Trustworthy federated learning enabled predictive analytics for sustainable nutrient and carbon management in intensive livestock systems

Intensive livestock farming is a major contributor to environmental degradation, including nutrient runoff, water pollution, and greenhouse gas emissions.The research will involve the development of predictive analytics using federated learning and trustworthy AI techniques, building on the X10AI AGRISMART digital twin platform, which integrates real-world data from anaerobic digestion, ammonia recovery, pyrolysis, and precision agriculture across UK farms. The project will focus on fusing hyperspectral drone imagery with structured (e.g., yield, weather) and unstructured (e.g., farm logs, regulatory reports) datasets through multimodal data harmonisation and summarisation using large language models.

A federated learning framework will be designed to enable collaborative model training across farms while preserving data privacy. The models will incorporate both physics-informed and data-driven components and will be validated through two case studies: (1) predicting grass growth to support phosphorus “geo-mining” and sustainable manure export, and (2) forecasting slurry spreading windows based on local soil and weather conditions.

Training will include advanced skills in machine learning, remote sensing, environmental modelling, and explainable AI. The PhD offers opportunities to work with academic experts and industry partners (x10AI, ABP, Sainsbury’s, NFU), access real-world datasets, and contribute to research with direct environmental protection policy and industry relevance.

The PhD project is ideal for students with strong programming and mathematical skills, and a passion for AI and sustainability.