SIMBACA

SIMBACA project: Summary

The abnormal mortality of honeybee colonies seen over the last 20 years could have multiple causes, such as climate, pathogens and parasites or pesticides. [1,2]. In the SIMBACA (Interactive Agent-Based Simulation of a Honeybee Colony) project, beekeepers, experts in bee biology and experts in modelling and interaction will collaborate to simulate a colony of bees in a Dadant hive to evaluate the impact of different factors and beekeeping practices on the health of the colony.

Project's objectives

In SIMBACA, we propose to simulate a colony of bees in a Dadant hive, a hive widely used in beekeeping, taking into account the ecosystem in which the hive is situated and its impact on the colony. In particular, we want to allow the user to interact with the simulation observable in 3D constructed from the model. The interaction with the simulator has two different uses:

1. An educational use, to allow new and experienced beekeepers to learn good beekeeping practices through predefined scenarios and realistic means of interaction;
2. A scientific use, aimed at biologists, to evaluate ecological phenomena such as the impact of Varroa destructor or other parasites or pathogens (Nosema, virus) or anthropogenic factors such as the effects of pesticides or apicultural practices (e.g. strategies for overwintering colonies and countering Varroa).

Propositions

The establishment of such an interaction will involve addressing two main scientific issues, in collaboration with our partners:

1. Construct a model of the colony that allows to interact locally at the level of bees, integrating complex emergent phenomena such as capabilities for auto-adaptation and auto-regulation of the colony, recruting, etc.;
2. Create a virtual representation of the hive and integrate visualization and interactions with the colony, which are realistic, intuitive and ergonomic.

Scientific Approach: Agent-Based Model of the Honeybee Colony

Agent-based (or individual-based) modelling is based on an algorithmic description of behaviour and interactions of individuals making up a complex system (the "micro" level). Population models, because they focus on the colony ("macro" level), do not take into account the "micro" level and its existing links with the "macro" [3] and therefore do not allow local interactions with the colony as we would wish. In contrast to these models, agent-based modelling focalises on the individual to obtain the overall behaviour by emergence. This approach is particularly suited to model a honeybee colony, where several tens of thousands of honeybees interact and live together to form a complex system.

In our case, the numerous, apparently simple, interactions between several tens of thousands of honeybees on the "micro" level can lead to the emergence at the macro level of complex phenomena such as capabilities for auto-adaptation and auto-regulation of the colony. For example, bees can change roles depending on the needs of the colony, and regulate brood temperature [4].

. In such a model, we concentrate on the honeybee as an individual, by modelling its biological cycle, behaviour, and the interactions it can have with the other bees and its environment, etc. This model still has to be calibrated and validated from data existing in the literature, and data collected from hives by partner biologists and beekeepers.

Example of possible applications

Integrated pest management strategy against Varroa destructor: the Drone Brood Removal method

The drone brood removal method is an alternative to pesticides, and is based on the observation that male (drone) cells attract Varroa significantly [5]. It involves regularly removing frames containing drone broods to destroy the Varroa present in the cells.

This method is promising (see for example [6]) but it can have multiple variants and parameters that are difficult to position due to variation in climate, honeybees or beekeeping practices. We can list the following parameters;

• Dates of removals
• Date of first operation
• Number of removals
• Frame reuse?

So how to find the best parameters?

Thanks to our virtual hive simulator, we could for example conduct experiments both in vivo and in virtuo (conducted with the simulator). We could propose the following experiment protocol:

2 in vivo experiments on hives under an oceanic climate (Finistère, with GDSA29) and a Mediterranean one (Vaucluse, with INRA):

• 1 control batch and 3 batches to evaluate one of the parameters, for example the optimal number of removals;
• Regular count of Varroa.

In the same time, in virtuo experiment with the virtual hive to identify the optimal parameters combination according to the context at short and long terms.