Learning about cloud-climate effects from machine-learning calibration of cloud parameters in climate models

• Applicant: Professor Dr. Johannes Quaas
• Subject Area: Atmospheric Science
• Term: since 2024
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 545086515

LACCMACC addresses the representation of cloud- and precipitation mechanisms in atmospheric models with the aim to improve the understanding of their role for simulated cloud-climate effects. Specifically, both the cloud response to global warming, and the effective forcing due to aerosol cloud interactions will be explored in revised simulations.The key idea is that new machine-learning-based approaches allow for an objective „tuning“ of the empirical parameters in the process representations. LACCMACC will adapt and apply such a tool,the „HighTune“ method, in cooperation with its developers at the Laboratoire de Météorologie Dynamique in Paris, France (Frédéric Hourdin). HighTune will be applied to the ICON atmospheric model in kilometre-resolution configuration in regional and subsequently global setting. As a reference to the parameter choices this scheme will determine, LACCMACC will make use of process-oriented parameterisation evaluation approaches developed by the proposing team in the past. The final key step is to analyse parameter choices in their regional and spatial variation. The hypothesis is that inconsistencies, discrepancies and implausibilities of the choices as found by themachine-learning scheme in comparison to default values and to process-oriented evaluation will allow to conclude about systematic problems in parameterisations and possibly remedy these. A revised, optimal, set of parameters will be compared to default and process-evaluation-based settings in terms of simulated cloud-climate feedbacks and aerosol-cloud effective forcings.

• DFG Programme: Research Grants