• Data augmentation;
  • point process;
  • hierarchical models;
  • dynamic systems
  • C11;
  • C22

The spatio-temporal dynamics of disease spread provide an area of acute scientific interest but substantial analytic challenges. Analytic approaches need to draw from current knowledge of disease systems yet be flexible enough to provide targeted inference without restrictive assumptions. We explore this balance in detail with respect to the observed pattern of spread of sugarcane yellow leaf virus, the causal agent of yellow leaf disease. This disease affects cultivated sugarcane plants and spreads via an aphid vector (Melanaphis sacchari). The system is of considerable economic interest, but presents several analytic complications. First, the plants begin as well separated individuals that grow into large plants with overlapping leaf canopies enabling increased vector access between neighbouring plants. The overlap between neighbouring plants first occurs within rows and then extends between rows, presenting another interesting spatial component to consider when modelling the propagation of infection. Second, the vector initially flies between plants but at some point sheds wings and shifts to walking between plants. We are particularly interested in how these two features of the system and their interrelation may impact the spatial spread of disease over time. In addition, we also explore the impact of potential immigration of infected vectors from outside of the study area. We treat our observed cane field as a spatial grid with regular spacing between rows and smaller spacing within rows. Our data consist of observed disease status of each of 1643 plants at each of six time points. We provide descriptive summaries of the temporal evolution of spatial pattern in infection, then propose hierarchical models to describe the pattern and capture the influence of the ecological and environmental factors listed above.