Microsoft AI Tool Can Predict Weather Up to 6 Weeks Ahead
Microsoft Research, in collaboration with MIT, has unveiled the Probabilistic Bias Correction (PBC) AI tool, detailed in a Nature paper published today, October 14, 2026. This open-source and cost-effective machine learning framework dramatically extends reliable weather forecast accuracy from two weeks to an unprecedented six weeks, offering substantial.
In a significant leap for meteorological science, a team led by Microsoft Research, with contributions from MIT, has introduced a pioneering machine learning tool named Probabilistic Bias Correction (PBC). Published today, October 14, 2026, in the esteemed journal Nature, this innovative framework promises to extend accurate weather forecasting from the traditional two-week limit to an extraordinary six weeks ahead. This development marks a pivotal moment for global communities and industries heavily reliant on long-range weather predictions.
Solving the Subseasonal Prediction Challenge
For years, both traditional physics-based numerical weather prediction (NWP) models and even advanced AI models have struggled with a steep decline in forecast skill beyond the two-week mark. This challenge, known as subseasonal prediction, has been a major barrier for effective planning against severe weather phenomena. Microsoft's PBC directly confronts this issue by operating on probabilistic forecasts rather than raw observations, reducing systematic errors and significantly improving reliability at lead times of two to six weeks. OpenAI Claims Ex-Employee Transferred Files on Unreleased Product to Apple.
The research, spearheaded by a 13-author team including Microsoft Research's Hannah Guan (who conducted the research as an intern), Soukayna Mouatadid, and Lester Mackey, alongside MIT's Judah Cohen, underscores a cost-effective approach that requires no supercomputer, making it accessible even for resource-constrained regions like Kenya.
Unprecedented Accuracy Gains
The impact of PBC on forecast accuracy is substantial. When applied to the European Centre for Medium-Range Weather Forecasts (ECMWF)'s dynamical ensemble, which is widely considered the global benchmark, the PBC framework showcased remarkable improvements:
- Precipitation Skill: Tripled the raw model's positive precipitation skill.
- Temperature & Pressure: Converted negative-skill temperature and pressure forecasts into skillful predictions.
- Extreme Events: For week-3 extreme highs (2016–2024), it boosted temperature skill by 51% and pressure skill by 39%.
- Flood Forecasting: Enhanced week-3 flood forecast skill by 76–92% across 2,435 flood events catalogued between 2016 and 2026.
This breakthrough also earned recognition in a major international forecasting competition. An ensemble combining PBC-ECMWF and PBC-PoET, named MicroDuet, won the 2025 ECMWF AI Weather Quest competition, placing first for all weather variables and lead times, outperforming 34 competing teams worldwide. The core idea involves applying two machine learning models – Debias++ and Persistence++ – in parallel to correct and blend forecasts with historical data and climatology, resulting in more robust predictions.
Microsoft's Expanding Horizon in AI Weather
This latest development complements Microsoft's broader commitment to leveraging AI for Earth system forecasting. The company previously introduced Microsoft Aurora, a 1.3 billion-parameter AI foundation model for the Earth system, developed by Microsoft Research and published in Nature in May 2025. Aurora focuses on medium-range forecasts (up to 10 days) and excels not just in general weather prediction but also in areas like ocean waves, air quality, and tropical cyclone tracking.
An updated version, Aurora 1.5, released by Microsoft Weather in July 2026, further enhanced capabilities with 22 new forecast variables, hourly resolution, and probabilistic ensemble forecasting for uncertainty quantification. Aurora is remarkably efficient, producing a 10-day global forecast in mere seconds on a single GPU, approximately 5,000 times faster than traditional numerical weather prediction systems. Epic Games Plans Fortnite Linux Release via Easy Anti-Cheat Overhaul.
Both Aurora and the new PBC tool are open-source, allowing researchers and developers globally to build upon these advancements. PBC’s low computational demand, coupled with its long-range accuracy, holds immense promise for nations like India, where agricultural planning, disaster management, and infrastructure resilience are critically dependent on reliable forecasts. The ability to predict extreme heatwaves, monsoons, or floods weeks in advance will empower communities to prepare more effectively, mitigate risks, and potentially save lives.
Disclaimer: AI tools assisted in compiling the foundational data and research for this report. The final content was reviewed, edited and verified by human editors at LatestLY.
(The above story first appeared on LatestLY on Oct 14, 2026 10:24 PM IST. For more news and updates on politics, world, sports, entertainment and lifestyle, log on to our website latestly.com).
