Key Findings
A new machine-learning approach from MIT has broken the “predictability desert,” significantly improving weather forecasts for the two-to-six-week window. In an international competition against the European Center for Medium-Range Weather Forecasts (ECMWF), the MIT model emerged as the winner, proving superior to climatological averages. The system specifically boosted the performance of debiased dynamical models, improving accuracy for 98% of precipitation targets, 92% of temperature targets, and 91% of pressure targets.
Technical Details
The researchers implemented a “probabilistic bias correction” method. Rather than replacing existing dynamic models—which simulate the ocean and atmosphere—this AI acts as a postprocessing layer or “wrapper” that learns from past errors to correct systematic biases. Unlike standard bias corrections that smooth forecasts toward an average value, this method adapts to and improves the prediction of extreme weather events such as cold snaps and floods. While the system was developed using the MIT Engaging supercomputing cluster, the final system does not require a supercomputer to operate.
Background & Context
Historically, meteorologists viewed predictions beyond 14 days as highly unreliable due to chaos theory, a concept pioneered by MIT professor Edward Lorenz. While AI models often outperform dynamic models in short-term ranges, they typically underperform in longer ranges. The subseasonal range (2–6 weeks) has traditionally been compared to “rolling dice” because atmospheric noise typically overwhelms the signal. However, researcher Judah Cohen suggests that the atmosphere tends to fall into recurring patterns, allowing for the prediction of subsequent “dominoes” once an event occurs.
Significance & Outlook
Improved subseasonal accuracy provides critical lead time for emergency planning and resource mobilization. Key applications include:
- Energy: Stabilizing power grids by planning for ice storms and heat waves.
- Agriculture: Optimizing planting, fertilizing, and harvesting schedules to increase yields; Kenya’s meteorological office has expressed interest in using the model for monsoon rains.
- Emergency Management: Mobilizing resources prior to hurricanes or floods to protect lives and property.
Source: https://www.techexplorist.com/ai-breakthrough-weather-forecasts-accurate-six-weeks/105782/
Get our weekly technology intelligence — free
Receive an infographic that lets you judge at a glance whether each field’s analysis report is worth reading.
Subscribe Free — Weekly Tech Intelligence
By subscribing, you’ll receive Troy-Technical’s weekly technology intelligence newsletter.
- Your email and selected fields are used only to deliver the newsletter.
- We never share your information with third parties.
- You can unsubscribe anytime via the link in each email.
See our Privacy Policy for details.
Takes about a minute · Unsubscribe anytime
