2018 ESA Annual Meeting (August 5 -- 10)

COS 26-1 - A Lagrangian tripartite perspective can unify network construction methods and integrate novel data to better forecast animal contact dynamics

Tuesday, August 7, 2018: 8:00 AM
335-336, New Orleans Ernest N. Morial Convention Center
Kezia Manlove, Department of Wildland Resources, Utah State University, Logan, UT, Christina Aiello, USGS Western Ecological Research Center, Henderson, NV, Pratha Sah, Biology, Georgetown University, Washington, DC, Bree Cummins, Mathematical Sciences, Montana State University, Bozeman, MT, Peter Hudson, Penn State University and Paul C. Cross, Northern Rocky Mountain Science Center, US Geological Survey, Bozeman, MT
Background/Question/Methods

Animal contacts underlie pathogen transmission, but contact dynamics are notoriously difficult to measure. As a consequence, disease ecologists and field epidemiologists must often attempt to understand contact patterns using limited and non-traditional datasets. Here, we apply ideas from movement ecology to formalize how different information streams relate to contact dynamics. We propose a unifying tripartite network with “individual”, “space”, and “time” node types. This tripartite structure can embed and relate information on each of the node types to place constraints on the range of contact network modularity and degree structure that could be present in the system, even when animal-to-animal contact information is not available per se. Conventional contact network construction methods are all lower-dimensional projections of this tripartite form, and animal behaviors generate predictable correlations between the node types. Group size and habitat utilization distributions enter the structure by defining inter-node degree distributions. Future work investigating relations among the three elements — individual identity, location in space, and position in time — could further facilitate use of non-traditional datasets to inform contact patterns relevant to understanding pathogen transmission.

Results/Conclusions

We find that many different kinds of data, including group size distributions, survey data on space use and occupancy, GPS tracks, and more conventional network-generating data like mark-recapture data from trap grids can all substantially constrain the plausible degree distribution and modularity of contact networks. We use simulation to estimate the magnitude of each constraint under different behavioral motifs.