Using playback and mist nets, we trapped 10 male European nightjars in June and July 2010 and 14 males in June and July 2011 in Northern Jutland, Denmark (57.06°N, 9.13°E). Birds were ringed with a metal ring and fitted with geolocators (Mk 10_S; 1.1 g; British Antarctic Survey, BAS [24]). The light sensor was placed on a 0.8 cm stalk, in order to raise the sensor above the plumage. We attached the loggers as a backpack using a full-body loop harness (comparable to the wing harness in [25]) made from a 2-mm-wide braided nylon string. Females were released without geolocators because in previous years we recaptured a lower proportion of females than males. Geolocators from seven male nightjars were retrieved by July 2013 and data were available from six (2010–2011: 4; 2011–2012: 2). In three devices, logging had stopped before the bird returned to the breeding site (on 28 March, 18 April and 7 May). An extra effort was devoted to catch birds from territories that were provided with geolocators, presumably resulting in a slightly higher recapture rate (29%) than the rate of 20% of ringed-only male nightjars (own data).
A total of 25 common swifts were trapped during breeding at two Danish locations. At Bjødstrup, eastern Jutland (56.29°N, 10.53°E), 10 and 7 individuals were fitted with geolocators in 2010 and 2011, respectively, of which 6 individuals were re-trapped (2010–2011: 3; 2011–2012: 3). At Nyborg, Fyn (55.30°N, 10.82°E), 8 individuals were fitted with geolocators in 2010 of which 3 returned in 2011 and 1 returned in 2012. The recapture rate (40%), was below the rate reported in a similar study (75% [22]). This is potentially due to birds switching between local nest sites from year to year (as we have observed in previous cases but no quantitative data are available), and we could not obtain permissions to recapture birds located at nearby nest sites located in adjacent private houses. We used geolocators (Mk 20, 0.6 g from BAS) fitted on the birds using a body loop harness made from 1-mm braided nylon string. Birds were trapped between sunset and full darkness in mist nets close to the nesting site or in the nest box.
The Copenhagen Bird Ringing Centre with permission from the Danish Nature Agency approved capturing and tagging of nightjars and swifts.
Location estimation
Data for nightjars were downloaded and analyzed using the BASTrak software suite [26]. We defined sunrises and sunsets using the threshold method with a sun angle estimated for each individual. There was an unusually high level of noise in the light measurements on many days, possibly caused by shading from either the vegetation that the birds hide in during day or from the feathers on the back of the birds. To account for this noise, we only used data from days with little noise (measured as standard deviation in the light measurements between sunrise and sunset). To decide the acceptable level of noise, we looked for the level under which the estimated latitudes were stable during breeding (Additional file 1: Fig. S1). Because we filtered out the days with most noise, we only needed to discard all latitudes within two weeks on each side of the equinoxes. The remaining data were calibrated to the breeding latitude during their breeding site attendance and the derived sun angles (−2.2° to −5.0°) were used to estimate positions (Additional file 2). We compared this to using Hill–Ekström calibration [27] during the longest non-breeding stopover, but presumably because of the amount of noise in the data this was only possible for two birds. Hill–Ekström calibration requires that there is a specific sun elevation angle that corresponds to a local minimum in latitude variation during a stopover. We tested angles between 0° and −8° and found no minimum in four birds; in the two birds where we found a minimum in latitude variation, the estimated sun elevation angle was within 0.5° of the one obtained using breeding latitude calibration.
The cuckoo data are from eight published satellite tracked birds with completed fall migration in 2010 of four males and two females, and three males and one female in spring 2011 ([23]; Fig. 2; Additional file 2). The common swift data from light-level-based geolocators were analyzed using a threshold of 2, a sun angle of −5° and excluding three weeks around the equinox. To calibrate the sun elevation angle we used Hill–Ekström calibration [27] during the longest non-breeding stopover. All resulting sun elevation angles were between −4.5° and −5.5°, and we used −5° for all individuals (Additional file 2).
Breeding, stopover and wintering periods: location, timing and distances
We used departure in one year and return in the following year to calculate duration of breeding in all three species.
For geolocator data (nightjars and swifts), we defined a stopover as when the birds interrupted their migration for more than 5 days and calculated an estimated stopover position by averaging the longitudes and latitudes during this period. We estimated stopover and movement periods based on changes in latitude and longitude except during equinox, when only longitude was considered. Dictated by the uncertainty of light-level data [27], we required consecutive stopovers to be separated by more than four degrees either latitude or longitude; otherwise, the two locations were considered part of the same stopover and the location was acquired as the average throughout both stops. Departure and arrival dates were taken as the last and first, respectively, days that were within two degrees longitude of the average location estimate (Additional file 3).
In the nightjar, some of the fall and spring stopovers occurred during equinox, making latitude estimation impossible. To estimate these stopovers for illustration (Fig. 1), we calculated theoretically realistic latitudinal spans based on travel speeds. These were calculated using the preceding or following stopover (if the preceding was not available), travel speeds of 200–500 km/day [3, 23] and the longitude estimated from the light data during the stopover. We calculated the overall migrating distance as the minimum distance between breeding site, stopover sites and winter area. For this calculation, we also included estimated stopovers during equinox which provided data on individual movement, despite being disregarded because of their positional uncertainty. Migration speed was calculated as migration segment distance divided by duration of the migration segment. All durations of segments include days stopping over. Some might only be a few days (shorter than what we consider a stopover), making some migration speeds close to travel speeds.
For satellite tracking data (cuckoos), we defined stopovers, dates and distances overall similarly. However, the higher accuracy of the satellite tracking data theoretically allows for separation of small-scale changes between sites. However, we grouped cuckoo stopovers in specific geographical regions at a spatial scale similar to that which we obtained for the other two species. Individual stopover locations were estimated as averages within these regions as described in [22]. Because of non-daily transmissions, departure and arrival dates are rarely identifiable and we used last and first, respectively, days within the region/site in question (Additional file 3).
Geolocators in most cases allow for accurate determination of stopover timing to within 1 day. Because of the 10 h on 48 h off duty cycle of the satellite transmitters, estimation of departures from stopovers was less accurate for satellite data and transmissions were occasionally missed, resulting in even less accurate estimates. We based stopover duration only on known locations resulting in potential underestimation of stopover duration (and overestimation of travel time). Missing transmissions only occurred relatively frequently in Europe during fall, potentially leading to underestimates of stopover duration of cuckoos here. However, we found that cuckoos staged for longer periods than the other species in this region anyway. Therefore, we conclude that our comparisons of stopover duration are likely conservative.
Comparison of European nightjar with common swift and common cuckoo
We compare the migration routes and the timing of the nightjars to those of common cuckoos and common swifts. The study of migration in all three species is challenging given their relatively small size, long-distance migration and difficult study conditions in the wintering area (ranging from being nocturnal to highly aerial often in highly remote areas). To make the data comparable among species, we included only adult birds (but were not able to separate sexes in analyses). Thus, our sample sizes are necessarily small. The different species were mostly tracked in the same year but some nightjars and swifts were from a year later than the cuckoos. As birds’ schedules could vary as a response to the variable conditions among years, comparing schedules from the same year would have been preferable. At least at the regional scale that we are considering, we do not expect any major differences among years. Similar consistency in stopover use has also been reported in other long-distance migrants, for example red-backed shrikes [10] although their timing appears more flexible.
The three species showed overall similarities in the routes and timing taken: all undertook stopovers in the Sahel after leaving Europe, subsequently spending the major part of the non-breeding season, i.e., wintering, in Central Africa and at stopover sites in West Africa before returning to the breeding sites. We therefore compared the timing of departure and arrival in these four places: breeding, Sahel (8–20°N), wintering (south of 5°N and east of 10°E) (for swifts this could include trips to eastern Africa) and West Africa (west of 10°E and 0–20°N) in spring. We also compared durations of stopovers as well as migration speed. Roaming behavior during the non-breeding season was quantified as the number of stationary periods as well as the distance traveled among stationary sites during winter. Because we aim to describe differences in spatiotemporal migration patterns in species with otherwise similar constraints, we focused our analyses of the wintering period to birds wintering in the same general area, thus using only the seven swifts traveling to Central Africa (three remained in West Africa all winter). Clearly, the short-stopping swifts saved time and potentially energy by using a shorter-distance migration strategy that potentially influences their ability to fatten up before travel and speed of migration depending on local seasonal conditions.
We compare the three species fitting separate ANOVAs for each of the parameters above as a function of species. Because of the large numbers of comparisons, Bonferroni-corrected significance levels are also indicated. Overall, results should be interpreted with caution due to generally low sample sizes. Because of varying sample sizes, tests were not directly comparable and in general, we only present comparisons from significant tests. All models were fitted using general linear model (GLM) in R 3.1.0 [28].