MIT scientists unveil an AI model that doubles forecast accuracy at 2-6 weeks
MIT researchers Judah Cohen and Lester Mackey published in Nature an AI-based weather prediction model with post-processing that learns from its errors by comparing them with historical data. In the subseasonal range of 2 to 6 weeks, known as the "desert of predictability", overall forecast accuracy doubled compared with current models. The results could be useful for farmers and energy providers, and for predicting extreme events such as floods or cold waves.
In 35 seconds
- The model uses AI post-processing that compares its inaccuracies with historical data to correct them.
- In the 2 to 6 week range, overall forecast accuracy doubled.
- Three versions of the model were tested with the ECMWF system, based on historical climate averages.
Read the full story at Meteored España
IASummary generated by artificial intelligence from a story by Meteored España. It may contain errors: the original source prevails.
Comments
Be respectful. Comments reported by several people are hidden until a moderator reviews them.
You do not need an account to read, only to comment. Sign in with a code sent to your email or with your Google account. We only show the name you choose, never your email.