CR2025_25 How can AI-based insect monitoring technologies enhance large-scale insect monitoring?
Lead Supervisor: Charlie Outhwaite, Institute of Zoology
Email: charlotte.outhwaite@ioz.ac.uk
Co-supervisors: Tom Breeze, Department of Sustainable Land Management, University of Reading; Alexa Varah, Natural History Museum; Tom August, UK Centre for Ecology and Hydrology.
CONTEXT. Recent studies have indicated a significant decline in insect biodiversity in the UK and globally1. Understanding the drivers and extent of these declines is limited by available data. This data is often collated as part of citizen science recording efforts, most of which are not structured schemes that can deliver consistent data. However, negative perceptions of insects2, and the complexity of collecting and identifying insects, have limited more structured monitoring efforts to a few high profile taxa such as butterflies (UK Butterfly Monitoring Scheme) and bees (UK Pollinator Monitoring Scheme).
Novel, AI-driven technologies offer new opportunities for scaling up and gap-filling in insect monitoring3. AI-driven identification apps, such as ObsIdentify and Seek, provide species identifications from photographs making engagement with species monitoring easy and satisfying as you need little to no knowledge on species taxonomy to be able to identify species. Similarly, automated insect camera traps, such as the AMI trap, are new developments that allow automated recording of insects in the field, collecting images of species that are used to develop AI-algorithms for their identification. If used in a structured manner, these traps could offer opportunities to fill gaps, scale up recording effort, and improve taxonomic capacity in insect monitoring.
PROBLEM STATEMENT. Despite this potential, it is not fully understood to what extent the data collected from these AI methods are comparable with existing monitoring efforts and historic data, nor how effective they may be at fostering engagement in insect monitoring and conservation among the public. If they are to be integrated to better understand large-scale patterns of change, we must first understand their similarities and differences and ensure their uptake by the public to maximise their potential for data collection and engagement with insect conservation.
PROJECT AIMS. This project will address these challenges by combining ecology, data science and social science to explore the potential for two AI-identification tools (apps and automated insect camera traps), to scale-up insect biodiversity monitoring.
During the studentship you will use novel data science methods to determine how compatible data generated by these technologies are with data collected using traditional techniques and identify biases within the data.
Using the outcomes of this data analysis, you will develop new hierarchical models to establish a data workflow that harmonizes these different data sources into a single model to understand patterns of change.
You will also work with social scientists to conduct surveys through ZSL London Zoo and the Natural History Museum, London, to explore how these AI technologies can facilitate changes in perceptions and knowledge around insects, encourage data collection for insect monitoring schemes, and evaluate the potential impact this could have on insect data modelling.
Training opportunities:
Members of the supervisory team are core members of a 4-year COST Action project InsectAI (https://insectai.eu/), which brings together stakeholders from ecology, computer science and other disciplines to understand how computer vision-based technologies can aid insect monitoring. You will benefit from the ongoing work and training of InsectAI and be supported to attend their events. This project brings opportunities for extensive network building, and potential funding for conference attendance and Short-Term Scientific Missions. In addition, through collaboration with the supervisory team and their networks, you will develop skills in data science methods, biodiversity monitoring techniques, and hierarchical modelling.
Student profile:
The studentship is suitable for a student with a first or upper second-class degree in quantitative biological/ecological sciences, with an interest in coding and data analysis. Experience of survey techniques or use of ecological modelling approaches would be beneficial; however, these skills are not essential because training opportunities will be provided. The student should demonstrate a capacity and willingness to learn new approaches and techniques.
Please note: Due to the nature of this project and to comply with visa regulations, only Home students should apply.
References:
- Wagner et a. (2021) https://doi.org/10.1073/pnas.2023989118
- Sumner et al. (2018) https://doi.org/10.1111/een.12676
- Klink et al. (2022) https://doi.org/10.1016/j.tree.2022.06.001