Conceived and designed the experiments: CS MM KF EM BS. Performed the experiments: CS MM KF EM BS. Analyzed the data: CS MM KF EM BS. Wrote the paper: CS MM KF EM BS.
The authors have declared that no competing interests exist.
Tracking repeat migratory journeys of individual animals is required to assess phenotypic plasticity of individual migration behaviour in space and time. We used light-level geolocators to track the long-distance journeys of migratory songbirds (wood thrush,
The degree to which long-distance migration is flexible in time and space is much debated
Inferences regarding the flexibility of migration schedules and routes have been largely restricted to observations at single breeding, winter, or stopover areas, providing only a snapshot of individual migratory behaviour. Observations of repeat migratory journeys of individual birds may yield important insights into the degree to which migratory programs are flexible, but data are rare due to the difficulty in continuous tracking of birds over large distances. Recent examples from birds large enough to carry satellite tags are illuminating; both osprey and marsh harriers showed relatively consistent migration timing, particularly in spring, but low route fidelity, suggesting strong endogenous control of schedules but relative flexibility to local conditions along migratory routes
This study was conducted in accordance with the recommendations of the Ornithological Council ‘Guidelines to the Use of Wild Birds in Research’ and was approved by the York University Animal Care Committee (Animal Care Protocol Number: 2009-2 W (R1)). Governmental scientific permits for capture, handling and geolocator attachment were obtained in the U.S., Belize, and Costa Rica.
We used data from light-level geolocators (MK14S, 1.6 g, British Antarctic Survey) retrieved between 2008 and 2011 at a breeding site in Pennsylvania, USA (‘PA’, 41.8°N, 79.9°W,
Light data were analyzed using BASTrak software package (British Antarctic Survey). Raw light data were adjusted for any clock drift (typically <3 min.). Sunrise and sunset were defined as light transitions where the light levels crossed a threshold of 16 (2008 model) or 5 (2009–2010 model). These thresholds represent similar light intensities, based on static calibration of geolocators in known locations. Light transitions were then visually inspected and edited using the program TransEdit to delete false sunrises and sunsets (e.g. transitions during daytime caused by shading) and to score the quality of true sunrise and sunset transitions. The slope of the light data at dawn or dusk was visually compared to transition slopes from static geolocators with a full sun exposure. Very shallow slopes were marked as low confidence, as were transitions that included small peaks in light intensity prior to reaching sunrise threshold, or after reaching sunset threshold. In these cases, the marked transition was unlikely to be within 10 minutes of the actual sunrise/sunset transition and so was excluded from subsequent analysis. Only the transitions with a high confidence score were used in further analyses. After each light data file was edited, we used the program Locator (BAS) to transform light data into latitudinal and longitudinal positions and used a sun elevation angle calculated using season-specific data
We relied primarily on longitude to determine timing of movements, since error in longitude is much smaller than error in latitude and longitude is not affected by the equinoxes, whereas latitude cannot be determined near the equinox (day length is the same everywhere). Position estimates may be influenced by topography, weather, seasonal changes in behaviour, and vegetation structure
Movements away from breeding or wintering sites were defined as shifts in longitude greater than 2° in a direction consistent with migration; such shifts were typically accompanied by strong shifts in latitude consistent with migration direction. Arrival dates at breeding and wintering sites were determined when longitudinal values no longer shifted in a direction consistent with migration, varied less than 2°, and remained similar throughout the breeding or wintering period. Autumn departure date was unobtainable for many birds because migration was due south (i.e. primarily shifting in latitude) and thus position was masked by the autumnal equinox. Therefore, we used the date birds crossed 23.4°N (entry to Tropics) as a measure of timing of migration as it occurred well after the equinox period
We examined three migration variables (date and longitude at cross of 23.4°N and arrival date) that are directly comparable between autumn and spring migration. The autumn equinox made it impossible to obtain departure dates for birds that did not substantially shift longitude on departure. Migration pace and duration is therefore not directly comparable between seasons. In autumn we measured pace and duration beginning at 23.4°N (i.e. the last leg of the trip) whereas in spring the pace and duration reflected the entire journey. However, timing of crossing 23.4°N in autumn is influenced by events at breeding sites
To explore factors influencing variation in spring and winter arrival date of all birds (
Variable | df | f | |
|
|||
date cross 23.4°N | 38,8 | 3.13 | 0.77 |
longitude crossing 23.4°N | 39,8 | 1.63 | 0.28 |
winter arrival date | 43,8 | 1.60 | 0.25 |
Duration | 42,8 | 1.58 | 0.25 |
|
|||
departure date | 42,9 | 4.27 | 0.01* |
date cross 23.4°N | 44,10 | 3.13 | 0.03* |
longitude crossing 23.4°N | 43,10 | 0.70 | 0.80 |
breeding arrival | 43,9 | 4.69 | 0.009** |
duration | 51,8 | 2.74 | 0.07 |
Total of 56 individual fall and spring migrations tracked, including 9 individuals tracked twice and one individual tracked three times. Significance level indicated by asterisks: *
The timing of spring departure from Central America explained much of the variation (71%) in arrival dates at breeding sites, along with sex and age factors (F4,45 = 31.47, R2 = 0.71, p<0.001) (
(a) Spring departure date versus breeding arrival date of 56 migration tracks of 45 different individual wood thrushes (line shows least-squares regression). Black data points indicate female, grey male. Migration timing for individuals tracked in two consecutive years: (b) winter arrival date, (c) spring departure date, (d) breeding arrival date. For b–d, lines show 1∶1 relationship.
Autumn date crossing 23.4°N, was the only factor retained in the minimum adequate model but explained only 25% of the variation in winter arrival date (F1,45 = 16.34, R2 = 0.25, p<0.001). Individual had a significant effect in all spring migration timing variables (spring departure date, date crossing 23.4°N, breeding arrival date) except for spring migration duration (
variable | df | f |
|
|
|
||||
date cross 23.4°N | 7, 8 | 0.44 | 0.05 | 0.44 |
longitude crossing 23.4°N | 9,10 | 1.28 | 0.12 | 0.43 |
winter arrival date | 8, 9 | 4.20 | 0.62 | 0.02* |
autumn migration duration | 6,7 | 2.82 | 0.48 | 0.10 |
|
||||
spring departure date | 9, 10 | 5.94 | 0.71 | 0.005** |
date cross 23.4°N | 9, 10 | 2.90 | 0.49 | 0.07 |
longitude crossing 23.4°N | 9, 10 | 1.28 | 0.12 | 0.35 |
breeding arrival date | 8, 9 | 4.96 | 0.66 | 0.01* |
spring migration duration | 8, 9 | 2.37 | 0.41 | 0.11 |
Our results, based on comparisons among and within individuals, suggest that the timing of songbird migration in spring is under strong endogenous control and highly repeatable from year to year. For all birds, spring departure date was a significant predictor of breeding arrival date; differing departure dates and spring routes did not uncouple the relationship between departure and arrival. Considering the 27-day range of departure dates from wintering sites, and that comparisons were made over multiple years and presumably variable environmental conditions, it is surprising that departure dates of individuals tracked in multiple years were highly consistent between years. The high repeatability in spring departure date suggests a stronger influence of endogenous schedules than local environmental conditions, in contrast to a recent study of a Neotropical migrant warbler with strong social dominance that influences individual access to food, where departure dates were only 38% repeatable
Timing was less repeatable at date of crossing 23.4°N, which implies flexibility in migration timing en route to breeding sites
Winter arrival date was also consistent for individuals and en route timing (date crossing 23.4°N) explained 25% of the variation in arrival date. Field and laboratory studies indicate that autumn departure date is heritable in songbirds and largely under endogenous control
In contrast to migration timing, migration route (as measured by longitude after crossing the 23.4°N), had relatively low repeatability in spring and autumn. This suggests that route is not under strong endogenous control, and may be influenced by individual energetic condition and weather patterns. A similar flexibility in migratory routes, but not timing, was found in migratory raptors
Our low repeatability estimates for routes compared with timing of migration may occur in part if there are large differences in measurement error between these two aspects of migration behaviour. The precision of light-level based geolocation data in estimating location, even for longitude, is likely low compared with timing of major migration movements. We quantified spatial error in longitude (55±18 km, mean ±95% CI) based on data obtained from wood thrushes carrying geolocators at known wintering sites (McKinnon et al. in prep.), although error estimates during migration may be higher because there are fewer days on which to base locations. Migration timing was also based largely on longitudinal shifts, defined using the same longitudinal error estimates from ground-truthing. Since both spatial and temporal measures of migration depend on longitude, error may be comparable. Unfortunately, it is not possible to ground truth timing of migration and estimate error since a bird’s movements can only be determined from the geolocators themselves. However, withinindividual route differences in longitude from year to year typically deviated by more than several hundred km (see
Birds were tracked from (a) Pennsylvania (b) Costa Rica. Yellow = spring migration, pink, pink = fall migration. Orange circle = breeding site, blue = winter site. Short-dashed lines indicate migration tracks in the second year and long-dashed lines a third year. Dotted lines indicate where migration route was unknown due to poor-quality light data, or geolocator battery failure.
Repeatability is a measure of individual consistency relative to other individuals in the population. Spring departure differed by only 3 days, on average, for individuals from one year to the next which is surprising considering that departure dates in the population spanned 30 days. In contrast, longitude of spring migration route at 23.4°N was highly flexible for some individuals (8–10° difference between years) and nearly spanned the population-level range in spring route (12° longitude). Within-individual differences in spring or fall route (
Overall, our results show that migration schedules are more consistent among individuals and more repeatable within individuals than migratory routes, particularly in spring. Consistent schedules, based on tracking of individual osprey
(DOC)
We thank E. Gow, T. Done, and many field assistants and volunteers. We thank Thomas Alerstam and two anonymous reviewers for their comments on an earlier draft of this manuscript.