In this paper we consider spatially heterogeneous epidemic systems, in which pathogen spread occurs through a landscape comprising favorable (e.g. susceptible or untreated) and less favorable (e.g. resistant or treated) sites. An important example is the deployment of resistant crop varieties in mixed species populations. It is well recognized that heterogeneity arising from the presence of multiple species or spatial variation in the interactions between individuals can endow population processes with a far more complex range of dynamics than would be exhibited in homogeneous settings. However, even though stochastic models for heterogeneous systems can be readily formulated, these models can only inform our understanding of any particular system if they can be parameterized for that system. We focus on processes and models with short-range interactions using data describing the spatio-temporal spread of Rhizoctonia solani, a soil-borne fungal plant pathogen, in mixed species populations as a convenient experimental system in which heterogeneity can be controlled. Such epidemics are frequently driven by an external source of infection (primary infection, the initiator of an epidemic and often associated with inoculum present in the soil), and secondary spread between infected and neighboring susceptibles in a population (secondary infection). We develop Bayesian methods to fit spatio-temporal percolation-based models, incorporating such features, to biological processes that evolve in homogeneous populations.