Climate change represents one of the most pressing challenges afflicting our societies, with governments worldwide, non-governmental agencies, and other stakeholders placing a greater emphasis on measures to successfully tackle and mitigate its adverse effects.
A pressing scientific challenge relates to the need to understand the interaction between aerosols, clouds, and precipitation processes because these currently represent one of the major sources of uncertainty for weather and climate prediction.
This project overarching objective is therefore to contribute to our understanding of aerosol-cloud-precipitation-climate interactions by leveraging emerging machine learning approaches.
In particular, this project has been developing machine learning models that combine information from multiple sources capable of capturing the key process of precipitation formation which greatly control cloud lifetime, namely autoconversion rates — the term used to describe the transformation from cloud droplets to raindrops by collision and coalescence of cloud droplets — that leverage a plethora of data sources such as satellite observations. Such machine learning models will also be able to deal with some 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.
To facilitate the development of innovative new methods, we have collaborated with our climate science partner at Leipzig University which provide us with access to real climate datasets.
