About this role
The continuous release of Earth Observation (satellite) data and the emergence of Machine Learning methods open up new possibilities for understanding forests. These large datasets provide complementary information on 3D structure (GEDI lidar, BIOMASS P-band radar, NISAR L-band radar) and high spatiotemporal resolution (Sentinel‑1 C-band radar, Sentinel‑2 multispectral). State-of-the-art foundation models (e.g., AlphaEarth, TerraMind) are currently being evaluated for different applications, but the inclusion of temporal components and newly available datasets in foundation models remains limited. There is also a need for accounting noise in Deep Learning models and quantifying uncertainty in real-world applications. This PhD studentship (scholarship) leverages large-scale Earth Observation data to evaluate and advance machine learning algorithms for one of the following application areas:
“Characterising forests variations in relation to distance from pre-Columbian earthworks in the Amazon forest, Brazil”, co-supervisor Prof Ted Feldpausch, University of Exeter, UK “Predicting mixed-forest composition and/or understanding intra-variability of same forest types at European level”, co-supervisor Dr Emily Lines, University of Cambridge, UK “Quantifying forest planation damage and supporting planning after a cyclone or tropical storm in New Zealand”, in collaboration with Interpine Group Ltd, NZ
The prospect candidate is requested to choose one application (listed or relevant) and write a 300 word proposal on how innovative algorithms can tackle it. Applicants are encouraged to reach out to the lead supervisor, Dr Milto Miltiadou ([email protected]), to gain insight into the specialised data available and the associated challenges of each proposed project.
The studentship will be awarded based on merit. Both Home and International Students are eligible. The PhD funding includes tuition fee coverage and stipend.