An investigation of social structure in housed dairy cows

The changing landscape of the UK’s dairy farms poses increasing challenges to farm staff in terms of monitoring individuals behaviour within increasing herds, and more intensive conditions. Failure to detect behavioural changes may be costly, both from a welfare and financial perspective, as such alterations may indicate underlying disease or other challenges with corresponding impacts on yield and animal well-being. Social behaviour may provide a useful indicator of normal animal activity, and subsequent changes with health status, particularly if automatically monitored to reduce labour. This thesis applies a local positioning system (LPS) to collect social proximities of dairy cows, to investigate the social structure of a housed herd via social network analysis, and any relationship with traits or health. The LPS was validated by comparing sensor reported, with human observed proximities, and accurately detected proximities at lying, feeding and in direct interactions. Use of this data to construct social networks indicated a highly connected structure, with some substructure becoming evident after filters were applied. An approaching significant effect of parity on sociality was found, but stage of lactation had no effect. Temporally, the network showed some stability but a much greater amount of variation. When divided into ‘functional area’ (feeding, non-feeding and milking), the non-feeding area of the shed yielded the most loosely connected network with likely most interest for further analysis due to its potential basis in choice, as opposed to forced proximity. In these functional area networks, some evidence exists for homophily (association with similar cows – based on parity and days in milk). Finally, sociality was investigated alongside health status, with evidence for a tendency for greater betweenness in lame cows than non-lame. The results suggest that sociality is a highly variable trait, and that further investigation is required to assess its suitability as a disease indicator.