Network Models: An Underutilized Tool in Wildlife Epidemiology?
Figure 2
Three examples of contact networks with identical number of nodes (with 100 nodes) and connectedness (where the mean number of effective contacts per node = 4), but different degree distributions. (a) Fully-connected network. Each node has a degree of 99, but a weight of 4/99 = .0404 is applied to each edge to keep the average connectedness of each node equal to four. Diseases spread through this network in an equivalent way as in a mass action model. For clarity, only 25 nodes out of 100 are represented here. (b) Random network with Poisson degree distribution and mean degree = 4, generated following the Erdos and Renyi model [5]. (c) Scale-free network generated using Barbasi-Albert’s preferential attachment algorithm [6], with mean degree = 4 and a power law degree distribution. The network is created by starting with one node and no edges. At each time step, a node is added and connected to two other vertices chosen in proportion to their current degree. This network is characterized by a few highly connected nodes, which may act as superspreaders during epidemics. (d) Stochastic SIR simulations of disease dynamics through the three networks (120 runs per network type). Squares, circles and triangles correspond to networks (a), (b), and (c), respectively. The final epidemic size (attack rate) is represented in relation to the intergroup transmission . The recovery rate is fixed at 0.1. Note that even when the mean connectivity is kept constant, disease impacts vary with network structure.
