Microsoft open-sources PBC tool to improve 2–6 week weather forecasts

Microsoft Research introduced Probabilistic Bias Correction (PBC), a machine learning method that corrects systematic bias in existing weather models. Published in Nature and released as open-source software, it improved ECMWF's physics-based model for 98% of precipitation and 92% of temperature targets and runs on CPUs.

• PBC improved ECMWF's model for 98% of precipitation and 92% of temperature targets
• The method doubled the accuracy of ECMWF's AI forecasting system in the study
• The tool runs on CPUs instead of supercomputers or GPUs
• The open-source method is being tested in Kenya with the national weather service