Data assimilation optimally combines geophysical observations with numerical models. Expanding the capabilities of current data assimilation systems to enable optimal ingestion of observation in the context of Earth System models (where several modeling components of the Earth system are coupled together) is of paramount importance for both climate reconstructions and short to long-range ocean and climate predictions. Within this research topic, the PhD candidate will assess the potential of deep learning in data assimilation, with a particular emphasis to the ocean. S/he will investigate the possibility to integrate traditional data assimilation schemes with new algorithms inherited from deep learning to enhance the exploitation of the current observing networks, e.g. neural network-based observation and cross-component operators.

Linear algebra and statistics; machine learning; programming skills; background in Earth System physics (oceanography and atmospheric physics) is a plus.

The PhD research project will be jointly supervised with prof. Buizza from Scuola Superiore Sant'Anna - Pisa. Additionally, the Rome branch of CNR ISMAR provides computational facilities (HPC) and a vibrant environment with several collegues expert in satellite oceanography and related disciplines, and several PhD students and postdocs.