eWEEK content and product recommendations are editorially independent. We may make money when you click on links to our partners. Learn More.
Researchers at the University of Cambridge have developed a new weather forecasting model that can match the accuracy of older, more expensive models but do it cheaper and faster. The latest model, called Aardvark Weather, can run a single desktop computer and generate forecasts in seconds instead of hours.
“Aardvark reimagines current weather prediction methods offering the potential to make weather forecasts faster, cheaper, more flexible and more accurate than ever before, helping to transform weather prediction in both developed and developing countries,” said Professor Richard Turner, who led the artificial intelligence research. “Aardvark is thousands of times faster than all previous weather forecasting methods.”
New AI model could revolutionize weather forecasting
Researchers at the University of Cambridge developed the Aardvark Weather AI model in collaboration with the Alan Turing Institute, Microsoft Research, and the European Centre for Medium Range Weather Forecasts. The model can be run on a single desktop computer instead of a supercomputer, which could greatly improve access to weather forecasting in locations all over the globe.
Aardvark Weather replaces all stages of the weather forecasting process with a simple machine learning model. Depending on the complexity, it can generate a forecast in seconds or minutes, using only about 10% of the weather data required by current forecasting systems. Researchers say that as they continue to improve the model, it will become capable of other kinds of forecasting, such as air quality, ocean dynamics, and sea ice prediction.
How traditional weather forecasting works
Aardvark Weather represents a huge improvement over traditional weather forecasting, which is expensive and time-consuming. It relies on weather data provided by weather stations, satellites, and balloons, which take money and time to gather. Then, the weather data is fed into physics-based forecasting models that make calculations called numerical weather prediction (NWP).
These calculations are so vast and complex that they must be run on supercomputers, which require significant power and energy. The supercomputers also need an entire team of trained staff members to operate them, which further increases the cost. Researchers also customize each individual forecast model over the course of years to improve the accuracy and keep up with long-term changes in weather patterns.
Turner said that developing these older models was crucial to training the Aardvark weather AI model. He also said that his team was “indebted” to the European Centre for Medium Range Weather Forecasts (ECMRWF) for providing essential weather data sets that were used to train the Aardvark model.