Avian influenza takes flight: host mobility, viral prevalence, and transmission at large spatial scales

Abstract

Many pathogens have large geographic distributions, but we currently have little ability to predict how they may change over time. Understanding what mechanisms drive the large-scale distributions and movements of pathogens is critical to designing effective surveillance programs, disease interventions, and predictive models of disease spread. Additionally, we lack information on what spatial scale these mechanisms are most important. In this dissertation we address one of the fundamental problems in ecology, the "problem of pattern and scale", in the context of disease prevalence and spatial transmission. Are the large-scale patterns we see emergent properties of many small-scale processes, or are they a product of large-scale processes themselves? We focused on the spatial distribution, prevalence and spatial transmission of influenza A virus in its endemic host, wild waterfowl. We used a zero-inflated Bayesian CAR model to determine if local environmental persistence of the virus or regional host migration were better predictors of large-scale patterns of prevalence. We found that an unweighted host migration network better predicted high and low values of prevalence than did local drivers. To understand how these factors impacted where IAV moves in the United States (US) we investigated how local-scale transmission and regional-scale host movements influence large-scale spatial transmission and our ability to detect these transmissions. We developed a Bayesian zero-inflated binomial network model to estimate the probability of spatial transmission between watersheds pairs. We found that regional host movement was the best predictor of spatial transmission and that Mallard ducks likely play a special role in moving the virus throughout the US. Viral movement patterns were closely associated with important waterfowl breeding and wintering habitats rather than flyways, as has been previously shown. In order to extend these analyses to other geographic areas and host species we need to construct continental scale host movement networks from movement data with differing spatial and temporal resolutions. We developed a method to simulate host movements from very few observations allowing us to match mark-recovery data to highly detailed satellite telemetry data. We used the biological information in the detailed movement data to estimate population posterior distributions of travel speed, turning angle, and direction. These quantities and an approximately Bayesian rejection scheme were used to simulate missing locations in the mark-recovery data with estimates of uncertainty. The method was validated with a telemetry dataset tracking the movement of Northern Pintail ducks, an important host of IAV. Movement networks constructed from simulated locations captured known population scale migration patterns of Pintail and exhibited similar higher order community structure. More broadly, this research contributes to our understanding of how host mobility impacts the prevalence and movement patterns of pathogens, and the spatial scale at which this mechanism is important. Our findings suggest that predicting the spread and spillover risk of IAV requires an understanding of where hosts move at the regional scale. In the future, as climate and land-use change alter the migration patterns of wild waterfowl, we can expect the distribution and movement patterns of IAV to shift as well.