Physics-aware and explainable machine learning for dust and cloud properties retrieval

Project

This project will develop advanced machine learning models for prediction, modeling,  emulation, detection and causal inference. Three main application and development domains will be:

1. Physics-aware and explainable machine learning for dust and cloud properties statistical retrieval of dust aerosols and cloud properties.;
2. Automatic detection of aerosol-cloud interactions in observations space by designing statistical emulators of high-resolution simulations, which will allow automatic sensitivity analysis and fast retrievals;
3. Detection of aerosol-cirrus cloud interactions in satellite data via unsupervised learning and observational causal inference to identify latent factors potentially impacting the detection.

Supervisors

Gustau Camps-Valls (supervisor), Philip Stier (co-supervisor), Ralf Giering (non-academic advisor)

Secondments

6 months at the University of Oxford and the Alan Turing Institute & 3 months at FastOpt

Enrolment in Doctoral degree

PhD in Image Processing at Universitat de Valencia