Quantum machine learning-based parameterizations for Earth System Models
Hedwig
Keller
German Aerospace Center, Institute of Atmospheric Physics
Poster
Quantum machine learning (QML) is a rapidly growing field that has promised various improvements over classical machine learning, and that the Earth system modelling community should position itself to explore [1]. Here, we show how to develop QML-based parameterizations for Earth system models (ESMs), for examples for cloud cover [2], turbulent fluxes [3] and convection [4]. Coupling these QML-based parameterizations to the ESM is challenging due to computational overhead, thus currently classical “shadows” [5] of the QML models are a good way forward. Finally, we explore the learning capabilities of the QML-models based on Explainable AI [6,7].
[1] Schwabe et al., Environmental Data Science 4, e35 (2025)
[2] Pastori et al, Machine Learning: Earth 2 (1), 015008 (2026)
[3] Dogra et al., In Prep.
[4] Londono Castillo, In Prep.
[5] Keller et al., In Prep.
[6] Schwabe et al., 2025 IEEE Int. Conf. QAI, p. 73-78 (2025)
[7] Sarandrea, Master’s Thesis LMU München (2025)
[1] Schwabe et al., Environmental Data Science 4, e35 (2025)
[2] Pastori et al, Machine Learning: Earth 2 (1), 015008 (2026)
[3] Dogra et al., In Prep.
[4] Londono Castillo, In Prep.
[5] Keller et al., In Prep.
[6] Schwabe et al., 2025 IEEE Int. Conf. QAI, p. 73-78 (2025)
[7] Sarandrea, Master’s Thesis LMU München (2025)
keller-hedwig-poster.pdf
(419.85 KB)