Three-dimensional trajectories and network analyses of group behaviour within chimney swift flocks during approaches to the roost

Abstract

Chimney swifts (Chaetura pelagica) are highly manoeuvrable birds notable for roosting overnight in chimneys, in groups of hundreds or thousands of birds, before and during their autumn migration. At dusk, birds gather in large numbers from surrounding areas near a roost site. The whole flock then employs an orderly, but dynamic, circling approach pattern before rapidly entering a small aperture en masse We recorded the three-dimensional trajectories of ≈1 800 individual birds during a 30 min period encompassing flock formation, circling, and landing, and used these trajectories to test several hypotheses relating to flock or group behaviour. Specifically, we investigated whether the swifts use local interaction rules based on topological distance (e.g. the n nearest neighbours, regardless of their distance) rather than physical distance (e.g. neighbours within x m, regardless of number) to guide interactions, whether the chimney entry zone is more or less cooperative than the surrounding flock, and whether the characteristic subgroup size is constant or varies with flock density. We found that the swift flock is structured around local rules based on physical distance, that subgroup size increases with density, and that there exist regions of the flock that are less cooperative than others, in particular the chimney entry zone.

Keywords: aggregation; chimney swift; flight biomechanics; flocking; network; swarming.

Figure 1.

(a) Example video frame shows the chimney, protruding from the building in the lower right, and chimney swifts. The image was processed with background subtraction to highlight the birds in yellow. See also electronic supplementary material, movie S1. (b) A schematic of the field set-up shows the cameras on the parking deck, approximate imaging volume in two dimensions, the chimney, and a typical late-evening flock pattern.

Figure 2.

Flock count (a), the number of individual birds in the recording volume at a particular time, (b) flock angular momentum, and (c) the cumulative entry of birds into the chimney. Flips in the angular momentum sign denote reversals in the circling direction. The four highlighted slices were used in the network analysis and related metrics given in table 1. Gaps in the data occurred when the cameras closed one movie file and began another during recording.

Figure 3.

(a) Blue lines correspond to the left axis and are the smoothed average relationship between pairwise physical distance and heading similarity for all birds in 62 frames pulled randomly from the four time slices. Smoothing was by spline with a local standard deviation tolerance. The right axis shows the probability density scale for the histogram of maximum weight distances and its fitted normal distribution (μ = 1.435183, σ = 0.354939) wingspan; normality was confirmed with an Anderson–Darling test (test statistic 0.313587, p > 0.15, null hypothesis of normality). The shaded region is the 95% CI for the mean. (b) The maximum weight pairwise distance did not vary with mean nearest neighbour distance (R² = 0.002315, p = 0.710362). However, as shown by the trend lines in (c), the average weight of the nth neighbour decreased with increases in nearest neighbour distance, (R = −0.353969, −0.378956, −0.394210 for the nearest, 5th nearest, and 10th nearest neighbours, respectively; p = 0.000112, 3.21440 ×10⁻⁵, and 1.42674 ×10⁻⁵). As expected, average weight also declined with neighbour index. Data in (c) were computed from 114 samples of 100 consecutive frames across the full dataset.

Figure 4.

Scaling of group sizes within the flock dendrogram. Panel (a) shows the summed transition probabilities reaching group size α computed for 24 different temporally close samples of dendrograms gathered from regions of 65 consecutive frames in all four time slices. Panel (b) shows a linear fit to the highest integer α such that the summed transition probabilities are <0.9, R = −0.540670, p = 0.006375; the fit produced residuals with an estimated mean μ = 1.421085 × 10⁻¹⁴ and an Anderson–Darling test for normality was performed which yielded a statistic of 0.268692, (null hypothesis of normality).

Figure 5.

The flock properties change over time and space. Each panel shows the local average per-bird value of a flock property during a time slice, shown along the average flock path for that slice. Time slice varies by row, ranging from flock marshalling (top) to chimney entry (bottom figure 2). Columns 1–4 show, from left to right: speed, nearest neighbour distance, summed thresholded edge weight, and spatial modularity. Flock circling direction is shown in column 1. The chimney building is shown in outline on the right side of the plots and the chimney itself is marked by a small open square. The width of the ribbon indicates the number of observations in each spatial bin.

Figure 6.

Properties of birds at landing. (a) Compares the network weights of pairs W_ij of landing and non-landing birds through time. Pairs either land or, for non-landing birds, enter a control volume 6 m above the chimney within 10 video frames of one another. Lines show the mean and 95% CIs; n = 65 pairs for landing and 171 for non-landing. Panel (b) shows the cumulative distribution of time intervals between successive landing events or birds flying through three different control regions. In this case, the volume of the control region target was adjusted to give equal number of individuals in all landing and non-landing cases. All results are from the last time slice, which contains the majority of landing events observed (figure 2).