Simulation modelling as a tool for evaluating surveillance programmes for detection of the Asian honeybee mite (Varroa destructor) in the South Island of New Zealand

AIM: To use a simulation model of the spread of the Asian honeybee mite (Varroa destructor) amongst apiaries, to evaluate a series of detection surveillance programmes for the South Island of New Zealand.

METHODS: Five potential incursion sites into the South Island were selected. A stochastic spatial simulation model, Varroa_sim, was adapted to simulate spread of the mite from these sites as a series of silent-phase propagating epidemics. The study population comprised all apiaries in the South Island registered in the Ministry of Agriculture and Forestry's (MAF's) apiary database in 2003. Six different surveillance programmes were simulated to try and detect the mite. Three of these were the actual multi-stage sampling plans conducted during the autumn (March–May) of 2001, 2002 and 2003, and the other three involved simple random sampling with sampling fractions equivalent to the actual numbers of apiaries tested in each of those years. The relative performances of the different surveillance plans were evaluated in terms of their ability to detect the mite early before it had spread too far and whilst there might still be a chance of eradication.

RESULTS: There were 13,798 registered apiaries in the South Island with valid map coordinates in the apiary database at the time of the study. The model generated 50 epidemics against which the various surveillance programmes were evaluated. The actual surveillance programmes conducted during the autumn of 2001 and 2002 generally performed fairly well in detecting the mite. The programme conducted in autumn 2003 detected the mite reasonably well in high-risk areas, but was very poor in low-risk areas. The simple random sampling strategies performed surprisingly well, and their relative rankings were proportional to the sampling fractions employed.

CONCLUSIONS: This study showed the value in using a spatial simulation model to generate plausible silent-phase epidemics, against which detection surveillance programmes could be evaluated, in ways that would otherwise not be possible.