Inferring social network structure in ecological systems from spatio-temporal data streams.

Area of Science:

• Ecology
• Network Science
• Behavioral Biology

Background:

• Social network structure is crucial for understanding animal behavior but often difficult to ascertain from observational data.
• Spatio-temporal datasets of individual occurrences offer rich information but require novel analytical methods for network inference.

Purpose of the Study:

• To develop a methodology for extracting social network structures from spatio-temporal data of individual co-occurrences.
• To apply this framework to understand the dynamics of social bonding, specifically pairing behavior in wild birds.

Main Methods:

• Identification of temporal regions of dense agent activity ('gathering events').
• Inference of social links based on co-occurrences during these events.
• Statistical testing of link significance against a null model.

Main Results:

• Successfully extracted social network structure from large-scale wild bird foraging datasets (>10^6 points).
• Observed that established pair bonds in great tits (Parus major) are maintained continuously.
• Demonstrated that new pair bonds form variably before breeding, marked by rapid increases in network proximity.

Conclusions:

• The proposed methodology effectively infers social networks from spatio-temporal co-occurrence data.
• Provides novel insights into the temporal dynamics of social bonding and pair formation in wild bird populations.
• The general framework is applicable to diverse systems involving interacting agents with temporal co-occurrence information.