Parameterization and evaluation of a spatiotemporal model of the potato late blight pathosystem

A spatiotemporal, integrodifference equation model of the potato late blight pathosystem is described. Formerly, the model was used in a theoretical context to analyze and predict epidemic dynamics in spatially heterogeneous mixtures of host genotypes. The model has now been modified to reflect a research interest in interactions between genotype, environment, landscape, and management. New parameter values describing host-pathogen interactions were determined and new environment-pathogen relationships included. A new analytical equation describing lesion expansion and associated necrosis has also been developed. These changes prompted a need to assess the quality of model predictions. Cultivar-isolate-specific interactions were characterized in the model using three quantitative components of resistance: infection efficiency, lesion growth rate, and sporulation intensity. These were measured on detached potato leaflets in the laboratory. Results of a sensitivity analysis illuminate the effect of different quantitative components of resistance and initial conditions on the shape of disease progress curves. Using the resistance components, the epidemic process of lesion expansion was separated from the epidemic process of lesion propagation providing two reference curves for diagnosing observed epidemics. The spatial component of the model was evaluated graphically in order to determine if realistic rates of focal expansion for potato late blight are produced. In accordance with theory, the radius of a predicted focus increased linearly with time and a constant focal velocity was reached that compared well with published experimental data. Validation data for the temporal model came from 20 late blight epidemics observed in field trials conducted in the Netherlands in 2002 and 2004. The field data and model were compared visually using disease progress curves, and numerically through a comparison of predicted and observed t(5) and t(50) points (time in days until 5 and 50% disease severity is reached, respectively) and relative areas under the disease progress curve values. Temporal model predictions were in close agreement with observational data and the ability of the model to translate measured resistance components, weather data, and initial conditions into realistic disease progress curves without the need for calibration confirms its utility as a tool in the analysis and diagnosis of epidemics.