Microsoft Unveils ML Tool That Extends Weather Forecast Accuracy to Six Weeks

Microsoft researchers have released a new machine‑learning tool that aims to extend the reliable window of weather forecasts from the usual two weeks to as far as six weeks.

The system builds on traditional physics‑based models but uses data‑driven techniques to correct systematic errors that usually grow over time. By training on historical weather patterns and satellite data, the model can adjust its predictions in a way that keeps accuracy higher than current methods.

A key benefit is cost. The approach requires fewer computational resources than running full‑scale numerical weather prediction systems, which could make longer‑term forecasts more affordable for regional agencies and private firms. Farmers, for instance, could use the extra lead time to schedule planting or irrigation, while emergency planners could better prepare for storms.

The technology is still in the research phase and has not yet been rolled out commercially. Microsoft says it plans to collaborate with meteorological organizations to validate the model in real‑world settings before a wider release.