Abstract

Socially structured wildlife populations, such as the African buffalo (Syncerus caffer), are poorly characterized by either traditional disease models that assume random mixing or spatial disease models that assume limited dispersal between fixed groups. Dynamic network models in combination with data on who spends time with whom, however, more accurately reflect connections within and between groups and the spread of disease between associating individuals. We used 2 years of radio-tracking data on 64 African buffalo to estimate monthly association matrices. These matrices were then used as a substrate to model disease dynamics in the buffalo system and investigate the importance of the topology of connections in the network as well as the variation in the frequency of contact between individuals. In agreement with previous studies on static networks, we found that topology was very important and that only a small proportion of connections between groups are necessary to create a ‘small world’ network. Increasing the variation in frequency of contact between individuals had little impact upon disease dynamics in this system. Cluster analyses of the association matrices demonstrated that herds are not as well defined as previously thought and are increasingly amorphous over time. Buffalo associations were more tightly clustered in 2002 than 2003, perhaps due to drier conditions in 2003 forcing herds out of previously habitable areas. As a result, we predict diseases to spread faster through the buffalo population during drought conditions due to both increased stress and increased population mixing.