If you are a student at the University of Tübingen in a degree program at the department for Computer Science or the Graduate Training Center Neuroscience, and are interested in working with us (e.g. for a masters, bachelor thesis, research internship, essay rotation or as an SHK/Hiwi), please send an enquiry to mls-office. Please include a current CV including a description of previous research/work experiences, a transcript of all previous courses and grades, and a brief statement of what research you want to do and why you want to join our group. A list of open topics is below— however, note that this list is not exhaustive, and we might have additional topics available. @inf.uni-tuebingen.de
We currently do not have open positions for internships for external students. Informal enquiries can still be sent to mls-office. However, as we receive a large number of applications, we may not be able to respond to your message, and we do not respond to ‘generic’ applications. @inf.uni-tuebingen.de
The Greenlandic and Antarctic ice sheets are important components in the Earth System impacting climate evolution with their imprints on the global radiation balance and ocean current formation. Satellite observations over the last decades have quantified significant mass loss of both ice sheets, but the underlying processes markedly differ. In Greenland, atmospheric warming results in increased surface melting during summer which is not compensated for by snowfall during winter. In Antarctica, snowfall rates have increased year-round. However, ice-stream acceleration and an increase in ocean-induced melting at the Antarctic perimeter more than offset this mass gain and result in a net ice-mass loss. Predicting the future evolution of both ice sheets in a warming world hence requires processed based on data-driven ice-sheet models.
Radar surveys across both ice sheets provide baseline measurements such as ice thickness and bed elevation, but also image the isochronal stratigraphy, i.e. allow measurements of the non-intersecting internal radar reflection horizons that were once deposited at the surface. With time, these surfaces get buried progressively through snowfall and also deformed during the ice-dynamic transport towards the coasts. The radar observations hence contain an integrated memory of the ice-dynamic, atmospheric and oceanic history that the ice sheet has experienced. This archive is currently not used by the modeling community because digitization of the internal layering is labor intensive and only done for selected sub-projects.
This project is an interdisciplinary collaboration between the Alfred-Wegener-Institute for Polar- and Marine research in Bremerhaven (Prof. O. Eisen) and the Machine Learning Cluster of Excellence & the Geoscience Department at the University of Tübingen. We will build on recent advances in deep learning for automated segmentation and classification which is optimized for the specific properties and challenges of radar-image data, and which allows inclusion of geophysical prior knowledge. Our goal is to establish a novel workflow for automatic extraction of internal layering from radar data, and for deriving functional uncertainty estimates. If successful, this framework will unearth an immense observational archive that will provide the sorely needed observational constraints to predict the future of polar ice sheets on societal relevant timescales.
This Master’s Thesis will work on specific aspects of this bigger project. We also anticipate further opportunities for research projects on similar topics. We will also be looking for a "Studentische Hilfskraft" to help with data-processing and organisation. Please get in touch with mls-office if you are interested. Include a CV with a description of previous research/work experiences and a transcript of previous courses and grades. @inf.uni-tuebingen.de