Dr. Christopher Irrgang
I work on the application and integration of machine learning (ML) in Earth System Modelling (ESM). This research has started with the development and training of neural networks for specific tasks, e.g., the inversion of Earth system observations to estimate geophysical quantities. Further ML applications were developed in the topics of data assimilation and physical downscaling. Now, I increasingly focus on a generalized development of self-validating, physically consistent, and interpretable hybrids of neural networks and our available Earth system models (Fusion of ML and ESM).
This work includes a diverse range of research questions, e.g.:
- Where and why should machine learning be applied instead of traditional ESM approaches?
- How can we train neural networks to produce physically consistent outputs?
- How can we use neural networks to improve the forecast skill of current state-of-the-art numerical models?
- Can neural networks exceed the forecast skill of numerical models?
- What are the limitations of observation- and model-data driven ML tools?
- How can we quantify and assure explainability and interpretability of machine learning in ESM?
Our work in public media
CarbonBrief Guest Post: How AI is fast becoming a key tool for climate science
Opportunities and limits of AI in climate modelling
Artificial Intelligence learns continental hydrology
Artificial intelligence and data assimilation: A successful marriage for Earth system research
An artificial neural network for monitoring ocean warming
Since July 2017: PostDoc in section 1.3 "Earth System Modelling"
- Numerical simulation and interpretation of global processes in the Earth system
- Usage and integration of machine learning (ML) in Earth System Modelling (ESM)
2014 - 2017: PhD student in the Helmholtz graduate school GeoSim, Freie Universität Berlin (PhD Thesis)
2010 - 2013: M.Sc. Mathematics at Christian-Albrechts-Universität zu Kiel
2007 - 2010: B.Sc. Mathematics at Philipps-Universität in Marburg