Development and Testing of a High Throughput Behavioural Monitoring System for the European Honeybee

Abstract

Behavioural monitoring of the European honeybee (Apis mellifera) has traditionally been achieved through manually marking and observing bees for long periods of time (Robinson, 1992). However, the labourious and time-consuming nature of this task, combined with the difficulty of observing multiple individual bees simultaneously makes longer-term studies extremely challenging.

The purpose of this research was to develop an automated system for tracking the movements of honeybees within the hive. Two versions of the tracking software were developed over the course of this project. The first version focussed on tracking and identifying large groups of bees, while the second version implemented a system capable of uniquely recognising and tracking up to 50 individual bees. The accuracy of the tracking software was tested against the traditional technique of manually observing and tracking bee movements. The experimental methodology described in this thesis provides researchers with a relatively cheap and easy to use tool to investigate honeybee behaviour. The software system developed is freely available online via GitHub as an open source project.

The tracking system was used to conduct a number of experiments investigating honeybee behaviour. These experiments were conducted in order to pilot the software, identify limitations, and develop new techniques and metrics to improve it. The first experiment investigated circadian rhythm utilising the first software version. It found evidence for socially acquired circadian rhythms amongst juvenile bees in the hive. However, the key limitation of this technique (it only tracked groups rather than individuals) spurred the development of version two of the software which added this capability.

Version two of the software was piloted in an experiment examining the impact of caffeine on juvenile honeybees. While there is evidence that the caffeine potentially had a short-term effect, there was no significant difference between the control and treatment groups over the extended period that the experiment was run. A number of behavioural metrics, software visualisation tools and techniques were then explored and tested based on the data gathered from this experiment. These provide a template that future research utilising this tracking software can build on.

Finally, future work and improvements to the tracking system are proposed, including what would be required to extend the capabilities of the tracking software. Show more