Building a cloud in the southeast Atlantic: understanding low-cloud controls based on satellite observations with machine learning

Understanding the processes that determine low-cloud properties and aerosol-cloud interactions (ACIs) is crucial for the estimation of their radiative effects. However, the covariation of meteorology and aerosols complicates the determination of cloud-relevant influences and the quantification of the aerosol-cloud relation. This study identifies and analyzes sensitivities of cloud fraction and cloud droplet effective radius to their meteorological and aerosol environment in the atmospherically stable southeast Atlantic during the biomass-burning season based on an 8-day-averaged data set. The effect of geophysical parameters on clouds is investigated based on a machine learning technique, gradient boosting regression trees (GBRTs), using a combination of satellite and reanalysis data as well as trajectory modeling of air-mass origins. A comprehensive, multivariate analysis of important drivers of cloud occurrence and properties is performed and evaluated. The statistical model reveals marked subregional differences of relevant drivers and processes determining low clouds in the southeast Atlantic. Cloud fraction is sensitive to changes of lower tropospheric stability in the oceanic, southwestern subregion, while in the northeastern subregion it is governed mostly by surface winds. In the pristine, oceanic subregion large-scale dynamics and aerosols seem to be more important for changes of cloud droplet effective radius than in the polluted, near-shore subregion, where free tropospheric temperature is more relevant. This study suggests the necessity to consider distinct ACI regimes in cloud studies in the southeast Atlantic.