Data science
MarDATA
[Stephan Juricke, Arne Biastoch]
GEOMAR and the Ocean Dynamics research unit are dedicated to promote the development of the rapidly expanding field of marine data science. As part of this effort, Ocean Dynamics is chairing the Helmholtz School for Marine Data Science MarDATA in close collaboration with the Kiel University, the Alfred Wegener Institute Helmholtz Centre for Polar and Marine Research Bremerhaven, the University Bremen and the Constructor University of Bremen. Through MarDATA, doctoral students are approaching marine data science projects from an interdisciplinary perspective, addressing topics of the marine science domain with advanced data science technologies.
Applications
[Nils Hutter, Yannick Wölker, Sweety Mohanty, Rajka Juhrbandt, Willi Rath]
In recent years, machine learning (ML) has transformed numerical weather prediction, enhancing both efficiency and accuracy through data-driven models. At GEOMAR, we develop efficient emulators for sea ice and ocean dynamics, designed to enable stable, long-term climate simulations at drastically reduced computational costs—paving the way for digital twins of the ocean. In addition to purely statistical surrogate models, our work focuses on building hybrid numerical-ML models that integrate traditional physics-based simulations with the flexibility and adaptability of machine learning. This agile framework allows us to replace individual physics components with simplified, interpretable machine learning models, providing a unique opportunity to extract new physical insights from observational data. Furthermore, we develop ML-based parameterizations for oceanic processes and statistical models for predicting and deriving key climate indices, such as the Atlantic Meridional Overturning Circulation (AMOC).
