The experimental releases were performed in August 2009 and June and July 2010, following a small pilot study in August 2007 and May 2008, with the three tracks of this pilot study included in the analysis.
We used two release sites situated in about equal distance in almost opposite directions; both sites lay in rural regions, with the nearest village in similar distances and directions from the home direction:
Oberlais (50°24′ N, 9°06′ E), 44.8 km northeast of the loft; home direction, 225°; within the Vogelsberg anomaly, an area with strong, irregular fluctuations in intensity and steep local magnetic gradients in varying directions, with the release point about 10 km from the edge of the anomaly in the homeward direction. At the release point, the intensity was 133 nT above the one at the loft.
Essenheim (49°47′ N, 8°08′ E), 42.2 km southwest of the loft; home direction, 61°; the control site in magnetically “quiet” terrain, where the magnetic intensity changes very little in all directions; at the release point, it was 5 nT above the one at the loft.
The local magnetic conditions around the two sites are illustrated in Fig. 1, based on magnetic intensity data for 100 × 100 m squares showing the deviations from the reference field (DGRF 1980.0) provided by the Leibniz Institute of Applied Geophysics. The two sites are identical to the sites A2 and C2 of our previous study (Wiltschko et al. 2010).
The test birds were adult pigeons from our Frankfurt loft (50°08′ N, 8°40′ E). The magnetic conditions at the loft, 57 nT below the reference field in a magnetic minimum, are given in detail in Wiltschko and Wiltschko (2003a). The birds were at least 1 year old. In their first year of life, they had completed a standard training program up to 40 km in the cardinal compass directions and had been additionally released for several training flights up to 30 km in spring each year. All had completed additional training flights up to 20 km carrying dummy weights to prepare them for carrying the GPS recorder, had participated in previous tracking experiments from various sites, and thus had ample experience in flying with the recorder. In the present study, all pigeons were unfamiliar with the release sites, i.e., they had never homed from these specific sites before.
GPS tracking devices
The GPS recorders used in this study were based on the prototype developed by von Hünerbein et al. (2000), with either an embedded patch antenna or a Y-antenna and a data logger as additional components. The weight including the battery ranged from 35 g in the pilot study to 23 g used later. The recorder was set to take a positional fix every second, with a precision of ±4 m in the older models and ±1.8 m in the recent models. After the receivers had contact with a sufficient number of satellites, they were wrapped in plastic to shield them from water and attached to the pigeons' backs by means of a harness made from Teflon tape (see Wiltschko et al. 2007 for details). Immediately before release, the recorder was placed on the dorsal plate of the harness and fixed with Velcro and additional sticky tape. The harness and coating added another 7 g to the load.
From 18 pigeons released within the anomaly and 17 pigeons released at the control site, we obtained 12 tracks and 13 evaluable tracks, respectively; the others had to be excluded from analysis because of recorder failure or pigeons flying only short distances (<3 km) and returning only after the battery had run empty.
Track analysis and statistics
In the present study, we focus on the beginning of the flight within 5 km from the release point. In a first step, we compared the behavior before and after the first “Points of Decision.” This first Point of Decision is defined by the highest increase in the steadiness of flight; it marks the moment when the pigeon begins to leave the release site, with steadiness and flying speed increasing. For identifying a Point of Decision, steadiness is determined as sliding means of the vector length calculated every 15 s from 60 consecutive headings of the tracked pigeon, with the headings being the direction between two consecutive positional fixes (Schiffner and Wiltschko 2009 for details). The first Point of Decision, usually close to the release point, separates the initial phase of flying around at the release site from the following departure phase. Over longer distances, most tracks include periods where the pigeons fly steadily towards home and periods where they do not increase the distance from the release site continuously. Hence, additional Points of Decision can be defined by the highest increase in steadiness following these stalling periods, using the method described above (see Schiffner et al. 2011).
We determined the positions of the first “Point of Decision” of all tracks from each site and calculated (1) their center of distribution and (2) their median distance from the release point. For the initial phase of each track, we determined (3) its duration, (4) the distance flown, (5) the mean vector of headings, based on all headings during this phase, (6) the absolute deviations of the mean heading from the home direction, and (7) the steadiness of flight, now represented by the mean vector lengths resulting from all headings of the track during the entire duration of the specific phase. The mean vector, the absolute deviation from the home direction, and the steadiness were also calculated for the departure phase, excluding data beyond 5 km from the release point.
From the data for the individual pigeons, we then determined second order means for the mean headings and the medians of the other variables for the tracks from both sites. In addition, we calculated mean vectors from the bearings 2.5 and 5 km from the release point. These variables were then compared between the two release sites. The distribution of the first Points of Decision were tested with Hotelling's one-sample test for bivariate samples for directional preferences and compared between the two sites with Hotelling's two-sample test. The second order mean vectors were tested for directional preference with the Rayleigh test and compared using the Mardia Whatson Wheeler test (Baschelet 1981). The other variables were compared using the Mann–Whitney U test.
In a second step, we analyzed the relationship of the individual tracks with the local distribution of magnetic intensity, in particular with the changes in intensity the pigeons experienced in the course of their flight. We recorded the difference in intensity between the consecutive 100 × 100 m squares a pigeon crossed, and calculated, for each track, (1) the maximum difference in total intensity the pigeon had experienced during the first minute of its flight and (2) the percentage of the various differences among the total differences between all segments the pigeons had visited until 5 km from the release point. For comparison, we generated two sets of random distributions derived from the local distribution of magnetic intensity, the first by determining the differences based on 12 trajectories, defined as straight lines 30° apart originating at the release point and ending at a distance of 5 km starting at the north and the second by 13 similar trajectories 15° apart within the homeward semicircle. The distribution of differences in the pigeons' tracks and the random distributions were then compared with two-way ANOVA with repeated measurements. This analysis was performed only for the anomaly data because, in the magnetically quiet region around the control site, the differences in intensity, mostly being 0 or 1 nT and never exceeding 2 nT, were too small for this type of approach.