methods and 540 Figure 2. A We used 5 classes to separate the data (green: “high difference in pressure between birds”, 541 red: “medium pressure difference, low correlation in raw pressure and altitudinal changes”, blue: 542 “medium pressure difference, high correlation in raw pressure and altitudinal changes”, yellow: “low 543 pressure difference, low correlation in raw pressure and altitudinal changes”, and black: “low pressure 544 difference, high correlation in raw pressure and altitudinal changes”). Birds were classified as 545 “together” from black points in panel A, when they had B) a high correlation in raw daily pressure 546 between birds (to determine if birds have similar pressure patterns), (C) a high correlation in 547 directional pressure changes (to determine whether they make the similar movement decisions at the 548 same time) and (D) low median absolute pressure difference between birds (so that birds in different 549 pressure zones are not classed as together). E and F represent the density distribution of pressure 550 ranges linked to background weather, foraging and migration. Throughout the day (panel E), weather 551 never causes a change bigger than 8 hPa, foraging 2-42 hPa and migration 3-331 hPa. At an hourly 552 basis (panel F), weather never causes a change bigger than 1 hPa, while foraging and migration always 553 change by at least 1 hPa ranging to 37 hPa during foraging and 205 hPa during migration. 554 555 556 557 Figure S3. Raw air pressure measurements in hectopascal for birds that were together (left 558 panel) or apart (right panel), Related to Figures 3 and 4. Left panels (A-F) illustrate fission-fusions 559 in the breeding and non-breeding grounds for (A) GA (group 7) with OF-PR (group 6), for (B) UT 560 with TY-UK (group 3), (C) SH with RH-TQ-UO (group 1) and (D) RZ with RH-TQ-UO (group 1). 561 For A-D, grey represents periods when the bird joins another group. Panels E and F represent birds 562 AA and UG who on some occasions are together with individuals from group 3 (cyan) and sometimes 563 group 4 (green). Note there are times when both groups overlap. Right panels (G-N) show examples of 564 raw air pressure measurements (P) in hectopascal (hPa) for all social groups during the non-breeding 565 season. The examples illustrate fission-fusion for 1: BL-RH-TQ-TW-UO (panel G), 2: SJ-SO (panel 566 H), 3: TY-UK (panel I), 4: TZ-UG (panel J), 5: OI-OO-OZ-QK (panel K), 6: OF-PR (panel L), 7: 567 GA-NR (panel M) and 8: PL-QK (panel N). Different pressure measurements indicate that birds are 568 no longer experiencing the same weather conditions while also no longer changing altitude in unison, 569 and are therefore no longer together (grey background). For example, in N, PL and QK roost together 570 at a higher altitude (910 hPa i.e. between 919 and 945m according to the range of known temperatures 571 in the region) than they forage (920 hPa i.e. 825-849m). During separation on January 25th, QK then 572 forages at a higher altitude, but returns to roost with PL. Note that G-J were tagged in 2016-2017, and 573 I-N in 2015-2016. 574 575 Figure S4. Geolocation and raw pressure measurements during periods of separations, Related 576 to Figures 3 and 4. Outer panels. Geographic location of birds during periods of separation as 577 estimated by geolocation by light. Group fission occurred during A south migration for BL-TH-TQ-578 TW-OU (group1), D south migration for TZ-UG (group 4), E south migration for SJ-SO (group 2), H 579 north migration for SJ-SO (group 2), J south migration for TY-UK (group 3), L north migration for 580 TY-UK (group 3), M south migration for OF-PR (group 6) and P south migration for GA-NR (group 581 7). Shading represents the 95 confidence intervals of geolocation estimates, and lines the median 582 geolocation estimate. Note that geolocation lacks precision, especially during equinox (relevant for a 583 and g), thus these figures only provide a rough estimate of bird locations when separating. Inner 584 panels: Raw air pressure measurements in hectopascals (hPa) for social groups which separated 585 during migration. B panel: BL-TH-TQ-TW-OU (group1), C panel: TZ-UG (group 4), F and G 586 panels: SJ-SO (group 2), J and K panels: TY-UK (group 3), N panel: OF-PR (group 6) and O panel: 587 GA-NR (group 7). Migration periods can be identified based on the relative change in ambient air 588 pressure. Indeed, changes in altitude during flight will cause pressure to decrease substantially relative 589 to breeding and non-breeding residency seasons, allowing us to identify migratory periods (horizontal 590 black bar). Different pressure measurements indicate that birds are no longer experiencing the same 591 weather conditions while also no longer changing altitude in unison, and are therefore no longer 592 migrating together (grey background). Note that A, B, C, E, H and G all separated during south 593 migration, while d and f during north migration. Finally, A, B, C, D, E and F were tagged in 2016-594 2017, and G and H in 2015-2016. 595 596 597 ID Sex Age Release Recapture Pressure last record Light last record Release nest Release colony Recapture nest Recapture colony Group RZ M 2 11/07/16 15/07/17 recapture recapture 21 B 51 A 0 SH M 2 11/07/16 16/07/17 21/06/17 21/06/17 28 A 3 A 0 OO F 2 19/07/16 15/07/17 recapture recapture 60 B 45 A 0 TV M 2 10/07/16 12/07/17 recapture recapture 21 A 19 A 0 UH M >2 11/07/16 13/07/17 recapture recapture 33 B 57 B 0 UT F 2 09/07/16 13/07/17 recapture recapture 1 B 72 B 0 AA F >2 11/07/16 11/07/17 02/03/17 02/03/17 32 B 13 B 0 BM F >2 12/07/16 13/07/17 recapture recapture 34 B 68 B 0 RH M 2 10/07/16 14/07/17 recapture recapture 25 A 27 A 1 UO M >2 11/07/16 14/07/17 recapture recapture 31 B 28 A 1 TQ F >2 12/07/16 17/07/17 14/05/17 14/05/17 41 A 28 A 1 TW M >2 13/07/16 14/07/17 25/02/17 25/02/17 63 B 37 A 1 BL F >2 11/07/16 14/07/17 22/06/17 22/06/17 18 A 26 A 1 SJ M 2 10/07/16 13/07/17 19/05/17 19/05/17 22 A 49 B 2 SO F 2 13/07/16 13/07/17 recapture recapture 61 B 54 B 2 TY F 2 10/07/16 13/07/17 recapture recapture 14 A 77 B 3 UK F >2 13/07/16 13/07/17 recapture recapture 51 B 85 B 3 TZ F 2 11/07/16 13/07/17 28/03/17 28/03/17 36 A 80 B 4 UG M 2 11/07/16 14/07/17 22/02/17 22/02/17 22 B 83 B 4 OI M 2 16/07/15 11/07/16 recapture recapture 50 A 41 A 5 OO F 2 13/07/15 12/07/16 recapture recapture 31 A 41 A 5 OZ M >2 12/07/15 12/07/16 08/05/16 08/05/16 28 A 47 A 5 SA F >2 13/07/15 15/07/16 recapture recapture 33 A 53 A 5 OF M >2 12/07/15 11/07/16 04/03/16 04/03/16 22 A 24 A 6 PR M 2 12/07/15 11/07/16 28/04/16 28/04/16 20 A 40 A 6 GA F 2 15/07/15 11/07/16 recapture recapture 39 A 24 B 7 NR F 2 15/07/15 19/07/16 recapture 12/07/16 41 A 50 A 7 PL M 2 11/07/15 10/07/16 recapture recapture no data A 4 A 8 QK M 2 15/07/15 10/07/16 recapture no data 52 A 7 A 8 598 Table S1. Information pertaining to each tagged bird, Related to STAR Methods. 599