Optimal pest detection networks for robust management decisions

About the Project

Sugar beet is a strategically important crop and its productivity is increasingly threatened by insect pests such as aphids that transmit virus yellows, and beet flea beetles. As climate change alters pest distributions and seasonal dynamics, there is a growing need for robust surveillance systems that can provide early warning of pest outbreaks and support timely management decisions. Monitoring networks are a cornerstone of Integrated Pest Management (IPM), providing the information needed to guide interventions such as spray thresholds.

However, existing monitoring systems are often spatially sparse, costly to maintain, and focused almost exclusively on pest detection. Beneficial insects, which play a critical role in natural pest suppression, are rarely considered when designing monitoring networks or decision-support tools. Significant knowledge gaps therefore remain regarding how monitoring networks can be optimised to jointly monitor pests and beneficial insects, maximise detection accuracy, minimise costs, and support better on-farm decision making.

Research methodology:

This PhD will use extensive historical datasets and field sites provided by the British Beet Research Organisation (BBRO) to develop and test next-generation framework to generate monitoring networks for pests. The student will develop a pipeline to optimising trap placement, accounting for spatial and temporal variability in pest and beneficial insect populations, and improving the translation of monitoring data into management actions. Using GIS, geospatial optimisation techniques, AI, and reinforcement learning, the project will develop monitoring networks that balance early detection, operational costs, and beneficial insect protection and translation to on farm actions. The research will evaluate approaches ranging from location allocation models (e.g. p-median and maximal coverage models) to adaptive monitoring frameworks based on Bayesian optimisation and reinforcement learning. A second core aspect of the of the PhD will be exploring approaches for turning data generated from the monitoring network to actions that can be taken by the farmer, comparing approaches from traditional to spray thresholds, to advanced forecasting systems that integrate trap captures, beneficial insect populations, weather forecasts, and projected pest dynamics, and end-to-end multicriteria deep learning approaches.

The final year will include testing and validating these tools within BBRO monitoring networks.

The student will spend time at BBRO in Norwich to gain a deeper understanding on agricultural pests, detection networks, and pest control. The student will also have the opportunity to become basis qualified in arable pest management.

Supervisors:

• Shaun Coutts: SCoutts@lincoln.ac.uk
• Alistair Wright: alistair.wright@bbro.co.uk
• Bhattarai, Binod: binod.bhattarai@abdn.ac.uk

Applications:

Our fully-funded studentship package includes:

• All PhD tuition fees paid.
• A tax-free stipend at UKRI rates to cover living costs.
• A Research Training Support Grant (RTSG) of £3,000 each year to support travel, training and consumables costs (up to £12,000 in total).
• Additional funding to support outreach and dissemination, attendance at summer schools, research events, and development projects.

Interested applicants should visit 'https://www.sustain-cdt.ai/how-to-apply' for full instructions on how to submit their application.

Requirements:

The candidate will have a background in (minimum 2:1 honours or a Master’s degree) Computer Science or AI or operations research or population ecology / pest management. A strong expertise or demonstrated interest in any of spatial analysis, AI, pest managements, or applied operations research, would be highly advantageous.

Enquiries:

Shaun Coutts SCoutts@lincoln.ac.uk

Application Deadline:

12:00 noon (UK time) Friday, 16 October 2026

References

1. Joarder, S., Sikdar, D., Akash, A. H., Bhattarai, B., & Gyawali, P. (2026). Two is better than one: A Collapse-free Multi-Reward RLIF Training Framework. arXiv preprint arXiv:2605.22620.
2. Howard, L. et al. (2022) A review of invasive species reporting apps for citizen science and opportunities for innovation. NeoBiota 71
3. Gao, X., Xue, W., Lennox, C., Stevens, M., & Gao, J. (2024). Developing a hybrid convolutional neural network for automatic aphid counting in sugar beet fields. Computers and Electronics in Agriculture, 220, 108910.
4. Regmi, S., Panthi, B., Dotel, S., Gyawali, P. K., Stoyanov, D., & Bhattarai, B. (2024, June). T2fnorm: Train-time feature normalization for ood detection in image classification. In 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) (pp. 153-162). IEEE.
5. Darbyshire, M., Coutts, S., Hammond, E., Gokbudak, F., Oztireli, C., Bosilj, P., ... & Parsons, S. (2025). Multispectral fine-grained classification of blackgrass in wheat and barley crops. Computers and Electronics in Agriculture, 237, 110484.