Data on cattle movements within the United Kingdom have recently become available. As part of the conditions for lifting an export ban on British beef following the bovine spongiform encephalopathy epidemic, the European Union required that the UK should have "An effective animal identification and movement recording system". The Cattle Tracing System (CTS) was introduced in September 1998, and the scheme was extended to include all cattle by the beginning of 2001.

Contact networks have proved valuable in studying the epidemiology of diseases in man, such as human immunodeficiency virus; the availability of CTS cattle movement data has enabled contact network analysis to be applied to diseases of farm livestock. The CTS data may be represented as a large network; cattle holdings are represented as nodes, with a movement of cattle between holdings being an edge.

To address concerns about the quality of this cattle movement data, a field study was conducted on Lewis, one of the Western Isles of Scotland. Farmers were recruited with the assistance of the local veterinary surgeon, and asked to record a range of potential risk behaviours relating to the transmission of infectious diseases (moving livestock, sharing pasture, etc.) for a one-month period. For the study area in question, movements of cattle not reported to CTS (especially to or from common grazing land) were a substantial contribution to the contact network during the study period.

A wide range of measures of network structure exist, but their relevance to the dynamics of infectious diseases on networks is unclear. To address this, a discrete-time stochastic SIR simulation model of disease on a network was designed and implemented in software. Using this simulation model, a network model with the key structural features of the CTS contact network was constructed, by considering a range of measures of network structure, and testing resulting model networks against CTS-derived networks. The resulting model was shown to predict the dynamics of a simulated disease model on that contact network more closely than existing models of global network structure.

Much work on the contact structure of the UK cattle herd has relied on relatively simple static network representations of movement data. By using simulated diseases, the serious shortcomings of static network representations compared to more complex dynamic network representations were demonstrated.

A substantial library of software for the generation and analysis of large networks, and the simulation of disease thereupon, has been produced, and has been made generally available. The design and implementation of this software is discussed, including the algorithms and data structures deployed, as well as validation of the software, and its portability to different computing platforms.