Optimizing Kilometer-Scale Climate Modeling: Refining Cloud Microphysics Using Machine Learning and Satellite Correlation

The modeling of the Earth Climate System has undergone outstanding advances to the point of resolving atmospheric and oceanic processes on kilometer-scale, thanks to the development of high-performance computing systems. In the preparation phase of the global kilometre-resolution coupled ICON climate model, there's a critical need to fine-tune cloud microphysical parameters. Our approach involves investigating the optimal calibration of these parameters using machine learning techniques. Our initial focus involves calibrating the autoconversion scaling parameter by correlating it with satellite-derived top-of-atmosphere and bottom-of-atmosphere radiation fluxes. This calibration process entails conducting limited area simulations specifically within the North Atlantic and South Pacific region using ICON. Through these simulations, various adjustments to cloud microphysical parameters are made, aiming to assess their potential impacts on radiation flux output.