Automatic feeder for monitoring and supporting bumblebees to counteract the effects of climate change and global insect decline

Motivation

Bumblebees are agriculturally important pollinators, but are currently declining in abundance in the UK and around the world due to climate change and other man-made stressors, such as the use of herbicides and pesticides. Understanding bumblebee declines requires better understanding of bumblebee behaviour across species. We have conducted preliminary investigations of 7 species of wild-caught bumblebees native to North-East England. We found that bumblebees of different species prefer to forage on different plants. The aim of this project is monitor climatic conditions and bumblebee abundance across different habitats, to establish sensory stimuli (smells) that bumblebees find most attractive, and to provide additional food resources to bumblebees in resource-poor areas.

Aims

This project will develop a novel setup, a scented feeder, to study innate smell preferences across bumblebee species, and to monitor bumblebee abundance. Additionally, the feeder will be used to provide additional resources to bumblebees in resource-poor areas.

Aim 1) To develop and build the scented feeder with video-recording capabilities

Aim 2) To develop the image analysis pipeline, to analyse feeder visitations

Aim 3) To use the feeder to investigate pollinator abundance and smell preferences

Aim 4) To investigate the effects of additional resources on bumblebee abundance

Methodology

A1) The scented feeder will be developed as a portable device designed for field deployment over 24/48h. The feeder will be equipped with a high-resolution camera system capable of continuous recording to enable detailed monitoring of insect interactions. The feeder will also incorporate temperature and humidity sensors to log local microclimatic conditions for better analysis of how these factors influence pollinator behavior.

A2) The image analysis pipeline will employ state-of-the-art deep machine learning models, such as convolutional neural networks and transformers, to automatically identify and classify insect visitors based on their species and behavioral patterns even under challenging conditions like variable lighting or overlapping individuals. Additionally, the pipeline will integrate behavioral analysis algorithms to quantify visitation frequency, duration, and interaction with different olfactory stimuli.

A3) We will deploy multiple copies of the feeder in different habitats (urban, semi-urban, field, forest) around Durham and Newcastle, both with and without sugar, and baited by different smells. This will allow us to investigate innate smell preferences of pollinators, as well as general presence/abundance of bumblebees and other pollinators in the area.

A4) We will compare bumblebee abundance in areas where we provided additional food resources (baited feeders) and the areas where we did not. This will indicate whether food supplementation is a feasible way to boost bumblebee populations.

Timetable

Year1: Aim1 through the year; Aims 3-4 in summer; conference ECRO/ESITO in Europe; BMVA Computer Vision Summer School; Year2: Aims 3-4 in summer; Aim2 through the year; ICML conference/Neuroethology Congress; International Computer Vision Summer School (ICVSS); Year3: Aims 3-4 in summer; ESA conference; Year 3.5: finishing up experiments and writing up.

Novelty and Impact

This project will focus on native UK bee species that are agriculturally important pollinators, are in decline and are poorly studied. The setup we will develop will allow us to monitor bumblebee abundance and the climatic parameters of their habitat, to investigate how areas with different temperatures affect bumblebee abundance.

The student will acquire knowledge and skills in:

1)     computer vision

2)     data science

3)     engineering and robotics

4)     pollinator ecology

5)     insect behaviour

6)     presentation and scientific writing;

7)     research supervision;

8)     impact and public outreach.

Additionally, the student will benefit from experience of interdisciplinary working in 4 collaborating labs.

Requirements: We are looking for an independent and enthusiastic student able to develop the project and drive it forward. Interest and demonstrable skills in computer science, machine learning, robotics and animal behaviour are a plus. Full training in the techniques required for this project will be provided. We specifically encourage applications from people of backgrounds, underrepresented in the UK academia.

How to apply

To apply for this studentship, applicants should submit their application using the online system: https://studyatdurham.microsoftcrmportals.com/en-US/. Please select PhD in Computer Science: Course Code G5A001.

Applications will be processed as they are received until the position is filled, but no later than the 22nd January 2025. Applicants are strongly advised to contact Dr Atapour-Abarghouei ([Email Address Removed]) and Dr Riabinina ([Email Address Removed] ) before submitting their application.