Fully-funded PhD: AI-Driven Remote Sensing for Species-Level Savannah Monitoring

Savannahs cover 50% of Africa and face rapid change from woody vegetation encroachment, which threatens biodiversity, rangeland productivity, and livelihoods. Current monitoring efforts treat woody cover as a single class, masking species-specific impacts. This PhD will pioneer the first species-level monitoring framework using drone-based multispectral data fused with very-high-resolution satellite imagery (Pleiades Neo), powered by cutting-edge geospatial AI.

You will develop a scalable training pipeline to map species-level encroachment across landscapes, combining drone data with satellite products (Sentinel1/2, EnMAP, GEDI). The project is co-designed with South African government agencies and supported by Airbus, providing premium satellite imagery and technical expertise. While the project includes methodological and applied components, the primary focus will be on developing and validating the scalable geospatial AI framework, with field and policy integration supported through established collaborations.

You’ll gain advanced skills in remote sensing, AI, ecological modelling, and policy engagement, working across disciplines and continents. The project includes an industrial supervisor to support non-academic training and skills development. You’ll contribute to open-source tools and decision-ready indicators for restoration and land management impacting savannahs.

We welcome applicants from diverse backgrounds excited by the challenge of combining Earth Observation, AI, Ecology and Sustainability to drive real-world change.

Project aims and objectives

Main Aim: To develop a scalable, species-level monitoring framework for woody vegetation encroachment in African savannahs using drone and satellite data fused with advanced geospatial AI.

Specific Objectives:

• Design and implement a hierarchical training pipeline linking UAV multispectral data with very-high-resolution satellite imagery (Pleiades Neo).
• Conduct field campaigns in South Africa using a multispectral UAS
• Apply self-supervised and interpretable deep learning models to upscale species-level mapping to regional satellite products.
• Organise co-creation workshops with local stakeholders and generate decision-ready indicators for restoration and land management, co-designed with South African government agencies.
• Develop open-source tools (QGIS plugin and web viewer) to support operational uptake and policy integration.