Climate change mitigation and adaptation depend on our ability to predict climate. Aerosols (small particles in the air that interact with sunlight and clouds) and clouds cause a lot of uncertainty in climate models. A large part of this uncertainty comes from unknowns in the inputs to the model that inform it about emissions, properties and processes that affect aerosol and cloud. To better understand aerosols and clouds and their effect on climate, observations have been made from instruments on the ground, on aircraft and on satellites. We will use observations from all these platforms to improve a climate model. To do this, we use a machine learning technique to examine model predictions taking into account the major unknowns in the model aerosol and cloud inputs. By comparing these predictions to the observations, we can rule out model versions that are incompatible with observational evidence. This will reduce the model uncertainty, improving our ability to predict climate.

2024

2026