Background
Heatwaves, floods and storms are increasing under climate change but there is uncertainty in the rate of increase. To help quantify the uncertainty, we use computer models and chaotic sensitivity to initial conditions to produce artificial but realistic scenarios that span the current and future changes in extreme events. These simulations produce events that are more extreme than anything we have yet experienced but still obey the laws of physics and could therefore occur. Plausible bounding scenarios for increasing risks of extremes along with dynamical mechanisms and precursors that give rise to unprecedented extremes are now a realistic prospect, to to provide clear scenarios for forecasters and decision makers. With the advent of AI models trained on past data, serious question are now also being asked about whether these models can generate unprecedented extremes by extrapolating beyond observed limits of the current climate.
PhD Opportunity
Large ensembles of climate simulations out to seasons or years ahead provide an invaluable database for analysis of extreme events. This project will use existing datasets and complement them with new AI/ML generated data to answer the following research questions:
1. How extreme could current weather and climate events be?
2. What is the uncertainty and reasonable worst case scenario for how extremes could change in the next few years?
2. What are the mechanisms and precursors behind the atmospheric dynamics that drive truly unprecedented events?
3. Are AI/ML models able to extrapolate out of sample to generate similar unprecedented extremes?
This kind of analysis, dubbed the UNSEEN method, has been used to inform government decision making in the past and these analyses will provide clear information to inform future decision makers and contingency planners. The analysis is not restricted to the UK and there is scope for the student to shape their own project and research to look at a range of global extremes. However, the datasets are large and you will need good computational analysis, statistical and geophysical fluid dynamical knowledge to carry out the research.
As well as the main supervisor Professor Adam Scaife, the project will also be cosupervised by Professor James Screen at Exeter University and Professor Amanda Maycock at Leeds University.
Applicant Profile
Students with a strong background in physics, maths and statistics would be well suited to this project. Good programming skills e.g. in Python or similar for data analysis and visualisation are also required.
Other Information
Recent papers in this area include:
Rapidly increasing chance of record UK summer temperatures (https://rmets.onlinelibrary.wiley.com/doi/10.1002/wea.7741)
and
Current chance of unprecedented monsoon rainfall over India using dynamical ensemble simulations (https://iopscience.iop.org/article/10.1088/1748-9326/ab7b98)
