Eliza Duncan(She/Her)University of Exeter
Eliza Duncan is completing a PhD at the University of Exeter, as part of the Centre for Doctoral Training in Environmental Intelligence. Her research focuses on untangling natural aerosol processes, using machine learning techniques and evaluation of climate models in a Lagrangian framework. |
 |
TalksUntangling Aerosol Processes: Exploiting Explainable Machine Learning Techniques in a Lagrangian Framework24-Apr-24Aerosols represent a major source of uncertainty in climate modelling, compounded by the inherent difficulty in accurately characterising their natural baseline. To investigate the complex relationships between sources, meteorology, and aerosol properties in different rural environments, this study utilises explainable machine learning techniques and a Lagrangian framework. We demonstrate the capability of predicting aerosol properties from airmass history using XGBoost regression models. By employing TreeSHAP to interrogate the models, we demonstrate the effectiveness of these model interrogation techniques to untangle the complex, non-linear processes that govern aerosol properties in rural and pristine regions. This framework allows us to further our understanding of aerosol processes and therefore improve representation in climate models. |