Researchers at the Department of Energy are using the world’s fastest supercomputer to bring up-to-the-minute, pinpoint weather forecasts weeks in advance to the general public’s finger tips. The project is called ORBIT, which stands for Oak Ridge Base Foundation Model for Earth System Predictability.
By training the world’s largest AI model for weather prediction, hyperlocal and hyperaccurate forecasts are now within reach. The model uses 113 billion parameters to predict weather up to 30 days in advance. For short term forecasts (days or weeks in advance), the model has seen accuracy rates as high as 95%. For long-term forecasts (more than two weeks in advance), it still achieves a 60-80% accuracy rate.
ORBIT not only produces highly accurate forecasts, but it also provides uncertainty values in its predictions. For example, it can indicate whether its forecast is made with 95% confidence, or 75% confidence, etc.
The research team ultimately hopes to train ORBIT to generate highly accurate forecasts months in advance, with the goal of benefiting government agencies, first responders, farmers, and the general public.
