Disease transmission in territorial populations: the small-world network of Serengeti lions

Abstract

Territoriality in animal populations creates spatial structure that is thought to naturally buffer disease invasion. Often, however, territorial populations also include highly mobile, non-residential individuals that potentially serve as disease superspreaders. Using long-term data from the Serengeti Lion Project, we characterize the contact network structure of a territorial wildlife population and address the epidemiological impact of nomadic individuals. As expected, pride contacts are dominated by interactions with neighbouring prides and interspersed by encounters with nomads as they wander throughout the ecosystem. Yet the pride-pride network also includes occasional long-range contacts between prides, making it surprisingly small world and vulnerable to epidemics, even without nomads. While nomads increase both the local and global connectivity of the network, their epidemiological impact is marginal, particularly for diseases with short infectious periods like canine distemper virus. Thus, territoriality in Serengeti lions may be less protective and non-residents less important for disease transmission than previously considered.

Figure 1.

Comparison of actual and simulated lion populations with illustrations of territory, contact and transmission networks. (a) Irregular background shapes represent 70% kernel pride territories; nodes represent pride centroids [58]; edges represent neighbouring prides. (b) Simulated lion territory network with nodes (prides) and edges indicating neighbouring territories. (c) Daily locations of a GPS-collared nomad in Serengeti National Park (SNP) and Ngorongoro Conservation Area (NCA) in 2006–2007. Spatio-temporal locations for two-week intervals (represented by month then day in the legend) are represented by shades of grey. (d) Simulated nomad with grey shades representing daily locations for the same temporal scale (two-week intervals). (e) Simulated territory network where the red box represents the same spatial scale as figure 1b. (f) Simulated contact network for both nomads and prides where dark blue edges represents nomad-mediated contacts, green, cyan and red edges represent contacts at territory distances one, two or greater than two, respectively. (g) Simulated transmission network showing only infected prides and contacts during which transmission occurred.

Figure 2.

Average Euclidean distance (kilometres) connecting prides involved in infectious contacts (figure 1f) across a range of different recovery rates; this is shown for pride–pride and nomad-mediated pride contacts. This figure is based on the Serengeti-sized ecosystem (180 prides) with 200 simulations at each of the eight recovery rates (total of 1600 runs). Values are averages across all edges in the network.

Figure 3.

(a) Global and (b) local efficiencies for both a Serengeti-sized (180 prides) and Enlarged hypothetical network of 900 prides, with and without nomads. For each network, and at each recovery rate, 200 simulations were run. The Neighbouring prides only network is Serengeti-sized and excludes all pride–pride contacts between non-neighbouring prides. It has a local efficiency identical to the Prides only networks and thus is not included in the local efficiency graph.

Figure 4.

Epidemiological impact of nomads. (a) Probability of an epidemic and (b) average prevalence in an epidemic versus recovery rates both with and without nomads at T = 0.04. Insets: differences between epidemic probabilities in runs with and without nomads. Each point in these figures is calculated from 325 random SEIR simulations at the stated recovery rate, assuming the point estimated parameter values given in the second column of table 1. Probabilities in (a) are fraction of simulated outbreaks that reach at least 30% of prides in the ecosystem; average prevalences in (b) are mean fractions of prides infected across such epidemics; and differences are simply each blue value minus the corresponding green value. (c) Estimated sources of transmission to prides. Values indicate proportion of transmission events to prides from nomads, neighbours (directly adjacent prides: d = 1), prides two steps away in the territory network (d = 2), prides three steps away in the territory network (d = 3), and ‘Other’ that consists of territorial males and prides at territory distances greater than 3.