Integrating AI into Earth-System Models: A Focus on Southern Ocean Clouds and Radiation

Supervisory Team:

• Dr Marc Mallet
• Dr Sonya Fiddes
• Dr Alain Protat

Brief project description:

Earth-system models struggle to accurately simulate cloud properties over the Southern Ocean, leading to subsequent challenges in accurately predicting the levels of shortwave and longwave radiation at the bottom and top of the atmosphere (Mallet et al., 2023a, Elementa). Recent studies have opened promising avenues by using machine learning to reduce biases in the prediction of shortwave radiation over the Southern Ocean (Mallet et al., 2023b, Artificial Intelligence for the Earth Systems). This project aims to leverage recent observations of radiation and cloud properties collected from satellites and ships in the Southern Ocean. It will involve training and testing machine learning models using predictor variables from reanalyses and various versions of the Australian ACCESS model. The outcome of this work will be to demonstrate the feasibility of integrating machine-learning approaches into the accurate prediction of Southern Ocean cloud and radiation properties.

Skills students will develop during this research project:

The student will develop skills in working with gridded output data from Earth-system models, as well as how to train, test, and evaluate models built using machine learning algorithms in conjunction with ship-based observations of atmospheric data. They will also develop skills in using cloud computing through the National Computing Infrastructure.