Microsoft's Aurora 1.5 adds 22 weather variables and ensemble forecasting
New open-source model improves hourly forecasts with probabilistic uncertainty quantification
Aurora 1.5 represents a significant upgrade to Microsoft's open Earth-system foundation model, originally introduced in 2024 and published in Nature. The model now covers 22 additional weather variables—including solar radiation, total cloud cover, and humidity fields—expanding from just 4 original variables. It also achieves hourly temporal resolution, enabling finer-grained forecasts for precipitation onset or storm tracking. A key new capability is probabilistic ensemble forecasting, which runs multiple simulations to represent model uncertainty and provide a range of possible outcomes. This addresses a frequent user request and improves decision-making under uncertainty. The model is released as open source on GitHub with checkpoints on Hugging Face, allowing researchers and developers to evaluate, adapt, and build upon it.
Beyond open access, Microsoft Weather connects Aurora 1.5 to enterprise-grade infrastructure, data, and operational support. This bridges frontier research with practical applications for sectors dependent on accurate weather intelligence—energy traders, agricultural planners, transport operators, and climate risk analysts. By offering managed access and decision-support capabilities, Microsoft positions Aurora 1.5 as a scalable solution for organizations that require reliable, confident weather predictions. The update underscores a trend toward making foundation models more useful in high-stakes environments, combining the transparency of open science with the reliability of cloud-based enterprise services. As climate risks grow, tools like Aurora 1.5 help improve preparedness and resilience.
- Adds 22 new weather variables (e.g., solar radiation, cloud cover, humidity) to the original 4, broadening relevance for energy, agriculture, and transport sectors.
- Introduces hourly temporal resolution and probabilistic ensemble forecasting to quantify forecast uncertainty, a top-requested feature.
- Released open-source on GitHub and Hugging Face, with enterprise-grade managed access and infrastructure via Microsoft Weather services.
Why It Matters
Makes advanced weather forecasting more accessible and practical for critical sectors like energy, agriculture, and climate resilience.