Algorithms for multimodal data-based inference and prediction for climate science applications

Project

This project will develop new multimodal deep learning algorithms that are able to capture the relationship between diverse data and image modalities – such as infrared imager information, hyperspectral infrared sounding information or active remote sensing – in order to deliver relevant insights in outstanding climate science challenges such as aerosol-cloud-precipitation-climate interactions.

The project will also develop robust deep learning algorithms that are able to deal with a number of issues arising in the climate science domain such as: i) uncertainty in the testing/training data, ii) different degrees of uncertainty in different data modalities, and iii) small data.

The algorithms will also be applied to real data to determine historical auto-conversation rates from actual satellite observations. These novel observations are critical to evaluate the representation of precipitation.

Supervisors

Miguel Rodrigues (supervisor), Johannes Quaas (co-supervisor), Alberto Arribas (MetOffice, non-academic advisor)

Secondments

3 months at University of Leipzig, 3 months at the MetOffice InformaticsLab

Enrolment in Doctoral degree

PhD in Electronic & Electrical Engineering at University College London