Dynamic Social Networks in Dairy Cows

Social relations have been shown to impact individual and group success in farm animal populations. Fundamental to addressing these relationships is an understanding of the social network structure resulting from the co-habitation and co-movement of relationships between individuals in a group. Here, we investigate the social network of a group of around 210 lactating dairy cows on a dutch farm during a 14 days period. A positioning system called \emph{Cowview} collected positional data for the whole period. We make the assumption that spatial proximity can be used as a proxy for social interaction. The data is processed to get adjacency matrices. Then social networks are identified based on these matrices. Community detection techniques are applied to the networks. We measure metrics of different dimensions to test community structure, centralization, and similarity of network structure over time. Our study show that there is no evidence that cows are subdivided into stable social communities when looking at interaction in the whole barn. We, however, notice relatively clear communities when dividing the barn into areas with different activities. The social network is characterized by significant centralization, low connectivity, and a hierarchy.