Researchers led by MIT have developed an AI-based method that makes weather forecasts two to six weeks ahead more accurate.

Added on top of existing forecasting systems, the open-source machine learning layer learns each model's past errors and corrects its predictions accordingly, doubling forecast accuracy for that time frame when applied to the European Centre for Medium-Range Weather Forecasts (ECMWF) model. Published in Nature, the work could support planning in areas such as agriculture, energy and emergency management.