Selection of timescales to study social network temporal dynamics in vultures

Animal social interaction patterns change over time, but the continuous nature of social interactions makes selecting a timescale for studying the temporal dynamics of social networks challenging. We applied both a heuristic approach and a multilayer reducibility analysis approach to study timescales of change in social networks of free-ranging griffon vultures, Gyps fulvus . We analysed social networks in two behavioural situations: in-flight interactions, which we expected to fluctuate seasonally but to exhibit a relatively constant pattern of change at subseasonal scales; and diurnal ground interactions, such as while feeding, which we expected to show a pulsed temporal pattern following carcass availability. The heuristic method confirmed the suitability of a 3–10-day aggregation window for studying temporal change in vulture social networks, which matches their feeding dynamics. It also highlighted how examining a variety of timescales of aggregation can offer different insights about network change patterns. Multilayer reducibility analysis confirmed that substantial change occurred at every aggregation timescale we tested, with no redundancy in network layers; that is, social interactions in this population were not oversampled. However, it revealed more similarity between chronologically nonadjacent layers in the feeding networks as compared to the flight networks, further supporting the influence of carcass availability as a driver of co-feeding network structure. Multilayer reducibility analysis over a multiseason timescale did not reveal structural similarities by season, likely due to substantial differences in population composition and tag coverage between seasons. We discuss the benefits and limitations of both the heuristic method and multilayer reducibility network analysis as tools for studying long-term animal social network structural change. • Determining timescales to study temporal interaction patterns can be challenging. • We tested two methods for studying temporal dynamics of social networks. • We used data from a GPS-tracked population of Eurasian griffon vultures. • Heuristic analysis confirmed suitability of a 3–10-day aggregation window. • Multilayer reducibility analysis revealed similarities between adjacent seasons.