YES-27-A-03 Machine learning to improve understanding of regional aerosol effects on climate

Most climate adaptation relies on Regional Climate Models (RCMs) because they provide climate information at a scale relevant to the decisions that they inform. However, much of the climate modelling focus has been on global, not regional climate modelling, with RCM representation of earth system processes often a generation behind global climate projections. As a consequence, most current National Climate Scenarios neglect the impact of regional dimming and brightening (aerosol driven trends). In continental conditions, downwind of aerosol emission sources, this can lead to under-estimation of regional warming of up to 50%, leaving countries under prepared for the potential magnitude of future regional change. This context is often lost to potential users of these National Climate Scenarios.

Global climate projections demonstrate that aerosols can drive a range of regional trends, but lack the resolution to inform adaptation. New RCM projections now account for historical and future aerosol trends, but do not sample uncertainty in the range. This Phd will assess how missing (aerosol) uncertainties in current RCM generation leads to overly confident picture of future climate change. The project will then explore of “aerosol aware” generation of RCM scale climate datasets (using “ML Downscaling tools”) can lead to climate projects consistent with current state-of-the-art global climate modelling. This work can form a pathway that National Climate Scenario groups could follow to translate current global climate projection uncertainties into impact relevant high resolution climate datasets.

Please visit the lead supervisor website to find out more about the aerosol-climate research group: https://environment.leeds.ac.uk/see/staff/1196/professor-ken-carslaw-frs