Disentangling patterns of collective escape in starling flocks

Complex patterns of collective behavior may emerge through self-organization, from local rules of interactions among individuals. The relation between local rules and global patterns is not trivial;computational models have been proven necessary for tracing emergent properties in animal groups.One of the most complex patterns in nature is the collective behavior of European starlings under attack,with their large flocks changing shape and internal structure rapidly and continuously. Even though their collective behavior is well studied, our understanding of how patterns of collective escape emerge is still limited; models of collective escape are rare. Here, by qualitatively analyzing footage of starling flocks pursued by a robotic predator, we observe that some patterns of collective escape often co-occur in a single flock (more than one pattern arise simultaneously at different parts of the flock). This phenomenon has never been studied before. Aiming to understand how this co-occurrence emerges, we develop anew three-dimensional agent-based model, inspired by our empirical observations. Our simulated flocks resemble large flocks of starlings in both the rules of motion and coordination at the individual level, as well as properties of the flock, its density and internal dynamics and structure. Patterns of collective escape emerge from one or few individuals performing a discrete escape maneuver (turning or diving)away from an attacking predator, and from their neighbors copying their behavior. We study the co-occurrence (in a single flock) of patterns such as vertical shape, collective turns, splits, and cordons. Our model is the first to realistically simulate several patterns of collective escape in bird flocks, enabling the future study of these patterns in detail across species