A Natural Disaster Exacerbates and Redistributes Disease Risk Among Free-Ranging Macaques by Altering Social Structure

Abstract

Climate change is intensifying extreme weather events, with severe implications for ecosystem dynamics. A key behavioural mechanism whereby animals may cope with such events is by altering their social structure, which in turn could influence epidemic risk. However, how and to what extent natural disasters affect disease risk via changes in sociality remains unexplored in animal populations. By simulating disease spread in free-living rhesus macaques (Macaca mulatta) before and after a hurricane, we demonstrate doubled pathogen transmission rates up to 5 years following the disaster, equivalent to an increase in pathogen infectivity from 10% to 20%. Moreover, the hurricane redistributed the risk of infection across the population by exacerbating sex-related differences. Overall, we demonstrate that natural disasters can amplify and redistribute epidemic risk in animals via changes in sociality. These observations provide unexpected further mechanisms by which extreme weather events can threaten wildlife health, population viability and spillover to humans.

Keywords: disease ecology; epidemic simulation; natural disaster; rhesus macaques.

Figure 1.. Year, social group, pathogen infectivity, and Hurricane Maria affect simulated epidemic risk.

Simulation outputs are grouped into group-by-year replicates. Each year is represented by a different colour, with orange shades indicating pre-hurricane years and purple shades post-hurricane years. (A) Each line represents one of 1000 simulations carried out on each group-year replicate (N=24,000), depicting the proportion of infected individuals at each simulation timestep. Post-hurricane years (purples) reached higher proportions of infected individuals markedly faster than pre-hurricane years (oranges). Each macaque social group is a different row and the group’s name is on the right of each row. NB: not all groups were observed in all years; as such, there are not all orange and purple combinations in all panels of A. (B) Distribution of mean individual infection timesteps separated by years. (C) Each point represents the mean simulation time step at which individuals become infected in a given simulation (N=24,000) as a function of pathogen infectivity. Solid lines represent models fitted to each group-by-year replicate, coloured by year as in Panel A. Pathogen infectivity was associated with substantially earlier mean infection timestep on a log₁₀-scale. (D) The effect estimate for a ten percentage point increase in infectivity is given next to the estimate for the effect of the hurricane. The blue violins represent the probability distribution of the posterior effect estimate; the error bars the 95% credibility interval and the points the mean.

Figure 2.. Hurricane Maria alters the distribution of epidemic risk across individuals.

Panel A displays the full model estimates from our models of individual infection, both without (“Base”) and with summed social connection strength (“+Strength”) as an explanatory variable (A). The other panels display raw data indicating the individual factors driving epidemic risk and the effects of the hurricane in terms of social status (B), age (C), and sex (D). Panel A, *: 95% CI does not include 0. Panel B-C, the line is a linear fitted model, with 95% confidence intervals displayed as a grey ribbon. The R and P value of the raw fit are depicted at the top; R: Pearson correlation coefficient. Panel D, ***: P<0.01; F: females; M: males.