Machine Learning for Improving Precipitation and Aerosol-Cloud Interactions

Clouds are critical for weather and climate prediction. The multiple scales of cloud processes and their dependence on the environment, turbulence and aerosols make cloud simulation difficult, and especially accurate representation of emergent processes (like cloud radiative effects). Often models and measurements are used to develop empirical relationships for large-scale models to be computationally efficient. Machine learning provides another potential tool to improve our empirical parameterizations of clouds and to ground precipitation processes and aerosol-cloud interactions more firmly in known and observed physics which is computationally expensive to represent at large scales. Here we describe approaches for using rain formation treatments from complex bin and even super-droplet models in bulk weather and climate models, and even directly using observations from cloud chamber experiments. We also describe approaches for using machine learning and observations to constrain the uncertain parts of cloud microphysics. Initial work has started with understanding rain formation and the effects of aerosols. We will illustrate use of ML for rain formation in two different climate models. We will also discuss the prospects for machine learning methods being used to improve interactions between cloud physics and turbulence and how this might lead to further advances.