Microsoft announced an open‑source machine‑learning tool that corrects systematic errors in existing weather models. The method, called Probabilistic Bias Correction (PBC), lets agencies produce more reliable forecasts two to six weeks out. It runs on ordinary CPUs, not on supercomputers or high‑end GPUs.
How PBC improves forecasts
PBC learns the typical bias of a weather model—like a tendency to predict colder temperatures because the model was trained on older climate data—and adjusts the output. In tests, it raised the accuracy of the European Centre for Medium‑Range Weather Forecasts (ECMWF) physics‑based model for 98 % of precipitation targets and 92 % of temperature targets. It also doubled the accuracy of ECMF’s newer AI forecasting system.
Why this matters
Because the software is lightweight and free, smaller weather services, especially in developing regions, can adopt it without costly hardware. Better medium‑range forecasts give farmers more time to protect crops and help emergency responders prepare for extreme weather events.
Why it matters
For everyday people, more accurate forecasts two to six weeks ahead can mean earlier warnings for storms or droughts, giving farmers and local officials a better chance to act. The low‑cost, open‑source nature of the tool means it could reach regions that previously lacked the computing power for advanced weather models.
