Oxford: Researchers have used artificial intelligence to uncover simplified mathematical relationships that could improve scientists’ understanding of long-term climate variability across the North Atlantic region.

The research focuses on interactions between the North Atlantic Ocean, the atmosphere and rainfall patterns. Scientists used AI-based equation discovery to analyse observational information and identify relationships that may be difficult to detect through conventional approaches.

Rather than using artificial intelligence simply to forecast weather, the researchers employed it to search for mathematical equations that describe important physical relationships within the climate system.

The approach produced three relatively simple equations that capture connections between different components of the North Atlantic climate. These relationships could provide researchers with a clearer way of studying how ocean and atmospheric conditions influence one another.

The North Atlantic plays an important role in the global climate system. Ocean temperatures, atmospheric circulation and rainfall patterns in the region can influence weather conditions across Europe, North America and other parts of the Northern Hemisphere.

Understanding these interactions is challenging because the climate system contains many interconnected processes operating across different timescales.

Traditional climate models use large numbers of equations and variables to represent these processes. While such models are extremely valuable, identifying the simplest underlying relationships can be difficult.

AI-based equation discovery offers a different approach. Instead of asking a computer only to predict future conditions, researchers can allow algorithms to search through observational data for mathematical structures that explain recurring patterns.

The resulting equations can then be tested against independent observations to determine whether they represent genuine physical relationships rather than statistical coincidences.

This distinction is particularly important in climate science. An equation that performs well on existing data is not necessarily a reliable description of the underlying climate system.

Researchers therefore need to compare AI-generated relationships with established physical principles and observations.

The new work could eventually help scientists improve their understanding of climate variability and potentially make some aspects of climate modelling more efficient.

The findings may also demonstrate how artificial intelligence can complement traditional scientific methods. Rather than replacing physical climate theory, AI can be used as a tool for identifying patterns and suggesting relationships that researchers can investigate further.

The North Atlantic is an especially important area for this type of research because its ocean circulation and atmospheric behaviour are linked to climate conditions over a much wider region.

Changes in the ocean can influence atmospheric temperature and rainfall, while atmospheric conditions can in turn affect the ocean through winds, heat exchange and precipitation.

The newly identified relationships provide researchers with additional tools for examining these feedbacks.

Further testing will be needed to establish how broadly the equations can be applied and whether they remain accurate under different climate conditions.

Scientists are also likely to investigate whether similar AI techniques can identify useful relationships in other parts of the global climate system.

The study highlights a growing role for artificial intelligence in scientific research. When combined with high-quality observations and established physical knowledge, AI could help researchers uncover hidden patterns while keeping the resulting scientific explanations relatively simple.

As climate science becomes increasingly data-intensive, such approaches could become an important complement to conventional modelling and analysis.