Analysis of pig trade movements in Austria: network representations and implication for infectious disease epidemiology

Understanding animal movement patterns between holdings is crucial for developing effective surveillance programs and outbreak mitigation in pork production. While social network analysis is commonly employed to model animal movement networks, the importance of weighing edges and selecting appropriate edge attributes has received limited attention. In this study, I investigated two representations of a static network of pig movements in Upper Austria 2015-2021, in which edges were weighted by trade frequency or batch volume. The distributions of edge weights and node centrality metrics between both networks were studied using density plots; moreover, node rankings through strength, betweenness, and closeness were compared using the Kendall correlation test. Finally, connection preferences in both networks were examined using the Leiden algorithm for community detection. The results revealed that most holdings in the networks exchanged at low frequency and volume, while few pairs of holdings exhibited intense connections. Both networks demonstrated similarity in node ranking across various centrality measures in which trading points were identified as influential nodes, potentially playing a significant role in infectious disease spread within the network. The frequency-based network showed clear boundaries between trade communities, which facilitates the implementation of disease control measures. In contrast, the volume-based network revealed a less defined spatial structure, which may pose challenges in implementing effective control measures. To conclude, the selection of weighting methods significantly impacts the interpretation and analysis of the network. Therefore, careful consideration should be given to choosing appropriate weighting methods based on the context and the specific disease being modeled.