Professor Corwin Wright will lead the three-year GW-LIFT project, which will use AI to combine decades of atmospheric measurements from satellites, balloons, aircraft and ground-based instruments. Funded by UK Research & Innovation (UKRI) through its new Turing AI Interdisciplinary Pioneer Fellows program, the project also includes a substantial allocation of computing time on the Isambard-AI national supercomputer, based at the Bristol and Bath Science Park.
The research will focus initially on atmospheric gravity waves: small-scale waves that transport energy and momentum through the atmosphere and influence processes including large-scale atmospheric circulation and climate. Despite their importance, gravity waves are poorly represented in current weather and climate models, partly because individual observing instruments capture only part of their behavior and effects.
GW-LIFT, short for a Gravity Wave Latent inFerence network for climate, will use AI to bring together measurements collected over several decades by a wide variety of satellite, balloon, airborne and ground-based instruments, including NASA and European Space Agency missions. By combining these different sources, the team aims to recover information that no individual instrument can provide. The researchers will then use the resulting information to develop and test improved representations of gravity waves in physical weather and climate models.
Unlike approaches which use AI to replace conventional weather models, the Bath project will use AI to improve the underlying physical models themselves. This is particularly important in data-limited environments such as the upper atmosphere, and for predicting future climate conditions where AI models can struggle to generalize beyond the data on which they were trained.
Professor Corwin Wright says: “This fellowship is a fantastic opportunity to tackle a problem that my group and I have been working on for many years. We have decades of high-resolution observations of the atmosphere, but each instrument only sees part of the picture, making it very difficult to use these measurements to improve our models.
“We're going to use AI to bring these observations together and turn them into better representations of atmospheric physics, ultimately helping to improve weather forecasts and climate predictions.”
Applications to Jet Streams, Aerosols and Clouds
The project builds on around a decade of research by Professor Wright’s group at the University of Bath using satellite observations to identify shortcomings in atmospheric models. GW-LIFT creates an opportunity to move from identifying those shortcomings towards correcting them.
Professor Wright, who is also a Co-Director of the Centre for Climate Adaptation & Environment Research (CAER), is leading the project at Bath, with Dr Juliana Jaen already working on it and further research and software engineering appointments planned. The principal external partner is Dr Jonathan Smith, Head of the AI Lab at the British Antarctic Survey, alongside collaborators at UK and international universities and research centers, including the US National Center for Atmospheric Research (NCAR).
Although the initial focus is gravity waves, the researchers believe the methods developed through GW-LIFT could be transferable to other atmospheric processes that are difficult to observe, including jet streams, aerosols and clouds. This could provide a more general approach for using observations and AI to improve environmental prediction.
Part of £39 Million UK Investment in AI Research
The Bath project is one of 20 new Turing AI Fellowships announced by UK Research and Innovation (UKRI), backed by more than £39 million in government funding. The fellowships are designed both to advance fundamental AI research and accelerate the use of AI across different disciplines.
Professor Charlotte Deane, Senior Responsible Owner for the UKRI IS8 AI Program and Executive Chair of EPSRC, said the fellowships would support researchers in advancing AI while applying it to challenges across a wide range of disciplines.
The Turing AI Pioneer Interdisciplinary Fellowships began in October 2026 and will run for three years.