Dr. Anneli Guthke’s junior research group at SimTech welcomes its newest member, Mehrdad Mohannazadeh. Having a background in applied mathematics, he conducted his doctoral studies in the field of computer science and machine learning with a thesis on trustworthy, probabilistic classification models at the University of Bielefeld. As a researcher at UFZ Leipzig, he has contributed to the development of an operational flood forecasting system for Germany.
At SimTech, he will work on the DFG-funded project “Diagnostic Evaluation of Hybrid Hydrological Models in the UNITE Framework (UNI-BENCH)”, a continuation of the group’s successful previous work on model diagnostics. The goal will be to extend the formal evaluation framework for comparing physics-based, data-driven, and hybrid hydrological models toward three core dimensions of model evaluation: performance, interpretability, and process consistency. Hybrid models, which combine physically-based and machine learning components, promise both predictive power and process insight - but often fail to deliver trustworthy, interpretable results.
UNI-BENCH addresses this by developing new diagnostics, including entropy- and mutual information-based metrics, to understand the internal functioning of hybrid models. The project aims to map model behaviour onto physical processes to ensure that models are “right for the right reasons.” Tools developed will be integrated into the existing open-source UNITE toolbox.
We wish Mehrdad all the best for his research at SimTech and are looking forward to a successful and inspirational collaboration!