Alpine karst spring discharge prediction in view of climate change using recent advances in Deep Learning

• Applicants: Dr. Stefan Broda; Professor Dr. Nico Goldscheider; Dr. Tanja Liesch
• Subject Area: Hydrogeology, Hydrology, Limnology, Urban Water Management, Water Chemistry, Integrated Water Resources Management
• Term: since 2023
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 529209885

Karst aquifers play an important role in the Alpine region. They cover about 56% of the area and a substantial part of the population is completely or partially dependent on drinking water from karst springs, which are often associated with valuable ecosystems, and contribute to hydro power generation. It has been stated that the Alps are amongst the most vulnerable areas towards climate change in Europe. As a result of increasing air temperature, the volumes of snow and ice stored will be greatly reduced resulting in a shift of water balance combined with a seasonal redistribution of precipitation. Further, high and low water flow events are expected to occur more frequently. State-of-the-art modelling of karst spring discharge, mostly with conventional distributed or lumped parameter models is still limited to site-specific, mostly research studies. These models require a lot of manual model tuning and calibration. As of to date, no easily transferable approach, which is simultaneously applicable to many karst spring catchments, is available. In this project, we will develop a modern Deep Learning based approach for modelling karst spring discharge, which is uniquely suited for building up more general models that take advantage of the information from different sites. Deep Learning is a subfield of machine learning on artificial neural networks, that has emerged as a highly successful field both in academic challenges and industrial applications. The proposed study region will be the Alps, comprising karst areas in Austria, Switzerland, Germany, France, Italy, and Slovenia, with a focus on the mountainous area delimited by the Alpine Convention, that is particularly affected by climate change. The study will take the World Karst Spring hydrograph (WoKaS) database as a basis. It will be supplemented with additional data from authorities and water suppliers during the course of the project, especially in regions with poor coverage of WoKaS. The work will include the creation of a comprehensive dataset including catchment attributes and meteorological forcings for about 150 springs. Classical lumped parameter models will be set up as benchmarks and compared to the newly developed Deep Learning based model results. The goal is to examine the suitability of new Deep-Learning modelling approaches for climate change impact assessment for a variety of short- and longterm predictive modelling tasks. Further, an in-depth case-study of the Dachstein area, whose large karst region contributes significantly to water supply, will expand the comparative examination with a physically based 3D model. Lastly, we will use the newly developed models to infer what likely impacts climate change will have on the alpine karst aquifers.

• DFG Programme: Research Grants
• International Connection: Austria
• Partner Organisation: Fonds zur Förderung der wissenschaftlichen Forschung (FWF)
• Co-Investigator: Professor Dr. Thomas Himmelsbach
• Cooperation Partners: Professor Dr. Josef Hochreiter; Dr. Daniel Klotz; Dr. Gerhard Schubert