In-field monitoring of ground-nesting insect aggregations using a scaleable multi-camera system

Insects provide essential ecosystem services, but are threatened by multiple anthropogenic stressors. Observing insect populations and behaviour is crucial to gain a better understanding of species’ interactions, and their responses to different stressors and conservation measures. However, the observation of insects can be challenging, especially, when observing large scale aggregations of ground nesting insects. Here, many individuals of the same species nest close together and interact with each other making the simultaneous observation difficult.

Camera based motion detection and neural networks have recently emerged for insect observations. They have the potential to make insect monitoring continuous and more precise, as well as more cost-efficient, compared to more traditional methods, such as manual observation or trapping.

We are presenting an automated multi-camera observation system for aggregations of ground-nesting insects. The system has been tested and improved over two seasons observing an aggregation of the ground-nesting bee species Andrena vaga Panzer, 1799 and is to our knowledge the first system with which long-term observation of an aggregation of ground-nesting insects has been conducted. It offers the following main advantages over existing systems:

The system is adaptable to different observation projects and able to detect insects of different sizes and shapes (e.g. parasites of Andrena vaga, or bumblebees) scaling the monitored area through height adjustments. Images from multiple cameras are stitched into an overview image with minimal overlap. The system can be used under different weather and environmental conditions (winter and summer, outdoor and laboratory). By only storing imagery if the detected motion in front of the camera is likely originated from an insect, it reduces post-processing work and required data storage capacity. In observing the natural environment, no attraction mechanism is employed, allowing for the monitoring of the insects’ natural behaviour. Our tests confirmed the capability of the system with motion detection reducing manual observation time of the Andrena vaga aggregation by 92.2 % providing new insights into their interactions and behaviour.