Microsoft researchers have released an open-source machine learning tool called Probabilistic Bias Correction that improves the accuracy of long-range weather forecasts by correcting systematic errors in existing models.
Key Points
- The new tool, published in the journal Nature, enhances forecast reliability for the two-to-six-week window.
- Probabilistic Bias Correction (PBC) improves the accuracy of European Centre for Medium-Range Weather Forecasts (ECMWF) models for 98 percent of precipitation targets.
- Unlike resource-heavy generative AI models, the PBC framework is lightweight and can run on standard CPUs.
- The software is open-source, aiming to provide affordable, high-quality forecasting tools to agencies in resource-constrained regions.
- Researchers from Microsoft, MIT, and the ECMWF collaborated on the project to address climate-related biases in historical weather data.