Constructing social networks from spatial data
To examine social interactions in different situations we focused on three social situations: co-flying, diurnal interactions on the ground (e.g., while co-feeding), and nocturnal interactions on the ground (namely while co-roosting). All social interactions were inferred from movement data obtained from the GPS tags, based on movement speed and spatial proximity. For each social situation, we created an undirected weighted interaction network (Croft et al. 2008; Wey et al. 2008). In each network, nodes denote individually tagged vultures and edges connect vultures that interacted in a particular social situation, based on the details below. Edge weights represented the association strength between each pair of vultures measured as simple ratio index (SRI) which divides the time two vultures were observed together (as defined below) by the total time they could have interacted - i.e., times in which they both had a GPS location recorded.
Co-flying : Vultures were considered flying if they were moving faster than 5 m/s (Spiegel et al. 2013b). Owing to their vision, vultures can see each other from afar (Pennycuick 1971; Spiegel et al. 2013a). We therefore deemed individuals flying within 1000m of each other during a 10-min window to have a co-flight interaction (Fig. 2).
Nocturnal ground interactions : Vultures nest and roost on high cliffs and tend to aggregate at communal roosts. Individuals were considered interacting on the ground during the night if they co-occurred at the same roost overnight. Roosts were spatially defined as polygons on a map (see example roost polygon in Fig. 2 and (Harelet al. 2017)) shaped according to geographic features, such as dry streams or cliffs, where vultures are known to roost. The area of the roost polygons was 783.5 ± 1751.3 m2 (mean ± sd), which falls well within the perceptual range of vultures. To account for poor reception within the canyons serving as roosts, and for the vultures’ diurnal activity, we associated vultures to their nightly roost polygons using either their last position of the day or their first position on the following morning; if neither of these two locations fell within a roost polygon we used the average Euclidean distance between these two positions to assign a vulture to a roost. Vultures occasionally roost outside of communal roosts, so if this average position did not fall within a roost we did not assign those locations to any roost.
Diurnal ground interactions : Vultures were considered interacting on the ground during the day if they were not flying (i.e., ground speed slower than 5 m/s), and their locations were within 50m of each other during a 10min time window. We excluded any interactions inside known roosts during the day. Therefore, these diurnal ground interactions likely represent interactions while feeding and joint sunbathing at feeding sites.
Aggregate networks : To combine interactions from all social situations in a single aggregate network, we summed the edge weights from all three social situations (co-flight, diurnal, and nocturnal ground interactions) for each pair of individuals.