The Climate Modelling research group at Oxford combines climate and computer science to build better models of the climate on Earth and other planets. We do research on numerical modelling, machine learning and high-performance computing for efficient predictions of future climates; data compression and information theory; and software engineering to build next-generation climate models.
Project Highlights
SpeedyWeather.jl
An interactive, flexible atmospheric model built to accelerate climate research.
Data compression and BitInformation
Compressing atmospheric data into its real information content
Testing generalisation of Machine Learning-based models
Machine learning has to respect the physics to generalise to future climates
News
- 25/10 - New paper on hybrid physics-machine learning model with Maximilian Gelbrecht and Niklas Boers has been published in JAMES!
- 25/10 - Milan was invited to give the annual lecture of Oxford’s Doctoral Training Centre on interactive climate modelling – live coding included, of course.
- 25/09 - SpeedyWeather was used for the hands-on sessions at the NCAS Climate Modelling Summer School this week! SpeedyWeather is taking over the teaching of the next generation of climate modellers. Thanks for everyone involved to make this possible.
- 25/09 - Valentin Churavy (U Mainz and U Augsburg and Enzyme developer) and Milan Klöwer gave a joint presentation on Differentiable programming for scientific computing with Enzyme and Julia at the ML coupling workshop in Cambridge UK last week! Pluto notebook of the talk is here.
- 25/06 - Juniper Tyree won the 3-minute thesis at the University of Helsinki for presenting their work on Fearlessly Compression Weather and Climate Data. Congratulations!