Quantifying Near-Cloud Aerosol Properties and Radiative Effects across Various Cloud Organizations Using Machine-Learning Techniques

Aerosols remain one of the largest sources of uncertainty in radiative forcing estimates. While reducing uncertainty is undoubtedly important, much less attention is paid to the appropriateness of the best estimate of aerosol direct forcing, which is largely constrained by aerosol properties in clear-sky regions and has ignored the contribution from near-cloud aerosols. Since near-cloud regions cover 20-30% of the globe and aerosols have distinctly different properties from those far away from clouds, ignoring near-cloud aerosols can bias our forcing calculations. However, it is challenging to retrieve near-cloud aerosol properties from passive satellite observations due to the adjacent three-dimensional (3D) cloud radiative effects. We will briefly introduce a machine-learning-based method and explain how we overcome this issue and retrieve near-cloud aerosol properties. With this advanced retrieval capability, we will contrast how near-cloud aerosol properties vary under four different cloud organizations dubbed Sugar, Gravel, Flowers, and Fish in the trade-wind region, and discuss the implied aerosol direct radiative effects for the current and future climate.