AI Weather System Delivers Fast, Accurate Forecasts Without Supercomputers

Image Credit: Lucy Chian | Splash

A team of researchers from the University of Cambridge, in collaboration with the Alan Turing Institute, Microsoft Research, and the European Centre for Medium-Range Weather Forecasts (ECMWF), has developed a groundbreaking artificial intelligence system called Aardvark Weather. This innovative technology, detailed in a study published in the journal Nature on March 20, 2025, offers highly accurate weather forecasts using minimal computing resources, such as a standard laptop. Unlike traditional forecasting methods that rely on supercomputers, Aardvark Weather aims to make weather forecasting faster, more affordable, and more accessible, particularly for regions with limited technological infrastructure.

[Read More: The Evolutionary Journey of AI]

How Aardvark Weather Works

Aardvark Weather replaces the complex, multi-stage process of traditional weather forecasting with a single AI model. Typically, weather forecasts involve gathering data from satellites, weather stations, and other sources, processing it through numerical models, and refining predictions with human expertise—a process that can take hours on specialized supercomputers. Aardvark, however, uses machine learning to analyze raw data directly, producing global and local forecasts in seconds or minutes. The system was trained on the ECMWF’s ERA5 dataset, which combines decades of weather observations with physical models, enabling it to learn patterns and make predictions efficiently.

The AI model’s design includes an encoder, processor, and decoder modules, with a total of about 87 million parameters. It can be trained in approximately 100 GPU hours and run on a desktop computer, using just 10% of the data required by conventional systems. This efficiency allows Aardvark to generate forecasts thousands of times faster than existing methods while maintaining or surpassing their accuracy. For instance, it outperforms the U.S. Global Forecast System (GFS) on several variables and competes with forecasts enhanced by human meteorologists.

[Read More: Google Cloud’s AI Predicts 15-Day Weather for Energy Sector]

Implications for Accessibility

One of Aardvark Weather’s most significant advantages is its accessibility. By eliminating the need for supercomputers, the system enables accurate forecasting in regions where advanced infrastructure is scarce, such as developing countries. Its flexibility allows it to be tailored quickly for specific needs, such as predicting rainfall for agriculture in Africa or wind speeds for renewable energy in Europe. This adaptability contrasts with traditional systems, which require years of development for customization. Dr. Scott Hosking from the Alan Turing Institute emphasized that Aardvark “democratizes forecasting”, making advanced weather predictions available to communities in the Global South and data-sparse regions.

The system’s potential to support climate resilience is notable. It can be applied to predict extreme weather events like hurricanes, wildfires, and floods, as well as broader environmental phenomena such as air quality and sea ice dynamics. This versatility could aid policymakers, emergency planners, and industries in preparing for climate-related challenges.

[Read More: AI-Powered NeuralGCM: Revolutionizing Climate and Weather Forecasting]

Collaborative Innovation

The development of Aardvark Weather highlights the power of collaboration between academia and industry. Led by Professor Richard Turner from the University of Cambridge, the project integrated expertise from the Alan Turing Institute, Microsoft Research, and ECMWF. Microsoft has committed to keeping Aardvark open-source, ensuring that the global scientific community can build upon and share its advancements. Matthew Chantry from ECMWF noted the importance of such partnerships in advancing AI-driven forecasting while openly sharing data for broader societal benefit.

[Read More: The Battle for AI's Future: Open vs. Closed Source]

Future Prospects

The Alan Turing Institute plans to establish a dedicated team, led by Professor Turner, to further develop Aardvark Weather. This team will focus on deploying the system in developing regions and expanding its capabilities for environmental forecasting, such as ocean dynamics. Researchers, including lead author Dr. Anna Allen, see Aardvark as a starting point for a new era of AI-driven forecasting, with potential applications beyond weather, such as in disaster preparedness and resource management. However, the system is still experimental and requires refinement to match the resolution of top-tier models like ECMWF’s Integrated Forecast System.

[Read More: The Forecast Revolution: How AI is Reshaping Weather Predictions]

License This Article

Source: The Alan Turing Institue, The Guardian, arXiv

3% Cover the Fee
TheDayAfterAI News

We are a leading AI-focused digital news platform, combining AI-generated reporting with human editorial oversight. By aggregating and synthesizing the latest developments in AI — spanning innovation, technology, ethics, policy and business — we deliver timely, accurate and thought-provoking content.

Previous
Previous

AI-Powered XenoBug by IISER Bhopal Identifies Bacterial Enzymes for Environmental Cleanup

Next
Next

AI Breakthrough in Biodiversity Conservation at McGill University: New Study Highlights Potential