Microsoft releases Aurora 1.5: a model for weather and climate forecasting

Microsoft Research has unveiled Aurora 1.5 — an expanded version of the open Aurora foundational model for weather prediction, climate modeling, and energy applications. The update adds 22 new meteorological variables, hourly temporal resolution (instead of daily), and probabilistic ensemble forecasting, making the model more accurate and useful for real-world applications in science and industry.

What Aurora 1.5 expanded

Aurora 1.5 builds on the foundational Aurora model, developed by Microsoft Research for modeling the Earth system. The new version improves functionality across three key parameters. First, the number of tracked meteorological variables has increased to 22 — sea level and altitude pressure, air temperature, humidity, precipitation, wind speed and direction, solar radiation, and other climate parameters. Each variable stores information about atmospheric conditions at specific points, allowing the model to capture local phenomena.

Key improvements:

• 22 additional meteorological variables for comprehensive local forecasting coverage
• Hourly temporal resolution instead of daily for high detail
• Probabilistic ensemble forecasting — multiple scenarios with uncertainty assessment
• Released as an open-source model for the research community

Second, Aurora 1.5 operates at hourly temporal resolution rather than daily. This fundamentally changes practical utility: it can predict local events hours in advance — thunderstorms, wind gusts, sudden frosts. Daily forecasts interest climatologists; hourly ones are needed by meteorologists, aviation, and energy companies. Third, Aurora 1.5 generates not one deterministic forecast, but a probabilistic ensemble: multiple possible weather scenarios with uncertainty assessment. This more realistically reflects the nature of forecasting — the future is uncertain, and the model shows a range of possible outcomes.

What Aurora 1.5 is needed for

The expanded model targets three main directions. In meteorology and natural disasters, Aurora 1.5 performs local forecasts with hourly resolution — critical for warning about hurricanes, tornadoes, floods, frosts. Aviation, maritime, and utility services can use such forecasts for planning and safety.

In climate research, Aurora helps study long-term climate change trends, the impact of greenhouse gases, glaciers, and ocean currents on the Earth system. Scientists can run multi-decade simulations using the open model, without dependence on corporate cloud services.

In energy, Aurora predicts solar and wind generation with hourly resolution. A grid operator knows how the sun will shine and the wind will blow over the next 24-48 hours, and optimizes activation/deactivation of renewable sources, backup capacity, and batteries. Against the backdrop of the transition to green energy, this is a critical task.

The open release of Aurora 1.5 is a decisive advantage. Researchers and companies can download the model, fine-tune it on local data, and deploy it without dependence on Microsoft cloud services. This is especially important for developing countries, where cloud infrastructure is expensive and climate risks are high.

What this means

Aurora 1.5 reflects a broader trend in AI: the shift from universal large language models (ChatGPT, Claude) to specialized foundational models for specific domains. If ChatGPT is a universal assistant, then Aurora is a deep specialist in Earth systems, capable of working with atmospheric physics and climate equations at a level accessible to researchers and engineers.

Microsoft's open approach allows industry and science to build their own solutions on a stable foundation: local forecasts in African countries, wind farm optimization in Scandinavia, climate modeling in Brazil. This accelerates AI adoption in climate and energy sectors, where accuracy and algorithm transparency are critical.