Dynamic network models of a continental epidemic: soybean rust in the USA

Abstract

With rapid global movement of epidemics, research efforts to characterize dynamics of epidemics have gained much focus. Traditional epidemiological models have focused on only temporal components of epidemics. Development of spatio-temporal models proved to be a notable achievement in epidemiology. Network-based epidemiological models enable better handling of spatial and temporal components of an epidemic. Early network models considered a binary level of contact between infected entities, which is an idealistic approach. A realistic approach would use weighted edges which signify the level of interaction between the nodes where the edge-weights change over time as a function of environmental factors. Estimation of edge weights from observed time series is a relatively less explored area for network modeling. Dynamic networks make the problem more complicated as edge weights change over time. Estimation of parameters for models describing the edge weights as a function of variables that change in time has the potential to provide better general models. Soybean rust (caused by Phakopsora pachyrhizi) is an important disease globally and its occurrence in the US has been studied extensively since its introduction in 2004. Rust is a fungal disease which propagates as a result of the fungal spores being carried by the wind. In this thesis, a network network based model is proposed to predict the intensity of spread of the disease in space and time. This model uses the host abundance and wind data and the observed rust incidence time series to compute the edge-weights. Also, the edge-weights in the model change over time thus following a dynamic approach. In order to cut costs involved with the establishment and maintenance of infection monitoring sites, the effect of removal of monitoring nodes using various strategies has also been analyzed in this thesis. The model has been tested with observed soybean rust data from sentinel plot network from across the United States.