AG DeLIGHT (Deep-learning-informed Glacio-Hydrological Threat)

The DELIGHT Framework International Junior Research Group, funded by the Elite Network of Bavaria, is conducting research into developing a fully coupled suite of models that trace the entire hydrological pathway of glacierised catchments, from precipitation, through glaciers, to rivers. The research group will work on developing this suite of models by combining several existing process models and augmenting or replacing certain parts with neural networks to make the whole run within a sensible timeframe.
Accurately predicting these glacio-hydrological hazards requires modelling of the entire hydrological process chain, something not before possible as it would take too long to run the simulations. With the advent of machine learning and the resulting possibility to greatly accelerate the modelling of physical systems, this fundamental barrier is now being lifted, making more accurate predictions feasible for the first time.
As a third of the world’s population live in catchments with glacierised headwaters and will consequently be exposed to changes in water quantity and quality as glaciers retreat, as well as the possibility of devastating glacial lake outburst floods (GLOFs) as water fills unstably dammed previously ice-covered depressions, we hope to contribute with our work to an improvement in accuracy of glacio-hydrological hazard predictions.
The work of this project will be conducted with a range of international partners:
- University of Lausanne
- Université Grenoble Alpes
- University of Bristol
- University of Zürich
- University of Innsbruck
- Bavarian Academy of Sciences
The research group is affiliated to the interdisciplinary International Doctoral College ‘Measuring and Modelling Mountain Glaciers and Ice Caps in a Changing Climate’ (M³OCCA) at FAU
We hope to start being able to show interesting results in the second year of the project, so check back in early 2026 when there might be more to see!
The project members will be present at:
-
Global Glacier Modelling Workshop, Oslo, 10-12 Feb. 2025
-
Alpine Glaciology Meeting, 27-28 Feb. 2025
-
Céline has a poster on identifying possible future GLOF sites
-
-
EGU 2025, 27 April-2 May 2025
-
Céline will be presenting the same poster here
-
Do feel free to come and talk to us about the project! Over the next year, we hope that all three of us will be presenting project-related material.
- Glacier modelling
- Glacier processes
- Glacier hydrology
- Machine learning
The project is funded by the Elite Network of Bavaria from 01.09.2024 to 31.08.2030.