January 28, 2026 14:00 CEST
Live on Microsoft Teams.
On 28 January at 14:00 CEST, the ESA Φ-Lab Collaborative Innovation Network will host a new Φ-talk. Details are below.
Meet the speaker
Jonathan Prexl holds an M.Sc in Physics from Philipps-Universität Marburg. Subsequently worked as a project scientist at the Technical University of Munich (TUM), focusing on large-scale processing of spaceborne Earth observation data. Recently completed a PhD under Prof. Dr. Michael Schmitt at the University of the Bundeswehr Munich, with a dissertation titled “Sensor-Informed Self-Supervised Representation Learning for Multi-Modal Earth Observation Data”.
Talk abstract
Synthetic Aperture Radar (SAR) data is increasingly important for Earth observation, yet its unique imaging geometry, modality-specific noise characteristics, and strong dependence on acquisition parameters pose significant challenges for deep learning. Naive approaches often ignore these sensor-dependent properties, resulting in representations that oversimplify or discard physically meaningful information. This presentation summarizes results from several prior publications that collectively address how to effectively leverage the physical sensor knowledge (inherently available for Earth observation data) to design better self-supervised learning methods for spaceborne SAR data in order to move towards more unifying SAR Foundation models.