Thu, 19 Jun 2025
The biggest tech firms are using AI to forecast the weather but is that better than existing models?
Machine-learning weather models have been developed by big businesses, challenging traditional physics-based computer forecasts.
The accuracy of these new models varies depending on the variable being forecasted.
Some machine-learning models performed better than traditional ones at predicting atmospheric pressure patterns, while others trailed behind.
However, they underperformed compared to traditional models at smaller scales and were unable to predict showers or specific weather events like the Boscastle floods.
AI models may struggle to forecast effects from rare events not seen in their 40-year training data, such as the 1991 eruption of Mount Pinatubo.
There are also concerns about how well AI models will perform in a warmer world due to climate change.
Traditional physics-based models are still necessary for accurate forecasts and will likely continue to run alongside machine-learning models.
>>
Terms of Use | Privacy Policy | Manage Cookies+ | Ad Choices | Accessibility & CC | About | Newsletters | Transcripts
Business News Top © 2024-2025