Details
Satellite remote sensing is revolutionising agricultural monitoring. However, smallholder farms - which produce a third of the world's food - remain challenging to monitor using Earth Observation (EO) because fields are small and irregular, cropping diverse and mixed, cloud cover persistent, and reference data scarce. Geospatial foundation models (GeoFMs) such as TESSERA and AlphaEarth present opportunities to overcome these challenges. These models reduce labelling requirements and improve transferability of EO-based sensing approaches. However, existing GeoFM benchmarks are skewed towards homogeneous, large-scale farming in Europe and North America, whereas we know little about whether GeoFMs improve mapping of smallholder farms.
Aim and Questions
This project will assess if GeoFM embeddings outperform conventional remote sensing features and models for mapping smallholder agriculture. The project will answer four questions:
- How robust is current evidence? How reliable are current evidence and benchmarking protocols for GeoFM applications for smallholder agricultural remote sensing?
- When and where do GFMs help? Under what conditions do GeoFM embeddings outperform conventional features and models for mapping smallholder agriculture, and do gains vary for different modes of use (e.g. frozen vs fine-tuned embeddings; single- vs multi-modal; annual vs sub-seasonal)?
- How effectively do GeoFMs transfer? Do GeoFMs improve EO model transferability across regions, years and farming systems while reducing reference data requirements?
- What do GeoFMs embeddings capture? What agronomic information do GeoFM embeddings capture, and does this explain where they succeed or fail to map smallholder agriculture?
Methodology
The PhD will begin with a systematic meta-analysis of GeoFM performance for agriculture, focusing on evaluation design, training and testing protocols, and reported gains. It will develop a benchmarking framework built on a unique set of field-level reference datasets collected by the suprvisory team including 15,000+ geolocated surveys with crop and irrigation histories for smallholder farms in Africa and Latin America from IrrEO and TRANSCEND projects. Experiments will vary GeoFM model, adaptation strategy, input modalities and volume of training labels. Evaluations will apply spatially and temporally held-out validation, learning-curve analysis and uncertainty assessment to assess generalisability and transferability and explainbility of GeoFM-based models modelling approaches.
Outcomes and training
The project will deliver 3+ high-impact publications, an open reference dataset and benchmarking protocol for smallholder agricultural mapping. It will produce evidence-based guidance for the EO community, national agencies and development partners on the effective use of GeoFMs for agricultural policy and investment monitoring, evaluation and learning. The PhD is closely integrated with the IrrEO project (2025-2029, led by Foster and Bispo). The student will gain in-demand skills at the interface of AI and EO, supported by an interdisciplinary supervisory team with established track records in geospatial AI and EO. The student will be integrated with existing CDT training programmes, including the iRisk multi-hazards informatics DFA and AI Fundamentals CDT, and benefit from research and training collaboration opportunities with leading GeoFM development groups (TESSERA - University of Cambridge, Clay Foundation - United States, VITO Remote Sensing - Belgium) and implementing partners (Gates Foundation, World Bank, and the European Space Agency).
This project is expected to start in September 2027.
Before you apply: You should contact the project supervisors to learn more about the project and discuss whether it may be suitable for you.
How to apply:
To be considered for this project, you must submit a formal application through our online application portal. If you already have an applicant account, use this link to open an application for FSE PhD President's Doctoral Scholarship. If you do not yet have an applicant account, follow the account setup instructions. Once your account is set up, use this link to open an application for this programme.
When applying, please specify the full title and supervisor/s of the project, details of your previous study, and names and contact details of two referees. You must also upload a Supporting Statement describing your motivation to apply to the project, your CV and transcripts of awarded and in-progress university qualifications. Please note late or incomplete applications will not be considered.
Equality, diversity and inclusion are fundamental to the success of The University of Manchester and central to all our activities. A diverse research community strengthens creativity, productivity and quality, while increasing the societal and economic impact of our work. We welcome applicants from all career paths, backgrounds and sections of the community, regardless of age, disability, ethnicity, gender, gender expression, sexual orientation or transgender status.
We welcome applications from candidates returning to study after a career break or experience in other roles. Flexible study arrangements may be available, including part-time study at 50%, 60% or 80%, subject to the requirements of the project and funder.
Eligibility: Applicants are expected to hold (or be about to obtain) a minimum upper second-class undergraduate honours degree (or equivalent) in computer science, environmental science, physical geography, environmental engineering, or a related discipline. An additional Masters level qualification (completed or soon to be obtained) in a similar or related discipline would be desirable. Skills in programming (Python, R, or equivalent) and data science are essential. Research experience in remote sensing or Earth Observation is desirable.
FSE_PDS
Funding Notes
Funding covers tuition fees, an annual stipend at the UKRI minimum rate (currently £21,805 per year) plus £1,000/annum enhancement, and a research training support grant of up to £20,000 for the full 4-year programme.
