Summary
Big companies are using machine-learning weather models, which are different from traditional models that rely on physics. While some of these machine-learning models have been more accurate recently, others have not performed as well. The potential for these models to improve quickly means we could see big changes in forecasting accuracy soon.
Key Facts
- Machine-learning weather models are being developed by large companies.
- Traditional weather forecasts rely on physics-based models that use supercomputers.
- These traditional models take hours to run and use a lot of computer power.
- Machine-learning models use past weather data to make predictions quickly, often in under a minute.
- In recent tests, some machine-learning models were more accurate than traditional models for certain forecasts.
- The accuracy of these models can vary, and improvements are happening fast.
- Weather forecasts are important for safety and planning, as shown by significant damage and deaths from severe weather events.
