How can clouds be described more accurately?

We can see clouds almost every day in the sky. They come in various shapes and sizes - sometimes white and puffy, sometimes dark grey and covering the entire sky. Clouds are usually categorized in three broad groups based on their altitude and then further split into types, classes or 'regimes' depending on their observable characteristics - presence of lightning, vertical and horizontal spread, pattern size… However it has proven difficult to define those cloud regimes objectively, which is especially of interest since different cloud types have very different radiative effects. This will be achieved by combining on-the-ground observations of cloud types - called synoptic observations - and satellite imagery. This data will then be fed to a Machine Learning model in order to build an interpretable and explainable cloud classification method. In fact clouds form thanks to particles present in the atmosphere called aerosols - like dust or pollutants -  on which water droplets or ice crystals develop. Hence clouds are really sensitive to such particles and the effects of their interactions are still hard to characterize and quantify as of today. The second part of the project is thus to study the relationship between the properties of clouds in each class and aerosol particles to better understand both those interactions and cloud adjustment in response to aerosols.