Sequential area interaction sampling for early plant disease detection

We develop general novel sequential point process models for plant disease surveillance, aiming to accelerate early detection of disease presence in a study domain. The framework adapts to the level of knowledge about the propagation parameters: with full information, it reduces to a sequential hard-core process, whereas under uncertainty it becomes a sequential area-interaction process. By incorporating past observations at each step, the proposed models optimize detection time in an adaptive setting. The dynamics of the proposed sequential point process models are analyzed, compared to alternative sampling strategies through a simulation study, and illustrated in a real surveillance context.