Data Science

The Institute for Computational Science is a major driver of research in Data Science at UZH. At ICS, we focus on Data Science for two major directions of research: Environmental sciences and geosciences (Jan Dirk Wegner) and astrophysics (Robert Feldmann, Joachim Stadel).

To solve scientific questions in the environmental sciences and geosciences, we research at the frontier of machine learning, computer vision, and remote sensing. The objective is to invent original, data science methods that analyze environmental data at very large scale automatically. We innovate on a very technical level and closely collaborate with our colleagues from, for example, ecology to jointly find new ways to protect our environment at global scale. Scientific projects include global mapping of vegetation parameters like canopy top height and carbon stocks at very high spatial and temporal resolution, monitoring of agricultural land, water- level prediction under flooding scenarios, or establishing a rapid-alert system that detects forest degradation. On the technical side, we investigate exciting topics like uncertainty quantification in deep learning, explainable and casual AI, graph neural networks, or time-series analysis with neural ordinary differential equations and transformers. We believe that interdisciplinary research is key to scientific breakthroughs and always aim at putting our research into practice by collaborating with NGOs, companies or public administration.

Astrophysics with its ever-increasing amounts of data collected with terrestrial and spaceborne sensors is another exciting research direction for data science. For instance, the James Webb Space Telescope launched at the end of 2021 produces 30 Gigabytes of raw data every day, with subsequent post-processing dramatically increasing the size of the resulting scientific data sets. The next generation Vera C. Rubin Observatory will image the entire sky every few nights generating 30TeraBytes/night or 10PetaBytes/year. The future Square Kilometre Array Observatory, which scans the sky at radio wavelengths from two locations in South Africa and Australia, is expected to produce 8.5 Exabytes during the first 15 years of operation. These new data sets offer exciting perspectives for new discoveries, tests for our current astrophysical models, and a rich playing ground for the development of new data science tools and techniques. At ICS, we are interested in exploring these data sets and comparing them to theoretical predictions to learn more about the formation of cosmic structure, the composition of intergalactic and interstellar gas, and the evolution of galaxies. In addition, we are working on developing new machine learning methods with applications in a broad range of research areas. These new methods not only complement traditional approaches based on high-performance computing but they often make it possible to tackle much harder problems that possible before. As such, data science can really be understood as the third pillar of astrophysical research on par with theoretical modelling and observational studies.