Job description
Forests play a central role in the land-use nexus by providing carbon storage, biodiversity, and renewable resources, while being increasingly affected by land-use change and climate extremes. Robust, spatially explicit and temporally consistent information on forest biomass and structure is essential for assessing long-term land-use trade-offs and informing integrated modelling and governance frameworks. Recently launched Synthetic Aperture Radar (SAR) satellite missions have been specifically designed to retrieve three-dimensional forest structure and above-ground biomass with high sensitivity, but their observational records are short. In contrast, established SAR missions such as Sentinel-1 offer dense and consistent time series extending back more than a decade, albeit with limited biomass sensitivity.
This PhD position is part of the International Research Training Group C4LaND and focuses on developing transfer learning approaches that link forest above-ground biomass and structure estimates from next-generation, biomass-oriented SAR missions to long-term SAR archives, thereby enabling spatially explicit reconstruction of forest dynamics at least back to the beginning of the Sentinel-1 era. The project will be hosted at KIT (Karlsruhe Institute of Technology), Institute for Photogrammetry and Remote Sensing (IPF), under the supervision of. Your Melbourne co-advisor will be
Lines of research include
1. Developing cross-sensor transfer learning frameworks based on multi-level SAR observables for above-ground biomass estimation by exploiting polarimetric SAR features across sensor generations and enriching them with higher-order structural information from Polarimetric InSAR and Tomographic SAR where available.
2. Exploration of multi-modal and multi-model support using optical and hyperspectral data to improve robustness and generalizability of transfer learning between SAR sensors.
3. Sensor-aware uncertainty characterization and error transfer, with emphasis on decomposing the error budget and estimating loss of precision associated with products derived from established missions compared to new biomass missions.
4. Validation across long-term forest observatories in multiple regions, including Europe (TERENO), Australia (CSIRO Permanent Rainforest Plots), and potentially Brazil, ensuring transferability across forest types, climatic zones, and land-use contexts relevant to C4LAND.
Starting date
October 2026
Personal qualification
5. An above-average M.Sc. degree (or equivalent) in remote sensing, environmental sciences, computer science, forestry or a related field.
6. Strong background and interest in Earth observation, remote sensing of forests and SAR data processing
7. Experience with machine learning, domain adaptation, or transfer learning methods is highly desirable.
8. Excellent communication skills and enthusiasm for working in an international and multidisciplinary research environment.
9. Fluency in English, written and spoken
10. Willingness to undertake a one-year research placement at the University of Melbourne, Australia and comply with formalities at both institutions.