In short, Aardvark Weather, an AI-based system, promises to significantly increase weather forecasts by providing forecasts several tens of times faster, while using thousands of times more computing power than current methods. The system was developed by researchers at the University of Cambridge and is supported by the Alan Turing Institute, Microsoft Research, and the European Centre for Medium-Range Weather Forecast.
Because traditional methods rely on powerful supercomputers and extensive teams of experts, the speed and efficiency of modern prediction systems is essential, and often takes several hours to produce predictions.
Recent innovations from tech giants such as Huawei, Google and Microsoft have demonstrated that AI can significantly improve certain aspects of the forecasting process. These companies have achieved faster and more accurate predictions by integrating AI into these solvers.
As an example, Google is developing an AI model for weather forecasting, and is currently marketing two models to enterprise cloud customers. The model developed by Google Deepmind uses historical weather data to predict future conditions 10-15 days ago.
Aardvark represents a major advancement by replacing traditional prediction processes with a single streamlined machine learning model. A standard desktop computer can be used to process data from a variety of sources, including satellites and weather stations, to generate global and local predictions in minutes.
“Aardvark rethinks current methods of weather forecasting, offering the possibility to make weather forecasts faster, cheaper, more flexible and more accurate than ever,” explained Professor Richard Turner of the Faculty of Engineering at Cambridge. “Aardvark is thousands of times faster than all previous weather forecasting methods.”
Despite operating with only a small portion of the data used in existing systems, Aardvark outperforms the US GFS forecast system with several key metrics, and maintains forecasting and competitiveness for the National Weather Service, typically with multiple models and expert analysis.
“These results are just the beginning of what Aardvark can achieve,” said Anna Allen, the first author of Cambridge’s Computer Science and Technology Bureau. She said the end-to-end learning approach can be easily applied to other weather forecasting issues such as hurricanes, wildfires and tornadoes. It can also be used for broader Earth System predictions, such as air quality, ocean dynamics, and sea ice predictions.
One of the most interesting aspects of Aardvark is its flexibility and simple design. Learn directly from the data, whether or not you forecast temperatures to support wind conditions for African agriculture or European renewable energy companies, you can quickly adapt to create bespoke forecasts for a particular industry or location. This contrasts with traditional systems that require years of work by large teams to customize.
This feature could change weather forecasts in developing countries where access to expertise and computational resources is limited. “By shifting weather forecasts from supercomputers to desktop computers, we can democratize forecasts and make these powerful technologies available in developing countries and in data sparse regions,” says Dr. Scott Hosking of the Alan Turing Institute.
Aardvark is expected to play an important role in expanding the scope of weather forecasts. Turner said the model can ultimately accurately predict the 8-day forecast, exceeding the current model’s capabilities for three days. This advance, along with Aardvark’s adaptability and efficiency, positions it as a transformative force in meteorology.
Aardvark’s next steps include developing a new team within the Alan Turing Institute, exploring technology deployments in the Global South and integrating it into a broader environmental forecasting initiative.
