Modelling spatio-temporal tree disease epidemics in Great Britain

Presently, tree populations worldwide face unprecedented threats from invasive pests and pathogens endangering biodiversity, timber production and human wellbeing. From first principles, this thesis incrementally extends a simple percolation model of forest-based epidemics into a more involved stochastic dispersal framework combined with tree canopy data. The approach developed here couples two spatially-explicit epidemic models at different scales. First, a non-local stochastic model of pathogen dispersal between trees is constructed. Second, the small-scale epidemic model is projected onto a large-scale distribution of host abundance, resulting in an $R_0$-map across Great Britain. Subsequently, a clustering algorithm is employed to identify high-risk regions in the $R_0$-map. Initial results indicate a global epidemic phase transition across the distribution, conditional on an infectivity parameter. The approach to `spatially scale-up' an epidemic model over the entire landscape is computationally efficient, flexible and adaptable to many pests and pathogens. In addition, numerous studies have sought to understand and optimise epidemic control in botanical populations. The mainstream control paradigm generally seeks to optimise an `eradication radius' about infected symptomatic trees over a relatively small spatial scale. However, large-scale epidemic control based solely on the spatial distribution of hosts has yet to be explored in depth. As such, this thesis will also examine how host heterogeneity, combined with targeted epidemic control, can give rise to natural `pinch-points' that may slow the epidemic spread between regions. Ultimately, this investigation intends to help policymakers reach informed decisions about where to focus control in the landscape of Great Britain