Modelling the spread of American Foulbrood in honey bees

Background/Question/Methods

The last two decades have seen a significant decline in the number of honey bees worldwide. The contribution that bees make to agriculture through pollination has been estimated to be in the billions of dollars, and hence these losses have had a large economic impact on crop production worldwide, leading to many affected countries investigating possible causes. One of the main factors in explaining the drop in numbers is increased incidence of honey bee diseases, such as Foulbrood, Deformed Wing Virus and the Varroa mite.

Our work involved the investigation of an epidemic of American Foulbrood (AFB) on the island of Jersey, off the Northwest coast of France, during the summer of 2010. A primary census of every managed honey bee colony was carried out in June, and follow-up visits to infected apiaries were conducted in August. The dataset included the times of all visits, the spatial coordinates and owner of each hive. We employed a rigorous likelihood-based framework and employing MCMC Bayesian methods to infer the underlying parameters, including model constants and infection times of both known and unknown (“occult”) infections. We then attempted to recreate the epidemic using a stochastic SIR model, to check the derived parameters. The stochastic model was then used to simulate a range of control strategies, to find the optimum inspection method for Jersey bee inspectors.

Results/Conclusions

The MCMC scheme gave results which indicated both distance- and owner-based transmission were important in the spread of infection. The location and time for the initial infection could also be inferred. Importantly, these parameter values were validated by the stochastic model, producing epidemics of comparable sizes to those observed. This validation is imperative in confirming that the model construction captures the basic infection process well, a step is lacking in many disease spread models. Given a reliable modelling framework, we were then able to test a range of alternative control strategies. The suite of control measures we simulate highlighted that, with limited resources available, the actions taken by bee inspectors on Jersey were appropriate; while many additional strategies would not be cost-effective in combating the disease. It is found that only extremely rigorous measures significantly increase the chance of eradicating AFB entirely from the island of Jersey. This is in agreement with practical understanding that AFB is extremely tenacious and can survive for decades.