Oxford team uses AI to find three simple equations behind North Atlantic climate swings
Using an AI method called equation discovery, Oxford physicists derived three readable equations that capture how North Atlantic ocean temperatures, air pressure and rainfall interact over decades.
By Diego Callegari, Editor
· United Kingdom · 2 min read
Source: University of Oxford Department of Physics ↗
Physicists at the University of Oxford have used an AI technique called equation discovery to derive three simple equations that describe how the North Atlantic's ocean, atmosphere and rainfall interact over decades, the university's physics department said. The study, with the UK Centre for Ecology & Hydrology and the Met Office, was chosen as a highlight paper by the journal Earth System Dynamics.
Two climate patterns shape weather from Europe to the Sahel
The North Atlantic's climate swings naturally over years and decades, driven by two main patterns. The Atlantic Multidecadal Variability is a slow rise and fall in sea surface temperatures, linked to Atlantic hurricanes, Sahel droughts and summer weather in Europe and North America. The North Atlantic Oscillation is a pressure see-saw between the Azores and Iceland that steers European winters.
Forecasters lean on both for predictions from seasons to decades ahead. How they interact, and what role rainfall plays, is still debated.
Unlike a black box, the AI produced equations people can read
Many machine-learning models make predictions without showing how. Equation discovery instead searches data for the simplest equations that describe how a system changes. Using an algorithm called SINDy, the team analyzed monthly observations from 1950 to 2022, picked three linked equations from hundreds of candidates, and then tested them on records from 1900 to 1939 that played no part in building them.
“Rather than starting with equations that we think should describe the climate system, we let the data guide us towards the equations that best capture how the North Atlantic evolves.” — Andrew Nicoll, DPhil student and lead author
The equations reproduced the decade-scale variability in the records and showed skill in predicting statistics of ocean temperatures and rainfall several years to a decade ahead. They also produced a 20-year cycle, which the researchers read as a slowly fading ocean oscillation, as the published paper describes.
Rainfall emerged as a key link between ocean and air
Rainfall terms were among the most consistent across all candidate models. Rain releases heat into the atmosphere as vapour condenses, and adds fresh water that lowers ocean salinity, which can slow the currents carrying heat north. On that basis the team proposes new feedbacks, including that the North Atlantic Oscillation responds differently to ocean temperatures depending on whether rainfall is above or below normal. Simple models are earning their place in climate science elsewhere; at NYU Abu Dhabi, a deliberately simple algorithm forecasts Arctic sea ice up to nine months ahead.
These are proposals, not settled findings. Professor Hannah Christensen, who leads the research group, said the team will spend the coming months checking the predictions against observational data and developing new theories for the region.
Sources
- University of Oxford Department of Physics · physics.ox.ac.uk
- esd.copernicus.org
How we choose and verify sources: editorial policy.










