Machine learning for climate extremes
About the project or challenge area
High-resolution predictions of climate risks from extreme weather events are very important for assessing risks to society, but these are much too expensive to produce for most applications. This project will focus on hurricane simulations. Machine learning methods have been trained to make high-resolution predictions given input from cheaper coarse-resolution models, but this has not been shown to be reliable for extreme events, due to a lack of high resolution data. This project will fill that gap by using high-resolution extreme hurricane simulations, produced using rare event sampling methods, to train neural networks to predict detail of extreme hurricanes’ winds and rainfall given low-resolution input. This would enable better climate predictions and be the first demonstration of the use of rare event sampling for developing AI tools.
Prof Dorian Abbot, Department of Geophysical Sciences, University of Chicago - will provide additional co-supervision and data for the project.
Why choose this opportunity?
You will have the opportunity to learn how to use climate model simulations, model extreme weather events and train machine learning algorithms to produce a scientifically useful tool. This will give you very valuable expertise in both climate science and AI. You will also have the chance to build international links with Prof Dorian Abbot’s group at the University of Chicago who are partners on this project.
You will have enthusiasm for developing scientifically useful machine learning applications and understanding high-impact weather events. The project would also best suit a student with experience of writing computer programs to carry out data analysis. Prior experience of machine learning is not essential, but would make an application more competitive.
How to apply
All students can apply using the button below, following the Cabot Masters by Research Admission Statement. Please note that this is an advertised project, which means you only have to complete Section A of the Research Statement.
Before applying, we recommend getting in touch with the project's supervisors. If you are interested in this project and would like to learn more about the research you will be undertaking, please use the contact details on this page.
Find out more about your prospective research community
The Environmental Change theme is a vibrant community of researchers who integrate expertise across multiple disciplines to provide the evidence base and solutions to tackle the world's most pressing environmental challenges. Find out more about the Environmental Change research theme.