3 min read
0 comments
A new artificial intelligence developed by researchers is revolutionizing meteorology and weather forecasting. Capable of predicting the weather faster than traditional methods, the AI works even on regular computers, making forecasts more accessible and efficient.
A new weather forecasting system is surprising researchers. It uses artificial intelligence, is faster than supercomputers and consumes much less energy. Its name is Aardvark Weather.
Developed in just 18 months, the system is already comparable to the best current meteorological models.
And most impressively, it runs on a regular computer, eliminating the need for expensive, powerful machines.
- Thailand has unveiled its first police robot — equipped with 360-degree cameras, facial recognition and real-time analysis of footage from drones and security cameras
- China makes AI education mandatory by 2025 — children as young as six will have to learn about the topic
- Artificial intelligence helps prevent fires in corn ethanol plants
- Bill Gates believes that these professions will be replaced by AI
Weather forecasting without supercomputers
Today, weather forecasts are made based on complex physical models. These models require a lot of data and hours of processing in supercomputers. It is an expensive, slow and energy-intensive process.
Aardvark Weather takes a different approach. It uses machine learning to predict the weather from raw data.
This includes satellite images, records from weather stations, ships and balloons. With this data, the system can generate forecasts directly, without going through traditional atmospheric models.
According to researchers, Aardvark is dozens of times faster than conventional systems.
And it uses thousands of times less energy. Instead of a room full of servers, it only needs one desktop computer.
Amazing performance
In tests, the Aardvark outperformed the Global Forecast System (GFS), being the national model for the United States. Even using only 8% of the data that the GFS uses, the new system was more accurate in several situations.
It also came close to the official predictions of the US Weather Service. This caught the attention of experts.
However, there is a limitation: Aardvark’s spatial resolution is still lower. It uses a grid of 1,5 degrees latitude by 1,5 degrees longitude. GFS, on the other hand, operates on a grid of 0,25 degrees.
This means the new system may be less accurate in very local predictions, such as rainfall in a specific neighborhood.
Potential for growth
Even with this limitation, the researchers see a great future for Aardvark. Because the system learns from the data it receives, it can be adjusted for different regions and needs.
One example would be using high-resolution data to improve local forecasts. This could benefit agriculture in parts of Africa or optimize renewable energy production in Europe.
Additionally, Aardvark can be adapted for other types of forecasts, such as hurricanes, wildfires, tornadoes, air quality, and even sea ice.
"These results are just the beginning“, said Anna Allen, co-author of the study. “The Aardvark approach can be applied to a variety of environmental challenges.”
Forecast for everyone
The impact of the new system goes beyond technology. It could help regions that currently lack access to supercomputers or advanced infrastructure.
Scott Hosking of The Alan Turing Institute echoes this idea. “Aardvark’s breakthrough isn’t just about speed, it’s about access. We can bring weather forecasting to places that previously couldn’t get this kind of technology."
With this, Aardvark can democratize weather forecasting. And at the same time, reduce costs, energy consumption and processing time.
The full Aardvark Weather study was published Thursday (March 20) in the journal Nature.
