Quantifying the Causal Effect of Cloud-Controlling Factors on Marine Stratocumulus Clouds

Abstract Our incomplete understanding of clouds and their role in cloud–climate feedbacks leads to large uncertainties in climate projections. By applying causal discovery, an unsupervised machine learning method, and afterward causal effect estimation (supervised method) to daily satellite and reanalysis data, we systematically analyze causal links between cloud properties and selected cloud-controlling factors with the aim of identifying and quantifying the sensitivity of cloud properties to different factors. This will contribute to improving our understanding of how clouds react to changes in controlling factors as a consequence of climate change. Here, we focus on marine stratocumulus (Sc) clouds off the coast of South America. Specifically, we analyze dynamical and thermodynamical cloud-controlling factors to quantify the causal effects on macrophysical properties of marine Sc clouds. In agreement with previous studies, we find that sea surface temperature, lower tropospheric stability, surface sensible heat flux, and 10-m wind speed are the main drivers influencing these clouds. While the causal links between these factors and the cloud properties—total cloud cover, total cloud water path, and cloud optical depth—show similar behavior, the cloud effective radius remains largely unexplained, suggesting that the background aerosol might play an important role. In contrast, the cloud-top pressure is influenced by all cloud-controlling factors investigated, except the surface sensible heat flux. Application of this approach to climate model output could be a step toward a process-oriented evaluation of the sensitivity of simulated clouds to climate change.