Transboundary livestock diseases are a threat to the United States’ (US) agricultural system, potentially affecting food security and the US economy. Many tools used to understand potential disease spread and the effect of response actions on that spread inherently require an underlying model. Livestock disease spread models at large spatial scales like the US are inherently complex, thus a common simplifying assumption is to use spatial kernels to capture the bulk of transmission events. However, long-distance transmission events are of particular management concern in livestock disease outbreaks because they provide the potential for establishment of new foci of infection. The importance and challenge of incorporating long-distance transmission into livestock disease models parallels basic ecological problems regarding how best to capture spatial and other network processes, such as dispersal, that arise due to tradeoffs in our ability to capture the tails of spatial distributions. Using lessons from extreme value theory, we create livestock disease simulations that use data on livestock movements to more accurately model the tails of the spatial transmission kernel. These predictive models are used to evaluate alternative control actions for managing livestock disease in order to inform policy.
Results/Conclusions
Our livestock disease results show, perhaps not surprisingly, that control actions focused on rare, long-distance transmission play an important role in addition to control actions that reduce short-range, local spread. However, using mechanistic modeling to predictively describe long-distance transmission provides more detailed decision-support for the exact types of control actions that are likely to be successful. The livestock disease study highlights the importance of understanding heterogeneity in transmission, which can also be represented by variability in the characteristics of contact networks. Thus, our livestock results motivate an additional study of wildlife contact networks. Using a literature survey, we found that wildlife contact networks differ in many ways from their theoretical and livestock counterparts. Given their differing properties, we explore the question of if detailed understanding of wildlife contact networks is necessary to characterize transmission or if phenomenologically capturing the bulk of transmission at local scales using samples of individual-level contacts is sufficient. We make recommendations regarding how preliminary data can be used to differentiate which level of characterization is needed. This distinction is important because the effort involved in capturing whole wildlife contact networks is substantially different than sampling individual-level contact data.